# Robozaps Blog Site: https://blog.robozaps.com Generated: 2026-10-04T04:34:14.703Z Posts: 141 This file is a plain-text export of every published post on this blog, structured for ingestion by AI crawlers and search assistants. Each post is delimited by a line of three dashes. # This Week in Humanoid Robot News: August 24–31, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-august-24-31-2026 Author: Radica Maneva Published: 2026-08-30T22:52:14.265Z Updated: 2026-08-30T22:58:45.120Z Last fact-checked: 2026-08-31T00:00:00.000Z Category: news Key takeaways: - XPENG's robotics unit raised over US$900 million at a valuation above US$6.3 billion (Aug 24), the largest single private embodied-AI round in China; IRON mass production is targeted for end-2026. - At Beijing's second World Humanoid Robot Games, Tiangong Ultra won the 100m in 8.64 seconds (Aug 26), an organizer-timed, controlled-course result, not a real-world capability benchmark. - The week's news was again China-heavy; no new verified US or European deployment, funding round or order-page change cleared a reputable source. - Context: reports this month put Chinese makers at about 97 percent of first-half 2026 global humanoid shipments, the backdrop to both stories. - Register status is unchanged: 98 platforms tracked, 13 purchasable today, 28 pre-ban FCC authorizations across 10 makers, 0 Conditional Approvals since the July 28 Covered List change. FAQ: Q: Did a humanoid robot really beat Usain Bolt's 100m record? A: In a controlled robot race at the second World Humanoid Robot Games in Beijing, Tiangong Ultra, built by X-Humanoid, won the 100-metre final in 8.64 seconds on August 26, 2026, faster on paper than Usain Bolt's 9.58-second human world record. It was an organizer-timed run on the games' own course, not a like-for-like comparison with human sprinting, and it says little about whether a robot can do useful work. Q: How much did XPENG raise for its robotics business? A: XPENG said on August 24, 2026 that its robotics unit raised over US$900 million at a post-money valuation above US$6.3 billion, which it called the largest single private financing round in China's embodied-AI sector. The round was led by IDG Capital, with Gaorong Ventures taking part and Tencent and Alibaba backing it as strategic investors. Q: Can you buy the XPENG IRON humanoid now? A: No. XPENG targets mass production of IRON by the end of 2026, with commercial deliveries starting in 2027. There is no public consumer price or order page yet. Q: Which humanoid robot won the 2026 World Humanoid Robot Games 100m? A: Tiangong Ultra, developed by Beijing-based X-Humanoid, in 8.64 seconds. The games' 100-metre mark fell from 9.39 to 8.86 to 8.64 seconds across the week. Q: Are any of these humanoids available to buy in the US today? A: Neither of this week's headline robots is. Across the RoboZaps legality register, 13 of 98 tracked platforms are purchasable today, and 28 robots across 10 makers hold an FCC authorization granted before the July 28 Covered List action. This week in humanoid robot news, both the money and the spectacle came out of China. XPENG's robotics unit closed the largest private funding round the country's embodied-AI sector has reported, and at Beijing's second World Humanoid Robot Games a bipedal robot ran 100 metres faster than any human ever has, at least by the games' own clock. Neither event changes what a buyer can order today, and no new US or European deployment cleared a reputable source in the week to August 31. XPENG's robotics unit raises over $900 million XPENG said on August 24 that its robotics business raised over US$900 million at a post-money valuation above US$6.3 billion, which the company called the largest single private financing round in China's embodied-AI industry. The round was led by IDG Capital, with Gaorong Ventures taking part and Tencent and Alibaba backing it as strategic investors, Electrek reported. XPENG says the money will fund its IRON humanoid and its physical-AI models, with mass production of IRON targeted for the end of 2026 and commercial deliveries starting in 2027. For buyers, nothing shipped and nothing opened for order. There is still no public IRON price or order page. This is capital and intent, not a robot you can put on a task. A humanoid wins the Beijing games 100m in 8.64 seconds At the second World Humanoid Robot Games in Beijing, Tiangong Ultra, built by the Beijing developer X-Humanoid, won the 100-metre final in 8.64 seconds as the games closed on August 26. The mark fell through the week, from 9.39 seconds on opening day to 8.86 and then 8.64, all faster on paper than Usain Bolt's 9.58-second human world record. The number needs its context. This was an organizer-timed run on the games' own course, not a like-for-like race against a person, and several robots crashed into the padded barrier past the line, with staff spraying fire extinguishers on the machines that sparked. Jonathan Hurst, a robotics professor at Oregon State University, put it plainly: real progress will be measured in warehouses and factories where humanoids work autonomously for hours, not on a sprint track. More than 2,000 robots competed, across events from sprinting to boxing and housework. More humanoid robot news in brief The backdrop to both stories is scale. Reports published earlier in August put Chinese makers at more than 97 percent of global humanoid shipments in the first half of 2026, on roughly 19,100 units, about triple a year earlier, with AgiBot and Unitree together taking about three quarters of the total.The West stayed quiet. No new verified US or European order page, funding round or named deployment cleared a reputable source this week. Tesla is still installing its first Optimus production lines at Fremont, with output that Elon Musk has said will be slow at first and every unit staying inside Tesla's own factories. From the register The RoboZaps legality register still tracks 98 humanoid and quadruped platforms, of which 13 are purchasable today. Checked against the FCC's own equipment-authorization records, 28 robots across 10 makers hold an authorization granted before the July 28 Covered List action, and the Department of War has granted zero Conditional Approvals for advanced robotic devices since that change, as of July 29, 2026. Nothing in this week's news moves those counts, because neither XPENG's IRON nor a games sprinter is offered for US sale. The full breakdown of which models are purchasable and which hold a pre-ban grant is in the free report, and the running humanoid robot news tracker keeps the current source-checked status. What to watch next week Whether XPENG attaches a public price or order page to IRON as it pushes toward end-of-2026 mass production.Unitree's first moves as a listed company, and whether its experimental Superman running platform, shown two days before the IPO, gets published specifications or third-party testing.Any first US or European deployment, funding round or order-page change, after a quiet week in the West.Progress on the next Chinese listings, with LimX Dynamics and Chery-backed AiMOGA reported to be preparing IPOs. RoboZaps runs buyer-side robot sourcing. A fixed-fee Sourcing Brief answers "which robot for our task" in five business days, with a public sample you can read first: https://robozaps.com/sourcing --- # This Week in Humanoid Robot News: August 17–24, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-august-17-24-2026 Author: Radica Maneva Published: 2026-08-24T09:00:00.000Z Updated: 2026-08-24T23:37:50.584Z Last fact-checked: 2026-08-25T00:00:00.000Z Category: news TL;DR: Unitree debuted on Shanghai's STAR Market on August 19 and jumped about 460 percent, but its robots' prices and order pages did not change. Beijing ran the World Robot Conference and opened the second World Humanoid Robot Games. No new verified US or European order-page change cleared this week. Key takeaways: - Unitree became the first pure-play humanoid maker on a mainland Chinese exchange, closing its August 19 STAR Market debut up about 460 percent from the 150.80-yuan IPO price (Reuters). - The debut is a capital and sentiment event. Unitree's robot prices and order routes did not change. - Beijing hosted the 2026 World Robot Conference from August 19 to 23 and opened the second World Humanoid Robot Games on August 22, with roughly 2,000 humanoids on display. - BYD's Xiao Di humanoid, unverified last week, appeared as a working showroom unit with no price or order route. - No new verified US or European humanoid deployment, funding round or order-page change cleared a reputable source this week. The register still shows 13 humanoids purchasable today and zero FCC Conditional Approvals. - What this means for buyers: fresh IPO capital does not shorten the FCC authorization line, so newly hyped models still cannot ship in the US yet. The 13 humanoids purchasable today stay the only near-term options, so if you are scoping a deployment, get quotes on those now rather than waiting on the headline names. FAQ: Q: Did Unitree's stock market debut change what its robots cost? A: No. Unitree's Shanghai listing on August 19, 2026 raised capital and lifted the share price, but the G1, H2 and R1 still sell through the same sales-assisted channel as before, with no new public price or order page. Q: What happened at the 2026 World Robot Conference and World Humanoid Robot Games? A: Beijing ran the World Robot Conference from August 19 to 23, with more than 300 exhibitors and over 3,000 products, then opened the second World Humanoid Robot Games on August 22 with 666 teams and about 2,056 robots from 16 countries. Both are demonstrations and benchmarks, not order pages. Q: Is BYD's Xiao Di humanoid available to buy? A: No. BYD showed a working Xiao Di unit at its Di Space centre in Zhengzhou, built to greet and sell to showroom visitors, but has published no price, no order route and no ship date. Q: How many humanoid robots can buyers order in the US right now? A: The RoboZaps register lists 13 humanoids purchasable today out of 98 platforms tracked. Twenty-eight robots hold an FCC authorization granted before the July 28, 2026 Covered List change and stay legal to import and sell; no Conditional Approval has been granted since. Humanoid robot news for August 17 to 24, 2026 was a China week. Unitree turned its Shanghai listing into one of the year's loudest market debuts, Beijing ran the World Robot Conference and opened the second World Humanoid Robot Games, and BYD's showroom humanoid finally appeared as a working unit. None of it changed what a buyer in the US or Europe can order today, which is the line this brief keeps drawing. Unitree's Shanghai debut jumps about 460 percent on day one Unitree began trading on Shanghai's STAR Market on August 19 under ticker 688836, the first pure-play humanoid maker to reach a mainland Chinese exchange. Against an IPO price of 150.80 yuan, the shares opened near 1,100 yuan and closed the first day at 845 yuan, up about 460 percent, according to Reuters. Intraday the stock touched 1,100 yuan, a gain of as much as 629 percent, which the South China Morning Post tied to a valuation near 66 billion US dollars, before it settled back toward 50 billion at the close. The sale itself raised about 6.1 billion yuan, or 904 million US dollars, per Bloomberg. What it means: this is a sentiment event, not a product event. Founder Wang Xingxing's paper stake passed 12 billion US dollars, and the listing hands Unitree fresh capital and a public currency for deals. It does not move a single price or order page. The G1, H2 and R1 still sell through the same sales-assisted channel they did last week, so the debut says more about how the market is pricing robotics than about what the robots can do. The buyer-facing detail sits in the humanoid robot stocks guide. Beijing's World Robot Conference fills the halls with humanoids The 2026 World Robot Conference ran in Beijing from August 19 to 23. State media reported more than 300 exhibitors from 26 countries, over 3,000 products and more than 300 debuts, with humanoids the main draw for overseas buyers on the floor (Xinhua). Chinese outlets showed machines pouring drinks, sorting parts and handing flowers to visitors. What it means: an expo hall is a demonstration environment, and several units visibly stumbled or froze during live runs. Treat the specifications and the choreography as vendor material until a third party tests a production unit off the show floor. A crowded conference is a signal about Chinese supply, not about what has a US order page, where the July FCC Covered List still blocks new foreign models without a pre-ban authorization. The second World Humanoid Robot Games open in Beijing Two days after the conference, Beijing opened the second World Humanoid Robot Games on August 22 at the National Speed Skating Oval, with the event scheduled through August 26. Organizers reported 666 teams and 2,056 robots from 16 countries across dozens of events, from sprints and football to new additions such as tug of war and weightlifting (CGTN). The 2025 edition drew about 280 teams and 500 robots, so the field roughly quadrupled in a year. What it means: the Games measure narrow athletic and task skills under staged conditions, not workplace readiness or safety. Some write-ups claimed a robot beat a human sprint world record. That figure traces only to secondary summaries and we could not confirm it against official results, so it is left out. The real signal is scale. China can field two thousand humanoids for a week of public benchmarking, which is a manufacturing story before it is a capability one. More humanoid robot news in brief BYD's Xiao Di humanoid, flagged here last week as an unverified plan, has now appeared as a working unit at BYD's Di Space centre in Zhengzhou. Reported specs are about 1.61 metres, 58.5 kilograms and 31 degrees of freedom, with real-time translation across several Chinese dialects and foreign languages (SCMP). Its stated job is to greet and sell on car showroom floors, and BYD has published no price or order route.No new verified US or European humanoid deployment, funding round or order-page change cleared a reputable source this week. Reports of a Samsung in-house humanoid programme and of an airline barring robots from cabins were single-sourced, so they are held for a second source.Several "August milestone" items on aggregator sites, including restated Apptronik and Figure numbers, trace back to earlier 2026 announcements rather than to this week. From the register The RoboZaps register tracks 98 humanoid platforms this week, of which 13 can be paid for today. Everything else is a preorder, a pilot, or an announcement. Checked against the FCC's own Equipment Authorization System, 28 robots in the register hold an authorization granted before the July 28 Covered List change and stay legal to import and sell, spread across 10 makers including Unitree, AgiBot, AgileX, Fourier, LimX, Petoi, Pudu, Robosen, Leju and Keenon. Conditional Approvals granted for advanced robotic devices since that change remain at zero. The free report lays out the same numbers with sources: https://robozaps.com/report. What to watch next week Whether Unitree's share price holds after the debut spike, and whether the listing changes how its robots are sold rather than only how the stock is priced.Official results and any credible performance data from the World Humanoid Robot Games once the event closes on August 26.Whether any producer files for, or the Department of War grants, the first Conditional Approval that would let a new foreign humanoid back onto the US market.Whether BYD attaches a price or a ship date to Xiao Di, which would move it from demo to product. RoboZaps runs buyer-side robot sourcing. A fixed-fee Sourcing Brief answers "which robot for our task" in five business days, with a public sample you can read first: https://robozaps.com/sourcing. --- # This Week in Humanoid Robot News: August 10–17, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-august-10-17-2026 Author: Radica Maneva Published: 2026-08-16T22:55:20.863Z Updated: 2026-08-16T23:05:39.305Z Last fact-checked: 2026-08-17T00:00:00.000Z Category: news Key takeaways: - Unitree opened its Shanghai STAR Market IPO subscription on August 10 under ticker 688836; the retail tranche drew record demand, reported by Bloomberg near 5,500 times covered and by Reuters more than 8,000 times oversubscribed. - The IPO priced on August 6 at 150.80 yuan a share, valuing Unitree near 61 billion yuan (about 9 billion US dollars) and raising about 6.1 billion yuan for roughly 10 percent of the company. The trading debut is expected between August 17 and 21 and had not happened as of this brief. - On the issue price Unitree is valued at about 219 times 2025 earnings, while its prospectus guides first-half 2026 revenue growth of 36 to 45 percent, down from 333 percent a year earlier, with adjusted net profit 6 to 22 percent lower. These are company figures. - BYD said it would show a Xiao Di service humanoid in early August at its Di Space centres, but as of August 17 RoboZaps could not verify an independent public unveiling or an official spec sheet; the floated specs remain company claims. - The RoboZaps register stands at 98 humanoid platforms tracked, 13 purchasable today, 5 pre-ban US authorizations across 3 makers, and zero Conditional Approvals since the July 28, 2026 Covered List change. FAQ: Q: What is the latest humanoid robot news? A: In the week of August 10 to 17, 2026, Unitree's Shanghai STAR Market IPO subscription opened on August 10 and drew record demand ahead of a trading debut expected between August 17 and 21. BYD's planned Xiao Di humanoid unveiling stayed unverified, and no new US humanoid deployment or funding round cleared a reputable source. Q: How oversubscribed was Unitree's IPO? A: Reports differ by tranche. Bloomberg put the retail portion near 5,500 times covered and Reuters reported the wider book more than 8,000 times oversubscribed, described as a STAR Market record for a technology company. The online lottery left about 19,414 winning numbers against roughly 9.78 million valid accounts. Q: When does Unitree start trading? A: The debut is expected between August 17 and 21, 2026, under ticker 688836 on Shanghai's STAR Market. As of this brief the shares had not started trading, so any first-day price move is not yet a fact. Q: Has BYD launched its Xiao Di humanoid? A: Not verifiably. BYD told Chinese outlets it would show Xiao Di in early August at its Di Space centres, with floated specs of 1.61 metres, 58.5 kilograms and 31 degrees of freedom, but as of August 17, 2026 RoboZaps could not confirm an independent public unveiling or an official BYD spec sheet. Q: How many humanoid robots can you actually buy in the US right now? A: The RoboZaps register lists 13 of 98 tracked platforms as purchasable today. Only models with a pre-ban FCC authorization (5 grants across 3 makers) or a Conditional Approval can enter the US market, and no Conditional Approval has been granted since the July 28, 2026 Covered List change. Humanoid robot news for August 10 to 17, 2026 turned on one event: the subscription phase of Unitree's Shanghai listing, which drew record demand before the shares have even started trading. It was otherwise a quiet week for hard product and deployment news, so this brief leads with the money and keeps the unverified claims clearly labelled. Unitree's Shanghai IPO subscription breaks a STAR Market record Unitree priced its STAR Market issue on August 6 at 150.80 yuan a share, valuing the company near 61 billion yuan, about 9 billion US dollars, and is selling 40,446,434 new shares, roughly 10 percent of its enlarged capital, to raise about 6.1 billion yuan. The subscription window opened on August 10 under ticker 688836, and demand set a record. Bloomberg put the retail tranche near 5,500 times covered, and Reuters reported the wider book more than 8,000 times oversubscribed, the strongest retail interest a technology company has drawn on the board. The online lottery that followed was correspondingly hard to win. Trade press reported about 19,414 winning numbers of 500 shares each against roughly 9.78 million valid accounts, so the allocation reads as heavily rationed rather than widely held. Strategic and core investors named across the offering include DeepSeek, Tencent's Qishan Investment, PetroChina's Kunlun Capital, Shoucheng and Meituan. What it means: this is the first pure-play humanoid maker to reach a mainland exchange, and the demand is a sentiment signal for the whole sector, not a product event. Nothing here changes what a buyer can order today or at what price. Check any figure against our humanoid robot price page and the Unitree H2 record before it goes into a budget. The trading debut is still ahead The listing itself had not happened as this brief went out. The debut is expected between August 17 and 21, and any first-day move belongs in next week's issue, not this one. The caution worth carrying into it is valuation. On the issue price the company is valued at about 219 times 2025 earnings, while its own prospectus guides first-half 2026 revenue growth of 36 to 45 percent, down from 333 percent a year earlier, with adjusted net profit 6 to 22 percent lower. Those are company and prospectus figures. The price is set on the story more than the current trend line. BYD's Xiao Di humanoid stays unverified BYD told Chinese business outlets, including the South China Morning Post, that it would show a service humanoid called Xiao Di in early August at its Di Space experience centres in Zhengzhou. Pre-launch descriptions put it at 1.61 metres, 58.5 kilograms and 31 degrees of freedom, with real-time translation across six Chinese dialects and six foreign languages. As of August 17, RoboZaps still cannot confirm an independently documented public unveiling or an official spec sheet from BYD. Read it as a company plan, not a shipment, and treat the specs as claims until BYD publishes them. More humanoid robot news in brief The week produced no verified new US humanoid deployment or funding round that cleared a reputable source. Several widely shared "August" milestones on aggregator sites trace back to earlier dates on inspection, so they are held out here rather than restated.The US legality picture did not move. No new model gained a US path this week, which keeps per-model FCC status the first question of any purchase. Our explainer on the FCC rule and the page on which humanoids are legal in the US carry the detail. From the register The RoboZaps register lists 98 humanoid platforms this week, of which 13 can be paid for today. Everything else is a preorder, a pilot, or an announcement. The record of pre-ban US authorizations covering full-scale humanoids still stands at 5 grants across 3 makers, and zero Conditional Approvals have been granted since the FCC added advanced robots to its Covered List on July 28, 2026. That last number is the one to watch, because a Conditional Approval is the only route a new foreign-made model has back into the US market. The full breakdown, with FCC status checked per model, is in the free report at robozaps.com/report. What to watch next week Unitree's first trading day, expected between August 17 and 21, and whether the debut price matches the subscription demand.Whether BYD produces an actual Xiao Di unveiling with a published spec sheet, rather than a teaser.The first Conditional Approval application or grant under the FCC's Covered List regime, which would change the register's zero. RoboZaps runs buyer-side robot sourcing. A fixed-fee Sourcing Brief answers "which robot for our task" in five business days, with a public sample you can read first: https://robozaps.com/sourcing --- # This Week in Humanoid Robot News: August 3–10, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-august-3-10-2026 Author: Radica Maneva Published: 2026-08-09T22:45:00.000Z Updated: 2026-08-09T22:54:11.714Z Last fact-checked: 2026-08-10T00:00:00.000Z Category: news TL;DR: Unitree priced its Shanghai IPO at 150.80 yuan on August 6, valuing it near 61 billion yuan, about 9 billion US dollars and 45 percent above the market's estimate, with subscriptions opening August 10. Legal analyses showed the FCC robot ban turns on a Buy American content test, not country of origin. Nucleus Robotics left stealth selling humanoid labour by the hour. Key takeaways: - Unitree priced its Shanghai STAR Market IPO at 150.80 yuan a share on August 6, 2026, valuing it near 61 billion yuan or about 9 billion US dollars, roughly 45 percent above the 104 yuan the market expected. Subscriptions open August 10 and listing is expected around August 20. - The price works out at about 219 times 2025 earnings, while Unitree's own prospectus guides to first-half 2026 revenue growth of 36 to 45 percent, down from 333 percent a year earlier, with adjusted net profit 6 to 22 percent lower. - The FCC's Covered List rule turns on a Buy American component-cost test rather than the maker's nationality, so a US-headquartered maker with an offshore bill of materials can be caught. Conditional Approval applications are due to the Department of War by January 1, 2028. - Nucleus Robotics left stealth on August 7 selling humanoid labour by the hour, with a first German factory deployment completed inside 90 days under intensive human supervision for the first three weeks. It has not disclosed its robot platform or how much of the work is autonomous. - HII committed up to 900 million dollars over seven years to Path Robotics and GrayMatter Robotics for autonomous welding and sanding in Navy shipyards, using quadrupeds rather than humanoids, with funding staged against milestones. FAQ: Q: What is the latest humanoid robot news? A: In the week of August 3 to 10, 2026, Unitree priced its Shanghai IPO at 150.80 yuan a share on August 6, legal analyses and IEEE Spectrum set out the real scope of the FCC's robot rule, Nucleus Robotics left stealth on August 7 selling humanoid labour by the hour, IEEE Spectrum profiled Toyota spinout Walden Robotics on August 3, and HII signed robot welding and sanding agreements worth up to 900 million dollars on August 6. Q: How much is Unitree worth after its IPO pricing? A: Unitree priced at 150.80 yuan a share on August 6, 2026, an issuance market capitalisation of about 61 billion yuan, roughly 9 billion US dollars. It is selling 40,446,434 new shares, 10 percent of its enlarged capital, to raise about 6.1 billion yuan. Subscriptions open August 10 and listing is expected around August 20. That is an offering price set with underwriters, not a traded market price. Q: Which robots does the FCC ban actually cover? A: A device is covered only if it combines ground mobility, remote operation, a combined weight over 4.4 pounds including any dock, environmental sensors, network connectivity of at least 200 kbps and controlling software. Connected vehicles, rail vehicles, drones, uncrewed underwater vehicles, FDA-regulated medical robots and fixed industrial arms are excluded. Whether a device counts as foreign-produced is decided by the Buy American component-cost test, not by the maker's nationality. Q: Has BYD launched its humanoid robot? A: Not verifiably. BYD's Zhengzhou Di Space teased a humanoid called Xiao Di for early August 2026, and pre-launch reports describe it as 1.61 m tall, 58.5 kg, with 31 degrees of freedom and real-time translation across six Chinese dialects and six foreign languages. As of August 10, 2026 RoboZaps could not verify that a public unveiling has taken place, and BYD has published no spec sheet. Q: Did Figure 03 really climb a ladder autonomously? A: Figure CEO Brett Adcock posted a video on August 1, 2026 describing Figure 03 climbing a ladder fully autonomously. That is a company claim. Figure has published no success rate and no technical breakdown, and the demonstration has not been independently verified. Humanoid robot news for August 3 to 10, 2026 was mostly about money and rules. Unitree priced its Shanghai listing well above what the market expected, the fine print of America's new robot import rule came into focus, and two companies put supervised robots on real factory floors. Vendor figures are labelled as claims throughout. The running humanoid robot news tracker carries the longer record. Unitree prices its IPO 45 percent above the market's estimate Unitree set its STAR Market issue price at 150.80 yuan a share on August 6, valuing the company at about 61 billion yuan, roughly 9 billion US dollars. That is about 45 percent above the 104 yuan consensus reported before book-building, and well above the 42 billion yuan valuation implied when the offering launched on July 31, which is what last week's brief recorded. The mechanics are 40,446,434 new shares, 10 percent of the enlarged capital, raising about 6.1 billion yuan. Subscriptions open today, August 10, and the shares are expected to list around August 20. Strategic placement drew DeepSeek, Tencent's Qishan Investment and PetroChina's Kunlun Capital. What it means: the price is about 219 times 2025 earnings. Unitree's prospectus guides to first-half 2026 revenue of 1.05 to 1.13 billion yuan, growth of 36 to 45 percent against 333 percent a year earlier, with adjusted net profit down 6 to 22 percent. The first listed humanoid pure-play is priced on the story, not the trend line, which is the caveat missing from most humanoid robot stock coverage. Full-year 2025 revenue was 1.708 billion yuan, and humanoids passed quadrupeds as the largest line at 867.8 million yuan. The FCC robot rule's fine print turns out to be a supply-chain test The Covered List addition itself took effect on July 28. This week the detail arrived, in law firm analyses published August 3 and August 7 and in IEEE Spectrum's August 4 report. A device is covered only if it meets every part of the test: ground mobility, remote operation, combined weight with any dock over 4.4 pounds, environmental sensors, at least 200 kbps of network connectivity, and controlling software. Connected vehicles, rail vehicles, drones, uncrewed underwater vehicles, FDA-regulated medical robots and fixed industrial arms sit outside it. The load-bearing phrase is "foreign-produced", which the FCC defines through the Buy American cost test rather than by country. On Sidley's reading, a device is covered unless domestic components exceed 65 percent of total component cost, a threshold that rises to 75 percent in 2029. It is a supply-chain rule: a US-headquartered maker can fail it and an offshore maker can pass. Conditional Approval applications go to the Department of War by January 1, 2028 and must carry ownership detail, a bill of materials, software provenance and a time-bound US onshoring plan. A separate blanket waiver lets already-authorised devices take qualifying software and firmware updates until at least January 1, 2029. Our FCC robot ban and humanoid robots legal in the US explainers are checked against this. IEEE Spectrum canvassed the industry on August 4. Boston Dynamics vice-president of policy Brendan Schulman called the rule "just the first round in a series of policies that will define the success and growth of the industry for decades to come". Ghost Robotics CEO Gavin Kenneally said active and purposeful spyware is deployed inside the US on Chinese robots, which is his company's position rather than an FCC finding. Nucleus Robotics sells humanoid labour by the hour German startup Nucleus Robotics came out of stealth on August 7, with a first deployment at an unnamed German industrial customer completed in under 90 days. Founder Melvin Schwarz frames the product as physical labour billed by the hour rather than a robot sale. The company's own site describes the goal as making humanoid labour an industrial commodity. The useful detail is the supervision. Robots ran under intensive human oversight for the first three weeks, and Nucleus treats that oversight as a data source rather than a temporary crutch. It has not disclosed which humanoid platform it deploys, how often a human intervenes, or what share of the work is autonomous today. Read the deployment as real and the autonomy as unquantified. Several humanoid robot companies now use the same model to get past the pilot stage. Walden explains why its Toyota-backed humanoid runs on wheels IEEE Spectrum published a close look at Walden Robotics on August 3, three weeks after the Toyota Research Institute spinout left stealth on 300 million dollars at a 1.1 billion dollar valuation. Its robot has arms and a sensor head on a wheeled base rather than legs, and two-finger grippers rather than dexterous hands. CEO Russ Tedrake's argument is that factories already run autonomous mobile robots, so a wheeled base inherits the existing safety framework along with stability and battery volume. The robots have been doing production work at a Toyota plant in North America since February. Specific tasks and customers stay confidential, the usual ceiling on this kind of claim. The trade-off recurs across the deployable humanoid field: legs buy terrain, wheels buy uptime. HII commits up to 900 million dollars to welding and sanding robots Shipbuilder HII signed performance-based production agreements worth up to 900 million dollars over seven years with Path Robotics and GrayMatter Robotics on August 6, covering autonomous welding and robotic sanding for Navy carriers, submarines, destroyers and unmanned surface vessels. Funding is staged against technology-readiness milestones, so the headline number is a ceiling, not a commitment. No humanoids are involved. Path Robotics pairs its Obsidian model with quadrupeds to weld outside fixed cells, which remains the more common shape for this work. HII says it plans to outsource more than 2.5 million shipbuilding hours in 2026. More humanoid robot news in brief August 7: Avatar Robotics raised a 6.5 million dollar seed led by AlleyCorp for teleoperated semi-humanoid mobile manipulators. It claims its robots have helped ship more than 900,000 products across live customer sites, a figure we could not independently check.August 7: The Robot Report published the first detailed write-up of Tacta Systems' TactaBot, a hand with 15 independently actuated joints and sub-grain-of-sand tactile sensors. The announcement was July 27 and shipping is slated for early 2027, so treat feeds dating it to this week with care.August 9: Analog Devices returned as a strategic sponsor of MassRobotics, the Boston non-profit whose residency has hosted much of the region's robotics startup base. What to watch next week Unitree's first trading day. Subscriptions open August 10 and listing is expected around August 20. The open is the first public mark on a humanoid pure-play, and the reference point every rival's next raise gets measured against.BYD's showroom robot. BYD's Zhengzhou Di Space teased a humanoid called Xiao Di for early August, reported at 1.61 m, 58.5 kg and 31 degrees of freedom. As of August 10 we could not verify a public unveiling, and the circulating specs are pre-launch media reports rather than a BYD spec sheet.Figure's ladder claim. Brett Adcock posted a video on August 1 showing Figure 03 climbing a ladder, described as fully autonomous. Figure has published no success rate or technical breakdown, and nothing has been verified independently since. Prices and availability move faster than most coverage admits. Check any figure here against our humanoid robot price page and the Unitree H2 record before it goes into a budget. --- # This Week in Humanoid Robot News: July 27–August 3, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-july-27-august-3-2026 Author: Radica Maneva Published: 2026-08-02T22:45:00.000Z Updated: 2026-08-02T22:55:50.445Z Last fact-checked: 2026-08-03T00:00:00.000Z Category: news TL;DR: The FCC added foreign-produced humanoids and quadrupeds to its Covered List on July 28, blocking new models from FCC authorization while leaving existing owners and already-authorized models untouched. Unitree launched a 4.2 billion yuan Shanghai IPO on July 31, Google DeepMind extended Gemini Robotics to whole-body control, and China took the top five humanoid patent portfolios. Key takeaways: - The FCC added foreign-produced humanoids and quadrupeds to its Covered List on July 28. New models cannot receive FCC equipment authorization; models authorized earlier, devices already owned and federal government use are unaffected. - Unitree launched its Shanghai STAR Market IPO on July 31, selling 40.45 million shares, 10 percent of its enlarged capital, to raise about 4.2 billion yuan, with pricing on August 6 and subscriptions from August 10. - Google DeepMind released Gemini Robotics 2 on July 30, the first model in the family to run legs, torso, arms and hands under one policy, demonstrated on Apptronik's Apollo 2. - A LexisNexis Patent Asset Index ranking published July 29 put Chinese startups in the top five humanoid patent portfolios, with Figure AI sixth, Apptronik eighth and Agility Robotics tenth. - Tau Robotics began a $30-an-hour humanoid cleaning service in San Francisco on July 28, with a background-checked remote operator driving the robot and every session recorded as training data. - BYD confirmed on July 28 that it will unveil its first humanoid robot in early August, with no name, price or specification published yet. FAQ: Q: What is the latest humanoid robot news? A: In the week of July 27 to August 3, 2026 the FCC added foreign-produced humanoids and quadrupeds to its Covered List on July 28, Unitree launched a 4.2 billion yuan Shanghai IPO on July 31, Google DeepMind released Gemini Robotics 2 with whole-body control on July 30, and a LexisNexis ranking published July 29 put Chinese startups in the top five humanoid patent portfolios. Q: Did the FCC ban humanoid robots in the United States? A: Not in the way the headlines suggest. The FCC added foreign-produced advanced robotic devices to its Covered List on July 28, 2026, which blocks new models from receiving FCC equipment authorization. Models authorized before that date can still be imported, marketed and sold, devices already purchased are unaffected, and federal government purchase and use is untouched. Producers can apply to the Department of War for a Conditional Approval that restores the path to authorization. Q: When is the Unitree IPO and how much is it raising? A: Unitree formally launched its STAR Market offering on July 31, 2026. Book-building begins August 5, pricing follows on August 6, subscriptions open August 10 and results are due August 14. It is selling 40.45 million new shares, 10 percent of its enlarged share capital, to raise about 4.2 billion yuan or roughly $620 million, which implies a valuation near 42 billion yuan. Q: What is Gemini Robotics 2? A: Google DeepMind released it on July 30, 2026 alongside Gemini Robotics ER 2 for embodied reasoning and Gemini Robotics On-Device 2 for local execution. It is the first model in the family to control legs, torso, arms and hands under a single policy rather than the upper body alone. DeepMind reports 45.7 to 76.3 percent success on general whole-body manipulation and 36 to 92 percent on multi-finger tasks. Q: Can you buy any of the robots in this week's news? A: No. BYD has published no name, price or specification for its August reveal, and Unitree's IPO does not change its catalogue or order routes. The clearest consumer purchase path remains the 1X NEO at $20,000 to own or $499 per month with a $200 refundable deposit, and no verified customer delivery has appeared yet. The humanoid robot news from July 27 to August 3, 2026 was written in Washington. On July 28 the FCC added foreign-produced humanoids and quadrupeds to its Covered List, closing new models out of the US market. Two days later Unitree, the maker with the most American exposure, launched its Shanghai IPO. For the current commercial status of each major robot, see our humanoid robot news tracker. The FCC blocks new foreign-made humanoids and quadrupeds The FCC updated its Covered List on July 28 to include advanced robotic devices, which it defines as mobile robots such as humanoids and quadrupeds, produced in foreign countries. Connected power inverters went on the same list. The Commission did not start this: under the Secure and Trusted Communications Networks Act it can only act on determinations made by national security agencies, and a White House-convened interagency body issued those first. The action is narrower than the headlines suggest. Covered equipment cannot receive new FCC equipment authorization, a precondition for importing, marketing or selling most electronics in the United States. It does not affect devices anyone already owns, models authorized before July 28, or federal government purchase and use. Producers can apply to the Department of War for a Conditional Approval that restores their path to authorization. No company and no country is named in the notice, and the determination covers foreign production regardless of the producer's nationality. China's foreign ministry answered on July 29, with spokeswoman Mao Ning saying protectionism does not make the United States more competitive. We covered the mechanics in what the FCC ban means for Unitree, and the buyer-side position in which humanoid robots stay legal in the US. Unitree launches a 4.2 billion yuan Shanghai IPO days later Unitree formally launched its STAR Market offering on July 31, two working days after the FCC acted. It is selling 40.45 million new shares, 10 percent of its enlarged share capital, to raise about 4.2 billion yuan or roughly $620 million, which implies a valuation near 42 billion yuan. Book-building starts August 5, pricing follows on August 6, subscriptions open August 10, and results are due August 14. Caixin reports 2025 revenue of 1.7 billion yuan and core net profit of 591 million yuan, humanoids at about 52 percent of revenue, and a registration that cleared in 104 days, the fastest in STAR Market history. A profitable humanoid maker with a public share price does not exist anywhere else yet. Nothing changed for buyers: Unitree's catalogue and order routes are the same, and the US restriction applies to new models rather than listings already authorized. The investor-side view sits in our humanoid robot stocks guide. Google DeepMind gives robots whole-body control Google DeepMind released three models on July 30. Gemini Robotics 2 is the vision-language-action model that drives the body, Gemini Robotics ER 2 handles embodied reasoning, multi-step planning and multi-robot coordination, and Gemini Robotics On-Device 2 runs locally and adapts to a new robot with a few hours of data. The change that matters is scope: this is the first version in the family to run legs, torso, arms and hands under a single policy instead of the upper body alone. DeepMind demonstrated it on Apptronik's Apollo 2 and names Franka Duo, Boston Dynamics and Agile Robots among supported platforms. The published success rates repay a close read: 45.7 to 76.3 percent on general whole-body manipulation, 74.2 to 89.6 percent on gripper pick-and-place, kitting and insertion, and 36 to 92 percent on multi-finger tasks. Unscrewing a light bulb scored 92 percent; screwing one back in scored 36 percent. ER 2 is in Google AI Studio, the other two go to early-access partners. Our Gemini Robotics 2 write-up has the full breakdown. China holds the five strongest humanoid patent portfolios LexisNexis Intellectual Property Solutions published a Patent Asset Index ranking on July 29 covering startups with at least one bipedal walking robot. Chinese companies took the top five places: Fourier, AgiBot, LimX Dynamics, Pudu Robotics and Unitree. Figure AI came sixth, Germany's Agile Robots seventh, Apptronik eighth, Leju ninth and Agility Robotics tenth. Over the past decade Chinese filings roughly doubled in interaction systems and grew fivefold in control and planning. The index weights patents by quality rather than counting them, although American portfolios still score higher per patent. A patent is a claim on future position, not evidence that anything ships. Our breakdown of the ranking and the company directory carry the detail. A San Francisco startup rents out a humanoid cleaner, with a person driving Tau Robotics opened an invite-only cleaning service in San Francisco on July 28, charging $30 an hour for one-hour visits covering vacuuming, surface wiping, trash and light clutter, with ladders, exterior work, biohazards and heavy furniture excluded. A background-checked operator at Tau's San Francisco facility drives the robot remotely, and the cameras record video, depth data, telemetry and every motion input the operator sends, all of which trains Tau's models. This is a data-collection business with a cleaning product attached, not an autonomy milestone. Ken Goldberg of UC Berkeley told ABC7 it is too early and that he would be surprised if a humanoid could usefully clean a house. File it with the other home humanoid options, where the first thing to check is how much human supervision the service assumes. BYD confirms its first humanoid robot for early August BYD told the China Securities Journal on July 28 that it will unveil its first humanoid robot in August, confirming a July 25 teaser from its Zhengzhou Di Space venue that read: early August, a new friend wants to meet you. Di Space sites are hands-on experience centres, so expect a physical prototype rather than a slide deck. The teaser says the robot will not arrive with a pile of specs and jargon, which is another way of saying no name, price or specification exists yet. BYD already runs UBTech Walker S1 units in its own plants; the open question is whether this one is for the factory or the showroom. More humanoid robot news in brief July 28: Humanoid Global, a minority investor in Agility Robotics, restated the Churchill Capital Corp XI combination at a $2.5 billion pre-money equity value, and said Digit holds more than $300 million in multi-year v5 orders subject to contractual milestones plus more than 65,000 operating hours across nine customer facilities. Those are company figures relayed by an investor, not audited results.Three items circulating as this week's news were not: Tesla's Optimus commentary came from the July 22 earnings call, the IEEE-RAS Humanoids conference runs in December, and aggregator feeds re-dated AgiBot's 15,000th unit (June 29) and the Mitsubishi Motors agreement with Highlanders (July 9). What to watch next week August 5 and 6: Unitree's book-building and final IPO price, the first public valuation of a profitable humanoid maker.August 10: Unitree subscriptions open, results on August 14, listing expected later in the month.Early August: BYD's first humanoid robot at Di Space, with no exact date announced.Ongoing: Conditional Approval applications to the Department of War, now the only route to FCC authorization for a new foreign-made humanoid or quadruped.Still unresolved: 1X NEO home deliveries, promised for 2026 at $20,000 to own or $499 per month, with no verified customer delivery yet. --- # China Takes the Top Five Humanoid Robot Patent Spots URL: https://blog.robozaps.com/b/china-humanoid-robot-patent-ranking-2026 Author: Radica Maneva Published: 2026-08-02T06:33:40.434Z Updated: 2026-08-02T06:36:39.984Z Last fact-checked: 2026-08-02T00:00:00.000Z Category: insights TL;DR: Chinese startups hold the first five places in LexisNexis’ humanoid-robot patent ranking and six of the top ten. China leads on portfolio scale and manufacturing volume, while the US retains the highest average patent quality. The ranking measures IP strength, not robot performance. Key takeaways: - Fourier, AgiBot, LimX Dynamics, Pudu Robotics and Unitree take the first five startup positions for humanoid-robot patent strength. - Six of the top ten startups are Chinese; one source-page summary incorrectly says seven. - China held 73% of active morphology-related patent families and 63% of global portfolio strength at the end of 2025. - The US held 5% of those patent families but 11% of portfolio strength and the highest average patent quality. - Patent strength and shipment volume matter, but buyers still need task, uptime, deployment, price and support evidence. FAQ: Q: Which company ranks first for humanoid robot patent strength? A: Fourier ranks first in LexisNexis’ startup-only Patent Asset Index, followed by AgiBot, LimX Dynamics, Pudu Robotics and Unitree. Q: How many Chinese companies are in the top ten? A: Six. The ranking table lists Fourier, AgiBot, LimX Dynamics, Pudu Robotics, Unitree and Leju Robotics. A separate summary line on the source page says seven, but that conflicts with its own table. Q: Does the ranking show which humanoid robot performs best? A: No. It measures the strength of related patent assets. It does not score task success, uptime, deployments, price, integration effort or support. Q: Why are Tesla, UBTECH and Boston Dynamics missing? A: The list is limited to startups. Tesla and UBTECH are public companies, while Boston Dynamics is owned by publicly listed Hyundai Motor Group. Chinese startups occupy the first five places in a new ranking of humanoid-robot patent strength. They take six of the top ten spots overall. That is a serious result, but it is not a league table for the best robot you can buy. The list comes from LexisNexis Intellectual Property Solutions, which analysed more than 26,000 patent families. Its Patent Asset Index rewards both portfolio size and patent quality. It does not score deployments, uptime, task success, price or customer support. The top ten humanoid robotics startups by patent strength The concentration at the top is the notable part. Fourier ranks first, while the Fourier GR-1 is already one of the better-documented Chinese humanoids in RoboZaps’ catalog. AgiBot, LimX, Pudu and Unitree complete an all-Chinese top five. What the ranking actually measures LexisNexis grouped the patents into three technical fields: interaction systems, morphology, and control and planning architectures. The Patent Asset Index then combines portfolio size with Technology Relevance and Market Coverage, also described as Competitive Impact. In plain English, it asks how much relevant intellectual property a company owns and how strong that property appears to be. The country-level numbers are even more lopsided. At the end of 2025, China held 73% of active morphology-related patent families and 63% of global portfolio strength. The US held just 5% of those families but 11% of portfolio strength, with the highest average patent quality. This helps explain why a fast-moving company such as AgiBot and its A2 humanoid can rank so highly while the US still retains a quality advantage per patent. China is producing far more relevant patent families; US portfolios carry more weight on average. There is one error worth flagging on the source page. A summary line says seven of the ten startups are Chinese. The published table and the accompanying detailed paragraph contain six: Fourier, AgiBot, LimX, Pudu, Unitree and Leju. RoboZaps uses the table result: six. What the list leaves out This is explicitly a startup ranking. Public companies and subsidiaries of public companies are outside it, which excludes Tesla, the publicly listed UBTECH and Hyundai-owned Boston Dynamics. LexisNexis separately notes that UBTECH holds about 370 relevant patent families and devotes more than a quarter of its portfolio to the three core humanoid fields. The ranking also says nothing directly about whether a robot can complete a paid shift. A strong patent portfolio may protect important technology or improve a company’s licensing position. It cannot tell a buyer whether the machine will meet a cycle-time target, recover from faults, integrate with existing software or receive competent local service. Patent strength is meeting manufacturing scale The patent result is not happening in isolation. TrendForce projects that Unitree and AgiBot will account for nearly 80% of global humanoid shipments in 2026. The Associated Press, citing Barclays and Omdia, reported that Chinese companies supplied about 85% of global humanoid shipments in 2025, with more than 13,000 units shipped and AgiBot and Unitree above 5,000 each. That makes the Unitree G1 more than a viral demo machine: it sits inside a company with both patent depth and rising production volume. Still, shipment counts can include development units, research platforms and robots not yet delivering a repeatable commercial return. A five-signal scorecard for buyers US companies remain competitive on several of those other signals. Figure is sixth in the patent ranking, and the Figure 03 is being developed around factory and household use cases rather than patent counts alone. Apptronik and Agility Robotics also make the top ten, giving US buyers three domestic startups in the list. The RoboZaps read China’s lead is real. Taking the top five startup positions, controlling most active morphology patent families and building most of the world’s current humanoid volume is more than a publicity cycle. It points to a deep link between research, supply chains and manufacturing. The US quality lead is real too. Its smaller patent pool carries disproportionate strength, and patent rankings do not capture the deployment evidence buyers ultimately need. The sensible conclusion is not that one country has already won. It is that China currently leads on patent scale and production, while individual purchasing decisions still have to be made robot by robot and task by task. --- # Google DeepMind's Gemini Robotics 2 Can Control a Humanoid From Feet to Fingertips URL: https://blog.robozaps.com/b/gemini-robotics-2-humanoid-robot-ai Author: Radica Maneva Published: 2026-07-31T23:22:38.000Z Updated: 2026-08-01T23:02:01.618Z Last fact-checked: 2026-08-01T00:00:00.000Z Category: news TL;DR: Google DeepMind's Gemini Robotics 2 controls a full humanoid, legs included, and Apptronik's Apollo 2 is the demonstration platform. Published success rates run from 92% on unscrewing a light bulb to 32% on sweeping with a dustpan. The strategic story is that Google wants to supply the intelligence layer rather than build the robot. Key takeaways: - Gemini Robotics 2 is DeepMind's first model to control a humanoid's legs, torso, arms and multi-finger hands under a single learned policy. The 2025 generation handled the upper body only. - Apptronik's Apollo 2 is the headline platform. DeepMind also demonstrated the model on bi-arm systems including a Franka Duo, plus Dexmate, Trossen and SO101 research robots. - Published dexterity success rates on Apollo 2 range from 92% for unscrewing a light bulb down to 32% for sweeping into a dustpan. Screwing a bulb back in scores 36%. - Only Gemini Robotics ER 2 is broadly reachable, through Google AI Studio. The whole-body model and the on-device model go to early-access partners. - The strategic read is platform, not hardware. Google supplies intelligence to Apptronik and Boston Dynamics while Figure and Tesla stay vertically integrated. FAQ: Q: What is Gemini Robotics 2? A: Gemini Robotics 2 is a vision-language-action model from Google DeepMind, announced on 30 July 2026. It converts camera input and spoken instructions into motor commands, and it is the first model in the family that can control a humanoid's legs, torso, arms and fingers under one learned policy rather than the upper body alone. Q: Which robot does Gemini Robotics 2 control? A: Apptronik's Apollo 2 is the headline demonstration platform, tested with both SharpaWave and Inspire five-fingered hands. DeepMind also showed the model driving a two-armed Franka Duo with a Robotiq gripper, along with Dexmate, Trossen and SO101 research robots. Boston Dynamics and Agile Robots are acknowledged partners. Q: Can you buy a humanoid robot that runs Gemini Robotics 2? A: Not yet. Gemini Robotics ER 2, the reasoning model, is available through Google AI Studio and in private preview on the Gemini Enterprise Agent Platform. The whole-body vision-language-action model and the on-device model are limited to early-access partners through a trusted tester programme. No commercial licensing terms for robot manufacturers have been published. Q: How reliable is Gemini Robotics 2? A: It depends heavily on the task. On Apollo 2 with SharpaWave hands, DeepMind reports 92% success unscrewing a light bulb but 36% screwing one in, 44% tying a rubbish bag, 40% sealing a ziplock bag and 32% sweeping into a dustpan. Picking objects ranges from 76.3% off a shelf to 45.7% off the floor. None of those figures describe uptime or intervention rates in a working building. Q: Is Google building its own humanoid robot? A: No. Google DeepMind builds the models and partners with hardware makers. Apptronik runs a 90,000 square foot data-collection facility called Robot Park in Austin that feeds Gemini Robotics training, and Boston Dynamics signed an AI partnership in January 2026 to put Gemini Robotics models into the next-generation Atlas. Google DeepMind released Gemini Robotics 2 on 30 July 2026, a vision-language-action model that controls an entire humanoid robot: walking, balancing, crouching, and handling objects with five-fingered hands. Apptronik's Apollo 2 is the demonstration platform. The previous generation drove a robot's upper body. This one runs legs, torso, arms and fingers under a single learned policy, which is what DeepMind means by whole-body control. The strategic story is bigger than the demo reel. Google is not trying to win the humanoid hardware race. Its play is to be the intelligence layer inside robots that other companies build. What Gemini Robotics 2 actually is DeepMind shipped three models, and they do different jobs. Gemini Robotics 2 is the vision-language-action model. It turns camera input and spoken instructions into motor commands, and it can drive full humanoids as well as two-armed robots using either multi-finger hands or conventional parallel grippers. Gemini Robotics ER 2 is the embodied reasoning model, built on Gemini 3.5 Flash. DeepMind describes it as the robot's high-level brain. It holds a conversation, plans multi-step tasks, tracks its own progress, calls tools, and coordinates more than one robot on a shared job. It accepts interleaved text, images, video and audio with a 128k context window. Gemini Robotics On-Device 2 runs locally, without a network connection. Built on Gemini Robotics 1.5 and Google's Gemma models, it adapts to an unfamiliar robot body in a few hours using fewer than 200 example demonstrations. What the demonstrations show The clearest example is mundane on purpose. Told to put the watering can into the green bin on the bottom shelf, Apollo 2 walks across the room, finds the can, picks it up, crouches, and places it where it was asked. Nobody scripted the sequence of movements. Other demonstrations cover unscrewing a light bulb, tying a knot in a rubbish bag, sealing a ziplock bag, sweeping debris into a dustpan, and packing items tightly using grippers. Some run for several minutes. DeepMind also shows two robots working the same task, coordinated by the ER model. The hardware on show includes Apollo 2 fitted with SharpaWave and Inspire hands, a Franka Duo with a Robotiq gripper, and smaller research platforms from Dexmate and Trossen plus the open-source SO101 arm. The success rates Google published DeepMind published task-level numbers, and they are candid about how uneven performance still is. The light bulb pair is the tell. Taking a bulb out succeeds 92% of the time. Putting one back in succeeds 36% of the time. Removal tolerates rough alignment, insertion does not, and the space between those two numbers is roughly where robot dexterity sits today. The other two models publish their own scorecards. ER 2 picks the right moment out of a video 91.3% of the time, with a mean error of 0.96 seconds, but it judges how far along a task is correctly only 57.4% of the time. On-Device 2 shows the sharpest generational jump: adapting to the open-source SO101 arm takes it from 6.7% to 53.3%, and on Dexmate hardware from 24.4% to 75.6%. The previous on-device model finished those same two adaptations at 6.7% and 33.3%. Why this matters more than the demo Humanoid hardware is fragmenting. Apptronik, Figure, Tesla, Unitree, Boston Dynamics, AgiBot and a dozen others are all building bodies. Foundation models are consolidating instead, because training them takes data and compute that most robot makers cannot fund alone. The result is an Android-shaped possibility: many robot bodies, a small number of intelligence suppliers underneath them. Google already demonstrates Gemini across humanoids, two-armed systems, multi-finger hands and grippers, and it signed an AI partnership with Boston Dynamics in January 2026 covering the next-generation Atlas. Read that as an interpretation of Google's multi-platform strategy rather than a proven commercial outcome. No licensing terms have been published and the whole-body model is not on sale. The reality check A demonstration is not a deployment. Several things are worth holding onto before treating this as solved. You cannot buy the thing in the videos. ER 2 is in Google AI Studio and in private preview on the Gemini Enterprise Agent Platform, but the whole-body model and the on-device model go only to early-access partners through a trusted tester programme.Three of the five hand tasks DeepMind published complete less than half the time.Speed is still a problem, and DeepMind says so.DeepMind also states that the on-device model struggles to generalise to unfamiliar tasks and to control robots with many degrees of freedom.Nothing here describes a real building. Uptime, intervention rate, safety incidents and cost per completed task are all absent. On safety, DeepMind introduced ASIMOV-Agentic, a benchmark for agentic safety orchestration and uncertainty resolution, and showed the model detecting a nearby human and bringing the robot to a safe stop. That is real engineering, but a benchmark is not a certification or an insurer's sign-off. What it means for humanoid manufacturers Apptronik gains the most immediately. Apollo is the platform every clip is filmed on, and Apptronik runs Robot Park, a nearly 90,000 square foot facility in Austin that it opened on 30 June 2026, collecting the training data that feeds Gemini Robotics. Boston Dynamics is the other named beneficiary. Its January 2026 partnership puts Gemini Robotics models into the new Atlas perception and reasoning stack, and DeepMind is one of only two customers taking 2026 Atlas units. Figure and Tesla sit on the opposite side of the bet. Both build their own hardware and their own models. If Google's layer wins, vertical integration starts to look expensive. If it does not, they keep the whole margin. Our Figure 03 versus Apollo 2 comparison covers how differently those two software stacks are being built. Smaller manufacturers get the most interesting option: license intelligence rather than fund a frontier model. None of that is real until Google publishes commercial terms. What buyers should take from this Gemini Robotics 2 is good evidence that general robot intelligence is improving faster than most operations teams expect. It is not evidence that you can buy a humanoid, install Gemini, and automate arbitrary work. If you are evaluating a humanoid this year, the questions have not changed. Ask for task-level completion rates on your task, not on a benchmark. Ask for intervention rate per shift. Ask for safety documentation and the deployment support model. Ask for total installed cost rather than unit price. A 92% score on a light bulb tells you nothing about your production line. For where the current machines actually stand, see our ranking of the best humanoid robots, and follow the humanoid robot news tracker for what lands next. Compare on RoboZaps: Apptronik Apollo, Figure 03 and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # How Much Does a Robot Dog Cost? Real Prices From $80 to Six Figures URL: https://blog.robozaps.com/b/robot-dog-cost Author: Radica Maneva Published: 2026-07-31T00:00:00.000Z Updated: 2026-08-01T10:45:27.598Z Last fact-checked: 2026-07-31T00:00:00.000Z Category: insights TL;DR: A robot dog costs $79.99 at the toy end, $499 to $2,899.99 for a home companion, and $1,600 for a Unitree Go2, the cheapest walking quadruped worth owning. Industrial machines run $15,000 to $106,500 through resellers, and Boston Dynamics Spot has no public price. The sticker is not the bill: Unitree excludes freight, and Sony's aibo needs $300 a year once its three included years end. Key takeaways: - The cheapest robot dog worth owning is the Unitree Go2 Air at $1,600, and Unitree lists its prices with tax and freight excluded, so budget past the sticker. - Sony's aibo is $2,899.99 and includes three years of the AI Cloud Plan. Renewal is $300 a year, and the plan is required, so five years of ownership is closer to $3,500. - Real robot pets start at $79.99 with the WowWee Dog-E. The $40 robot dogs in social ads are drop-shipped plush toys, not a genuine budget tier. - Boston Dynamics Spot has no public price. The $74,500 everyone quotes was the 2020 online launch price; the NYPD paid $750,000 for two units in 2023. - Industrial quadrupeds run from about $15,199 for a Unitree Go2-W to roughly $106,500 for the B2 family through US resellers, and the top of the market is quote-only. FAQ: Q: How much does a robot dog cost? A: It depends entirely on whether you want a toy or a machine. Robot dog toys cost $79.99 to $179.99. Home companions run $499 for the KEYi Loona to $2,899.99 for Sony's aibo. A real walking quadruped starts at $1,600 with the Unitree Go2 Air. Industrial models run from about $15,199 to $106,500 through resellers, and the top tier is sold by quote only. Q: What is the cheapest robot dog? A: The cheapest robot pet from a real company is the WowWee Dog-E at $79.99, though it is a toy rather than a robot in the engineering sense. The cheapest robot dog that actually walks and balances itself is the Unitree Go2 Air at $1,600, before tax and freight. Q: How much does Boston Dynamics Spot cost? A: There is no public price. Boston Dynamics sold Spot online for $74,500 in 2020 and that figure is still repeated everywhere, but the company moved to enterprise sales and no longer lists it. What real buyers pay is easier to trace: the NYPD bought two units for $750,000 in 2023, and the LAPD fields a donated Spot valued at about $280,000. Configured units are a six-figure purchase. Q: Does a robot dog need a subscription? A: Sony's aibo does. The AI Cloud Plan is required to set it up and use it, and Sony includes three years with the purchase. After that, renewal is $300 a year. KEYi does not currently require a subscription for Loona, and the Joy for All Pup has no app or account at all. Q: Why are robot dogs so expensive? A: Legs. A walking quadruped needs twelve coordinated actuators, a balance controller running continuously, and enough sensing to place each foot on ground it has not seen before. That hardware and the engineering behind it is what separates a $1,600 Unitree Go2 from a $99 toy that rolls on hidden wheels. Above that, price is driven by payload, ingress protection, battery life and the support contract. Q: How much does the Unitree Go2 cost in total? A: The Go2 Air lists at $1,600, the Pro at $2,800 and the X at $4,500, and Unitree states those prices exclude tax and freight. US buyers have reported shipping in the $399 to $1,000 range, so a Go2 Air realistically lands somewhere around $2,000 to $2,600 once it reaches your door. Q: Are the $40 robot dogs in Facebook and TikTok ads real? A: No. Wuffy and Nicoo follow a documented drop-shipping pattern: anonymous storefronts, AI-generated advertising, and a battery-powered plush toy worth a few dollars wholesale. One buyer found the identical Nicoo item on Temu for under $6. They are not a budget tier of the robot dog market, they are a different product being sold under a robotics brand. A robot dog costs anywhere from $79.99 to well over $150,000, and the gap is not marketing. It is the difference between a toy that rolls on hidden wheels and a machine that places each foot on ground it has never seen. This guide gives the real robot dog cost at every tier, taken from manufacturer stores and named US resellers rather than aggregators, and then covers the part most price pages skip: what you keep paying after the box is open. Prices were checked on 31 July 2026. Where a machine is sold by quote rather than listed, we say so instead of repeating a retail figure that has been dead for years. Robot dog prices at a glance Seven tiers, and the jumps between them are large. The one most guides miss sits between the toys and the petbots: programmable kits like the $289 Petoi Bittle and the $649 Mini Pupper 2, which walk properly and are built to be coded. How much does each robot dog cost? How much does the WowWee Dog-E cost? $79.99. WowWee does not sell it directly, so that figure is the retail price through Best Buy, Macy's and similar. It is an app-connected toy with a light-up face and a personality setup that differs per unit, and stock has been thinning through 2026. It rolls rather than walks. For a child, that distinction matters far less than the price does. How much does the Joy for All Companion Pet Pup cost? $179.99, direct from Ageless Innovation. There is no app, no account and no subscription. It barks, responds to touch and sound, and has a heartbeat you can feel through its side. It was designed for older adults rather than for technology buyers, and it is the one product here that is routinely bought by care homes. How much does KEYi Loona cost? $499, against a $529 list price. Worth knowing that this went up: Loona sold for $429 earlier in 2026, so round-ups still quoting the lower figure are out of date. It is not shaped like a dog and it moves on wheels, but it recognises faces, plays fetch-style games and holds a conversation through a GPT-4o voice mode. KEYi does not currently charge a subscription on top. How much does Tombot Jennie cost? Around $1,500, and you cannot buy one yet. Jennie is a therapy Labrador puppy developed with Jim Henson's Creature Shop, aimed at dementia care, and it is waitlist-only with more than 18,000 pre-orders. Tombot does not list a price on its own site; the $1,500 figure comes from its earlier public statements. Treat it as an indication rather than a quote. How much does Sony aibo cost? $2,899.99, and this is the one where the sticker genuinely misleads. That price includes three years of Sony's AI Cloud Plan. The plan is not optional, because aibo needs it to be set up and to function, and once the three years are gone renewal runs $300 a year. Five years of aibo ownership is therefore about $3,500, not $2,900. Sony also announced in June 2026 that Japanese sales are ending as stock runs out, so this generation is closing. How much does the Unitree Go2 cost? $1,600 for the Air, $2,800 for the Pro and $4,500 for the X. Unitree states plainly that those prices exclude tax and freight, and US buyers have reported shipping in the $399 to $1,000 band, so a Go2 Air lands nearer $2,000 to $2,600 in practice. It is still the cheapest genuinely capable walking quadruped anyone sells, and it is the reason the category stopped being research-only. How much does Deep Robotics Lite3 cost? Roughly $5,000 to $19,000 depending on configuration, through US resellers rather than direct. The spread is mostly sensing: a basic remote-controlled unit sits at the bottom, a LiDAR-equipped research platform at the top. Deep Robotics is the most serious Chinese rival to Unitree, and it competes on payload and ruggedness rather than price. How much do the Unitree Go2-W and B2 cost? The wheeled Go2-W is reseller-listed from about $15,199. The B2 family runs roughly $85,000 to $106,500 through the same channels. The B2 is the one from the viral off-road videos: IP67-rated, around 20 kg of continuous payload, built for inspection and logistics rather than demonstrations. How much does Boston Dynamics Spot cost? There is no public price, and the figure everyone repeats is wrong. Boston Dynamics sold Spot online for $74,500 in 2020, then moved to enterprise sales and stopped listing it. Its shop now sells merchandise. What buyers actually pay is traceable through public procurement: the NYPD bought two units for $750,000 in 2023, and the LAPD fields a donated Spot valued at about $280,000. A configured Spot is a six-figure purchase. Full breakdown of what Spot actually costs buyers: our Boston Dynamics robot dog guide. How much does Deep Robotics X30 cost? Quote only. The X30 is Deep Robotics' industrial flagship, rated IP67 and from minus 20C to 55C, and it is sold into power-grid and tunnel inspection. Reseller estimates put it somewhere above $65,000, but no one publishes a number you can rely on. How much does Ghost Robotics Vision 60 cost? Press reporting puts it around $150,000 to $165,000. It is not sold to civilians. Vision 60 units patrol US Air Force bases, and Ghost Robotics sells into defence and security rather than to the public. What the sticker price leaves out Every tier has a running cost, and they are not the same shape. Freight, tax and duty This is the biggest single gap between listed and landed cost. Unitree's prices carry an explicit note that tax and freight are excluded, and the machines ship from China. On a $1,600 Go2 Air, freight alone can add a quarter to a half of the purchase price again. Anyone budgeting from the sticker will be short. Subscriptions Only one product here has a mandatory one, and it is the expensive one. Sony's AI Cloud Plan costs $300 a year after the included three years, and aibo will not set up without it. KEYi does not currently charge for Loona. The Joy for All Pup has nothing to subscribe to. If you are comparing a $2,899.99 aibo against a $499 Loona, the honest comparison over five years is roughly $3,500 against $499. Batteries and wear parts The toy tier runs on replaceable batteries and that is the whole story. The walking tier is different: a quadruped's battery is a consumable with a finite cycle life, and legs have joints that wear. Budget for a replacement battery within the ownership period on anything you actually walk, and treat feet and joint covers as parts that get used up. Repairs and support At the consumer end you have a warranty and not much else. At the industrial end you are buying a relationship: support contracts, spares availability and response times are part of the price, and they are a large part of why an inspection-grade quadruped costs what it does. This is also where quote-only pricing comes from, because the machine and the support are sold together. Why robot dog prices vary so much Legs are the answer to most of it. A walking quadruped needs twelve coordinated actuators, a balance controller running constantly, and enough sensing to handle ground it has not mapped. That is a genuinely hard engineering problem, and it is what a $1,600 Go2 is charging for that a $99 toy is not. Above the walking threshold, price tracks four things: how much the machine can carry, how well it is sealed against water and dust, how long it runs between charges, and what support comes attached. A B2 costs more than a Go2 because it carries 20 kg all day in weather, not because it walks better. The $40 robot dogs in social ads are not a budget tier Two of the most-searched robot dog names in America are not robotics products at all. Wuffy and Nicoo are promoted through AI-generated TikTok and Facebook advertising and follow a pattern consumer-protection researchers have documented in detail: an anonymous storefront, a countdown discount, and a battery-powered plush toy worth a few dollars wholesale. One Australian buyer found the identical Nicoo item listed on Temu for under $6. They matter to a cost guide only because they distort what people expect to pay. A walking, sensing robot dog cannot be built for $40. Real robot pets start around $80 from named companies with support and retail distribution, and anything genuinely lifelike starts in the four figures. The full scam checklist is in our robot dog toys guide. So what should you budget? For a child or a desk pet, $80 to $180 buys something real. For an older relative, the $179.99 Pup ships today and the roughly $1,500 Jennie is worth watching. For the most lifelike thing money buys, plan on about $3,500 over five years for an aibo rather than the $2,899.99 on the page. For anyone who actually wants a robot that walks, the Unitree Go2 at $1,600 plus freight is the entry point, and it is not close. Above that, you are no longer shopping. You are procuring, and the number comes from a conversation rather than a checkout page. Once you know the budget, the next question is which machine fits it: see our ranking of the best robot dogs. For the humanoid side of the market, we track what humanoid robots cost separately. --- # Humanoid Robots vs Cobots: Key Differences and Which to Choose URL: https://blog.robozaps.com/b/humanoid-robots-vs-cobots Author: Radica Maneva Published: 2026-07-30T10:09:54.032Z Updated: 2026-08-01T10:45:30.860Z Last fact-checked: 2026-07-30T00:00:00.000Z Category: comparisons TL;DR: Humanoid robots vs cobots solve different problems. Choose a cobot for stable work at a defined station; consider a humanoid when mobility, human-shaped reach and frequent task changes are essential. Validate either option against the complete task, safety case and deployed cost. Key takeaways: - Cobots describe a collaborative robot application; humanoids describe a human-like body form. - Cobots are usually the practical choice for repeatable work in a known station. - Humanoids become compelling when one robot must move between human-designed work areas and use existing interfaces. - Neither a cobot label nor a humanoid form makes an application safe without task-specific risk assessment. - Compare complete deployed cost, intervention rate, uptime and accepted output instead of robot price alone. FAQ: Q: What is the main difference between cobots and humanoid robots? A: A cobot is normally an industrial robot application designed for collaborative operation near people, often using a single articulated arm. A humanoid robot is defined by its human-like body, usually with two arms and a mobile base or legs. The terms describe different properties, so they are not direct opposites. Q: Are humanoid robots cobots? A: Not automatically. A humanoid could be integrated into a collaborative application, but its body shape does not make it a cobot or prove that it is safe near people. The complete application, including the tool, task, speeds, environment and foreseeable contact, needs a risk assessment. Q: Can humanoid robots replace cobots? A: Humanoids may replace some mobile or multi-station cobot concepts, but they are unlikely to replace cobots as a category. A fixed cobot remains simpler for repeatable station work. Humanoids are better candidates when mobility, human-shaped reach and frequent task changes create measurable value. Q: Which is safer, a cobot or a humanoid robot? A: Neither is inherently safer. Safety depends on the installed application. Cobots have mature industrial safety standards and integration patterns, while humanoids add risks such as whole-body motion, falling and carried loads. Both require a task-specific risk assessment and validation. Q: Which costs less, a cobot or a humanoid robot? A: A cobot usually costs less and is easier to estimate for one defined station. Compare complete deployed costs, though, including tooling, fixtures, facility changes, safety engineering, software, support and human intervention. A humanoid can make economic sense when it covers several locations or avoids expensive site changes. Q: When should a manufacturer choose a cobot? A: Choose a cobot when the work is repeatable, stays within a known area and benefits from purpose-built tooling. Machine tending, palletizing, inspection, dispensing and assembly are common candidates. The final cell may still need guarding or other protective measures. Q: When should a manufacturer consider a humanoid robot? A: Consider a humanoid when the job moves between work areas, uses equipment designed for people or combines several low-utilization tasks. Require a site-specific pilot that measures successful cycles, interventions, recovery, quality, uptime and the cost of the full deployment. Humanoid robots vs cobots is not a choice between an advanced robot and a basic one. It is a choice between two different approaches to automation. Cobots are usually built to perform a defined task close to people. Humanoid robots are built to move through human spaces and use equipment designed for the human body. For most repeatable factory stations, a cobot is the practical choice today. A humanoid becomes interesting when the work moves between stations, depends on human-shaped reach or changes too often to justify rebuilding the workspace. Humanoid robots vs cobots: the short answer Choose a cobot when the task is stable, the work envelope is known and you want a production system with mature tooling and integration support. Consider a humanoid when one robot must travel through a human-designed site, work at several locations or use doors, carts, shelves and controls without major facility changes. In either case, buy against a measured task and a validated safety case, not the category label. The distinction matters because the words describe different things. A humanoid is defined mainly by its body and movement. A cobot is defined by how an industrial robot application is designed to work near people. A humanoid could eventually operate in a collaborative application, but that would not automatically make every humanoid a cobot. The difference most comparisons miss A collaborative robot is commonly a six-axis arm with force limiting, monitored stops and other safety functions that help an integrator build a collaborative application. The arm may be mounted on a bench, pedestal, rail or mobile base. It normally uses a task-specific end effector such as a gripper, screwdriver, sander or welding torch. A humanoid robot is a general body plan. Most current industrial candidates have a torso, two arms, a head or sensor mast and either legs or a wheeled base. The design is meant to match spaces, tools and workflows originally created for people. That can reduce some facility changes, but it adds locomotion, balance, whole-body control, battery and perception problems that a fixed arm does not have to solve. The safety standards make the first point explicit. ISO describes collaborative operation as a property of the complete application, not a guarantee attached to a particular brand or robot. ISO 10218 covers industrial robot and system safety, while ISO/TS 15066 adds guidance for collaborative industrial robot systems. The tool, workpiece, speed, layout, foreseeable misuse and contact hazards all belong in the risk assessment. So a cobot holding a sharp knife is not automatically safe beside a person. A humanoid walking slowly is not automatically safe either. The installed application is what must be evaluated. Humanoid robots vs cobots at a glance Where are cobots the better choice? Cobots are strongest when the task can be described in a stable sequence: pick this part, move it here, apply this force, inspect this surface or load this machine. The task does not have to be simple, but its location, inputs and acceptable outputs need to be controlled. One process, one work area If work arrives at the same station every cycle, a fixed arm avoids the cost and failure modes of locomotion. The integrator can position the robot for the required reach, choose a purpose-built tool and control the presentation of parts. That usually produces a shorter route from proof of concept to acceptable cycle time. Common examples include machine tending, screwdriving, palletizing, dispensing, inspection and small-parts assembly. The exact application may still require guarding, scanners or separation from people. Collaborative operation is an engineering option, not a promise that every task will be fenceless. Precision and repeatability matter more than versatility A robot arm mounted to a known reference point has fewer variables to manage. That is useful when a process depends on consistent paths, placement or force. A humanoid may be able to reach the same machine, but its mobile base or leg position, torso motion and hand pose create a larger control problem. Extra flexibility has value only if the workflow uses it. You need an established integration path Cobot buyers can choose from a broad market of end effectors, vision packages, training, distributors and integrators. That does not make deployment effortless. It does make it easier to find people who have solved a similar station before and to estimate commissioning, spares and support. This is why a cobot is usually the default candidate for a small or midsize manufacturer automating its first process. A constrained problem is easier to validate. If the business case works, the same cell pattern can often be repeated. Where do humanoid robots have an advantage? A humanoid earns a serious look when the bottleneck is not just manipulation. The work may require movement between locations, access to infrastructure built around human height and reach, or frequent switching among tasks that do not justify a dedicated cell each. The job moves through a human workspace Factories, warehouses and back-of-house facilities contain stairs, narrow aisles, doors, carts, shelving and controls placed for people. A mobile humanoid is intended to operate in that environment without rebuilding every interface. That promise is most valuable in brownfield sites where changing the building, line or equipment would be expensive. Legs are not automatically the answer. If every route is flat and accessible, a wheeled robot or an autonomous mobile robot carrying an arm may be simpler and more efficient. The humanoid case gets stronger when the actual route, reach and task changes use the human-like form. One worker covers several low-utilization tasks Some jobs are difficult to automate because each individual action occupies only part of a shift. A dedicated arm at every location would sit idle. A general-purpose mobile robot could, in principle, move between tote handling, line-side replenishment, inspection and simple machine interaction. That is also where buyers should be strict. A demonstration of three tasks is not the same as reliable production across three tasks. Ask how often a human intervenes, whether the robot can recover after a failed grasp, how long task changeover takes and what happens when the environment differs from the training setup. Facility changes dominate the economics A dedicated cell can be cheaper at the robot level but expensive once conveyors, fixtures, guarding and layout changes are included. A humanoid proposal may justify a higher machine cost if it can use the existing process with fewer modifications. The calculation must include both sides. Claims about avoiding infrastructure changes are only useful when backed by a site survey and a defined acceptance test. Current deployments show that humanoids have moved beyond stage demonstrations, but the evidence is still narrower than the cobot market. Figure reports that its Figure 02 fleet worked in BMW production for eleven months and handled more than 90,000 parts. Agility Robotics reports that Digit moved more than 100,000 totes in a GXO operation. These are meaningful operational results. They do not prove that the same robots are ready for every plant or task. Which is safer: a cobot or a humanoid robot? Neither category is safer in the abstract. Safety depends on the installed application and the harm that can occur when something fails or a person enters the operating space. For a cobot application, the assessment covers the arm, tool, workpiece, fixtures, speeds, forces and the rest of the cell. Universal Robots, for example, states that an application risk assessment is mandatory and must include normal operation, setup, maintenance and foreseeable misuse. A force-limited arm can still create crushing, cutting, impact or ejection hazards. A humanoid adds whole-body motion. Assessors must consider falling, balance loss, carried objects, moving mass, feet or wheels, pinching between the robot and the environment, battery hazards, degraded sensing and recovery after a fault. A slow hand does not make the moving body irrelevant. The standards environment is also at a different stage. Industrial robot applications have established routes through ISO 10218 and ISO/TS 15066. In 2026, NIST introduced a humanoid robot baseline performance benchmark because shared measures for locomotion, manipulation and system performance are still developing. That is a reason to demand clear test evidence, not a reason to reject the category. Before purchase, ask the supplier to show the proposed safety functions, risk-assessment responsibilities, applicable standards, residual risks and validation plan for your actual tool and task. If the answer is only that the robot is a cobot or moves at human speed, the safety case is incomplete. Which costs less: a cobot or a humanoid? A cobot will usually have the lower purchase price and the more predictable integration path for one defined station. That does not mean the arm price is the project price. Tooling, vision, fixtures, safety equipment, engineering, commissioning, operator training, maintenance and production downtime all belong in the estimate. Humanoid pricing is harder to compare because the category mixes research platforms, pilot contracts, robots sold with software subscriptions and commercial deployments where public pricing is unavailable. Our guide to humanoid robot cost tracks disclosed prices, but a sticker price still says little about productive output at a particular site. Use a total-cost model built around accepted production work: Define the unit of value. For a cobot, that may be acceptable parts per hour. For a humanoid, it may be successful task cycles across several locations.Include the deployment boundary. Count tools, fixtures, facility changes, safety work, integration, software and support.Measure human assistance. Remote operation, reset time, exception handling and supervision are operating costs.Use achieved availability. Do not calculate payback from a demonstration cycle or a supplier maximum.Price the fallback. Understand what production does when the robot, network or cloud service is unavailable. A humanoid can win this calculation if it replaces several low-utilization automation projects or avoids costly changes to a brownfield site. A cobot can win by doing one high-volume process faster, more accurately and with less technical risk. The cheaper robot is the one that delivers the required output at the lower verified cost. How should you choose between a humanoid robot and a cobot? Start with a task study, not a product demo. Record where the work happens, how often it changes, what the worker touches, the required reach and payload, acceptable cycle time, quality checks, exceptions and hazards. Video several real cycles, including the awkward ones. 1. Does the robot need to move between work areas? If no, start with a cobot or conventional industrial arm. A fixed system is simpler to power, locate, guard and maintain. If yes, check whether an autonomous mobile robot, mobile manipulator or rail-mounted arm can cover the route before paying for a full humanoid body. 2. Is the environment expensive to redesign? Human-compatible tools and layouts are a genuine reason to test a humanoid. List the specific interfaces it must use: handles, shelves, carts, bins, control panels and stairs. Then make those interfaces part of the acceptance test. Do not give credit for general human compatibility without demonstrating the site-specific actions. 3. Is the task stable enough for dedicated tooling? High-volume, stable work normally rewards a task-specific end effector and controlled part presentation. Variable, lower-frequency work may reward a more general hand and perception system. The dividing line is economic, not philosophical: calculate utilization, changeover and the cost of exceptions. 4. What evidence matches your operating conditions? Ask for continuous runtime, intervention rate, successful cycles, recovery behavior and output quality on a task close to yours. Separate supervised pilot time from autonomous productive time. For a mobile system, include navigation and task switching in the cycle. For a cobot cell, include loading, tool changes and operator interaction. 5. Who owns integration and support? Clarify who designs the tool, completes the risk assessment, validates the application, trains operators, monitors the fleet, supplies spares and responds to failures. A technically impressive robot with an unclear support boundary is a poor production asset. A practical procurement sequence is to shortlist by task fit, run an offline or lab feasibility test, conduct a time-limited site pilot with written acceptance criteria, and scale only after the measured economics survive normal production variability. Cobot models in the RoboZaps robot database The models below show the range of cobot arms available for evaluation. Status reflects each canonical RoboZaps record at publication and can change as deployments develop. Open a model to see the full record, sources and current specifications. Payload, reach and deployment status are useful filters, but they are not a substitute for checking the tool, cycle and risk assessment. Compare these models against the same task specification. Humanoid models in the RoboZaps robot database Humanoid records span commercially available platforms, paid deployments and industrial pilots. That difference matters. A robot that can be ordered is not necessarily validated for a specific factory process, while a piloted machine may have strong task evidence without general availability. Use the database status as a starting point for diligence. Confirm regional availability, delivery timing, support, task evidence and commercial terms directly with the manufacturer. Will humanoid robots replace cobots? Not as a category. Humanoids and cobots solve overlapping but different automation problems. A fixed collaborative arm remains a strong design for repetitive station work. Giving that job legs, a second arm and a larger perception stack usually adds cost without adding output. Humanoids may take some jobs that would otherwise require a cobot on a mobile base, several dedicated stations or extensive facility modification. They may also work with cobots. A humanoid could replenish a cobot cell, move work in progress, change simple fixtures or handle exceptions between automated stations. The likely factory is mixed. Conventional industrial robots handle high-speed isolated work. Cobots handle constrained processes where people and robots need closer coordination. Mobile robots move goods. Humanoids cover selected gaps where mobility, reach and task variety justify the extra system complexity. Which should you choose? For a defined, repeatable task in one area, begin with a cobot. It offers the more mature supply chain, tooling and integration path. For work that genuinely depends on moving through a human site and using several human-oriented interfaces, test a humanoid against strict production criteria. Do not buy a body shape. Buy a verified result: safe cycles, acceptable quality, known intervention rates, supportable uptime and an economic case that includes the complete deployment. That standard favors cobots for many projects today, while leaving a clear and valuable role for humanoids where flexibility is the actual requirement. --- # Humanoid Robots Legal in the US After the FCC Covered List Ban URL: https://blog.robozaps.com/b/humanoid-robots-legal-in-us Author: Radica Maneva Published: 2026-07-29T06:00:00.000Z Updated: 2026-08-06T01:06:36.569Z Last fact-checked: 2026-08-06T00:00:00.000Z Category: insights TL;DR: Humanoid robots legal in the US are the ones that already held an FCC equipment authorization on 27 July 2026. Unitree's G1, R1, H2, Go2 and A2 all qualify. New foreign-made models are blocked until the Department of War grants a Conditional Approval. US-produced robots are outside the rule. Key takeaways: - Any robot with an FCC equipment authorization granted before 28 July 2026 stays legal to import, sell, own and use. - Unitree's entire current lineup is grandfathered. The R1, A2 and H2 were authorized in June 2026, four to five weeks before the ban. - The rule covers mobile ground robots over 4.4 lb. Fixed industrial arms, drones, road vehicles and medical devices are explicitly excluded. - "Foreign-produced" is defined by the Buy American domestic-end-product test, not by where a brand is headquartered. - Authorized robots keep receiving security and firmware updates until at least 1 January 2029 under a separate FCC waiver. FAQ: Q: Which humanoid robots can I legally buy in the US right now? A: Any humanoid that already held an FCC equipment authorization granted before 28 July 2026, plus any US-produced humanoid. That includes Unitree's G1, R1 and H2, all of which hold authorizations under grantee code 2A5PE. It also includes US-built machines such as Agility Robotics' Digit. What you cannot buy is a new foreign-made model that was never authorized before the cut-off, unless the Department of War grants it a Conditional Approval. Q: Is the Unitree G1 illegal in the US now? A: No. Unitree's humanoid authorization on file, FCC ID 2A5PE-YUSHU008, was granted on 3 March 2025, well before the 28 July 2026 cut-off. The grant record describes the device only as “Humanoid robot”, but the user manual filed with that same application, and hosted in the FCC's own exhibit record, is the G1 manual. The FCC action blocks new equipment authorizations; it does not revoke existing ones. Importing, selling, owning and using an already-authorized unit remains lawful. Check the FCC ID printed on the specific unit before you buy, because a variant sold under a different authorization may not carry the same status. Q: Are robot dogs covered by the FCC ban too? A: Yes. The National Security Determination names quadrupeds alongside humanoid robots and autonomous mobile robots. Unitree's Go2 and A2 quadrupeds hold pre-ban authorizations, so they stay legal. A new foreign-made robot dog with no prior authorization cannot receive one while the category sits on the Covered List. Q: What is a Conditional Approval and who grants it? A: It is the only route back onto the US market for a covered device. For advanced robotic devices the Department of War evaluates submissions and, if satisfied, transmits a specific determination to the FCC that a given device or class of devices does not pose the identified risks. That device is then exempt from the Covered List and can receive an FCC equipment authorization. Producers must supply the information requested in the Advanced Robotic Devices Guidance Document. Q: Can my existing robot still get software updates? A: Yes, until at least 1 January 2029. A separate FCC order released the same day, DA 26-789, waives the rules that would otherwise bar permissive changes to covered equipment. Every advanced robotic device already authorized for use in the US may keep receiving software and firmware updates, including security patches and operating-system compatibility fixes. The humanoid robots legal in the US after 28 July 2026 are the ones that already held an FCC equipment authorization on the day the rule landed. The Federal Communications Commission added foreign-produced advanced robotic devices to its Covered List, which blocks new authorizations. It does not cancel old ones. Unitree's G1, R1, H2, Go2 and A2 all cleared the FCC before the cut-off and remain legal to import, sell, own and use. This page tracks the status of each major model against the FCC's own equipment authorization records, and is updated when the Covered List changes or the Department of War grants a Conditional Approval. Last checked 29 July 2026. Which humanoid robots are legal in the US right now The test is simple and public. If a model holds an FCC equipment authorization with a grant date before 28 July 2026, it is grandfathered. If it does not, and it is foreign-produced, it cannot get one until the Department of War clears it. The table below shows what the FCC's Equipment Authorization System actually holds. The "Filed as" column is the FCC's own description of the device on the grant, which matters because manufacturers rarely file under the marketing name. That table covers the models people ask about most. If yours is not in it, check it directly below: every robot in the RoboZaps database against the FCC equipment authorization register, searchable by model, maker, or the FCC ID printed on the machine itself. Two cautions on reading this table. First, an absent record is not proof that no authorization exists, because manufacturers file under their legal entity rather than their brand. Unitree files as YuShu Technology Co., Ltd. under grantee code 2A5PE, which is why a search for “Unitree” in the FCC database returns nothing at all. Second, where the grant itself does not name the model, we matched it using the user manual filed with that same application: the Go1, Go2, B2 and G1 manuals are all hosted in the FCC’s own exhibit record, so every row above traces to an FCC document rather than to our guesswork. Where a row reads “No authorization found”, that is the limit of what the record shows and not a finding that the robot is barred: whether a device is caught by the rule turns on the domestic-end-product test described below, which no public source lets anyone run from the outside. The H1 has no separate authorization visible under the Unitree grantee code. If you are buying one, read the FCC ID from the label on the actual unit and check it against the FCC database using the method further down this page. What the FCC actually banned on 28 July 2026 The FCC's Public Safety and Homeland Security Bureau released DA 26-786, adding two new categories to the Covered List maintained under the Secure and Trusted Communications Networks Act. The robotics entry reads, in full: "Foreign-produced advanced robotic devices, except advanced robotic devices which have been granted a Conditional Approval by DoW." Under section 2.903(a) of the Commission's rules, equipment on the Covered List is barred from receiving an equipment authorization, and under section 2.911 every applicant must certify that its equipment is not covered. Without an authorization a radio device cannot lawfully be marketed or imported. That is the whole mechanism. There is no separate import ban, no customs seizure list and no penalty aimed at owners. The notice names no company and no country. The trigger was a National Security Determination sent to the FCC on 27 July 2026 by a White House-convened interagency body, which cited supply chain concentration and a run of real cybersecurity failures, including the September 2025 exploit that allowed takeover of fleets of Unitree robots. We covered the announcement and the Unitree fallout in our report on the FCC robot ban. The legal definition of an advanced robotic device The scope sits in Appendix C of the notice, as a scanned image, which is why most coverage has paraphrased it. An advanced robotic device is a mechanical mobile device, including autonomous mobile robots, humanoid robots and quadrupeds, that meets all of the following: is capable of locomotion, obstacle avoidance, navigation or movement on the ground;operates at a distance from a human operator based on commands or sensor data;has a combined weight, including any ground station or docking station, over 4.4 lb;contains a sensor capable of perceiving its environment, a network connection of at least 200 kbps in either direction, and software that controls autonomous navigation, movement perception, data collection or remote command-and-control. The determination then excludes connected road vehicles, rail-only vehicles, uncrewed aircraft, unmanned underwater vehicles, medical devices under section 513 of the Food, Drug and Cosmetic Act including surgical robots, limb prostheses and mobility aids such as wheelchairs, and fixed industrial arms of every kind: articulating, delta, gantry and SCARA. The weight threshold matters more than it looks. A small desktop or toy robot under 4.4 lb including its dock falls outside the rule entirely. What "foreign-produced" actually means This is the part most coverage has skipped, and it is where the rule differs from a straightforward China ban. The determination defines foreign-produced as "any article that would not qualify as a domestic end product, as that term is defined in 48 CFR 25.101(a)." That is the Buy American test. It turns on where the article was manufactured and on the cost share of its components, not on where the brand is headquartered or where the company was founded. A robot assembled in the United States from a majority of foreign components can fail the test. A robot from a foreign brand that clears the component threshold can pass it. Nobody outside the agencies has published a list of which specific models fall on which side, and any page that claims otherwise is guessing. For that reason the US-produced rows in the table above are marked as outside the rule on the basis of US manufacturing, not as a certified compliance determination. Robot dogs and quadrupeds face the same rule The determination names quadrupeds explicitly, and the supporting evidence leans on quadruped incidents, including the April 2025 disclosure of a pre-installed backdoor on foreign-produced robotic quadrupeds that exposed the camera feed and allowed full takeover. In practice the effect on the robot dog market is narrower than the headlines suggest, because the popular Chinese quadrupeds already hold authorizations. Unitree's Go2 cleared in May 2024 and the A2 cleared in June 2026. Newer entrants without a US filing are the ones shut out. Our guide to the best robot dogs covers the models still available. Conditional Approvals are the only way back in A producer of covered devices can apply for a Conditional Approval. For robots, the Department of War alone evaluates submissions, using the information requested in the Advanced Robotic Devices Guidance Document. If the Department transmits a specific determination to the FCC that a device or class of devices does not pose the identified risks, that device drops off the Covered List and can be authorized. The determination frames this as a transition mechanism, saying producers are encouraged to apply so they can continue to receive FCC authorization "while they work to onshore manufacturing." Conditional Approvals granted for advanced robotic devices as of 29 July 2026: none. The drone precedent shows what to expect. Uncrewed aircraft systems went onto the Covered List in December 2025, and that single entry has picked up four carve-outs since: a Blue UAS cleared-list exemption, a Buy American exemption added in January 2026, a Conditional Approvals clause in March 2026 and a toy drone exemption in June 2026. The robotics entry will almost certainly accumulate the same kind of exceptions, which is why this page is maintained rather than published once. If you already own one Nothing changes for you. Previously authorized devices remain legal to import, market, sell and use. The FCC did not revoke a single existing authorization, although it kept the power to revisit them later. Updates also keep flowing. A companion order released the same day, DA 26-789, waives the rules that would otherwise block permissive changes to covered equipment. In the Commission's words, all advanced robotic devices authorized for use in the United States "may continue to receive software and firmware updates that mitigate harm to U.S. consumers at least until January 1, 2029," including security patches and operating-system compatibility fixes. The one real risk for buyers is the second-hand and grey import market. A unit that was never authorized in the first place does not become legal because it is already inside the country. How to check any robot's FCC status yourself You do not have to take our word for any row in that table. The FCC's Equipment Authorization System is public. Find the FCC ID on the robot's label, in its manual, or in the compliance section of the product page. It looks like 2A5PE-YUSHU012.Open the FCC OET Authorization Search.Split the ID: the first three to five characters are the grantee code, the rest is the product code. Enter them in the two matching fields and search.Open the grant document. Check the Date of Grant. Before 28 July 2026 means the model is grandfathered.Read the Notes field on the grant. That is the FCC's own description of the device, and it is the only reliable way to tell which product an ID belongs to. If the seller cannot give you an FCC ID, treat that as the answer. If you are checking because you are about to buy, we do this as a paid product. The Robot Sourcing Brief is a fixed-fee shortlist for your task: three platforms with sources, authorization status per model with grant references, and what a pilot actually costs. It is US$1,950, delivered in five business days, and the fee is credited in full against any procurement through RoboZaps. Changelog 29 July 2026 — Page published. Covered List entry, scope definition and per-model authorization status verified against DA 26-786, DA 26-789 and the FCC Equipment Authorization System. Unitree Go1, Go2, B2 and G1 rows matched to their FCC-filed user manuals. Status wording for US-made models changed to reflect that no agency has published a model-by-model classification. No Conditional Approvals granted for advanced robotic devices at this date. Updated 30 July 2026: two status cells changed from “Blocked unless cleared” to “No authorization found” after a full sweep of all 81 manufacturers in the RoboZaps database, because the stronger wording asserted more than the public record supports. Updated 30 July 2026: added an interactive checker covering every robot in the database, 179 in total, after a full sweep of the FCC register across all 81 manufacturers. Compare on RoboZaps: Walker S2, Unitree H2, Unitree R1, Figure 03, Xpeng Iron and Apptronik Apollo. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # FCC Blocks New Approvals for Foreign-Made Robots: What It Means for Unitree URL: https://blog.robozaps.com/b/fcc-robot-ban-unitree-foreign-humanoids Author: Dean Fankhauser Published: 2026-07-29T04:08:04.136Z Updated: 2026-08-23T06:39:30.888Z Last fact-checked: 2026-08-22T00:00:00.000Z Category: news TL;DR: The FCC has blocked new authorizations for qualifying foreign-made mobile robots, including humanoids, quadrupeds and AMRs. New or previously unauthorized models are affected; previously authorized models remain legal under this action, and fixed arms fall outside the mobile-robot definition. Key takeaways: - The FCC added foreign-produced advanced robotic devices to its Covered List on July 28, 2026, preventing covered new or previously unauthorized models from receiving equipment authorization unless the Department of War grants Conditional Approval. - The rule covers mobile humanoids, quadrupeds and autonomous mobile robots over 4.4 lb that include sensing, networking and control software. - Previously authorized models remain legal to import, market, sell and use under this action, and the FCC allowed their security, software and firmware updates through at least January 1, 2029. The agency can still revisit earlier authorizations later. - The notice is country-neutral and does not name Unitree, although Unitree's product range is likely to be heavily affected. - Fixed stationary industrial robot arms fall outside the mobile-ground-robot definition; the separate inverter action concerns grid-connected power equipment, not robot motor drives. FAQ: Q: Did the FCC ban all Chinese robots? A: No. The FCC blocked new equipment authorizations for qualifying foreign-produced mobile robot models unless the Department of War grants Conditional Approval. The text applies by place of production rather than naming China or individual companies. Fixed industrial arms fall outside the mobile-robot definition. Q: Is the Unitree G1 now illegal in the United States? A: The FCC did not name the Unitree G1 or automatically make every existing unit illegal. Previously authorized models remain legal under this action, although the FCC can revisit earlier authorizations later. Buyers should verify the exact model and FCC ID with the seller. Q: Can existing Unitree robots still receive updates? A: Previously authorized covered robots may continue receiving security, software and firmware updates under an FCC waiver lasting through at least January 1, 2029. Other FCC rules still apply. Q: Are fixed industrial robot arms covered by the FCC robot ban? A: No. The covered definition is limited to mobile ground robots, so fixed industrial and medical robots fall outside its scope. That includes articulated, delta, gantry and SCARA arms. Q: Does the inverter ban cover the drives inside robots? A: The inverter determination addresses connected power equipment used to convert DC electricity from sources such as solar panels and batteries into AC electricity for grids and facilities. It is separate from robot motor drives and servo electronics. On July 28, 2026, the Federal Communications Commission added foreign-produced “advanced robotic devices” to its Covered List. In practical terms, that blocks covered new or previously unauthorized robot models from receiving the FCC equipment authorization generally required for import and sale in the United States, unless the Department of War grants Conditional Approval. Which specific machines that leaves you able to buy is tracked model by model in our guide to humanoid robots legal in the US, including the FCC ID and grant date behind each one. So the viral headline is broadly true. It is also easy to overread. This is not a recall, it does not make every foreign robot already in the country illegal, and the FCC did not publish a list naming Unitree or other Chinese manufacturers one by one. Update, August 22, 2026: Unitree is now a public company. It listed on Shanghai's STAR Market on August 19 under ticker 688836, opened about 629 percent above its 150.80 yuan issue price, and closed its first day near 845 yuan, up more than 460 percent. The listing changes none of the analysis below. The Covered List entry still blocks new US equipment authorizations for qualifying foreign-made mobile robots, and Unitree models that already held grants remain grandfathered. What the roughly 6.1 billion yuan raise does change is scale, so checking each model's US authorization before ordering matters more, not less. What the FCC actually changed The FCC acted after a White House-convened interagency group issued two national-security determinations: one covering advanced mobile robots and another covering power inverters. The agency then added both foreign-produced categories to the Covered List with immediate effect. Equipment on that list cannot receive a new FCC authorization. Because qualifying electronic devices generally need that authorization before they can be imported or marketed in the US, the result is an effective block on covered models that did not already have approval. The legal language is country-neutral. It covers foreign-produced devices regardless of the manufacturer’s nationality. “Foreign-produced” is tied to the federal domestic-end-product standard, so a US headquarters or American branding would not settle the question by itself. A manufacturer can seek Conditional Approval from the Department of War for a particular robot or class of robots. Which robots are covered? The definition is narrower than “all robots.” A covered device must be mobile, weigh more than 4.4 lb including any ground station or dock, and be able to move, navigate or avoid obstacles. It must also operate at a distance from a human operator or in response to sensor data, and include environmental sensing, network connectivity and control software. The FCC explicitly includes autonomous mobile robots, humanoids and quadrupeds. Its supporting determination also refers to wheeled and tracked ground vehicles. That captures the categories most people associate with embodied AI: warehouse AMRs, robot dogs and machines such as the Unitree G1. The definition’s boundaries matter just as much. It leaves out connected road vehicles, rail-only vehicles, drones, uncrewed underwater vehicles, regulated medical devices and mobility aids. Because the covered category is limited to mobile ground robots, fixed industrial and medical robots fall outside its scope. That includes articulated, delta, gantry and SCARA arms. What does the FCC robot ban mean for Unitree? Unitree is the obvious company to watch because its humanoids and quadrupeds fit the covered categories and it has become the price benchmark for both. Reuters reported that the action was aimed chiefly at Chinese competition and identified Unitree as a central target. The FCC’s actual notice, however, names no company or country. That distinction prevents a common mistake: it is not accurate to declare every current Unitree model banned or illegal. The dividing line is authorization, not whether a model had been publicly released. Models with FCC authorization before July 28 remain legal to import, market, sell and use under this action. Whether a particular SKU qualifies depends on its exact authorization and whether a hardware change creates a new model requiring fresh approval. The more certain effect is on models that lack prior authorization. A new Unitree humanoid, quadruped or materially changed model that needs fresh FCC approval now faces a block unless the Department of War grants Conditional Approval. The same rule technically applies to qualifying robots produced in Europe, Japan, South Korea and elsewhere, although US officials told Reuters that many non-Chinese suppliers were expected to obtain exemptions in practice. For a wider view of the company’s current product line, see our Unitree Robotics guide. What should existing owners and US buyers do? Existing owners do not need to stop using a robot solely because of this announcement. Previously authorized models remain legal to import, market, sell and use under this action. On the same day, the FCC issued a waiver allowing covered robots authorized before July 28 to continue receiving security, software and firmware updates through at least January 1, 2029. That protection is not necessarily permanent. The FCC retains authority to revoke or restrict earlier equipment authorizations later. The July 28 action does not do that to previously authorized robot models, but buyers should not treat the current position as an indefinite guarantee. For new purchases, the practical question is no longer just “Does the seller ship to the US?” Ask for the exact FCC ID and model number, confirm that the authorization predates the Covered List addition, and get the seller’s position on warranty service, replacement hardware and future updates in writing. A listing page or checkout button is not proof that a particular configuration remains importable. This could also change the economics. Buyers comparing humanoid robot costs may find that low-priced Chinese platforms are harder to replace than the headline suggests. US alternatives are generally sold through enterprise pilots rather than consumer checkout, and Conditional Approvals could produce a patchwork of allowed and blocked models. Why did the FCC cite cybersecurity? The government’s case has two parts: supply-chain dependence and the risk created when a networked machine can both collect data and move in the physical world. The determination points to cameras, microphones, LiDAR and other sensors that can map sensitive locations, as well as the possibility of remote disruption or control. That concern is not purely hypothetical. Security researchers disclosed a serious flaw in 2025 affecting Unitree’s G1 and H1 humanoids and Go2 and B2 quadrupeds. According to IEEE reporting, hardcoded Bluetooth keys could enable root-level compromise and allow an infected robot to attack other nearby Unitree robots. Unitree later said it had completed most fixes and would roll out updates. One vulnerability does not prove that every foreign robot is compromised, or that a manufacturer intentionally installed a backdoor. The FCC’s decision is a category-wide risk judgment, not a public finding of misconduct against every covered company. Why do fixed industrial robot arms fall outside the rule? The rule is deliberately about mobile systems. A fixed arm bolted to a factory floor falls outside the definition even if it has cameras, a network connection and sophisticated control software. That means the action does not currently block conventional foreign-made industrial arms from companies such as FANUC, ABB, KUKA or Yaskawa. Evan Beard’s LinkedIn post calls this an “industrial-arm gap” and argues that it should be closed. That is an advocacy position, not what the July 28 determination does. The inverter language in the same FCC notice also needs separating from the robot story. It concerns power equipment that converts DC electricity from sources such as solar panels and batteries into AC electricity for grids and facilities. It is not a ban on the motor drives or servo electronics inside industrial robot arms. What happens next? The immediate work shifts to model-level paperwork. Manufacturers will need to decide whether to seek Conditional Approval from the Department of War, redesign where products are made, or hold back US launches. Distributors and research labs will need much better records of FCC IDs and hardware revisions. For buyers, the market is likely to become less transparent before it becomes clearer. “Available globally” will no longer be enough. Our guide to where to buy humanoid robots will track which models remain verifiably orderable in the US as sellers respond. Bottom line Yes, the FCC has imposed a real and unusually broad restriction on new or previously unauthorized foreign-made mobile robot models. The strongest version of the claim is still wrong: this is not a blanket ban on every imported robot, current Unitree owners have not been ordered to stop using their machines, and fixed industrial arms fall outside the mobile-robot definition. The lasting significance is prospective. Foreign humanoids, quadrupeds and AMRs without prior authorization now need Conditional Approval from the Department of War before they can clear the FCC authorization gate. For Unitree and other Chinese manufacturers, that is a major new barrier to launching their next generation of robots in the American market. Earlier authorizations remain valid for now, but the FCC can revisit them later. Compare on RoboZaps: Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # What jobs will humanoid robots take from me? URL: https://blog.robozaps.com/b/what-jobs-will-humanoid-robots-take Author: Dean Fankhauser Published: 2026-07-28T12:10:16.556Z Updated: 2026-08-01T23:02:15.153Z Category: insights TL;DR: Morgan Stanley's model projects 62.7 million US jobs substituted by humanoid robots by 2050, and stresses that isn't a forecast of mass joblessness. The verified 2026 Western fleet is a few hundred machines. The gap is paced by the payback period, not the technology: physical work is substituted first, from about 2028; office work isn't modelled until the 2040s. Key takeaways: - 62.7 million is a projection, not a headcount. Morgan Stanley models it as the cumulative US jobs substituted by 2050, and says explicitly it doesn't imply that many people jobless. - The order of arrival is economic, not technical: high wages, long shifts and structured tasks clear a finance committee first. - Today's reality is small and counted: the most-deployed Western humanoid has an installed base in the low hundreds. - Forecasters disagree by roughly nine times for 2035 because they count different things (units shipped, units in service, or TAM). - The task, not the job, is the right unit of analysis. Roles get rewritten before headcounts move. FAQ: Q: Will humanoid robots take my job in 2026? A: Almost certainly not. The verified Western fleet in mid-2026 is in the low hundreds of machines, battery life is 1 to 4 hours, and no layoffs have been publicly attributed to humanoids. Substitution is a decade-plus story, not a this-year one. Q: Which jobs are most exposed? A: Physical, repetitive, hard-to-staff work: construction, production, warehousing, cleaning and food prep score highest in Morgan Stanley's model and are modelled to be substituted first, from around 2028. Management, legal and computing score lowest. Q: Does the 62.7 million figure mean 62.7 million people will lose their jobs? A: No. It is Morgan Stanley's cumulative modelled count of US jobs substituted by 2050, and Morgan Stanley states it does not imply that many people will be jobless. The model doesn't account for humanoids expanding existing industries or creating new ones. Q: What decides whether a role actually gets automated? A: The payback period. A robot is bought when its all-in cost (purchase, integration, maintenance and operating) pays back against the labour it displaces fast enough to clear a finance committee, which favours high wages, long shifts and structured tasks. Two numbers frame an honest answer. Morgan Stanley's model projects 62.7 million US jobs affected by humanoid robots — its cumulative count of workers substituted by 2050, across 831 occupations. Morgan Stanley stresses this is not a forecast that 62.7 million people go jobless. As of mid-2026 the most commercially deployed humanoid in the Western world, Agility Robotics' Digit, has an installed base estimated at roughly 75 to 100 units, and Agility doesn't publish an exact figure. Both numbers are real. The distance between them is the story, and it is measured in years, not months. Today and the forecast, on one axis A linear scale can't hold both ends of this subject. Placed side by side, today's verified fleet would occupy less than one eight-hundred-thousandth of the bar. Here is the same range on a logarithmic axis, where every step to the right is a tenfold increase. Find your occupation Morgan Stanley scored all 831 US civilian occupations for “humanoidability”: how much of the work is physical, repetitive, dangerous, or hard to staff. The scores roll up into the 22 standard occupation groups below. Bar length is the total workforce in a group; the filled portion is the share the model treats as substitutable. Select any row. The order of arrival The model doesn't run everything at once. Physical, low-credential work is substituted first and takes about a decade to build momentum. Office and professional work isn't touched until the 2040s. The number that actually decides it No job is automated because a robot can do it. It's automated when the payback period clears a finance committee. That calculation is the real timer on every occupation above, and it rewards high wages, long shifts, and structured tasks. Move the inputs to see where your own workplace lands. The model assumes 300 operating days a year, maintenance at 12% of purchase price annually, integration at 20% of purchase price, and a robot operating cost near $2 an hour (Roland Berger, April 2026). It's the shape of the curve that matters, not the decimal. The forecasters do not agree Anyone quoting a single number for 2035 is quoting one bank. Published estimates for that year span more than an order of magnitude, because they count different things: units shipped in a year, units in service, or the total addressable market. What is verified right now Against those projections, this is the counted, company-reported state of play in mid-2026. For the wider economic picture, from displacement versus job creation to sector shifts and wages, see the economic impact of humanoid robots on the job market. For where humanoids are actually deployed today, across manufacturing, logistics and offices, see humanoid robots in the workplace. Sources and method Occupation scoring, tiering, the adoption curve and the payback analysis come from Morgan Stanley, Humanoids: Investment Implications of Embodied AI (June 2024), with the 2035 and 2050 fleet figures from its A $5 Trillion Global Market note (April 2025). The 2035 spread also draws on Goldman Sachs (about $38bn TAM and 1.4m shipments by 2035) and Bank of America Global Research (about 10m units shipping annually by 2035). Deployment counts are company disclosures from Agility Robotics, Figure AI, Unitree and BMW, mid-2026, with a Technology.org deployment review from July 2026. The $2-an-hour operating cost is Roland Berger's Humanoid Robots 2026 (April 2026); payback ranges are from IDTechEx's Humanoid Robots 2026–2036 (May 2026). Task-level framing follows the ILO exposure index and the World Economic Forum's Future of Jobs 2025. Exposure percentages are modelled estimates of technical substitutability, not forecasts of unemployment. Compare on RoboZaps: Digit, Figure 03, Unitree G1, Tesla Optimus Gen 3 and Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Foundation Chooses AMD for Phantom MK-2—and Reveals a $100K-a-Year Industrial Lease URL: https://blog.robozaps.com/b/foundation-amd-phantom-mk2-industrial-lease Author: Radica Maneva Published: 2026-07-24T23:55:17.785Z Last fact-checked: 2026-07-24T00:00:00.000Z Category: news TL;DR: Foundation selected AMD Ryzen AI Embedded X100 for the announced Phantom MK-2 and says Phantom MK-1 leases for about USD 100,000 per robot per year. Its automotive deployment and manufacturing figures remain company claims. Key takeaways: - Foundation says Phantom MK-2 will use AMD Ryzen AI Embedded X100. - The approximately USD 100,000 figure is an annual industrial lease, not Phantom MK-1’s purchase price. - The 24,000-car deployment figure and factory-capacity targets are company claims without named-customer confirmation. - RoboZaps is not creating a Phantom MK-2 record until model-specific primary evidence exists. Foundation Future Industries says it has selected AMD’s Ryzen AI Embedded X100 platform for Phantom MK-2, an announced successor to its Phantom humanoid. The July 23 disclosure also put a useful commercial number on the current robot: Foundation told Reuters that industrial units are leased for approximately USD 100,000 per robot per year. That annual figure is a lease term, not a purchase price. RoboZaps has updated the Phantom MK-1 record accordingly and keeps the base robot’s purchase price listed as undisclosed. What the AMD partnership changes Reuters reported that Phantom MK-2 will use AMD Ryzen AI Embedded X100. AMD describes X100 as an embedded platform combining “Zen 5” CPU cores, RDNA 3.5 graphics, XDNA 2 NPU acceleration and unified memory for demanding edge workloads. That architecture is relevant to humanoid robotics because perception, model inference, planning and control compete for compute, memory bandwidth, power and thermal headroom inside the robot. AMD also listed Foundation founder and CEO Sankaet Pathak as a speaker in its July 23 session on AI-enhanced robotics for factories, warehouses and other physical-AI environments. Neither the AMD session page nor Foundation’s current product page provides a model-specific Phantom MK-2 datasheet, price or availability date. The deployment claim needs qualification Foundation told Reuters that Phantom MK-1 helped build more than 24,000 cars in 2025. That is a company claim: the automotive customer, number of robots, sites, tasks and operating hours were not disclosed, and RoboZaps has not found independent customer confirmation. The evidence supports qualified paid-deployment language, not an independently verified fleet-size claim. The company also told Reuters that its first factory, planned for October, would have capacity for 5,000 robots a year, followed by a second factory targeting 50,000 a year. Those figures are planned capacity targets rather than completed output or booked demand. Why the lease number matters An approximately USD 100,000 annual lease gives industrial buyers a concrete commercial reference point, even though the contract duration, service coverage, uptime terms, integration costs and minimum commitment remain undisclosed. It also changes how Phantom should be compared with robots sold through a list price: an annual lease shifts support and residual-value assumptions, but it does not by itself establish lower total cost. For buyers, the next useful disclosures would be named customer evidence, repeatable task and uptime data, service-level terms, and a model-specific Phantom MK-2 product page or datasheet. RoboZaps’ database decision Phantom MK-2 remains an announced successor, not a separate RoboZaps robot record. The AMD selection is material news, but it does not establish model-specific dimensions, payload, degrees of freedom, pricing or commercial availability. RoboZaps will not copy Phantom MK-1 specifications to the successor and will wait for primary model-specific evidence before adding it. --- # UWORLD U1 Robot: Price, Models, Specs and Release Date URL: https://blog.robozaps.com/b/uworld-u1-robot Author: Dean Fankhauser Published: 2026-07-20T23:02:33.470Z Updated: 2026-08-01T23:02:19.750Z Last fact-checked: 2026-07-21T00:00:00.000Z Category: insights TL;DR: UWORLD U1 starts at RMB 119,800 in China. UBTECH reports 13,361 orders, but customer deliveries were still unverified on July 21, 2026. First China deliveries are reported for September; no global release date has been announced. Key takeaways: - The U1 Lite starts at RMB 119,800; reported prices reach RMB 990,000 for the most expensive U1 Ultra configuration. - U1 launched as a product line on June 30, 2026. September is a reported China delivery window, not a confirmed global launch. - UBTECH's 13,361-order figure is company-reported and was announced before final payments opened. - Lite is a semi-torso robot, Ultra is presented with autonomous movement, and Pro locomotion remains unconfirmed. - RoboZaps found no verified customer delivery, independent long-duration test or official international support program by July 21, 2026. FAQ: Q: How much does the UWORLD U1 cost? A: The official starting price is RMB 119,800 in China. Chinese launch coverage lists U1 Pro at RMB 169,800, U1 Ultra in a female configuration at RMB 880,000 and the male Ultra at RMB 990,000. The US-dollar figures in the article are approximate conversions, not official US prices. Q: When will the UWORLD U1 be released? A: The U1 line was formally launched on June 30, 2026, after preorders reportedly opened on June 2. First-batch deliveries in China are expected in mid-September according to launch reporting. UBTECH has not announced a confirmed consumer release date outside China. Q: Can the UWORLD U1 walk? A: The U1 Lite is a semi-torso robot and does not walk. U1 Ultra is the high-dynamic model presented with autonomous movement. UBTECH's English launch announcement does not say whether U1 Pro can walk autonomously, so treat that capability as unconfirmed. Q: Is UWORLD U1 available in the United States, Europe, the UK or Australia? A: RoboZaps found no official consumer ordering route, regional price or delivery date for those markets as of July 21, 2026. The reported preorder and September delivery information applies to China. Q: Is the UWORLD U1 autonomous? A: UBTECH promotes autonomous conversation, memory, emotional inference and proactive interaction. U1 Ultra has been shown moving on two legs. Broad, unscripted home autonomy has not been independently demonstrated, and UBTECH is not currently selling the basic U1 as a robot that cooks or cleans. Q: How long does the U1 battery last? A: Contemporary reporting gives a runtime of two to four hours. A model-level battery capacity, standardized runtime test, recharge time and charging method were not available in the English company announcement RoboZaps checked. Q: Can U1 copy a real person's face and voice? A: UBTECH promotes appearance and voice customization. It also plans to donate 100 units using 3D facial reconstruction and voiceprint-based replication of designated people. The company has not published enough detail on consent, ownership, revocation or whether this program includes deceased people. Q: Is U1 a therapy robot? A: No clinical validation specific to U1 was found. Research on relational agents suggests possible benefits for loneliness, but the evidence remains limited and does not establish U1 as a treatment. It should not replace human care or professional mental-health support. Q: Has UWORLD delivered any U1 robots? A: No independently verified customer delivery had been found by the July 21, 2026 research cutoff. UBTECH says it had more than 13,000 orders at launch, while reporting puts the first China deliveries in September. Orders, manufactured units and delivered units are different evidence states. The UWORLD U1's silicone skin, fixed gaze and carefully styled hair will get most of the attention. Its order terms deserve more. UBTECH unveiled three U1 companion robots on June 30, 2026, with a starting price of RMB 119,800. The company says it had received 13,361 orders by launch day. Chinese reporting puts the first deliveries in September. Those facts make U1 more substantial than a concept video. They do not make it a proven consumer product. The published order total has not been independently verified, the initial reservations reportedly required a refundable RMB 3,000 deposit, and no customer delivery had been verified when we checked on July 21. This is not a hands-on review. It is a record of what UBTECH has announced, what reporters observed at the Shenzhen launch, and what remains unknown. UWORLD U1 quick facts UBTECH calls U1 the world's first “full-size ultra-bionic humanoid robot designed for mass production.” That is the company's description, not an independently established category record. Realbotix already offers life-size conversational robots, for example, although its full-body model moves on a wheeled base rather than walking. What is the UWORLD U1 robot? UWORLD is UBTECH's new consumer-facing brand. The U1 is its first product family, and it is aimed at companionship rather than factory work. Most humanoid companies sell the idea of physical labor: moving boxes, folding laundry, fetching objects or working a production line. U1 is being sold around presence. It looks at you, speaks, remembers previous interactions and changes its responses based on what it believes you are feeling. UBTECH lists companionship, reception, hospitality, elder support, tourism, research, education and premium domestic services among the possible applications. These are proposed uses, not documented deployments. At the launch, reporters were told that the basic U1 does not clean, cook or perform other household chores. U1 therefore sits at the expensive, humanlike end of the companion robots market. It is quite different from a small rolling robot pet, and it is not a direct substitute for the task-focused humanoid robots being developed for homes. “Ultra-bionic” has no standard technical definition. It is UBTECH's term for the combination of human proportions, silicone skin, facial actuation, voice interaction and body motion. UWORLD U1 Lite vs Pro vs Ultra The range spans a semi-torso model costing about as much as a small car and a walking full-body model priced like a house in some markets. *Conversions are approximate, based on exchange rates used in contemporary coverage. They are not official US prices. The model prices above come from IT Home's report from the launch. UBTECH's English-language company announcement confirms the three models and the RMB 119,800 starting price, but does not list every configuration. The jump between Pro and Ultra needs an explanation that the available English material does not provide. Chinese coverage says Ultra adds autonomous movement and more computing power. UBTECH's announcement only calls Pro “high-performance” and Ultra “high-dynamic”; it does not say whether Pro can walk autonomously. Until a model-level specification or an unscripted demonstration settles the point, Pro's locomotion belongs in the unknown column. Reported male and female configurations also differ in size. IT Home lists the male version at 183 cm and 42 kg, and the female version at 168 cm and 35.2 kg. Buyers still need a proper configuration sheet explaining which faces, voices, bodies and mobility packages can be combined. Price, preorder terms and the September delivery claim The U1 story has three separate dates that are easy to collapse into one: Preorders reportedly opened on JD.com on June 2, 2026.UBTECH formally launched the brand and product line on June 30.First-batch deliveries in China are reported for mid-September. Calling September the “launch” is therefore misleading. September is a delivery target, and there is no confirmed global rollout attached to it. According to IT Home, the initial preorder period ran until July 15 and a RMB 3,000 deposit secured a place in the first batch. Gasgoo reported a shipment target no later than September 15. AFP's report from the Shenzhen event also says deliveries are due to start in September. The accessible purchasing evidence is China-specific. We found no official consumer order page, regional price, warranty program or release date for the United States, Europe, the United Kingdom or Australia. That also makes the US-dollar figures above reference conversions, not prices a US customer can pay. International resellers may advertise their own offers, but those should not be mistaken for manufacturer pricing or official regional support. For context, our guide to what humanoid robots cost tracks the unusually wide gap between headline prices and the cost of a usable, supported system. U1 leaves many of those costs unstated: delivery, installation, training, repairs, replacement skin, software subscriptions and long-term support. What can the UWORLD U1 actually do? The launch material mixes observed behavior, engineering claims and future applications. Keeping them separate changes the picture. What was shown or observed The robots have animated faces, eye movement, lip movement, speech and configurable humanlike exteriors. AFP reported eye cameras, chest sensors and microphones at the launch. The U1 Ultra was presented as the walking version, while the basic model could move its head, eyes and mouth. So the hardware can perform those actions in a launch setting. We still do not know how reliably it handles a noisy room, a new home, an unexpected obstacle or hours of unscripted conversation. What UBTECH claims UBTECH uses 88 degrees of freedom as the series headline, though its English announcement does not provide a joint-by-joint breakdown for each model. The company says U1 can reproduce up to 90% of fundamental human movements. Its conversational system is promoted with similarly precise numbers: a 500-millisecond initial response, lip synchronization within 20 milliseconds, recognition of more than 20 emotional states at better than 90% accuracy, and deeper reasoning through models with hundreds of billions of parameters. Those are manufacturer claims. UBTECH has not published the test protocol, dataset, model-by-model result or independent benchmark needed to judge them. A fast response in a controlled demo is not the same as a useful conversation with a particular person over six months. The company also describes an “Agent Memory OS” that maintains long-term memories and a proactive care system that can respond without a wake word. These may be more important to the product than its walking ability. A companion that forgets every conversation is a kiosk; a companion that remembers everything creates a different problem entirely. What has not been demonstrated There is no good evidence yet that U1 can perform general household work, manipulate everyday objects reliably or navigate a home without supervision. UBTECH is not currently claiming that the basic model cooks or cleans. The launch material mentions premium domestic service as an application, but it does not document the tasks, autonomy level or human assistance involved. Reported battery life is two to four hours. The public material does not give us a useful model-by-model runtime test, recharge time, charging method or duty cycle. Even the upper end would require careful scheduling in a care or hospitality setting. Launch videos should also be read for what they prove. A robot completing a prepared sequence proves that the machine can complete that sequence. It does not prove general autonomy. AFP noted that many of the industry's most impressive movement displays remain preprogrammed or remotely operated, while real-world use is still limited. What do 13,361 orders actually tell us? UBTECH's announcement says cumulative U1 orders had passed 13,361 by June 30. That is a large number for a full-size humanoid robot, especially one with a top price near RMB 1 million. It is not the same as 13,361 robots delivered or even 13,361 robots paid for in full. 36Kr reported that the first-batch reservation required a fully refundable RMB 3,000 deposit and that final payments did not open until July 16, after UBTECH announced the 13,361 figure. UBTECH has not publicly broken that headline number into refundable reservations, binding purchase contracts or completed payments. We also do not know the cancellation rate, the mix of Lite, Pro and Ultra orders, or how many buyers are consumers rather than businesses or institutions. None of this proves the number is wrong. It tells us exactly what evidence state we have: company-reported demand at the order or reservation stage. The next meaningful numbers are units manufactured, final payments accepted and robots delivered to identifiable customers. Can a robot companion reduce loneliness? There is a sensible reason to study this category. The World Health Organization's Commission on Social Connection treats loneliness and social isolation as widespread health concerns, not a niche lifestyle issue. The early evidence on relational agents is cautiously encouraging. A 2024 systematic review and meta-analysis covering 14 studies and 286 participants found a moderate reduction in loneliness. The authors also found small samples and a high risk of bias. That evidence applies to relational agents as a broad category. It does not validate U1, establish long-term benefit or make a humanoid robot a clinical treatment. UBTECH's claim that U1 recognizes more than 20 emotional states with accuracy above 90% needs similar restraint. Emotions do not have one dependable facial signature. A major review published in Psychological Science in the Public Interest found that facial movements vary substantially by context, culture and individual. A benchmark score cannot show that a robot understands someone's inner state in the way a human caregiver might. The humanlike body creates another risk. People forgive a small robot for sounding mechanical. A realistic face raises expectations about timing, gaze and social judgment, then makes every mismatch more obvious. This is the uncanny valley as a product problem, not just a reaction to unusual launch footage. U1 may eventually prove useful in structured companionship programs. It should not be sold as a replacement for friends, family, carers or mental-health professionals. Custom faces and voices raise hard privacy questions UBTECH says buyers can customize U1's appearance, voice and personality. Its company announcement also describes a plan to donate 100 customized U1 robots during 2026. Those units would use 3D facial reconstruction and voiceprint-based identity replication to recreate “designated individuals.” That wording has produced headlines about recreating dead relatives. UBTECH's announcement does not say that. A designated person might be living, fictional, famous or personally known to the user. Until the company publishes eligibility and consent rules, anything more specific is speculation. The underlying questions are serious enough without embellishment. Who owns a face scan and voiceprint? Can the person represented withdraw permission? What happens when the robot is sold, repaired, inherited or disconnected from its account? Can a likeness be copied to another unit? What process applies after the person dies? Privacy extends beyond the replica. A U1 may combine cameras, microphones, emotional inferences, long-term memory and health-related reminders inside a home. UBTECH says its architecture favors local processing, minimizes cloud dependence and gives users hardware-level controls. The company also told AFP that data is encrypted and will not be used to train its models. Those are useful commitments. We have not found an independent privacy audit or a detailed public policy covering retention, deletion, human access, security updates and the exact data sent off the robot. For a machine designed to remember private conversations, that documentation is part of the product. UWORLD U1 vs Realbotix and 1X NEO “Companion robot” now covers very different machines. U1's closest visual competitor is Realbotix. Its closest consumer-humanoid comparison may be NEO, but the two products are trying to do different jobs. Realbotix's official range is useful context for UBTECH's “world first” language. Realbotix already offers custom faces, voices and full-body conversational robots. Its F Series starts at US$125,000, has a reported four-to-eight-hour battery and moves on a wheeled base. The company explicitly says its robots cannot walk. NEO makes almost the opposite design choice. It looks unmistakably robotic and is sold around utility in the home. 1X publishes a US$20,000 purchase price, a US$499 monthly option and a US$200 refundable deposit. It also admits that complex tasks may require a scheduled human operator. Our 1X NEO review keeps its delivery and autonomy claims separate for the same reason we do here. The comparison is not about which robot looks most advanced. A buyer choosing U1 wants believable social presence. A NEO buyer wants useful physical work. A Realbotix buyer may value a custom character without accepting the risk and complexity of bipedal walking. Who should consider a U1? The strongest early cases are commercial and research settings where the robot's presence is itself valuable: robotics labs, museums, showrooms, hotels, events and controlled human-robot-interaction studies. A care organization might also run a supervised pilot, provided it treats the robot as an experimental tool rather than a staff replacement. Private buyers would be taking a much larger gamble. Before paying the balance, they should ask for written answers on: The exact autonomous abilities of their chosen modelDelivery date and locationWarranty, repairs and replacement partsSoftware-support period and subscription costsFall protection, emergency stop and child or pet safetyBattery life under a defined workloadRemote-operation capability and human accessData storage, deletion and biometric-consent rulesRefund rights after final payment No independently tested consumer unit, global support program or detailed English model sheet was available when we checked. At these prices, that is enough reason to wait. The bottom line UWORLD U1 is real in the commercial sense: UBTECH has named the models, announced prices, accepted reservations in China and set a delivery target. It is not yet real in the more demanding sense of a product living with customers and surviving ordinary use. The line is still worth following. Lite brings a humanlike conversational robot down to RMB 119,800. Ultra pairs that social design with a walking body. If UBTECH converts its order book into reliable deliveries, U1 could become the first widely visible test of whether people want a companion robot to look this human. For now, the sensible verdict is watch, do not rush. The September delivery window will tell us more than the launch video did. How we researched this article RoboZaps checked UBTECH's company-supplied launch announcement against reporting from AFP and Chinese technology coverage, then used official Realbotix and 1X pages for the competitor comparison. Manufacturer performance and privacy statements are described as claims unless an independent test supports them. Prices, availability and delivery status were last checked on July 21, 2026. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # We found 33 commercial humanoid robot families. Just one full-size model targets the home URL: https://blog.robozaps.com/b/commercial-humanoid-robots-work-not-home-2026 Author: Dean Fankhauser Published: 2026-07-20T04:06:44.543Z Updated: 2026-08-01T10:53:49.442Z Last fact-checked: 2026-07-20T00:00:00.000Z Category: insights TL;DR: RoboZaps audited 74 catalogue listings and found 33 current, commercially addressable bipedal families. Twenty-nine target work or development. Four serve home, education or consumer uses, including just one full-size home robot, which was still taking preorders. Key takeaways: - The audit found 33 qualifying current bipedal families among 74 RoboZaps catalogue listings. - Twenty-nine of the 33 families were built for work or development. - Only four served home, education or consumer uses, including one full-size home robot that was still taking preorders. The commercial humanoid robot boom is not coming to your house. It is going to work. The industry's viral image is an all-purpose machine folding laundry, serving drinks or moving through a home. The commercial market taking shape in 2026 looks very different: factory deployments, warehouse work, robotics programs and managed trials with large companies. RoboZaps reviewed 74 listings in its humanoid robot catalogue and found 33 current bipedal families with a manufacturer-backed route to a customer. Twenty-nine were built for work or development. Only four served the home, education or broader consumer market. Just one was a full-size humanoid explicitly pitched for the home, and it was taking preorders rather than making immediate deliveries. Humanoids are becoming commercial products. Their first real customers are overwhelmingly businesses and developers, not households. Commercial humanoid robots are going to factories first Figure has put its humanoid on a BMW production line and signed another commercial deployment agreement. UBTECH says Walker S2 has entered mass production and begun deliveries. Agility Robotics offers Digit through enterprise deployments and Robots-as-a-Service agreements. Boston Dynamics says its initial 2026 Atlas production is committed to Hyundai and Google DeepMind. Those are meaningful commercial milestones, but they describe a market that looks more like industrial automation than consumer electronics. Customers negotiate deployments, prepare sites and rely on the manufacturer for support. They are buying an operational system, not taking home an appliance. Factories and warehouses also offer repeatable tasks and controlled environments. Homes contain stairs, clutter, children and pets, and a household robot has to handle many unrelated jobs at a price ordinary buyers can accept. University of Washington robotics professor Maya Cakmak has noted that optimistic studies of home humanoids often assume extensive safety testing, regulatory approval and insurance, conditions that may remain far from reality. The question is no longer whether a humanoid can produce an impressive demonstration. It is whether the machine can earn a place in a real workflow. Manufacturers are finding that place at work first. A shopping cart, a reservation or a sales call The word “available” hides three different markets. The most familiar route is direct ordering. Unitree lists the G1 and R1 through its online store. Even there, configuration, stock, shipping and support affect what a buyer can actually receive. The second route is a paid place in line. 1X NEO can be preordered with a $200 deposit. NEURA Robotics describes its €100 reservation for 4NE1 as refundable and says availability is expected at the end of 2026. Money changing hands is stronger evidence than a waitlist, but a reservation is not a robot ready to ship. The third, and most common among work-focused machines, is enterprise procurement. A buyer requests a quote, discusses the job and site, negotiates support and may lease the robot or pay for it as a service. Digit and Atlas fit this model, as do many machines entering factories through named partnerships. Humanoid robot prices can be just as slippery as availability. Unitree's H1 shop showed a $90,000 reference price, then told buyers to contact the company for the real price. Kepler announced the K2 at RMB 248,000. A deposit, a reference price and a negotiated deployment cost are not interchangeable. For buyers, the useful verb matters more than the headline claim: order, reserve, request a quote, lease or deploy. Each describes a different level of access and market maturity. The number that matters is four Only four qualifying families sat outside the work-and-development market: 1X NEO, Maxtronics NAO6, Noetix Dora and Robosen K1 Pro. Even that group stretches the meaning of “consumer.” NAO6 is aimed at education and research and is sold through a quote. Dora uses an intended-order process. K1 Pro is available through a manufacturer cart, but at consumer and STEM scale. Only NEO is presented as a full-size humanoid for the home, and its commercial route currently begins with a preorder. The robots are real. Manufacturers are taking deposits, shipping selected products, signing deployment agreements and committing fleets. But those advances are concentrated where businesses can absorb high costs, narrow the task and support the machine. For households, the humanoid future remains mostly a promise. For employers and developers, an early market is already here. The near-term humanoid revolution will arrive a few robots at a time, inside factories, warehouses and research programs, where the race to make them useful has already begun. Methodology RoboZaps, which publishes the catalogue used in this analysis, froze the 74 rows returned by its live Humanoids filter on July 16, 2026. The review counted one current bipedal family when first-party evidence showed a manufacturer-backed order, paid reservation or preorder, product-specific quote, lease or Robots-as-a-Service route, commercial delivery or committed external fleet. Duplicate listings, configurations and predecessor generations were consolidated. Thirty-three families qualified: 29 work or developer families and four home, education or consumer families. Thirty-eight rows did not clear the test, while three remained unresolved. Sources were rechecked July 20, 2026. The result describes the frozen RoboZaps catalogue, not the entire global market. The family-level evidence table lists all 33 inclusions, their commercial routes, primary sources and caveats. --- # This Week in Humanoid Robot News: July 13–20, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-july-13-20-2026 Author: Radica Maneva Published: 2026-07-19T22:43:15.034Z Updated: 2026-08-01T23:02:25.029Z Last fact-checked: 2026-07-20T00:00:00.000Z Category: news TL;DR: Hyundai is buying SoftBank's remaining Boston Dynamics stake after a put-option notice, WAIC 2026 brought new humanoid debuts from AgiBot and Unitree with no public checkouts, EngineAI opened the first freestyle humanoid fighting league in Shenzhen, and Unitree's Shanghai IPO remained unpriced as of July 20. Key takeaways: - Hyundai is moving to full ownership of Boston Dynamics after SoftBank exercised its put option; Reuters and Bloomberg report about $325 million for the remaining stake of roughly 10 percent, a price Hyundai has not confirmed. - WAIC 2026 in Shanghai (July 17-20) was the largest yet; AgiBot debuted four products led by the A3 Ultra and Unitree showed the GD01 transformable robot, all company claims with no public checkout. - EngineAI opened URKL in Shenzhen on July 17, the first freestyle fighting league for full-size humanoids, with 32 teams on standardized T800s and a 10 million yuan prize pool. - Walden Robotics left stealth on July 15 with about $300 million in seed funding at a $1.1 billion valuation, and LimX Dynamics added a $200 million pre-IPO round as China's listing rush accelerates. - Unitree's Shanghai IPO remained unpriced as of July 20, with a STAR Market debut still possible before the end of July. FAQ: Q: Who owns Boston Dynamics after the SoftBank deal? A: Hyundai Motor Group affiliates hold most of Boston Dynamics and are moving to acquire SoftBank's remaining stake of about 10 percent after SoftBank exercised its put option. Bloomberg and Reuters report a price of about $325 million; Hyundai's July 16 statement confirms the plan but calls final terms undecided. Completion would make Boston Dynamics wholly Hyundai-owned. Q: What new humanoid robots debuted at WAIC 2026? A: AgiBot debuted the A3 Ultra full-size humanoid, the X2 Edu education platform, the G2 Max industrial robot and the OmniHand 3 Ultra-M hand on July 18. Unitree showed the GD01, a 2.7 meter transformable robot it says can carry a person. All performance figures are company claims, and none of the debuts has a public price or order page. Q: Has Unitree's IPO priced yet? A: No. As of July 20, 2026, Unitree had not published a price or subscription date for its Shanghai STAR Market listing. Registration was approved on July 3, and reports point to a possible debut before the end of July. Q: What is the EngineAI robot fighting league? A: The Ultimate Robot Knock-out Legend opened in Shenzhen on July 17. Thirty-two teams fight with standardized EngineAI T800 humanoids for a prize pool of 10 million yuan, about $1.4 million. EngineAI wants the league to become a commercial fighting IP that funds robot development. Q: Can you buy any of the robots announced this week? A: No. None of the WAIC debuts is checkout-ready and Atlas has no public price. The clearest consumer order remains the 1X NEO at $20,000 to own or $499 per month, with first home deliveries still promised for later in 2026. The humanoid robot news from July 13–20, 2026 was about ownership and spectacle. Hyundai moved to buy out SoftBank's remaining stake in Boston Dynamics, China staged its biggest AI conference yet, and a Shenzhen robot fighting league put full-size machines in a ring. For the current commercial status of every major robot, see our humanoid robot news tracker. Hyundai moves to full ownership of Boston Dynamics SoftBank has exercised the put option it secured when Hyundai Motor Group bought Boston Dynamics in 2021, and Hyundai-side shareholders are now moving to buy SoftBank's entire remaining stake, roughly 10 percent of the company. Bloomberg and Reuters put the price at about $325 million. Hyundai's July 16 statement confirms the put-option notice and the plan to acquire the full stake, but says final terms, including price, are still going through internal governance and approval processes. The put option existed because Boston Dynamics was not taken public within the window set in the 2021 agreement. Full Hyundai ownership removes that IPO deadline, and it lands six months after the production Atlas was unveiled at CES on January 5 with deployments committed for 2026. Nothing changes for buyers yet: Atlas still has no public price or order page. WAIC 2026 fills Shanghai with new humanoids The World Artificial Intelligence Conference ran July 17–20 in Shanghai, and state media counted more than 1,100 exhibitors and over 300 first-time product showings, the largest edition so far. AgiBot used the show to debut four products on July 18: the A3 Ultra full-size humanoid, the X2 Edu education platform, the G2 Max industrial robot and the OmniHand 3 Ultra-M hand. AgiBot's claimed A3 Ultra numbers, including Nvidia Thor compute at 700 TOPS and eight-hour runtime with hot-swap batteries, are company figures with no third-party testing yet. Unitree showed the GD01, which it calls the world's first mass-produced transformable robot: 2.7 meters tall, 500 kilograms, able to switch between two-legged and four-legged movement and carry a person. Those are also company claims. Unitree also gave the first public look at a factory it says is run entirely by autonomous robots. None of the WAIC debuts came with a public price or order page. EngineAI stages the first freestyle humanoid fighting league On July 17, Shenzhen firm EngineAI opened the Ultimate Robot Knock-out Legend at the Nanshan Cultural and Sports Center, billed as the first freestyle combat tournament for full-size humanoids. Thirty-two teams compete on a standardized EngineAI T800 platform for a prize pool of 10 million yuan, about $1.4 million. Opening night produced the week's most shared clip: a robot called Matador kept fighting after a head kick left its head hanging loose. EngineAI founder Zhao Tongyang says the league is meant to become a commercial fighting IP whose revenue feeds robot development. Combat is a stress test in public: balance recovery, impact tolerance and repairability all get exercised in ways factory pilots avoid. The money kept moving: Walden's debut and China's IPO queue Walden Robotics left stealth on July 15 with about $300 million in seed funding at a $1.1 billion valuation. The Toyota Research Institute spinout is led by MIT's Russ Tedrake, counts Toyota, Nvidia, Boeing, Samsung Ventures and CoreWeave among its backers, and says a robot is already working shifts at a North American Toyota plant. We covered the full $1.2 billion funding week in last week's brief. The listing rush in China is accelerating. CNBC reported on July 13 that LimX Dynamics closed a $200 million pre-IPO round at about a $2.21 billion valuation, its second $200 million raise this year. Unitree's Shanghai IPO, registered on July 3, had still not published a price or subscription date as of July 20. A debut before the end of July remains possible, and its pricing will set the benchmark for every humanoid maker that follows. More humanoid robot news in brief July 13: Yaskawa and SoftBank validated a vision-language-action system that lets a MOTOMAN NEXT industrial robot pick irregularly placed wire harnesses and pack them into boxes, trained on SoftBank's GPU cloud with Nvidia Cosmos and Omniverse.July 15: Bloomberg reported that Chinese makers are pushing thousands of humanoids into logistics hubs and factories mainly to collect training data, widening the deployment gap with the United States.July 19: the FIFA World Cup final closed out Hyundai's match-day Atlas activation that began on July 5, the robot's largest public appearance to date. What to watch next week July 22: Tesla's Q2 earnings call. Optimus production at Fremont was promised for late July or August and had not started as of mid-July. Optimus still has no price and no order page.Any day: Unitree's STAR Market price and subscription dates.Ongoing: the Agility Robotics and Churchill Capital XI SPAC review, with the S-4 filed July 14.Later in 2026: first 1X NEO home deliveries. Order terms stay live at $20,000 to own or $499 per month, and no verified customer delivery has appeared yet. Compare on RoboZaps: Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 7 Best Unitree G1 Alternatives in 2026 (Compared) URL: https://blog.robozaps.com/b/unitree-g1-alternatives Author: Radica Maneva Published: 2026-07-17T07:59:10.292Z Updated: 2026-08-01T23:02:31.031Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: comparisons TL;DR: Booster T2 is the best direct cross-brand alternative to the Unitree G1. Booster K1 Geek is the budget development pick, R1 is the cheapest Unitree, PM01 favors openness, and H2 is the strongest full-size upgrade. For most labs, the programmable G1 EDU remains the safer all-round buy. Key takeaways: - Booster T2 is the closest cross-brand G1 alternative for a serious embodied-AI development team, but its price is quote-only. - Booster K1 Geek starts at $5,999 and supports secondary development, making it the strongest budget developer option. - Unitree's $4,900 R1 Air and $5,900 R1 do not support secondary development; researchers need the quote-only R1 EDU. - Unitree H2 starts at $29,900 with 31 DOF and 360 N·m maximum leg-joint torque, but only the H2 EDU supports secondary development and configurable high-power compute. - Compare the delivered configuration, developer rights, support and acceptance test rather than relying on the cheapest headline price. FAQ: Q: What is the best alternative to the Unitree G1? A: Booster T2 is the best direct cross-brand alternative for a serious development team. It has 31 degrees of freedom, two hours of stated walking time and a higher upper-body payload. Its price is quote-only, so the G1 EDU remains the safer all-round choice for many labs. Q: What is the cheapest programmable alternative to the Unitree G1? A: Booster K1 Geek starts at $5,999 and supports secondary development. It is smaller and less capable than the G1, but it is a better budget development platform than the locked base versions of the G1 and R1. Q: Is the Unitree R1 better than the G1? A: The R1 is cheaper and lighter, but the G1 has longer battery life, stronger perception options and more research headroom. The standard R1 and R1 Air do not support secondary development; researchers need the quote-only R1 EDU. Q: Is there an open-source alternative to the Unitree G1? A: EngineAI PM01 and Booster's development platforms are the strongest commercial options if openness is the priority. Confirm the exact access level before ordering because open tools, an open SDK and fully open firmware or hardware are different commitments. Q: Can I buy a Tesla Optimus instead of a Unitree G1? A: No. Tesla Optimus is not on general sale. There is no public customer order path or confirmed retail price, so it is not a practical alternative to a G1 today. Q: Is the Unitree H2 better than the G1? A: The H2 has more size, joint torque, degrees of freedom and high-power compute options, but it is not automatically the better research buy. Only the quote-only H2 EDU supports secondary development. The G1 remains easier to contain, transport and support in a typical lab. The best Unitree G1 alternative depends on why the G1 does not fit. Choose the Booster T2 for a more powerful cross-brand development platform, the Booster K1 Geek for programmable hardware under $10,000, the Unitree R1 for the cheapest entry into Unitree's ecosystem, or the Unitree H2 for a full-size performance upgrade. For most serious labs, however, the Unitree G1 EDU is still the best all-round choice. It is buyable, widely used and backed by a mature developer ecosystem. The catch is that the $13,500 base model is not the programmable version. A research team normally needs the quote-only EDU configuration, which starts at roughly $44,000 depending on hardware. The mistake is comparing only sticker prices. A $5,000 or $13,500 humanoid may not include the developer access, compute, hands, sensors, warranty or local support your project requires. Compare the delivered configuration, not the number in the headline. Unitree G1 alternatives at a glance Prices exclude some combination of hands, compute upgrades, shipping, duties, integration and service. See our humanoid robot cost guide before treating any base price as a project budget. How we chose these alternatives RoboZaps screened the 48 published humanoids in our production robot database on July 17, 2026, then checked current manufacturer pages for new releases and configuration details. Every robot had to pass three eligibility gates: a current manufacturer page, a store or sales-inquiry path, and a plausible overlap with a G1 use case. We then assessed developer access (35%), procurement certainty (30%), capability fit (20%) and price transparency (15%). Manufacturer specifications and recently verified database records took priority over reseller copy. The numbered order is for reading, not a claim that one score fits every buyer. Each “best” label names a use case. A full-size locomotion lab and a classroom buying a portable developer kit should not get the same winner. We also judged the exact configuration: a cheap display model without an SDK is not equivalent to a programmable research robot. That is why Tesla Optimus and Figure 03 do not make the shortlist. They may become competitors later, but they are not substitutes a lab can order on normal commercial terms today. 1. Booster T2: best direct cross-brand alternative The Booster T2 is the strongest direct G1 alternative for a team that wants a current embodied-AI development platform and is willing to request a quote. Booster lists three T2 configurations, all standing about 1.4 m tall with 31 degrees of freedom, two hours of continuous walking and a 2 m/s walking speed. The platform supports up to 10 kg across both arms, optional grippers or six-DOF dexterous hands, multiple cameras and Thor T5000 compute. Booster also presents it as an open development platform rather than a pre-programmed show robot. Those specifications put it closer to a well-equipped G1 EDU than to the locked $13,500 G1 base. The catch is commercial maturity. Booster has a live buy/inquiry path, but it does not publish the T2's price, standard warranty package or regional service coverage. The G1 has a larger installed base and a more familiar ecosystem for labs. Choose the Booster T2 over the G1 if: you want more upper-body capacity, 31 body DOF and a newer high-compute platform, and you can validate support directly with Booster. Stay with the G1 if: you value ecosystem maturity, existing research examples and a clearer international buying path. Source: Booster T2 specifications 2. Booster K1 Geek: best programmable alternative under $10,000 At $5,999, the Booster K1 Geek is a more honest budget development option than the cheap base versions of the G1 or R1. Booster explicitly supports secondary development and offers open-source tools; Unitree reserves secondary development for its EDU configurations. The K1 is much smaller: 95 cm tall and 19.5 kg, with 22 degrees of freedom and 48 TOPS of onboard AI compute in the Geek edition. That makes it easier to carry, safer to use in a classroom and cheaper to recover from inevitable falls. It is a sensible platform for kinematics, reinforcement learning, perception, human-robot interaction and student projects. Its limits are equally clear. The Geek battery is rated for about 30 minutes of walking at 0.4 m/s. It has four degrees of freedom per arm and is not a substitute for a G1 EDU configured for dexterous manipulation. The standard warranty is also only three months. Choose the K1 Geek over the G1 if: your priority is affordable, portable humanoid development rather than full-size reach or manipulation. Stay with the G1 if: you need longer runtime, faster locomotion, LiDAR or a stronger path to dexterous-hand research. Source: Booster K1 specifications and current price 3. Unitree R1: best cheaper option in the same ecosystem The Unitree R1 is the obvious alternative when the G1's hardware ecosystem appeals to you but its price does not. Unitree lists the R1 Air at $4,900 and the standard R1 at $5,900, before tax and shipping. The R1 stands 123 cm tall and weighs roughly 27 to 29 kg depending on configuration. The Air has 20 degrees of freedom; the standard R1 has 26. Both run for about one hour. The standard version adds a binocular camera and extra waist and head movement, while the EDU can be configured with more joints, dexterous hands and higher-performance compute. There is one major trap: the $4,900 and $5,900 versions do not support secondary development. If you want to write control software, run your own embodied-AI stack or add research hardware, you need the quote-only R1 EDU. The low headline price is therefore most relevant to demonstrations and basic experimentation, not a serious lab build. Choose the R1 over the G1 if: the lowest purchase price and easier transport matter more than sensors, runtime and development headroom. It also belongs on any shortlist of the cheapest humanoid robots. Stay with the G1 if: you need the better-developed research platform, longer battery life or stronger perception options. Source: Unitree R1 configurations and pricing 4. EngineAI PM01: best for low-level openness The EngineAI PM01 is the most interesting alternative for a team that distrusts closed hardware and wants access closer to the control layer. EngineAI describes the PM01 as a fully open embodied-intelligence platform. It provides low-level hardware interfaces, an ROS deployment framework, and open training and deployment code. The business edition is roughly 1.4 m tall and 42 kg, with 23 degrees of freedom, nearly two hours of stated battery life and hardware support for speeds above 2 m/s. The education edition adds one degree of freedom and secondary development support. Procurement is the problem. The manufacturer provides a purchase route but no stable public international price, and buyers should confirm what “fully open” covers in the exact hardware version: motor control, firmware, safety limits, simulation assets and redistribution rights are not the same thing. Choose the PM01 over the G1 if: low-level control access and open training/deployment code matter more than a mature global sales channel. Stay with the G1 if: you want more third-party research examples, clearer configuration documentation and a larger installed base. Source: EngineAI PM01 platform and open-code details 5. Noetix N2: best compact alternative with more payload The Noetix N2 sits between the tiny K1 and the G1. It stands 118 cm tall, weighs about 30 kg and carries roughly 5 kg while walking, according to Noetix's current specification table. That is more than the G1 base robot's roughly 2 kg arm payload, although the two figures are not measured under identical postures. The N2 has 18 degrees of freedom, a depth camera, IMU and a quick-swap 48 V battery rated for one to two hours. Noetix targets university research, education, performances and interactive use. An EDU version adds higher compute and secondary development support. The trade-off is articulation and commercial transparency. Eighteen degrees of freedom is modest, and Noetix does not publish pricing. A buyer should also confirm whether the quote covers the standard N2 or the EDU configuration, because the development access and compute are different. Choose the N2 over the G1 if: you want a compact platform with a higher stated carrying capacity for education, interaction or dynamic-motion work. Stay with the G1 if: you need more joints, a better-known SDK ecosystem or clearer configuration pricing. Source: Noetix N2 specifications and configurations 6. Unitree H2: best full-size upgrade The Unitree H2 is the clearest step up when you want more size, joint torque and configurable compute without jumping straight to H1 pricing. Unitree lists the base H2 at $29,900. It stands about 180 cm tall and weighs about 70 kg, with 31 degrees of freedom and maximum leg-joint torque of 360 N·m. The quick-release 0.972 kWh battery is rated for about three hours. That makes the H2 a much larger, higher-torque platform while keeping its published base price below the G1 EDU starting point. There is an important configuration catch. Unitree says only the H2 EDU supports secondary development. The $29,900 base model uses Intel Core i5 compute and does not include a high-power compute module. Unitree's headline material advertises up to 2,070 TOPS, while the detailed table lists configurable high-power modules such as Nvidia Thor on the EDU. Treat 2,070 TOPS as a configured-system claim, not a guaranteed base specification. Choose the H2 over the G1 if: you need a full-height platform, 31 degrees of freedom, much higher joint torque or room for a high-power compute configuration. Stay with the G1 if: you need a compact robot, easier lab containment or a better-known research ecosystem at lower physical risk. Source: Unitree H2 price, configurations and specifications 7. Unitree H1: best for high-dynamic locomotion research The Unitree H1 is not a cheaper G1. It is the step up for a team that needs a full-size machine with much more torque and locomotion performance. The H1 stands about 1.8 m tall, weighs 47 kg and has a manufacturer-rated moving speed of 3.3 m/s. Its knee joints reach about 360 N·m, compared with 90 N·m on the G1 base and 120 N·m on the G1 EDU. It uses a quick-swappable 864 Wh battery and combines 3D LiDAR with a depth camera. That performance comes with a very different budget and risk profile. RoboZaps' verified record places the H1 at about $90,000, depending on configuration. It needs more test space, more careful safety controls and a team prepared to handle the energy of a full-height robot. Choose the H1 over the G1 if: your research centers on high-dynamic locomotion, full-size biomechanics or high-torque control. Stay with the G1 if: you want a compact lab platform, hand research at a lower cost or easier transport and containment. Source: Unitree H1 official specifications Which Unitree G1 alternative should you choose? If two models still look close, ask each vendor for the same written package: exact configuration, developer access, hands, compute, sensors, spare batteries, shipping, duty, lead time, warranty, local support, remote-access policy and acceptance test. That comparison is more valuable than another demo video. Robots that are not real G1 alternatives yet Three famous names appear in almost every humanoid comparison. None is a practical G1 substitute for a normal buyer today. Tesla Optimus: not on general sale and has no confirmed customer price.Figure 03: limited to company-led pilots and partnerships; no public order path or price.Boston Dynamics Atlas: an enterprise industrial program, not an open research platform available through ordinary procurement. They are competitors in the long-term humanoid race. They are not the machines to put on a lab purchase order this quarter. Final verdict The Unitree G1 remains the benchmark because it is capable, orderable and relatively affordable. Its weakness is the gap between the $13,500 base and the programmable G1 EDU a lab actually wants. That is why the alternatives split by use case. The Booster K1 Geek is the budget development pick. The Booster T2 is the strongest direct cross-brand alternative. The EngineAI PM01 is the openness play. The R1 lowers the cost of entry but does not solve the developer-access problem unless you buy the EDU. The H2 is the most credible full-size upgrade; the H1 is the specialist choice when high-dynamic locomotion justifies a much larger budget and safety envelope. Choose the robot around the job, the development rights and the delivered system, not the most impressive video. Compare on RoboZaps: Unitree G1, Unitree R1, N2, Unitree H2, Unitree H1 and PM01. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 7 Boston Dynamics Atlas Alternatives and Competitors (2026) URL: https://blog.robozaps.com/b/boston-dynamics-atlas-alternatives Author: Radica Maneva Published: 2026-07-17T07:47:36.323Z Updated: 2026-08-01T10:53:54.095Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: alternatives-competitors TL;DR: Atlas publishes a 50 kg instantaneous and 30 kg sustained load figure plus IP67 protection, but Digit has the clearest commercial deployment record. Walker S2 targets continuous factory work; Figure 03 targets dexterous tasks; AGIBOT and Longcheer report G2 working on electronics lines. Apollo 2, Unitree H2 and Phoenix suit narrower pilots. Tesla Optimus Gen 3 is a watchlist item. Key takeaways: - Choose Digit if deployment evidence matters more than maximum payload. - Choose Walker S2 if automated battery changes and continuous factory operation are central to the use case. - Evaluate Figure 03 where tactile fingertips, palm cameras and variable parts handling matter. - Consider AGIBOT G2 when a wheeled robot can do the job on a flat production floor. - Unitree H2 has a public price, while Apollo 2 and Phoenix are better suited to narrower development pilots. For Atlas pricing and availability, start with our Boston Dynamics Atlas review. The broader humanoid robot cost guide explains why hardware price is only one part of a deployment budget. What an Atlas alternative has to replace The current electric Boston Dynamics Atlas is a 1.9 m, 90 kg enterprise humanoid. Boston Dynamics publishes a 50 kg instantaneous payload, 30 kg sustained payload, four-hour battery, 56 degrees of freedom, 2.3 m reach and IP67 protection. Atlas can navigate to a station and swap its own battery. That is a demanding benchmark. No rival on this list matches Atlas across payload, weather protection, mobility and published specifications at once. Atlas also has limits. Boston Dynamics does not publish a price, and its sales process starts with selected early adopters and a use-case evaluation. The company documents a Hyundai customer pilot and is building skills for part sequencing, machine tending, order building and related material-handling jobs. Buyers looking for a robot they can price online or deploy into a proven narrow workflow may find a better fit elsewhere. How we compared Atlas competitors We weighted five things: Evidence of work in a named customer environmentWhether the current model is deployed, piloting, publicly priced or only announcedFit for industrial material handling, logistics or manipulationPublished current-model specificationsThe amount of integration, supervision and custom development a buyer should expect A production target is not a deployment. A factory demonstration is not a paid rollout. A robot with a public price is not necessarily ready to work autonomously. Those distinctions decide the order below. Boston Dynamics Atlas alternatives compared Specifications and status were checked against current manufacturer material on 17 July 2026. Where a current model lacks a public specification, the table says so instead of borrowing a number from an older generation. The load figures are not directly comparable. Manufacturers variously report instantaneous, sustained, carrying, lifting, generic payload and per-arm peak or rated loads. Any procurement comparison needs the same task, reach and duty-cycle test across the shortlisted robots. 1. Agility Robotics Digit: best proven commercial alternative Digit is the safest first call when the requirement is a robot with real commercial history rather than the highest published payload. Agility says Digit is operating with Schaeffler, GXO, Toyota Motor Manufacturing Canada and Mercado Libre. In June 2026, the company reported more than 65,000 operating hours across commitments at nine customer facilities and more than $300 million in multi-year Digit v5 orders, subject to contractual milestones. Its current published hardware figures include a 16 kg carrying capacity and a four-hour battery. Digit is not a strength-for-strength Atlas replacement. It is designed around repetitive material workflows such as moving totes and connecting separate islands of warehouse automation. That narrower focus is its advantage. A buyer can evaluate an established workflow and support stack instead of funding an open-ended general-purpose experiment. Best fit: warehouses, distribution centers and manufacturing sites moving totes, bins or partsCurrent status: commercial deploymentsPrice: enterprise or Robot-as-a-Service terms are not publicMain limit: lower payload and a narrower manipulation envelope than Atlas Our full Digit review tracks its current customer and deployment record. 2. UBTECH Walker S2: best for continuous factory operation Walker S2 is the clearest Atlas alternative for operations worried about battery downtime. UBTECH says the robot can swap its own battery in about three minutes and continue operating around the clock. Its published payload is 15 kg across a working range from floor level to 1.8 m. UBTECH describes Walker S2 as being in mass production and delivery. Its 2025 annual report says production and delivery reached the thousand-unit level. Those are manufacturer-reported figures, so a buyer should still ask for customer references, installed-fleet uptime and service response times in the target region. The hardware is less powerful than Atlas on published payload. For repetitive factory work, however, automated power management may matter more than another 15 kg of carrying capacity. Best fit: long shifts, material handling and production-line tasks on flat factory floorsCurrent status: manufacturer-reported mass production and deliveryPrice: not disclosedMain limit: customer-level economics and performance data are not public 3. AGIBOT G2: best evidence in electronics manufacturing AGIBOT G2 is not a bipedal Atlas clone. It is a wheeled humanoid designed for precise industrial manipulation. On a smooth factory floor, that can be the more sensible architecture. AGIBOT and Longcheer say multiple G2 robots are integrated into tablet production lines for loading, unloading, inspection and sorting. The companies report up to 310 units per hour, a 19 to 20 second cycle, more than 99% success, integration within 36 hours and roughly 3,000 units per shift. Those figures come from the supplier and its deployment partner, but they are far more useful than an edited demo with no operating context. G2 uses automotive-grade components, IP42 protection and force-controlled operation. AGIBOT does not publish a full current spec sheet or price, so procurement still begins with a direct enquiry. Best fit: high-throughput electronics inspection and mixed-model productionCurrent status: documented production-line deploymentPrice: not disclosedMain limit: wheels rule out stairs and rough terrain; core physical specs remain sparse 4. Figure 03: best for dexterous variable tasks Figure 03 is worth shortlisting when the job changes from simple transport to difficult manipulation. Figure publishes a 20 kg payload, five-hour runtime, 61 kg weight and 1.2 m/s speed. Figure's launch material says its palm cameras and tactile fingertips are designed to keep working when the main cameras lose sight of the object. Figure said in April 2026 that it had produced more than 350 Figure 03 robots and increased production from one robot per day to one per hour. In June it showed Figure 03 at BMW handling a sequencing workflow that required picking variable parts, adjusting its stance and pulling a cart. The company accurately called that a demonstration. It should not be upgraded into a commercial deployment in retelling. Figure 03 is a strong candidate for a tightly scoped manipulation pilot. It is not publicly orderable, and Figure has not disclosed enterprise pricing. Best fit: variable parts, bimanual manipulation and AI-heavy task developmentCurrent status: production ramp and documented BMW demonstrationPrice: not disclosedMain limit: the current evidence is promising but earlier than Digit's commercial record 5. Apptronik Apollo 2: best modular US pilot platform Apollo 2 comes in bipedal and wheeled configurations. Apptronik is using fleets at its Robot Park facilities and at partner sites, including Mercedes-Benz and GXO, to collect task data through teleoperation and autonomous execution. The original Apollo specification listed a 25 kg payload and four-hour hot-swappable battery. Apollo 2 is the current robot, and Apptronik has not published an equivalent complete spec sheet for it. A buyer should ask which original figures still apply rather than letting an older table settle the question. Apollo 2 makes sense for organizations prepared to develop a task with Apptronik. It is less convincing as an off-the-shelf answer for a fixed launch date. Best fit: industrial pilots, task-data collection and teams that value a bipedal or wheeled choiceCurrent status: operational data-collection fleets and partner pilotsPrice: not disclosedMain limit: current-model specifications and commercial deployment terms are not public 6. Unitree H2: best publicly priced full-size platform Unitree H2 is the easiest model here to put into an initial hardware budget. Unitree publishes a $29,900 price before tax and shipping, 31 degrees of freedom, a roughly three-hour runtime, about 70 kg weight and 15 kg peak or 7 kg rated single-arm payload. That price does not make H2 a $29,900 Atlas. Unitree presents it as a platform for model development and multiple work scenarios, but RoboZaps did not find named industrial deployment evidence comparable with Digit, Walker S2 or G2 in the current record. H2 is a credible choice for a robotics team that wants full-size hardware to develop on. A plant manager buying an outcome should demand a task demo, safety package, integrator plan and service commitment before treating the list price as a deployment cost. Best fit: research, secondary development and well-resourced internal robotics teamsCurrent status: current product with public pricingPublic price: $29,900 before tax and shippingMain limit: limited public evidence of customer production work If public availability is the priority, compare the wider list of humanoid robots you can buy. 7. Sanctuary AI Phoenix Gen 8: best manipulation-data platform Phoenix Gen 8 uses a wheeled base. Sanctuary AI says customer feedback pushed it away from bipedal legs because wheels could support a stronger torso for precise, useful work. Gen 8 improves cameras, telemetry, sensor coverage and commissioning speed, with a strong focus on collecting data for its Carbon control system. That makes Phoenix relevant when hands, touch, human demonstrations and physical-AI training matter more than walking. Sanctuary also has an automotive development partnership with Magna. Current Gen 8 dimensions, payload, runtime and price are not public. Older Phoenix specifications belong to earlier generations and should not be copied into a Gen 8 comparison. Best fit: dexterous manipulation research, task-data capture and physical-AI developmentCurrent status: current Gen 8 data-capture platform with industrial development partnershipsPrice: not disclosedMain limit: not bipedal, and current hardware specifications are incomplete What about Tesla Optimus Gen 3? Tesla Optimus belongs on the watchlist, not in the current top seven. Tesla described Gen 3 as its first design intended for mass production. Its Q1 2026 update says the first production lines are being installed. Tesla has not published a current Gen 3 specification sheet, a final price, a public order page or evidence of an outside customer deployment. Optimus could become a serious Atlas competitor if Tesla converts its manufacturing plans into reliable robots and external support. A future production target is not a reason to delay a viable deployment today. The best Atlas alternative may not be a humanoid Buy the workflow, not the silhouette. If the task is fixed, repetitive and fenced, a conventional robot arm may be faster, cheaper and easier to validate. If goods move across a smooth floor, an autonomous mobile robot with a manipulator may avoid the cost and risk of bipedal locomotion. For truck unloading, Boston Dynamics Stretch is purpose-built. For inspection, Spot is the company's established commercial platform. A humanoid earns its complexity when the work genuinely requires human-scale reach, existing tools, stairs, narrow access or frequent task changes. If none of those conditions applies, Atlas and its lookalikes may be solving the wrong problem. Which Atlas competitor should you choose? For the strongest commercial deployment evidence: DigitFor continuous factory operation: Walker S2For high-throughput electronics work on flat floors: AGIBOT G2For dexterous, variable manipulation: Figure 03For a modular development partnership in the US: Apollo 2For publicly priced full-size development hardware: Unitree H2For manipulation data and physical-AI research: Phoenix Gen 8For a future mass-production bet: keep watching Tesla Optimus Gen 3 Before signing a pilot, ask the manufacturer to define the exact model revision, task, success rate, intervention rate, duty cycle, payload definition, battery process, safety boundary, integration work, acceptance test, support response and total cost. A robot that succeeds in a ten-minute demonstration may still fail the economics of a ten-hour shift. Frequently asked questions What is the best alternative to Boston Dynamics Atlas? Agility Robotics Digit is the best proven alternative for buyers who prioritize commercial deployment evidence. Walker S2 is stronger for continuous factory operation, while Figure 03 is more compelling for dexterous, variable tasks. No single robot matches Atlas across payload, weather protection, mobility and published specifications. What is the closest competitor to Boston Dynamics Atlas? For full-size industrial humanoid hardware, Figure 03 and Apptronik Apollo 2 are close US competitors. Walker S2 is the more direct factory-operations rival, while Digit is further ahead in documented commercial deployment. Is there a cheaper alternative to Boston Dynamics Atlas? Unitree lists H2 at $29,900 before tax and shipping. Boston Dynamics does not publish an Atlas price, so the size of the discount cannot be calculated from official list prices. H2 is a development platform with a much lighter published payload and less industrial deployment evidence. Can you buy a Boston Dynamics Atlas alternative now? Unitree publishes a price for H2, while enterprise buyers can contact Agility, UBTECH, AGIBOT, Figure, Apptronik and Sanctuary about deployments or pilots. Availability, lead time and support vary by region. Public pricing is rare outside Unitree. Which Atlas alternative has the best deployment record? Digit has the clearest commercial record in this group. Agility names deployments with Schaeffler, GXO, Toyota Motor Manufacturing Canada and Mercado Libre. AGIBOT G2 also has unusually detailed production-line evidence from Longcheer, although G2 uses a wheeled base. Is Tesla Optimus better than Boston Dynamics Atlas? There is not enough current evidence for that conclusion. Atlas has a published enterprise specification sheet and customer-pilot path. Tesla has not published final Gen 3 specifications, a price, a public order page or an outside customer deployment. --- # 8 Apptronik Apollo alternatives for industrial work in 2026 URL: https://blog.robozaps.com/b/apptronik-apollo-alternatives Author: Radica Maneva Published: 2026-07-17T06:44:47.779Z Updated: 2026-08-01T10:52:26.843Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: comparisons TL;DR: Digit is the closest Apollo alternative for a buyer who wants a commercial program now. Walker S2 is the strongest automotive-factory option, while Atlas leads on hardware but its 2026 allocations are committed. Figure 03 has the best public production ramp. Unitree H1 is the easiest full-size platform to source. HMND 01, KR1 and Phoenix are earlier-stage bets. Key takeaways: - Digit is the closest practical Apollo alternative because it combines enterprise access with sourced commercial agreements and deployment evidence. - Walker S2 is the strongest automotive-factory option, with a 15 kg payload and an autonomous battery-swap system. - Atlas publishes the strongest hardware specifications, but all 2026 deployments are committed. - Figure reports more than 350 Figure 03 units across internal and commercial-development programs, not 350 customer deliveries. - Only Digit and Walker S2 had source-verified deployment evidence attached among the 20 paid-deployment humanoids in the 17 July RoboZaps snapshot. FAQ: Q: What is the closest alternative to Apptronik Apollo? A: Agility Robotics Digit is the closest practical alternative for logistics and manufacturing buyers. It is available through enterprise Robots-as-a-Service agreements and has verified commercial deployment evidence. Its 16 kg carrying capacity is below Apollo's original 25 kg figure, but its buying and deployment path is clearer. Q: Which Apptronik Apollo alternative can you buy today? A: Unitree H1 has the clearest direct hardware access path, although Unitree asks buyers to contact sales for the real price. Digit can be acquired through an enterprise RaaS agreement. Most other close alternatives, including Apollo, Walker S2, Atlas and Figure 03, require direct enterprise discussions and do not have public standard terms. Q: Is Digit better than Apptronik Apollo? A: Digit is better documented as a commercial logistics deployment. Apollo has a higher original published payload, modular bipedal or wheeled configurations, and a major Google DeepMind partnership. The better choice depends on whether you want a defined logistics program now or a broader pilot and data-collection platform. Q: Is Walker S2 a real Apollo competitor? A: Yes. Walker S2 targets industrial and automotive work, handles 15 kg, and can swap its own battery in about three minutes. UBTECH says it entered mass production and delivery in 2025. Buyers should still request customer throughput, uptime, price and local service terms before treating it as proven at scale. Q: Is Boston Dynamics Atlas available to buy? A: Atlas is a production enterprise robot, but Boston Dynamics says every 2026 deployment is committed to Hyundai and Google DeepMind. There is no public price. Additional customers are expected from 2027. Q: Is Tesla Optimus an Apptronik Apollo alternative? A: It is a strategic competitor, not a current procurement alternative. Tesla has not opened Optimus sales to outside enterprises and does not publish standard price or delivery terms. Keep it on the watchlist unless Tesla provides a real external buying path. Apptronik Apollo has credible partners, strong hardware lineage and one of the better industrial stories in humanoid robotics. It still has no public price, no normal order path and no customer-published performance data. Apollo 2 is a scaled-pilot and data-collection platform; Apptronik describes the coming Apollo 3 as its first true commercial product. That makes “What are the best Apptronik Apollo alternatives?” a practical buying question, not a popularity contest. Buyers comparing Apptronik Apollo competitors need to know which vendors can support a real industrial program, not which robot has the most impressive demo. How we chose these Apollo alternatives We started with 48 published humanoid records in the RoboZaps database. The first filter was job fit: manufacturing, logistics, material handling or dexterous industrial work. The second was evidence. We looked for sourced specifications, named customer work, a credible buying or pilot path, and a current product generation. We did not rank a robot highly because it can dance, run quickly or appears in a polished demo. A useful Apollo alternative has to answer harder questions: Can a buyer engage the vendor? What exact job has the robot performed? Who confirmed it? What happens after the pilot? This is desk research. RoboZaps has not operated these robots. We compared the current RoboZaps database with manufacturer specifications, customer announcements and public buying paths, then marked claims that still rely on a vendor rather than an independent customer. In this article, “verified deployment evidence” means a cited deployment source passed the RoboZaps source check. It does not mean RoboZaps independently audited the customer site, throughput or uptime. One warning applies to Apollo itself. The familiar 25 kg payload and four-hour battery figures belong to the original robot unveiled in 2023. Apptronik's current Apollo 2 page confirms bipedal and wheeled configurations, swappable batteries and its Artemis/Fleet Connect software stack, but it does not publish a new numeric spec sheet. Our full Apptronik Apollo review separates the original specifications from the current product. Apollo alternatives at a glance 1. Agility Robotics Digit: the closest commercial alternative Digit is the clearest answer for an operator who likes Apollo's logistics story but does not want to wait for Apollo 3. Agility publishes a 16 kg carrying capacity and four-hour battery life. More important, it has moved beyond vague pilot language. Toyota Motor Manufacturing Canada signed a Robots-as-a-Service agreement after a successful pilot and plans to deploy Digit. Toyota should not yet be counted as an active deployment. Agility names GXO and Schaeffler among its active commercial environments. Its Arc platform connects Digit to warehouse and manufacturing systems and manages fleets. Apollo has the stronger published original payload at 25 kg and a more modular mobility concept. Digit is narrower. It is built around repetitive logistics and material-movement work rather than promising to become a general-purpose pair of hands. That narrower scope is an advantage if your first task is moving totes, transferring materials or connecting two existing islands of automation. Read the Agility Robotics Digit review for the full deployment and cost context. Choose Digit instead of Apollo when: you want a defined RaaS conversation and a task that already resembles Digit's commercial work. Do not choose Digit when: the job needs 25 kg handling, fine five-finger manipulation or a wheeled configuration. 2. UBTECH Walker S2: the strongest automotive-factory option Walker S2 belongs near the top because UBTECH has published both an industrial specification and a manufacturing claim. The robot handles up to 15 kg across a working range of 0 to 1.8 metres, has 52 degrees of freedom, and can replace its own battery in about three minutes. UBTECH markets the system for continuous operation using dual batteries and a swap station. UBTECH says Walker S2 entered mass production and delivery in November 2025. Its 2025 annual report describes production and delivery at the thousand-unit level. Those are company-reported figures, not an independent audit, but they are still more concrete than the “commercial scale is coming” language common in this market. Walker S2's industrial case is strongest in automotive production, where UBTECH reports work on inspection, material handling and production-line tasks. Apollo has a stronger North American partner set and a higher original payload figure. Walker S2 has the better public uptime design and a clearer manufacturing statement. See the RoboZaps Walker S2 record for the current sourced specification set. Choose Walker S2 instead of Apollo when: you run an automotive or factory workflow, need a published battery-swap strategy, and can support an enterprise deployment involving UBTECH. Do not choose Walker S2 when: you need independently published customer throughput, a public price or established local service terms. Ask for all three. 3. Boston Dynamics Atlas: the strongest robot you probably cannot get this year The production Atlas is the hardware benchmark in this group. Boston Dynamics publishes a 50 kg instant lift capacity, 30 kg sustained capacity, 20 kg one-handed capacity, four-hour battery life, a three-minute autonomous battery swap, IP67 protection and operation from -20°C to 40°C. Orbit connects the robot to manufacturing and warehouse systems. That makes Atlas a closer industrial substitute for Apollo than the old research videos suggest. It can handle heavier work, has a more complete public spec sheet and comes from a vendor with an established field-service operation. Access is the problem. Boston Dynamics says every 2026 Atlas deployment is committed to Hyundai and Google DeepMind, with additional customers following in 2027. Hyundai is the industrial customer; the Google DeepMind fleet supports a joint AI research program. There is no public MSRP. Choose Atlas instead of Apollo when: heavy handling, harsh conditions, system integration and serviceability matter more than near-term availability. Do not choose Atlas when: you need a pilot slot this year or a budget you can put into a normal procurement process today. 4. Figure 03: the strongest production and AI alternative Figure 03 has the most substantial public production story among Apollo's AI-first peers. Figure reported in April 2026 that BotQ had delivered more than 350 Figure 03 robots and demonstrated a one-unit-per-hour production cycle. Those were not 350 customer deliveries. Figure says the units were allocated across internal research, data collection, housework development and commercial-use development. It also published an end-of-line first-pass yield above 80%, plus battery and actuator production figures. The robot is 1.73 metres tall, weighs 61 kg, carries 20 kg, runs for five hours and uses the Helix 02 whole-body AI stack. At BMW's Spartanburg plant, Figure has shown it handling a sequencing workflow that combines part placement, body repositioning and cart movement. Apollo 2 offers more mobility choice because it can use legs or a wheeled base. Figure 03 is bipedal only. Figure's advantage is a more vertically integrated hardware, AI and manufacturing story, backed by specific production numbers and a defined current factory task. It still is not a normal purchase. Figure has no public price, checkout or standard delivery terms. The right conclusion is not “Figure 03 is available”; it is “Figure 03 has the stronger public scale case.” Choose Figure 03 instead of Apollo when: you want to evaluate dexterous bipedal work and are comfortable joining a vendor-led AI and fleet-development program. Do not choose Figure 03 when: you need a wheeled configuration, public commercial terms or an off-the-shelf delivery date. 5. Unitree H1: the best platform for teams that can build the application themselves Unitree H1 is different from the enterprise programs above. You can approach Unitree to buy the hardware. Its official shop shows a $90,000 listing, while also telling buyers to contact the company for the real price. Treat $90,000 as a public reference point, not a guaranteed quote. The H1 is 1.8 metres tall, weighs 47 kg and reaches 3.3 m/s. Unitree publishes an 864 Wh quick-swappable battery but does not turn that figure into a job-specific runtime guarantee. The base robot is known for locomotion, not for arriving with a finished warehouse workflow, acceptance test and enterprise service package. That distinction matters. Apollo is sold as a program around a job. H1 is a platform for a capable lab, integrator or internal robotics team. The Unitree H1 review covers its strengths and manipulation limits in more detail. Choose H1 instead of Apollo when: you need full-size hardware now, have robotics engineers, and want control over the software and integration work. Do not choose H1 when: you expect a vendor to own the task, safety case, deployment and support outcome end to end. 6. HMND 01: an emerging European enterprise alternative HMND 01 is an early but increasingly commercial option. Humanoid publishes a 15 kg payload, three-hour average runtime, 1.5 m/s speed and 29 body degrees of freedom excluding end effectors for the Alpha Bipedal. Buyers can choose modular five-finger hands or parallel grippers. That modularity makes it relevant to an Apollo shortlist. Both companies are trying to avoid a single fixed robot configuration and build around different job requirements. Humanoid's commercial position also changed materially in May 2026. It signed a binding, phased RaaS agreement with Schaeffler targeting a four-digit number of wheeled units by 2032, with the first systems due in German production environments in late 2026. Bosch became Humanoid's contract-manufacturing partner after a successful intralogistics proof of concept. The distinction is important: those agreements strengthen the HMND 01 family and vendor, but they do not prove the Alpha Bipedal at production scale. The Schaeffler rollout is for wheeled units, while the current biped was unveiled in December 2025. Buyers should ask which chassis, gripper and software configuration is covered by any proposal. Choose HMND 01 instead of Apollo when: you want a European industrial program, value interchangeable hands and grippers, and want a vendor with disclosed RaaS and manufacturing partners. Do not choose HMND 01 when: you specifically need a production-proven biped with customer-verified uptime and throughput. 7. Kinisi KR1: the most practical wheeled alternative on paper KR1 is a wheeled humanoid aimed at logistics and manufacturing. Its omnidirectional base should be easier to stabilize, service and certify on flat floors than a biped. The published record points to a 25 kg payload and an eight-hour battery, which would put it close to Apollo's original payload while doubling the original runtime. The words “on paper” are doing work here. Kinisi's official site confirms the wheeled platform, but several headline figures in the current RoboZaps record come from press or third-party sources. The database classified KR1 as a paid deployment, yet no verified customer deployment record was attached in the 17 July snapshot. Kinisi now says it is joining Bear Robotics, bringing KR1 and its engineering team into Bear's deployed fleet, enterprise customer base and cloud infrastructure. That could improve procurement and support, but the integration is still new. Treat access as a direct commercial inquiry until Kinisi or Bear publishes standard terms. KR1 should be on a discovery call, not at the top of a procurement recommendation. Ask Kinisi to confirm the payload at full reach, battery runtime under task load, customer site, service coverage and delivery terms in writing. Choose KR1 instead of Apollo when: your facility is flat, wheels are acceptable, and runtime matters more than stair access. Do not choose KR1 when: you need a mature evidence pack before the first vendor call. 8. Sanctuary AI Phoenix: an AI and dexterity alternative, not a simple chassis swap Phoenix has long been one of Apollo's closest conceptual rivals. Sanctuary AI built the program around dexterous manipulation and its Carbon AI system rather than athletic locomotion. Generation 8 moved to a wheeled base after customer feedback that bipedal legs were too fragile for a strong, precise torso. Be careful with its specifications. Sanctuary's earlier Phoenix release published a 25 kg payload, but the Generation 8 announcement does not repeat that number. It is safer to treat 25 kg as family history than as a current Gen 8 guarantee. Sanctuary's current positioning has also shifted. The company now sells production-ready Physical AI across commercially available robotic systems, with a longer path back to humanoid embodiments. That can be attractive if your hard problem is task learning and manipulation policy. It is less useful if your procurement brief says “deliver one humanoid body with fixed specs and support terms.” Choose Phoenix or Sanctuary instead of Apollo when: dexterity, training data and task-policy performance matter more than a particular locomotion platform. Do not choose it when: you need a straightforward like-for-like Apollo hardware purchase. Which Apollo alternative should you choose? If the requirement is “a humanoid robot we can buy from a public product page,” the shortlist gets much shorter. Our guide to humanoid robots for sale separates listed hardware from pilot-only programs. Robots we left off the main shortlist Tesla Optimus: a major Apollo competitor, but not an alternative a third-party enterprise can procure. Tesla's public program remains internal and announced rather than externally available. Figure 02: it has the best published shift data in Figure's history, including 1,250 hours at BMW, but Figure has retired it in favor of Figure 03. A retired generation is evidence for the vendor, not a new shortlist product. Unitree G1: affordable and easy to source compared with full-size enterprise humanoids, but its 2 kg payload and developer focus put it in a different job class from Apollo. AgiBot A2, Dobot Atom and other “paid deployment” records: they may deserve a later shortlist, but their current RoboZaps records did not have the same verified customer-evidence trail as Digit or Walker S2. A deployment label by itself is not enough for a top recommendation. What to demand in an Apollo alternative pilot Do not let vendors choose different success measures. Give every candidate the same work cell, objects, shift pattern and pass/fail rules. Ask for: Task completion rate over a full shift, not a selected demo run.Human interventions per hour and the exact definition of an intervention.Uptime after charging, battery swaps, safety stops and fault recovery.Payload at the actual reach and posture the job requires.Cycle time at the required placement tolerance.Safety architecture, certification status and responsibility for the site safety case.Integration scope for WMS, MES, conveyors, doors, scanners and existing automation.Service response time, spare-parts location and who performs field repairs.Pilot price, production price, software fees and the condition for moving from pilot to fleet.A written plan for what happens if the robot misses the acceptance target. The best robot is the one that clears that test with the least hidden integration work. A higher payload or faster walking speed does not rescue a system that needs constant human recovery. Final verdict Digit is the closest practical Apptronik Apollo alternative in July 2026. It gives buyers the strongest combination of commercial access, named deployments and a defined logistics job. Walker S2 is the better first call for automotive factories, while Atlas is the hardware leader for teams planning beyond the current year's allocation. Figure 03 belongs in a strategic pilot conversation, not a standard purchase order. Unitree H1 is the better route for teams that want hardware and intend to build the application themselves. HMND 01 now has a credible enterprise route, although its biped remains early. KR1 and Phoenix each solve a narrower problem. None should be treated as a drop-in Apollo replacement without a site test. Apollo itself remains credible. The mistake is assuming credible means purchasable or proven at scale. Compare every vendor on the same job, with the same acceptance test, and make the unpublished numbers part of the procurement process. --- # 6 Best AgiBot A2 Alternatives in 2026 (By Use Case) URL: https://blog.robozaps.com/b/agibot-a2-alternatives Author: Radica Maneva Published: 2026-07-17T06:44:17.370Z Updated: 2026-08-01T23:02:37.918Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: comparisons TL;DR: UBTECH Walker X is the closest external A2 alternative for customer-facing service. AgiBot X2 is the lower-cost event option, Unitree G1 suits developers, Mirokai fits hospitality and healthcare, and Digit is stronger for logistics. PUDU D9 remains a watchlist candidate. Key takeaways: - UBTECH Walker X is the closest external alternative for reception, guidance and other customer-facing service work. - AgiBot X2 starts at $24,240 for compact interactive use, while Unitree G1 starts at $13,500 for developer-led projects. - Mirokai is the strongest fit for healthcare and hospitality when social character matters more than bipedal walking. - Digit has the best commercial evidence for logistics, but it is not a reception or customer-service robot. - PUDU D9 is a promising full-size watchlist option whose delivered functions and support package still need contract-level confirmation. FAQ: Q: What is the closest alternative to the AgiBot A2? A: UBTECH Walker X is the closest external alternative for reception, guidance and public-facing service. It is a commercial service humanoid with multimodal interaction and autonomous navigation, although its current pricing is not public. Q: What is the cheapest AgiBot A2 alternative? A: Unitree G1 has the lowest clear public price in this shortlist at $13,500. It is a development platform rather than a turnkey service robot. AgiBot X2 starts at $24,240 and is a closer fit for events and interactive commercial use. Q: Which A2 alternative is best for events? A: AgiBot X2 is the strongest event and entertainment alternative because it is compact, easier to transport and built around performance and multimodal interaction. Mirokai is a better fit when the experience needs a distinctive social character and autonomous indoor movement. Q: Which alternative is best for warehouse work? A: Digit is the strongest warehouse alternative because it has commercial logistics deployments and is designed around repetitive material movement. PUDU D9 is a broader full-size manipulation platform, but its customer-ready functions need to be confirmed. Q: Should you choose Walker X or AgiBot X2 instead of A2? A: Choose Walker X when you want the closest external service-humanoid comparison and can work through a contact-sales process. Choose AgiBot X2 when a smaller, lower-priced robot for events or interactive displays fits the job better than an adult-size service robot. Q: Is Tesla Optimus a real AgiBot A2 alternative? A: It is a long-term competitor, not a practical current substitute. Tesla has not published a public order path, MSRP, delivery commitment or current commercial specification sheet for Optimus. The best AgiBot A2 alternative depends on the job. UBTECH Walker X is the closest external option for customer-facing service. AgiBot X2 is the lower-cost choice for events and interactive displays. Unitree G1 makes more sense for developers. Mirokai is better suited to healthcare and hospitality, while Digit is the stronger choice for repetitive logistics work. Those robots are not interchangeable. The A2 is primarily an interactive service humanoid, not a general answer to every reception, research and factory workflow. A good shortlist starts with the task, then checks whether the robot can actually be bought and supported in your region. AgiBot A2 alternatives at a glance Prices and buying paths were checked on 17 July 2026. Shipping, taxes, configuration, integration and local support can materially change the cost. What are you replacing when you replace an AgiBot A2? AGIBOT positions the base A2 for reception, customer service, exhibition presentations, retail guidance and business enquiries. The base model is quote-only. A2 Lite's $44,560 store price belongs to that separate variant, as our AgiBot A2 price and specifications review explains. That is the job this article uses as its baseline. A development platform or warehouse robot may be impressive, but it is not a direct substitute for a customer-facing service machine. 1. AgiBot X2: best lower-cost interactive alternative Best for: events, entertainment, branded displays and smaller public-facing installations. AgiBot X2 is the easiest place to start if you like the A2's interaction style but do not need a full-size service robot. AGIBOT lists the standard X2 at $24,240. It stands about 131 cm tall, weighs about 35 kg and has 25 degrees of freedom. The standard X2 is built around performance and interaction. It can handle speech, gestures, facial expressions and pre-set full-body motions. The X2 Ultra adds more sensors, an Orin compute option and support for secondary development, but it remains a smaller platform with a maximum three-kilogram payload in specific postures and no more than one kilogram across its full range of motion. Where X2 is better: it has a lower published entry price, is easier to transport and is a natural fit for stages, exhibitions and branded experiences. Where A2 is better: the A2 has adult height, greater load capacity and a stronger fit for a formal reception or service role where physical presence matters. Buying note: the standard X2 has a public store price, but buyers still need to confirm shipping, customs, software packages, navigation options and local support. 2. UBTECH Walker X: closest external service-humanoid alternative Best for: reception, guidance, demonstrations and service environments that need a bipedal humanoid. Walker X is the closest like-for-like external alternative in this group. UBTECH describes it as a commercial service humanoid and promotes visual navigation, hand-eye coordination, facial recognition, multimodal interaction and smart-device control. It is smaller than the A2 at 130 cm, although its 63 kg weight is close. UBTECH publishes 41 degrees of freedom, a maximum walking speed of 3 km/h, a two-hour combined-duty runtime and a 1.5 kg load per arm in the extended position. Where Walker X is better: it is purpose-built around service interaction and has a detailed public specification set from an established commercial robotics company. Where A2 is better: A2 is taller and publishes a much higher headline payload. AGIBOT also provides a clearer current family and global-store path. Buying note: UBTECH says it began commercializing Walker X in 2022, but that manufacturer statement does not prove current availability or support in a particular region. The current purchase route is contact sales. Ask for lead time, supported territories, language performance, local certification and field-service coverage in writing. 3. Mirokai: best for healthcare and hospitality Best for: hospitals, senior living, hotels, airports, museums and other environments where approachability matters more than bipedal walking. Mirokai takes a different approach. It is a social service robot that moves on an omnidirectional base, uses an animated face and is designed to interact without looking like an industrial machine. Enchanted Tools says its first commercial deployments are open across healthcare, hospitality, retail and public venues. Enchanted Tools says the current robot can navigate carpets, cables and indoor thresholds, dock for charging, monitor its surroundings and handle light everyday objects. These are manufacturer claims, not independent performance results. A site pilot still needs to measure navigation failures, intervention rate and safe stopping around staff and visitors. Where Mirokai is better: it has a friendlier character, a stronger healthcare and hospitality focus, and a mobility system suited to smooth indoor floors. Where A2 is better: A2 has an adult humanoid form, greater stated manipulation capacity and a better fit when human-like physical presence is part of the brief. Buying note: pricing is not public. Enchanted Tools has documented pilots and opened its first commercial deployments, but a buyer should still ask for supported tasks, uptime, charging behavior, staff intervention, data handling and service coverage. 4. Unitree G1: best developer alternative Best for: robotics teams, universities, embodied-AI developers and buyers who want a comparatively affordable humanoid platform. Unitree G1 starts at $13,500, making it the least expensive robot in this shortlist with a clear public price. It is 132 cm tall, weighs about 35 kg and runs for roughly two hours. The standard model has 23 joints, while EDU configurations can reach 43 with optional hands and added waist motion. The G1 is a platform, not a finished customer-service product. Its value is the hardware, motion system and development path. The EDU version supports secondary development, and Unitree publishes SDK and ROS material for technical teams. Where G1 is better: it costs less, has a stronger developer ecosystem and is better suited to custom robotics research. See our current Unitree G1 price and configuration review before comparing the base and EDU versions. Where A2 is better: A2 is built around service interaction, carries more and presents at adult height. A non-technical buyer should not assume a stock G1 arrives ready to run reception or guided-tour workflows. Buying note: Unitree itself warns that humanoid robotics remains early and that some demonstrated functions are still being developed. Budget for hands, compute, integration, safety work, training and support, not just the base robot. 5. PUDU D9: best full-size watchlist alternative Best for: teams comparing full-size platforms for manipulation, mobility and customer-engagement use cases. PUDU's current D9 page describes a robot close to the A2 in physical scale. PUDU publishes a height of 170 cm, weight of 65 kg, 42 degrees of freedom and a maximum test payload of 20 kg. It also pairs a bipedal body with dexterous hands and a sensor stack that includes RGB, RGB-D, tactile sensing and an IMU. The attraction is straightforward: D9 combines a human-scale body with a stronger manipulation brief than most social-service robots. PUDU also has an established commercial service-robot business, which could matter if D9 inherits its deployment and support discipline. Where D9 may be better: higher published test payload, a full-size body and a design aimed at both object manipulation and customer engagement. Where A2 is better today: A2 has a clearer defined service role and a more established procurement story for its family. Buying note: PUDU explicitly warns that some promotional functions are not yet available to customers and that shipped versions may differ. Keep D9 on the shortlist only after the exact configuration, deliverable functions, acceptance tests and support package are written into the offer. 6. Digit: best industrial alternative with commercial evidence Best for: warehouses, distribution centers and manufacturing workflows built around repetitive material movement. Digit is the right alternative when the real job is moving totes or materials rather than talking to visitors. Agility Robotics documents a commercial deployment at GXO, while its 2026 Toyota agreement followed a pilot and covers planned deployment. The current platform combines Digit with Arc fleet-management software and a deployment service. The current RoboZaps database records a 16 kg payload and a four-hour stated battery life. More important than those headline numbers is the GXO evidence: Digit entered regular commercial logistics operations rather than appearing only in a demonstration. Where Digit is better: it has stronger commercial evidence for repetitive industrial work, a dedicated fleet platform and a deployment model built around enterprise operations. Where A2 is better: A2 is designed for human-facing interaction. Digit's form, end effectors and software are optimized for material handling, not reception, guided presentations or brand engagement. Buying note: Digit is an enterprise deployment, not a public checkout purchase. Expect a site assessment, integration work, a managed rollout and contract pricing. Which AgiBot A2 alternative should you choose? For reception, events and guided experiences Choose Walker X if you want the closest external full-service humanoid. Choose AgiBot X2 when price, transport and entertainment value matter more than adult height or payload. Choose Mirokai when warmth, character and safe indoor mobility are central to the experience. The A2 itself remains a sensible candidate for a supervised service-robot pilot, provided AGIBOT confirms the exact variant, territory, support model and delivered software. For development and research Choose Unitree G1 EDU. It has the clearest development story in this group and a much lower entry price than enterprise service robots. Do not choose it to avoid integration work; choose it because your team wants to do that work. For warehouse or factory work Choose Digit when commercial operating evidence matters more than human-like presentation. Keep PUDU D9 on the watchlist if your workflow needs a full-size robot with more general manipulation, but require a customer-ready feature list before treating it as deployable. For home use Choose none of these. The A2 is a commercial service robot, while the strongest alternatives above are enterprise or developer platforms. A home buyer needs a separate shortlist built around domestic safety, privacy, remote supervision, support and actual home-task capability. Why Tesla Optimus is not in the main six Tesla Optimus is an obvious name, but it is not a practical A2 replacement for a buyer acting now. Tesla does not offer a public order page, MSRP, current commercial data sheet or delivery commitment. It may become a major general-purpose platform, but today it is a watchlist product rather than a procurement alternative. This distinction keeps the shortlist useful. A robot does not become an alternative because it appears in the same demo reel. It needs a defined job, a credible buying path and enough commercial detail to write an acceptance plan. What to ask before buying any A2 alternative The base price is rarely the number that decides whether a humanoid deployment works. Our humanoid robot cost guide covers the wider cost stack. For this shortlist, ask each vendor for the following in writing: the exact model, hands, sensors, compute and charging hardware;the tasks included at delivery, rather than shown in a demonstration;runtime under your workflow and the charging or battery-swap process;supported languages, noise conditions and interaction limits;the autonomous operating boundary and every human-intervention trigger;local certification, site-safety documentation and risk-assessment support;warranty, spare parts, field service, software licensing and security updates;lead time, shipping, customs, installation, training and acceptance tests. A pilot should measure task success, intervention rate, uptime and staff effort. A smooth ten-minute demonstration does not tell you what the robot will do during a full shift. Bottom line UBTECH Walker X is the closest external AgiBot A2 alternative for customer-facing service. AgiBot X2 is the better lower-cost choice for events and compact interactive installations. Unitree G1 is the best developer platform, Mirokai is the most focused healthcare and hospitality option, and Digit has the strongest commercial case for logistics work. PUDU D9 may become a serious full-size rival, but it belongs on a conditional watchlist until the customer-ready configuration and supported functions are clear. The most important choice is not the robot brand. It is whether you need conversation, development hardware or productive physical work. Define that first, then compare price, maturity and support inside the right category. Compare on RoboZaps: AgiBot A2, Lingxi X2, Walker X, Mirokai, Unitree G1 and Pudu D9. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # The 7 Best Unitree R1 Alternatives in 2026 URL: https://blog.robozaps.com/b/unitree-r1-alternatives Author: Radica Maneva Published: 2026-07-17T06:16:18.926Z Updated: 2026-08-01T23:02:44.256Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: alternatives-competitors TL;DR: No robot matches the R1's $4,900 entry price, size and brand support. Unitree G1 is the best overall upgrade, Booster K1 Geek is the closest priced developer option, and 1X NEO is the better home-focused choice. Choose by SDK access, hands, sensing and delivery status, not demo footage. Key takeaways: - There is no one-for-one R1 replacement at its $4,900 to $5,900 headline price. - Unitree G1 is the best overall upgrade for stronger sensing and roughly twice the stated runtime. - Booster K1 Geek is the closest comparable biped outside Unitree near the R1's price and includes secondary-development support. - Booster T1 is a stronger locomotion and competition platform, but its current price is not publicly disclosed. - 1X NEO is the clearer home-focused option, but it is an early-access service that may involve remote expert supervision. FAQ: Q: What is the best alternative to the Unitree R1? A: The Unitree G1 is the best overall alternative. It adds a depth camera, 3D LiDAR, about two hours of battery life and an EDU path with optional dexterous hands. It starts at $13,500, more than twice the R1 AIR's price. Q: What is the closest alternative to the R1 under $10,000? A: Booster K1 Geek is the closest comparable biped outside Unitree. It starts at $5,999 and supports secondary development, but it is smaller at 95 cm and its Geek battery is rated for about 30 minutes of walking. Q: Is the Unitree G1 programmable? A: Only the G1 EDU supports secondary development. Unitree's $13,500 base G1 does not. Buyers should request an EDU quote if custom software is required. Q: Which R1 alternative is open source? A: AgiBot X1 has the broadest published open-source package, including a development guide, bill of materials, mechanical files, training code and inference code. EngineAI also positions PM01 as a fully open embodied-intelligence platform. Q: Is there a better humanoid than the R1 for home use? A: 1X NEO is the clearer home-focused product. It is offered for $20,000 or $499 per month and is designed around chores, self-charging and voice interaction. It is an early-access service that may use remote expert supervision for unfamiliar tasks. The Unitree R1 is hard to beat on headline price. The R1 AIR starts at $4,900, the standard R1 at $5,900, and both put a 1.23 m biped in a price bracket that did not exist a year earlier. There is a catch. The cheaper R1 models do not support secondary development, have no hands, run for about one hour and use a simpler camera setup than more expensive research platforms. If any of those limits defeats your reason for buying a humanoid, the cheapest robot may not be the cheapest useful robot. We screened the current RoboZaps database, then checked each shortlisted model against its manufacturer's current product material. The result is a use-case list, not seven robots pretending to be interchangeable. TL;DR No robot matches the R1's $4,900 entry price, size and brand support. Unitree G1 is the best overall upgrade, Booster K1 Geek is the closest priced developer option, and 1X NEO is the better home-focused choice. Choose by SDK access, hands, sensing and delivery status, not demo footage. The best Unitree R1 alternatives at a glance Prices exclude shipping, tax, customs and optional hardware unless the manufacturer says otherwise. “Orderable,” “taking deposits” and “open source” are different statuses. Why look for an R1 alternative? The R1 remains the value benchmark. Unitree's current specification table lists the R1 AIR at $4,900 and the standard R1 at $5,900, with the R1 EDU sold by quotation. All three are roughly 1.23 m tall and use a quick-release battery rated for about one hour. The deciding differences sit lower in the table: The R1 AIR and standard R1 do not support secondary development. That is reserved for the R1 EDU.The cheap versions do not have dexterous hands.They use monocular or binocular cameras rather than the depth-camera-and-LiDAR setup found on the G1.One hour is enough for classes and demonstrations, but short for repeated experiments or events. Read the full Unitree R1 price and configuration review if those edition differences are still unclear. If you already know why the R1 falls short, use that missing capability to choose from the list below. How we chose these alternatives We started with RoboZaps' current robot records, then checked price, size, runtime, development access and availability against manufacturer pages. A robot received a place only if it offered a clear answer to one of the R1's limitations. We did not rank a prototype above an orderable robot because its demo looked better. Nor did we treat an open hardware repository as proof that a complete robot can be delivered next week. Where a manufacturer does not publish a price, the table says so. 1. Unitree G1: best overall R1 alternative The Unitree G1 is the safest upgrade for buyers who like Unitree's hardware and support ecosystem but need a more capable platform. It is 1.32 m tall, weighs about 35 kg and starts at $13,500. Unitree lists a depth camera, 3D LiDAR and about two hours of battery life, twice the R1's stated runtime. The G1 also offers a clearer path to manipulation. Its EDU configurations can reach 43 degrees of freedom and add three-fingered dexterous hands. That is a real difference from the handless R1 AIR and standard R1. There is still an edition trap. The $13,500 G1 does not support secondary development. Buyers who need custom software must request G1 EDU pricing, just as R1 developers need the R1 EDU. The Unitree G1 review covers those tiers in detail. Our humanoid robot cost guide explains why the delivered budget can extend well beyond the base robot price. Best for: teams that want stronger sensing, longer runtime and a supported upgrade pathPublic price: $13,500 for the base G1; G1 EDU is quote-onlyCurrent status: publicly orderable, with stock and delivery timing to confirmChoose it over R1 when: sensing and future manipulation matter more than the lowest entry priceKeep the R1 when: portability and budget matter more than LiDAR or longer runtime 2. Booster K1 Geek: closest alternative near the R1's price Booster's K1 Geek is the most interesting cross-manufacturer answer because it starts at $5,999, only $99 above the standard R1's headline price. It is smaller at about 95 cm and lighter at 19.5 kg, but its base proposition is more developer-friendly. Booster lists a stereo depth camera, 9-axis IMU, 48 TOPS of AI performance and support for secondary development. Those features make the K1 Geek a better classroom or introductory embodied-AI platform when writing and testing software is the point of the purchase. The compromise is endurance and scale. Booster rates the Geek edition for about 30 minutes of walking at 0.4 m/s. Its smaller body is easier to transport and safer to manage in a lab, but it is not a full-size stand-in for teams that specifically need the R1's reach or proportions. Best for: university classes, RoboCup teams and first embodied-AI projectsPublic price: from $5,999Current status: orderable from BoosterChoose it over R1 when: development access matters more than body sizeKeep the R1 when: you want the larger robot and can live with built-in behaviors or buy the EDU tier 3. EngineAI PM01: best open commercial research alternative The PM01 is the strongest alternative for a lab that wants an open platform without shrinking to a sub-one-metre robot. EngineAI describes it as a fully open embodied-intelligence platform. The business edition stands 1.4 m tall, weighs about 42 kg, has 23 degrees of freedom and uses a 10,000 mAh quick-release battery rated for nearly two hours. Its official specifications also list hardware support for movement above 2 m/s. That is useful for locomotion research, but the more important buyer point is the open development posture. The PM01 is designed to be worked on rather than simply demonstrated. EngineAI currently routes buyers to a purchase or enquiry flow but does not show a stable public price on the product page. Get a written quote that identifies the exact edition, compute module, warranty, software access and delivery schedule before comparing it with an R1 EDU quote. Best for: research labs that want an open, near-full-size humanoidPublic price: not disclosed on the current manufacturer pageCurrent status: commercial purchase route available; confirm the edition and lead timeChoose it over R1 when: openness, runtime and locomotion research justify more weight and costKeep the R1 when: you need a lighter platform with transparent entry pricing 4. AgiBot X1: best fully open-source alternative AgiBot X1 is less a retail substitute and more a complete open research route. AgiBot publishes a development guide, bill of materials, mechanical drawings, assembly instructions, inference code and training code. The official product page lists 34 active degrees of freedom, a 1.3 m height, 33 kg weight, two-hour runtime and 1 m/s maximum walking speed. That makes X1 a serious option for a team that wants to inspect and modify the stack instead of accepting a vendor-controlled platform. The two grippers and open materials also make it more relevant to manipulation research than a handless base R1. Openness does not remove procurement work. AgiBot does not publish a simple complete-system price on the X1 page. Confirm whether you are buying a complete robot, sourcing a kit or building from published materials. Best for: teams that value source access and hardware controlPublic price: not disclosedCurrent status: open-source platform; delivery of a complete system requires confirmationChoose it over R1 when: access to designs and code is the main requirementKeep the R1 when: you want a simpler commercial purchase and support path 5. Booster T1: best for locomotion and competition research Booster T1 is what the K1 grows into when the research program needs a larger, more durable platform. It stands about 1.18 m tall, weighs roughly 30 kg and offers 23 degrees of freedom in the base configuration, rising to 31 with grippers or 41 with dexterous hands. Booster says the T1 supports comprehensive APIs, an open-source framework and open tools. It rates the platform for about two hours of walking or four hours of standing, and says more than 50 robotics teams and research institutes use it. Those details matter more than the soccer videos: they point to a platform built for repeated experimentation. Booster does not publish a current T1 price on its product page, so buyers need a written quote. The T1 only makes sense if the lab will use its software stack, endurance and competition ecosystem enough to justify moving beyond the R1's transparent budget category. Best for: reinforcement learning, dynamic locomotion and robot competitionsPublic price: not disclosed on the current manufacturer pageCurrent status: enquiry-based; confirm configuration, price and lead timeChoose it over R1 when: research maturity and endurance justify a much larger budgetKeep the R1 when: you need an affordable platform for demonstrations or lighter experimentation 6. Noetix N2 EDU: best compact alternative with hand options The Noetix N2 is physically close to the R1: 1.18 m tall, about 30 kg and rated for one to two hours of battery life. Its product table lists depth cameras, an IMU and either a spherical hand or five-fingered hand configuration. As with Unitree, edition choice matters. The standard N2 does not include secondary development. The N2 EDU does, and adds a higher-compute configuration with an NVIDIA Jetson Orin Nano Super. Noetix also lists about 5 kg of continuous walking carrying capacity, although buyers should ask how that figure changes with posture and end effector. Noetix does not publish a price for either edition. That prevents it from beating the R1 on value, but it belongs on the shortlist for teams that want a compact robot with hand and developer options. Best for: compact manipulation, education and researchPublic price: contact consultantCurrent status: enquiry-based commercial productChoose it over R1 when: hand options and an EDU development tier matterKeep the R1 when: transparent pricing and a larger support ecosystem matter more 7. 1X NEO: best home-focused alternative NEO is the only robot here sold around a specific home-service proposition. 1X offers early-access ownership for $20,000 or a $499 monthly subscription, with a $200 refundable deposit and US deliveries starting in 2026. It is not a cheaper research platform. NEO is designed around voice interaction, scheduled chores, self-charging and a mobile app. 1X says it arrives with basic autonomy, while unfamiliar tasks can use scheduled remote expert supervision. That supervision model is central to the offer and should be treated as part of the service, privacy and operating-cost decision. Buyers considering the R1 for household chores should look at NEO because the R1 is not a home worker. Buyers who want to write locomotion policies or own a low-cost biped should stay with the developer platforms above. Best for: early adopters who want a home robot rather than a research platformPublic price: $20,000 ownership or $499 per month; $200 refundable depositCurrent status: taking deposits for US delivery starting in 2026Choose it over R1 when: home tasks and a managed service matter more than open developmentKeep the R1 when: ownership cost, portability and direct hardware experimentation matter more Which R1 alternative should you choose? When the Unitree R1 is still the better buy Keep the R1 on top of the shortlist if your main goal is to own a portable, full-bodied humanoid for demonstrations, movement experiments or education at the lowest public price. The AIR and standard editions make sense when built-in behaviors are enough. The EDU edition deserves a quote when you want to stay inside Unitree's ecosystem and add secondary development. The Booster K1 Geek is close on price but much shorter. Everything else either costs more, hides the price behind a sales call or asks the buyer to accept a less mature procurement path. That is why the R1 remains prominent in our guide to the cheapest humanoid robots. Robots that are not practical R1 alternatives Tesla Optimus, Figure 03 and Boston Dynamics Atlas appear in many humanoid roundups, but they are poor answers to this query. Optimus is not publicly orderable. Figure 03 is a vertically integrated platform without a public retail offer. Atlas is an industrial pilot robot with no public price. They may become important competitors in the broader humanoid market. They do not solve the immediate problem faced by someone comparing an orderable $4,900 to $5,900 R1 with another research, education or home robot. Unitree's R1-D starts at $4,290, but it is not a cheaper walking R1. It is a desktop or mobile dual-arm platform rather than a biped, so it belongs in a different comparison. Final verdict The best Unitree R1 alternative is the Unitree G1 if you can spend more and want better sensing, longer runtime and a clearer upgrade path. The Booster K1 Geek is the sharper choice when the budget needs to stay near $6,000 and secondary development is non-negotiable. Beyond those two, choose by job. PM01 and AgiBot X1 are the open-development picks. Booster T1 is the serious locomotion platform. Noetix N2 EDU adds compact manipulation options. NEO is the home bet. None of them erases the R1's price advantage. The right question is not “Which robot has the best demo?” It is “Which robot includes the capability I am actually paying for?” Frequently asked questions What is the best alternative to the Unitree R1? The Unitree G1 is the best overall alternative. It adds a depth camera, 3D LiDAR, about two hours of battery life and an EDU path with optional dexterous hands. It starts at $13,500, more than twice the R1 AIR's price. What is the closest alternative to the R1 under $10,000? Booster K1 Geek is the closest comparable biped outside Unitree. It starts at $5,999 and supports secondary development, but it is smaller at 95 cm and its Geek battery is rated for about 30 minutes of walking. Is the Unitree G1 programmable? Only the G1 EDU supports secondary development. Unitree's $13,500 base G1 does not. Buyers should request an EDU quote if custom software is required. Which R1 alternative is open source? AgiBot X1 has the broadest published open-source package, including a development guide, bill of materials, mechanical files, training code and inference code. EngineAI also positions PM01 as a fully open embodied-intelligence platform. Is there a better humanoid than the R1 for home use? 1X NEO is the clearer home-focused product. It is offered for $20,000 or $499 per month and is designed around chores, self-charging and voice interaction. It is an early-access service that may use remote expert supervision for unfamiliar tasks. Sources and references Unitree R1 product page — Unitree Robotics, accessed July 17, 2026Unitree G1 product page — Unitree Robotics, accessed July 17, 2026Unitree G1 official store page — Unitree Robotics, accessed July 17, 2026Booster K1 product page — Booster Robotics, accessed July 17, 2026EngineAI PM01 product page — EngineAI, accessed July 17, 2026AgiBot X1 product page — AgiBot, accessed July 17, 2026AgiBot X1 development guide — AgiBot, accessed July 17, 2026Booster T1 product page — Booster Robotics, accessed July 17, 2026Noetix N2 product page — Noetix Robotics, accessed July 17, 20261X NEO order page — 1X Technologies, accessed July 17, 2026RoboZaps Unitree R1 API record — RoboZaps Robot Database, retrieved July 17, 2026Unitree R1-D product page — Unitree Robotics, accessed July 17, 2026 Compare on RoboZaps: Unitree R1, Unitree G1, Booster K1, PM01, T1 Standard and N2. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 7 Best Agility Robotics Digit Alternatives in 2026 URL: https://blog.robozaps.com/b/agility-robotics-digit-alternatives Author: Radica Maneva Published: 2026-07-17T06:07:14.677Z Updated: 2026-08-01T10:54:01.581Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: alternatives-competitors TL;DR: UBTECH Walker S2 is the closest direct Digit alternative. Atlas has the strongest industrial hardware, while Apollo 2 and Figure 03 suit joint-development programs. KR1 is the better flat-floor bet. Digit still has the strongest disclosed commercial logistics record, so switching only makes sense when another robot fits the task or operating model better. Key takeaways: - UBTECH Walker S2 is the closest direct Digit alternative, with comparable payload and a three-minute autonomous battery swap. - Boston Dynamics Atlas has the strongest published industrial hardware, but its 2026 customer access remains limited. - Digit still has the strongest disclosed commercial logistics evidence, including more than 100,000 totes moved at GXO. - A wheeled mobile manipulator such as KR1 can be a better choice when a facility has flat floors and does not need bipedal movement. FAQ: Q: What is the best alternative to Agility Robotics Digit? A: UBTECH Walker S2 is the closest direct alternative. It is an industrial biped in mass production with a 15 kg payload and an autonomous battery-swap system. Boston Dynamics Atlas is the stronger option for high-payload industrial work, while Apptronik Apollo 2 and Figure 03 suit joint-development programs. Q: Which robot is most similar to Digit? A: Walker S2 is the most similar current product in use case and commercial positioning. Both are full-size industrial bipeds for factory or logistics work. Walker S2 emphasizes continuous operation through battery swapping; Digit has stronger public evidence from named commercial logistics deployments. Q: Is Agility Robotics Digit commercially available? A: Yes. Digit is offered through enterprise contracts, including Robots-as-a-Service arrangements. Agility has named commercial customers and integrator partners, but it does not publish a conventional unit price. Q: Is Walker S2 better than Digit? A: It depends on the facility. Walker S2 has a better published uptime design because it can replace its own battery. Digit has stronger disclosed site-level evidence, including more than 100,000 totes moved at GXO and a clearer North American service and integration model. Q: Can a company buy Apptronik Apollo 2? A: Apptronik is working with customers and operates Apollo 2 fleets at Robot Park and customer sites, but it has not published a standard price or conventional order page. Treat it as an enterprise pilot or joint-development conversation until Apptronik discloses broader commercial terms. Q: Is Unitree H1 a cheaper alternative to Digit? A: H1 has a public $90,000 store listing, but that is a hardware price. Digit is sold as a deployed solution with software, integration and support. H1 can be cheaper for a qualified robotics team that wants a development platform; it is not automatically cheaper for a production warehouse workflow. Q: Should a warehouse choose a wheeled robot instead of Digit? A: Often, yes. If the floor is flat and the robot does not need stairs, curbs or human-style foot placement, a wheeled mobile manipulator such as KR1 can offer longer runtime, greater stability and simpler maintenance. The site and task should decide the form factor. If you are comparing Agility Robotics Digit alternatives, start with the job, not the robot. Digit is unusually specific. It moves totes and materials through warehouses and factories, works with existing automation, and has public evidence from paid deployments. Agility says Digit has moved more than 100,000 totes at GXO, while its current published specifications include a 35 lb (16 kg) payload and four-hour battery life. That makes this a harder comparison than the usual “best humanoid robots” list. A faster prototype is not automatically a better alternative. Neither is a robot with dexterous hands but no service network, public price or customer deployment. The closest direct alternative is UBTECH Walker S2. It is an industrial biped in mass production, carries up to 15 kg and can swap its own battery in about three minutes. Boston Dynamics Atlas has much stronger published payload and environmental specifications, but its 2026 deployments are still limited to early customers. Apptronik Apollo 2 and Figure 03 are credible development partners for manufacturers that want broader manipulation and AI capability. For the full case for Digit itself, read our Agility Robotics Digit review or check the current Digit specifications and deployment record. Agility Robotics Digit alternatives compared The labels in that table matter. “Orderable,” “pilot,” “demonstration” and “paid deployment” describe different levels of risk. We checked the live RoboZaps robot database first, then used current manufacturer material to resolve generation changes and gaps. How we chose the best Digit alternatives We ranked these robots by how well they could replace Digit in an actual industrial decision. The order reflects: evidence from customer facilities, not demo videos alone;overlap with Digit's material-handling and logistics work;uptime, charging and fleet-operation design;integration, safety and support maturity;whether a buyer can start a real procurement or pilot conversation;the quality and recency of the public sources. Raw speed and degrees of freedom are secondary. A warehouse buyer cannot deploy a benchmark. The Digit baseline Agility Robotics Digit is a 175 cm biped built around industrial material handling. The current robot carries 16 kg, runs for up to four hours and connects to warehouse systems through Agility Arc. Pricing is not public; deployments are sold through commercial contracts that can include the robot, software, workcell integration and support. The proof is stronger than the spec sheet. Agility reports more than 100,000 totes moved at GXO, and Toyota Motor Manufacturing Canada signed a commercial Robots-as-a-Service agreement after its pilot. That is the standard every alternative has to meet. Digit is not the answer to every job. Its current task range is narrower than manipulation-first competitors, its public payload is modest, and the absence of a sticker price makes desktop ROI comparisons difficult. Those are valid reasons to look elsewhere. “A different robot looks more human” is not. 1. UBTECH Walker S2: the closest direct Digit alternative Best for: industrial operators that care most about uptime and want a bipedal alternative already in production. Walker S2 is the cleanest one-for-one competitor to Digit. UBTECH positions it for industrial manufacturing, and the company says mass production and delivery began in late 2025. The RoboZaps database records Walker S2 as a verified paid-deployment robot. Its standout feature is power management. Walker S2 carries two batteries and can replace a depleted pack at a battery station in roughly three minutes. It can continue operating on one battery during the swap. That design attacks a practical weakness of current humanoids: a four-hour battery does not cover a warehouse shift without charging or fleet overhead. The rest of the published hardware is close to Digit's job. Walker S2 handles up to 15 kg, works from floor level to a 1.8 m reach envelope and has 52 degrees of freedom. Digit carries 16 kg, so payload alone does not separate them. Where Walker S2 beats Digit Autonomous battery swapping offers a clearer route to round-the-clock operation.UBTECH reports 1,000-unit-level production and delivery, giving it a manufacturing-scale argument.Its broader articulated body and reach may suit factory tasks beyond tote transfer. Where Digit still wins Digit has better public, site-specific operating evidence. GXO's tote count and Agility's named commercial agreements are easier for a Western buyer to audit than fleet-scale claims without the same task-level detail. Agility also has a clearer North American integration and service story. Choose Walker S2 when autonomous battery swapping and Chinese industrial deployment matter more than Digit's Western logistics record. Start with the Walker S2 database record, then ask UBTECH for customer references, intervention rates, safety documentation and local service terms. 2. Boston Dynamics Atlas: the strongest hardware-led alternative Best for: demanding manufacturing tasks where payload, environmental protection and serviceability outweigh immediate fleet availability. The production version of Boston Dynamics Atlas is no longer just a research robot. Boston Dynamics now publishes a full industrial specification: 30 kg sustained payload, 50 kg instantaneous payload, 20 kg one-handed capacity, four-hour battery life, a three-minute autonomous battery swap, 56 degrees of freedom and IP67 protection. On paper, Atlas is in another payload class from the current 16 kg Digit. It is also designed for dirty industrial environments, field-replaceable components and integration with warehouse or manufacturing systems through Orbit. The constraint is access. Boston Dynamics says its 2026 deployments are committed to Hyundai and Google DeepMind, with a select group of early adopters following. Atlas is a real product and an active sales conversation, but it does not yet have Digit's open commercial path or disclosed logistics mileage. Where Atlas beats Digit Much higher published lifting capacity.IP67 protection and a wider operating-temperature range.Autonomous battery swapping and field-serviceable limbs.Stronger fit for heavy part sequencing and machine-tending work. Where Digit still wins Digit has the mature customer record. A buyer can inspect paid work, tote throughput, service partners and a repeatable deployment model today. Atlas offers stronger hardware with less commercial history. Choose Atlas when the task genuinely needs its strength or environmental hardening and your timeline allows an early-adopter program. Do not pay for a 50 kg lift if the workflow only moves 10 kg totes. 3. Apptronik Apollo 2: the best US joint-development option Best for: 3PLs and manufacturers willing to shape a deployment with the vendor. Apptronik introduced Apollo 2 in June 2026 in both bipedal and wheeled configurations. Fleets are collecting work data at Apptronik's Robot Park facilities and customer sites, with Google DeepMind involved in the AI program. The company markets Apollo for goods-to-person, person-to-goods, packout, kitting, inspection, sorting and machine tending. That combination makes Apollo 2 a serious Digit alternative for companies that want to co-develop a workflow. The optional wheeled base is especially sensible in flat facilities where legs add cost and failure points without adding much value. The current public record is still light on the numbers a procurement team needs. Apptronik has not published Apollo 2 pricing, throughput, intervention rates or a final production specification comparable with Atlas. The existing RoboZaps Apptronik Apollo review covers the earlier platform, so buyers should confirm which claims carry into Apollo 2. Where Apollo 2 beats Digit Bipedal and wheeled configurations let the body match the facility.The task roadmap is broader than Digit's current tote-handling focus.Apptronik's Robot Park program may suit companies that want to train a new workflow with the vendor. Where Digit still wins Digit's commercial evidence is ahead. Apollo 2 is collecting data and developing customer tasks; Digit already has named paid deployments and disclosed throughput. Choose Apollo 2 when you have a strong internal automation team, a clearly defined task and patience for a joint-development program. If you need a proven tote-transfer deployment, Digit remains the safer choice. 4. Figure 03: the best AI-first manufacturing alternative Best for: manufacturers prioritizing variable parts, vision-based manipulation and a broad AI roadmap. Figure 03 is built around Helix, Figure's vision-language-action system. The RoboZaps database records a 20 kg payload, five-hour battery and 1.2 m/s maximum speed, but the more important difference is how Figure approaches the work. At BMW's Spartanburg plant, Figure demonstrated Figure 03 sequencing parts that do not arrive in identical positions. The robot had to grasp thin components, reposition its body, pull a cart and correct for variation. That is a more dexterous, perception-heavy problem than Digit's best-proven tote workflow. The evidence label still needs discipline. Figure 02 accumulated meaningful production-line hours before retirement. Figure 03's June 2026 BMW work was described by Figure as a demonstration. It should not be presented as equivalent to a paid, multi-year Digit deployment. Where Figure 03 beats Digit Better fit for varied parts and dexterous manipulation.An AI stack designed to learn from visual input and demonstrations.A hardware design intended for both industrial work and eventual home use. Where Digit still wins Digit has the clearer commercial product, customer contract and operating history. Figure 03 has a higher capability ceiling, but a buyer is still underwriting more execution risk. Choose Figure 03 when the bottleneck is task variability rather than bulk material movement. If your workflow is stable and repeatable, the extra generality may not produce a better business case. 5. Kinisi KR1: the best alternative for flat warehouse floors Best for: warehouses and storerooms that need mobile manipulation but not bipedal locomotion. Kinisi KR1, also called Kinisi 01, puts a humanoid-style upper body on an omnidirectional wheeled base. That is not a compromise if the site is flat. Wheels are simpler, more stable and more energy-efficient than legs for long-distance movement on finished floors. The RoboZaps database records a 25 kg payload, eight-hour runtime and 2.4 m/s speed. Kinisi describes the platform as an edge-compute robot for picking, sorting, replenishment and transport, with autonomous, semi-autonomous and teleoperated modes. The company also announced that it is joining Bear Robotics. That could strengthen manufacturing, service and fleet infrastructure, but buyers should confirm what the transition means for product ownership, contracts and support before treating the current KR1 offer as fixed. Where KR1 beats Digit Longer published runtime and higher payload.Faster, more stable travel on flat floors.Onboard processing can reduce dependence on cloud connectivity.The wheeled base may be cheaper to maintain than a dynamically balanced biped. Where Digit still wins Digit can step over obstacles and move through spaces where wheels are a limitation. More importantly, its paid deployment evidence is clearer. KR1's public material emphasizes pilots and scaling rather than a GXO-style operating ledger. Choose KR1 when a site audit shows that legs add no value. It is the most useful alternative in this guide because it forces the right question: do you need a humanoid, or do you need a mobile manipulator? 6. Sanctuary AI Phoenix Gen 8: the manipulation-first alternative Best for: companies developing dexterous tasks where teleoperation and high-quality training data are part of the plan. Sanctuary AI built Phoenix around dexterous hands, tactile sensing and its Carbon control system. Its current Gen 8 robot uses a wheeled base, a deliberate response to customer feedback that legs were too fragile for the strong, precise upper body they wanted. That makes Phoenix a poor choice for sites that need stairs or rough floor transitions. It also makes Phoenix a credible alternative for repetitive manipulation at a workstation, where walking is secondary and reliable hands matter more. Be careful with specifications. Older Phoenix generations were bipedal, and their height, weight and payload figures still circulate. Those numbers should not be copied onto Gen 8 without confirmation. The current manufacturer material emphasizes data capture, field of view, telemetry and faster commissioning rather than a new public spec table. Where Phoenix beats Digit A stronger manipulation and tactile-sensing focus.Human-in-the-loop supervision and teleoperation are built into the development model.The wheeled form may be more stable for precise upper-body work. Where Digit still wins Digit is the better mobile logistics product. It has a commercial deployment model, fleet software and proof from a defined warehouse workflow. Phoenix is more attractive when the job is manipulation development, not tote transport. Choose Phoenix when your team wants to build and supervise complex hand tasks. Ask Sanctuary which work is autonomous, which requires teleoperation and what intervention rate it has measured on your task. 7. Unitree H1: the best alternative for teams that want to own the hardware Best for: robotics labs, integrators and developers that need a full-size platform with a public purchase path. Unitree H1 is the easiest robot here to put into a conventional hardware budget. Unitree's official store lists it at $90,000, excluding duties and integration costs. The RoboZaps database records a 1.8 m height, 47 kg weight and maximum speed of 3.3 m/s. The price does not make H1 a $90,000 replacement for Digit. A Digit contract includes a developed workflow, fleet controls, deployment engineering and support. H1 is a platform. Turning it into a reliable warehouse worker may require perception, task training, grippers, safety engineering, integration and ongoing maintenance that cost more than the robot. Where H1 beats Digit Public pricing and a direct hardware purchase path.High mobility and a mature developer ecosystem.Better fit for teams that want to own and modify the platform. Where Digit still wins Digit is sold as an operational solution. H1 is better treated as a development asset unless an integrator can prove the full workflow, safety case and support package. Choose H1 if you have robotics engineers and a research or integration objective. Do not choose it because the sticker price looks lower than an undisclosed RaaS contract. What about Tesla Optimus? Tesla Optimus belongs on the watchlist, not this ranked procurement list. The RoboZaps Optimus Gen 2 record currently labels it announced, and Tesla has not opened a normal external buying path with final public specifications and commercial support terms. Optimus may become a major Digit competitor. It is not a deployable alternative for an external warehouse buyer today. When the best Digit alternative is not a humanoid A humanoid earns its cost when the work requires human reach, human tools or movement through a site built for people. If the workflow is fixed, purpose-built automation often wins. For trailer unloading, Boston Dynamics Stretch handles cases up to 50 lb and already targets that exact task. For long warehouse transport, an AMR is usually more efficient than a walking robot. For a stationary pick-and-place cell, a fixed industrial arm may be faster, cheaper and easier to certify. Before shortlisting any humanoid, compare it with the simplest machine that can complete the job. Digit itself often works with AMRs rather than replacing them. Which Digit alternative should you choose? If two robots still look equal, ask for the same evidence from both vendors. A useful evaluation includes throughput on your objects, human interventions per shift, uptime, charging overhead, failed-grasp recovery, safety boundaries, integration work, local service response and the exact commercial terms. Our humanoid robot cost guide explains why hardware price is only one part of the comparison. If you need a sourceable shortlist rather than a research platform, compare the most advanced humanoids you can buy. Final verdict Walker S2 is the best direct Agility Robotics Digit alternative in 2026. It targets the same industrial buyer, has comparable payload and solves charging downtime with an autonomous battery swap. Atlas is the stronger machine for high-payload industrial work, but its customer access is earlier and tighter. Apollo 2 and Figure 03 are better choices when the task demands more general manipulation and the buyer is prepared to co-develop the deployment. Digit remains the benchmark for commercial logistics evidence. The right reason to reject it is a mismatch in payload, task range, facility design, uptime strategy or commercial model. The wrong reason is that a competitor has a better demo. RoboZaps can help scope the job, compare vendors and plan the deployment. See our robot procurement service. Frequently asked questions What is the best alternative to Agility Robotics Digit? UBTECH Walker S2 is the closest direct alternative. It is an industrial biped in mass production with a 15 kg payload and an autonomous battery-swap system. Boston Dynamics Atlas is the stronger option for high-payload industrial work, while Apptronik Apollo 2 and Figure 03 suit joint-development programs. Which robot is most similar to Digit? Walker S2 is the most similar current product in use case and commercial positioning. Both are full-size industrial bipeds for factory or logistics work. Walker S2 emphasizes continuous operation through battery swapping; Digit has stronger public evidence from named commercial logistics deployments. Is Agility Robotics Digit commercially available? Yes. Digit is offered through enterprise contracts, including Robots-as-a-Service arrangements. Agility has named commercial customers and integrator partners, but it does not publish a conventional unit price. Is Walker S2 better than Digit? It depends on the facility. Walker S2 has a better published uptime design because it can replace its own battery. Digit has stronger disclosed site-level evidence, including more than 100,000 totes moved at GXO and a clearer North American service and integration model. Can a company buy Apptronik Apollo 2? Apptronik is working with customers and operates Apollo 2 fleets at Robot Park and customer sites, but it has not published a standard price or conventional order page. Treat it as an enterprise pilot or joint-development conversation until Apptronik discloses broader commercial terms. Is Unitree H1 a cheaper alternative to Digit? H1 has a public $90,000 store listing, but that is a hardware price. Digit is sold as a deployed solution with software, integration and support. H1 can be cheaper for a qualified robotics team that wants a development platform; it is not automatically cheaper for a production warehouse workflow. Should a warehouse choose a wheeled robot instead of Digit? Often, yes. If the floor is flat and the robot does not need stairs, curbs or human-style foot placement, a wheeled mobile manipulator such as KR1 can offer longer runtime, greater stability and simpler maintenance. The site and task should decide the form factor. Disclosure: This is desk research based on RoboZaps data, manufacturer material and named deployment evidence. It is not a hands-on test. Product availability, pricing and support terms can change and should be confirmed with the manufacturer before procurement. --- # 7 Best Figure 03 Alternatives in 2026 (Compared) URL: https://blog.robozaps.com/b/figure-03-alternatives Author: Radica Maneva Published: 2026-07-17T05:58:47.826Z Updated: 2026-08-01T23:02:50.336Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: alternatives-competitors TL;DR: There is no drop-in Figure 03 replacement. Choose 1X NEO for home early access, Unitree H2 for buyable full-size hardware, G1 for development, Digit for logistics, Walker S2 for factory uptime, Apollo 2 for enterprise pilots, or Atlas for heavy industrial work. Figure 03 itself has no public order path. Key takeaways: - Closest home alternative: 1X NEO, but its early-access model includes remote expert supervision for unfamiliar tasks. - Closest buyable full-size alternative: Unitree H2, listed at $29,900 before tax and shipping. - Best lower-cost developer platform: Unitree G1, priced from $13,500. - Strongest real logistics evidence: Agility Robotics Digit, deployed under a multi-year Robots-as-a-Service agreement with GXO. - Best heavy-duty industrial option: Boston Dynamics Atlas, with a published 50 kg instantaneous payload and a first customer pilot at Hyundai. FAQ: Q: What is the best alternative to Figure 03? A: There is no single best replacement. 1X NEO is the closest home alternative, Unitree H2 is the clearest buyable full-size platform, Digit has the strongest logistics deployment evidence, and Atlas has the strongest published heavy-industrial specifications. Q: Can you buy Figure 03? A: No public Figure 03 price, preorder or normal order page was available when this guide was checked on 17 July 2026. Figure has produced more than 350 units for development, data collection and commercial use-case work, but that is not the same as public sale. Q: What is the cheapest Figure 03 alternative? A: Among the robots in this guide with public pricing, Unitree G1 is the cheapest at a starting price of $13,500. It is much smaller than Figure 03 and should be treated as a development platform, not a direct replacement for a full-size general-purpose worker. Q: Which Figure 03 alternative is best for the home? A: 1X NEO has the clearest consumer offer. It is available under a $20,000 ownership option or a $499 monthly subscription, with a refundable deposit. Buyers should understand that unfamiliar tasks may involve scheduled remote expert supervision. Q: Which alternative has the strongest real-world deployment? A: Agility Digit has the clearest commercial evidence through its multi-year Robots-as-a-Service deployment with GXO. Figure 03, Atlas and Apollo 2 have meaningful customer-site demonstrations or pilots, but those are different commercial stages. Q: Is Unitree H2 better than Figure 03? A: H2 is easier to buy and has a public $29,900 price. Figure 03 has stronger public evidence around general-purpose autonomy, dexterous manipulation and its BMW workflow. H2 is the better choice for teams that need full-size development hardware now; Figure 03 is the stronger program if autonomy is the main comparison and immediate access is not required. Figure 03 has one of the strongest combinations of humanoid hardware, dexterous manipulation and general-purpose AI in the market. It also has no public price, preorder or normal buying process. That makes the best Figure 03 alternative less about finding an identical robot and more about choosing the option that fits the job you need done. For a home robot, 1X NEO is the closest alternative with an order path. Unitree H2 is the clearest full-size option with a public price. Unitree G1 is the lower-cost development choice. Digit has the strongest commercial logistics evidence, Walker S2 is built around factory uptime, Apollo 2 is an enterprise learning platform, and Atlas is the heavy-duty industrial option. Figure 03 alternatives compared Prices and status were checked on 17 July 2026. A deposit, pilot, sales enquiry and commercial deployment are different things. None should be treated as proof that a robot can be delivered for your use case on a normal retail timeline. What a Figure 03 alternative needs to replace Figure describes Figure 03 as a general-purpose robot for the home, but the same platform is also being developed for manufacturing and logistics. Its official specifications include a height of 5 ft 8 in, a weight of 61 kg, a 20 kg payload, five-hour runtime and 1.2 m/s speed. Figure also says it has produced more than 350 units at BotQ and demonstrated Figure 03 performing a sequencing workflow at BMW’s Spartanburg plant. That is a difficult combination to match. Some rivals are easier to buy but less autonomous. Others have stronger commercial deployments but are designed for one narrow job. The full Figure 03 review covers its price record, Helix AI, hardware and deployment evidence. This guide focuses only on what to choose instead. Best Figure 03 alternatives 1. 1X NEO: the closest home alternative NEO is the obvious first candidate if you want the clearest public route to getting a full-size robot into a home. It is a poor fit if remote human supervision is a deal-breaker. 1X offers NEO under two early-access options: $20,000 for ownership or $499 per month, with a $200 refundable deposit. The company says US deliveries start in 2026. That is more concrete than Figure 03, which has no consumer order page or published price. NEO is also designed around home safety. It weighs 66 lb (about 30 kg), has a soft body, covered joints and low-inertia tendon drives. Its published runtime is four hours. The lower weight and padded exterior make it a more natural household proposition than most industrial humanoids. The catch is autonomy. 1X says NEO arrives with basic autonomy, while an expert can remotely supervise complex tasks it does not yet know. That support model may help the robot complete useful work earlier, but it also raises privacy, connectivity and service-dependence questions that buyers need to understand before placing a deposit. Read the deeper Figure 03 vs 1X NEO comparison for the privacy, autonomy and home-safety tradeoffs. 2. Unitree H2: the closest buyable full-size alternative H2 makes sense when you want human-scale hardware with a real public price. It is less compelling when the requirement is proven autonomous work rather than a platform your team can develop or integrate. Unitree lists the standard H2 at $29,900 before tax and shipping. The official specification gives a height of roughly 1.8 m, a weight of about 70 kg, 31 degrees of freedom and a 0.972 kWh quick-release battery. The EDU version is available through Unitree’s sales team for customers that need secondary development. On physical scale, H2 is much closer to Figure 03 than the smaller G1. It is a credible choice for labs, developers and organizations that need a full-size humanoid body now. What buyers should not assume is that a public purchase price also buys Figure-level autonomy. Unitree’s current H2 material emphasizes motion, computing and model support, but offers much less public evidence of autonomous production work. Our Figure 03 vs Unitree H2 comparison covers the hardware, hands, safety and buyer-fit differences in more detail. 3. Unitree G1: the lower-cost development option G1 belongs on the shortlist when price, availability and developer access matter more than human scale. It drops away when the job needs Figure 03’s reach, payload or home-oriented design. Unitree prices the G1 from $13,500. Its published specifications include a height of about 1.32 m, weight of about 35 kg, more than 2 m/s maximum speed, roughly two hours of battery life and 23 to 43 joint motors depending on configuration. Those numbers make G1 attractive for university labs, robotics teams and developers who need hardware to train policies, test locomotion or build demonstrations. Its smaller frame is easier to transport and cheaper to replace than a full-size industrial system. It is not a like-for-like Figure 03 substitute. The RoboZaps database records a 2 kg payload for G1, compared with Figure 03’s manufacturer-stated 20 kg. Buyers should also distinguish the base configuration from EDU and dexterous-hand variants, because the headline starting price does not include every capability shown in Unitree demonstrations. 4. Agility Robotics Digit: the strongest logistics alternative Digit is the practical option when the job is repetitive tote movement or warehouse material handling. It is not the answer if you want a five-fingered general-purpose robot for varied manipulation. Digit has the cleanest commercial deployment evidence in this group. GXO signed a multi-year Robots-as-a-Service agreement with Agility Robotics after a proof-of-concept pilot. At the SPANX facility, Digit moves totes from other robots and places them onto conveyors inside a live warehouse workflow. Agility’s published specifications put Digit at about 5 ft 9 in and 140 lb, with a 35 lb payload and up to four hours of battery life. It can autonomously dock, and Agility Arc handles fleet deployment and management. The specialization is the point. Digit is not trying to be a household assistant or a universal dexterous platform. It is a bipedal mobile manipulation robot built around logistics. That makes it a weaker match to Figure 03’s general-purpose ambition but a stronger option for a buyer with one well-defined warehouse task. 5. UBTECH Walker S2: the factory-uptime alternative Walker S2 deserves a look when long operating windows and industrial material handling are central to the project. Buyers who need public pricing or detailed customer-level performance data before contacting a vendor will find the record thinner. Walker S2’s standout feature is autonomous battery swapping. UBTECH says the robot can replace a battery in about three minutes, using a dual-battery design that lets it manage charging and swapping around task priorities. Its published payload is 15 kg across a working range of up to 1.8 m. That design addresses a practical problem that demo videos often ignore: a robot that spends hours charging is difficult to use across a production shift. UBTECH presents Walker S2 as in mass production and delivery, but it does not publish a list price. Buyers still need to confirm lead time, service coverage, safety certification and operating evidence for their country and task. 6. Apptronik Apollo 2: the enterprise learning-platform alternative Apollo 2 fits companies that want to participate in an enterprise pilot and build task data with a vendor. It does not fit buyers looking for an off-the-shelf commercial fleet today. Apptronik unveiled Apollo 2 in June 2026 in bipedal and wheeled-base configurations. The company says operational fleets are collecting data at Robot Park locations and at partner sites including Mercedes-Benz and GXO. The work combines teleoperation and autonomous execution across manufacturing, logistics and retail tasks. That makes Apollo 2 relevant to organizations that want to shape a future deployment. It also creates an important limitation: Apptronik describes Apollo 2 as the learning and data-collection platform feeding the next commercial product, Apollo 3. Buyers should not mistake activity at partner sites for a broadly available Apollo 2 product with a settled price and service package. The modular base is useful. A wheeled configuration may fit current industrial safety standards and predictable floors, while the bipedal version can collect data in spaces that require human-like mobility. Whether that flexibility beats Figure 03 depends on the site, not on a generic robot ranking. 7. Boston Dynamics Atlas: the heavy industrial alternative Atlas is the high-end option when payload, reach and industrial hardening matter most. It makes little sense as a household assistant or generally available developer platform. Atlas has the strongest published industrial hardware specifications in this set: a 50 kg instantaneous payload, 30 kg sustained payload, 2.3 m reach, 56 degrees of freedom, four-hour battery and IP67 protection. It can also swap its own battery. Boston Dynamics says Atlas is in its first customer pilot at Hyundai, working on real-world sequencing tasks. The company is opening discussions with a select number of early adopters for applications such as machine tending, part sequencing and order building. This is a serious industrial program, but it is not a normal product purchase. There is no public list price, and Atlas is larger and heavier than Figure 03. It is best viewed as the high-payload option for qualified industrial teams, not as a general alternative for every Figure 03 buyer. What about Tesla Optimus? Tesla Optimus is one of Figure’s most important strategic competitors, but it is not a practical Figure 03 alternative today. Tesla has no public order page, final retail price or complete Gen 3 specification sheet. Its advantage is potential manufacturing scale, not present buyer access. If you are comparing programs rather than choosing hardware, Optimus belongs on the shortlist. If you need to place a deposit, buy a development robot or start a qualified industrial pilot, the seven options above have clearer paths. Which Figure 03 alternative should you choose? For a home: choose 1X NEO if you accept early access and scheduled remote expert support.For full-size development hardware: choose Unitree H2.For lower-cost research: choose Unitree G1.For a live warehouse workflow: evaluate Agility Digit.For factory uptime: evaluate UBTECH Walker S2.For a co-developed enterprise pilot: talk to Apptronik about Apollo 2.For high-payload industrial work: talk to Boston Dynamics about Atlas.For a normal retail purchase with proven general autonomy: choose none yet. Before signing an order or pilot, ask for the exact model revision, configuration, payload definition, duty cycle, autonomy boundary, supervision requirement, safety certification, service coverage, spare-parts plan and acceptance test. A polished demo is not an uptime guarantee. If availability is the deciding factor, use our guide to the most advanced humanoids you can buy to separate retail listings, deposits, vendor enquiries, pilots and commercial deployments. How we chose these alternatives We screened the live RoboZaps robot database for models that overlap with at least one part of Figure 03’s intended role: home assistance, general-purpose manipulation, development, logistics or industrial work. We then weighted five factors: Commercial status: orderable, deposit, pilot or deployment.Task fit: what the robot is actually designed to do.Deployment evidence: named customers and real workflows, not only demonstrations.Price transparency: public price, subscription, vendor quote or undisclosed.Published specifications: only where the figure changes a buying decision. Manufacturer claims remain claims unless a named customer or commercial agreement supports them. The robots have not been tested against one shared benchmark, so this guide does not manufacture an overall score. Compare on RoboZaps: NEO, Unitree H2, Unitree G1, Digit, Walker S2 and Apptronik Apollo. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs XPENG IRON (2026): Which Humanoid Is Ahead? URL: https://blog.robozaps.com/b/tesla-optimus-vs-xpeng-iron Author: Radica Maneva Published: 2026-07-16T23:06:50.562Z Updated: 2026-08-01T23:02:59.411Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: comparisons TL;DR: XPENG IRON leads on disclosed current hardware and a specific Q1 2027 retail deployment plan. Tesla Optimus leads on manufacturing scale ambition, but Tesla has not published a current product sheet, buyer terms or external operating data. XPENG's official releases also conflict on IRON's aggregate compute. Neither robot is for sale. Key takeaways: - XPENG IRON leads this disclosure-based comparison because XPENG has published more hardware detail and a dated first deployment plan. - Tesla is installing Optimus production lines and has the larger stated scale ambition, but capacity plans are not buyer evidence. - Neither robot has an official MSRP, public order channel, delivery schedule or customer-grade operating record. - There is no fair autonomy winner because Tesla and XPENG publish no common task, intervention or uptime benchmark. - For a 2026 buyer, the verdict is watch both, procure neither. FAQ: Q: Is XPENG IRON better than Tesla Optimus? A: XPENG IRON is ahead on public evidence: XPENG has disclosed the current robot's 82 body degrees of freedom, 22-DoF hands, on-robot compute and a Q1 2027 retail deployment plan. Tesla has the larger manufacturing ambition, but it has not published a complete current product sheet or external operating record. That makes IRON better documented, not proven better at work. Q: Can you buy Tesla Optimus or XPENG IRON in 2026? A: No. Neither robot has a public order page, official MSRP, published buyer terms or delivery schedule. XPENG has not announced an official IRON price. Q: Which robot has better hardware? A: XPENG publishes more current hardware detail for IRON, including 82 body degrees of freedom, 22-DoF hands, three Turing AI chips and an all-solid-state battery claim. Its official releases conflict on aggregate compute, citing either 2,250 or 3,000 effective TOPS. Tesla has not released an equivalent complete specification for its current production-intended Optimus design, so a direct performance verdict is not possible. Q: Which company is closer to mass production? A: Both target 2026 production progress. Tesla says first-generation Optimus production lines are being installed. XPENG says its mass-production-ready IRON is poised to enter joint hardware-software integration, targets formal mass production by year-end and plans in-store guide deployments from Q1 2027. Neither company has published a verified current output count for the compared robot. Q: What evidence would prove one robot is ahead? A: The strongest evidence would be a defined customer task with units deployed, task success, intervention rate, uptime, runtime under load, safety documentation, support terms and total installed cost. Demonstrations and production targets do not establish those results. XPENG IRON leads this 2026 disclosure-based comparison. XPENG has disclosed more of its current robot and attached it to a dated deployment plan. Tesla Optimus is the bigger manufacturing bet, but Tesla still provides no current product sheet, public order channel, customer uptime data or firm delivery date. Neither robot is ready for procurement. That distinction matters. A published demonstration and a production target are evidence of progress, but they are not the same as a supported product. This comparison prioritizes current-generation disclosure and dated deployment evidence over demonstrations, theoretical capability or brand-scale ambition. It scores only what Tesla and XPENG had published by July 17, 2026. Quick verdict What exactly are we comparing? This is not a clean specification-sheet contest. Tesla's Optimus review tracks the evergreen program, while the separate Gen 3 page owns the version chronology. Tesla described Gen 3 as its first design intended for mass production, yet it has not published a complete current-generation specification sheet. XPENG's side is clearer. The XPENG IRON review now centers on the next-generation model revealed in November 2025, not the first-generation 2024 robot. XPENG published specific hardware details for that design and followed with a June 2026 rollout update. The comparison therefore asks a more useful question: which company has provided the stronger evidence that its current humanoid can become a real, supported product? Hardware: IRON wins on disclosure XPENG has put numbers behind its latest design. At its 2025 AI Day, the company said Next-Gen IRON has 82 degrees of freedom across the body, 22 degrees of freedom in each human-scale hand, three Turing AI chips, a human-like spine, bionic muscles, a flexible outer skin and an all-solid-state battery. XPENG's official materials conflict on aggregate compute: the November 2025 release says 3,000 effective TOPS, while its January 2026 Singapore release and February 2026 Indonesia release say 2,250 TOPS. RoboZaps treats the figure as unresolved. Those are manufacturer claims, not independent performance results. They do, however, give an operator something concrete to question. Buyers can ask how the 82 degrees of freedom are allocated, what loads the hands can sustain, how runtime changes by task, and whether the battery has passed the relevant safety tests. Tesla has published walking and manipulation demonstrations, but its current program page and 2026 investor updates do not provide an equivalent product sheet. The last publicly shown Gen 2/2.5 design is not a safe substitute for an unreleased production design. On disclosed current hardware, IRON wins by default. AI and autonomy: no defensible winner yet Both companies are trying to reuse an automotive AI stack in a humanoid body. Tesla describes Optimus as a general-purpose autonomous biped that depends on vision, planning, balance, navigation and physical-world interaction. At CVPR 2026, Tesla said its robotics work uses large-scale multimodal foundation models in an end-to-end pixels-to-actuation system. XPENG says IRON combines its physical-world model with VLT, VLA and VLM components. Its stated advantage is vertical integration across chips, models and hardware. The company also plans to open an IRON SDK to developers. Raw TOPS does not decide this round. Neither company publishes a common task set, intervention rate, recovery rate or full-shift result that would allow a fair comparison. Tesla describes a large-scale multimodal training stack. XPENG publishes a more explicit on-robot compute architecture. Neither disclosure establishes superior autonomy. Hands and movement: IRON is the clearer design IRON's 22-DoF hands and 82-DoF body make dexterity central to the product, not a side feature. XPENG designed its gait and human-like form for retail and human-facing settings, but appearance does not validate operational reliability. More joints can increase control complexity and maintenance burden. A high degree-of-freedom count is valuable only when the robot can repeat useful work without frequent intervention. Tesla has made hand development a visible priority and described the latest hand as a major part of the Gen 3 upgrade. What is missing is a current, comparable disclosure. Until Tesla publishes the production hand design and task data, IRON leads this category on evidence rather than proven productivity. Production and deployment: two different bets Tesla has the more aggressive scale story. Its Q4 2025 update called Gen 3 its first design meant for mass production. It said preparations for the first production line were underway, including supply-chain readiness, with production planned before the end of 2026 and eventual capacity of one million robots a year. The Q1 2026 update said first-generation lines were being installed, the Fremont line was designed for one million robots a year, and preparations for the first large-scale Optimus factory would begin shortly in Q2. These are line, factory and designed-capacity milestones, not current output. XPENG has published the clearer first deployment path. In June 2026, the company said the mass-production-ready IRON version was "poised to enter" joint hardware-software integration, with formal mass production targeted by year-end. It also said IRON robots are set to work as shopping guides in XPENG stores from Q1 2027. XPENG has separately named Baosteel as an industrial inspection partner. Neither statement proves output. Tesla has not disclosed a current production count or an external customer operating record. XPENG's retail and industrial plans remain forward-looking until robots are doing the work. Today, Tesla leads on stated scale ambition; IRON leads on the specificity of the next deployment. Price and availability: neither is a buyer option Tesla Optimus is not for sale. There is no MSRP, deposit, reservation page, delivery schedule or published buyer terms. XPENG IRON is also not for sale and has no official price. XPENG's current robot page describes the program but provides no order or price path. Third-party estimates should not be treated as quotations from XPENG. For realistic ranges, integration costs and support requirements, use the separate humanoid robot cost guide. If you need a robot now, neither belongs on a procurement shortlist. Compare models with a current order channel, documented support and a defined task envelope in the guide to humanoid robots buyers can actually procure. Which program is more relevant to watch by use case? Factory automation program to watch: Optimus. This is an inference from Tesla's manufacturing footprint, not evidence of a customer deployment. Its internal factories could provide a proving ground.Near-term retail or service pilot to watch: IRON. XPENG has named the job, location type and Q1 2027 starting window.Developer access: IRON. XPENG has said it will open an SDK; Tesla has not announced a public Optimus developer platform.Published hardware detail: IRON. XPENG provides a current body, hand, compute and battery description.Long-term scale ambition: Optimus. Tesla's stated eventual capacity is larger, although capacity is not production.2026 procurement: Neither. There is no buyer-ready offer to compare. What would change the verdict? The next useful evidence is not another stage walk. It is operating data. A serious evaluation needs a defined task, units deployed, paid or internal deployment status, task success rate, human interventions, uptime, runtime under load, safety controls, service coverage and total installed cost. Tesla can close the gap by publishing the current production design and a repeatable operating record. XPENG can convert its disclosure lead into a commercial lead by meeting the 2026 production target and showing that the Q1 2027 retail robots work reliably without hidden human assistance. Final verdict XPENG IRON leads this disclosure-based comparison today. It has the more complete current hardware disclosure, a named developer strategy and a dated first deployment plan. Tesla Optimus remains the larger manufacturing and AI-scale bet, yet buyers still have to bridge too many gaps with assumptions. That does not make IRON the better working robot. It makes IRON the better-documented program. Until either company publishes customer-grade performance and an orderable configuration, the practical decision is the same: watch both, procure neither. Compare on RoboZaps: Tesla Optimus Gen 2, Xpeng Iron and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs AgiBot A2 (2026): Which Is Ready Now? URL: https://blog.robozaps.com/b/tesla-optimus-vs-agibot-a2 Author: Radica Maneva Published: 2026-07-16T14:20:18.887Z Updated: 2026-08-01T23:03:18.564Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: A sales-assisted AgiBot A2 pilot is the more practical route in 2026, but only after AGIBOT confirms the exact variant, delivery territory, local support and compliance. A2 Lite pricing and A2 Ultra endurance do not transfer to the base A2. Tesla Optimus has the broader ambition, but no public order page, MSRP, delivery commitment or current product data sheet. Key takeaways: - AgiBot A2 has the stronger procurement path in this matchup, subject to a written quote for the exact variant, delivery territory and support package; Tesla Optimus is not publicly orderable. - AGIBOT publishes A2 specifications including 169 cm height, 69 kg weight, 700 Wh battery, two-hour stated runtime and 40+ active degrees of freedom. - Tesla publishes the Optimus goal and production preparation, but no current commercial specification sheet, price, warranty or delivery terms. - A2 is aimed first at reception, guidance and exhibitions; Optimus targets broader general-purpose work. - A demo, company autonomy label or production-line claim is not a substitute for customer uptime, intervention and acceptance data. FAQ: Q: Which is better in 2026, Tesla Optimus or AgiBot A2? A: AgiBot A2 has the stronger procurement case for a supervised service-robot pilot because AGIBOT publishes base-A2 specifications and accepts commercial enquiries. That is not proof of delivery, support or certification in every market, so buyers still need a variant-specific written quote. Optimus remains a development program with no public order page or current product data sheet. Q: Can you buy Tesla Optimus or AgiBot A2? A: Tesla does not take public Optimus orders. AGIBOT accepts A2 sales enquiries, while its official store separately lists the A2 Lite variant at $44,560. The base A2 and A2 Ultra are quote-only, and the store does not establish delivery or support in every country. Q: How do Tesla Optimus and AgiBot A2 differ? A: Optimus is intended as a general-purpose robot for unsafe, repetitive or boring tasks. AgiBot A2 is positioned more narrowly for reception, customer service, retail guidance and exhibitions. Q: What are the AgiBot A2 specifications? A: AGIBOT publishes 169 cm height, 69 kg weight, a 700 Wh battery, two hours of stated runtime with battery swapping, and more than 40 active degrees of freedom. These are manufacturer specifications, not independent performance results. Q: Does Tesla publish official Optimus specifications? A: Tesla does not currently publish a commercial Optimus product specification sheet. Older Gen 2 demonstration figures should not be assumed to describe the unreleased production robot. Tesla Optimus and AgiBot A2 are often grouped together as full-size humanoid robots. They are not at the same commercial stage, and they are not aimed at the same first job. Verdict: AgiBot A2 has the stronger procurement path for a supervised service-robot pilot. AGIBOT publishes base-A2 specifications and accepts commercial enquiries; the separate A2 Lite variant has a listed store price. That does not establish base-A2 delivery, certification or support in every region. Optimus is the more ambitious general-purpose platform, but Tesla still has no public order page, MSRP, delivery commitment or current product specification sheet. A buyer cannot procure it. This comparison uses manufacturer sources available on July 16, 2026. Company specifications and production statements are labelled as claims. A demonstration is not treated as a customer deployment. Tesla Optimus vs AgiBot A2 at a glance The table deliberately leaves Tesla hardware cells blank where Tesla does not publish a current value. Older Gen 2 demonstration figures should not be presented as specifications for an unreleased production robot. The two robots solve different first problems Tesla describes Optimus as a general-purpose autonomous biped intended to handle unsafe, repetitive or boring work. Its public AI page emphasizes balance, navigation, perception and interaction with the physical world. The goal is broad. The product offer is not. AgiBot A2 has a narrower brief. AGIBOT positions it for human-facing service work: reception, guided presentations, customer service, retail assistance and exhibitions. Its product page describes full-duplex conversation, facial recognition, sound-source localization and lip-reading in noisy environments. Those are vendor-published capabilities rather than independent performance results, but they show what the A2 is being sold to do. That distinction matters. An operator choosing a reception or exhibition robot can write an A2 pilot around a defined environment and task. An operator choosing Optimus would be waiting for Tesla to publish the product, support model and deployment terms. Availability: AgiBot has a sales path; Tesla does not AGIBOT provides a product enquiry form for the base A2 and runs an official global store. The store lists the separate A2 Lite at $44,560. The base A2 and A2 Ultra remain quote-only, so the Lite price should not be copied across the family. An enquiry form is not a delivery commitment. AGIBOT does not publish a universal lead time, country-by-country delivery list, local certification matrix or field-service coverage for the base A2. Before treating it as available, get the exact variant, delivery territory, lead time, warranty, spare-parts route and local support entity in writing. Tesla has not opened Optimus reservations or public sales. Its Q1 2026 update said first-generation production lines were being installed in anticipation of volume production. Tesla's annual report and quarterly filing still describe Optimus as a robot being developed and commercialization as future work with execution risk. Production preparation is useful evidence. It is not a customer order process. Until Tesla publishes an order page, exact configuration, delivery market, warranty, support terms and a current data sheet, Optimus belongs on a watchlist rather than a procurement shortlist. For the broader price context, see our humanoid robot cost guide. It separates published catalog prices from quotes, targets and unsupported estimates. Hardware comparison: A2 is measurable; Optimus is not yet comparable AGIBOT publishes 169 cm height, 69 kg weight, a 700 Wh battery, two hours of runtime and more than 40 active degrees of freedom for the base A2. It also says the battery can be swapped. These are manufacturer specifications, not a guarantee of two useful operating hours under every workload. The A2 perception stack is more explicit. AGIBOT lists a 360-degree lidar and six cameras, with its HIMUS mapping and VectorFlux planning software used for 3D-SLAM navigation. The company describes the resulting mobility as L4-level. There is no neutral humanoid autonomy standard that makes this label directly comparable with an automotive L4 rating, so buyers should ask for the operating design domain and intervention policy. Tesla's public material explains the software direction but does not give buyers a current product data sheet. Tesla says it is extending vision and planning work into bipedal robotics. That may become a strong advantage because Tesla can combine AI development with manufacturing scale. It does not let a buyer compare payload, runtime, hand configuration, duty cycle or safety limits today. Deployment evidence: do not mix variants or demos AgiBot has shown the A2 series in public service settings. At a February 2026 event at United Nations Headquarters, AGIBOT said A2 robots performed Chinese calligraphy and held natural voice interactions with guests. That is relevant front-of-house evidence, although it does not publish uptime, intervention rate or task success over a normal working week. The A2 Ultra variant also completed a 106.286 km walk that AGIBOT says received a Guinness World Record. The result belongs to A2 Ultra. It should not be used as proof that every A2 configuration has the same endurance, autonomy or thermal performance. Tesla's evidence is earlier in the commercial ladder. The company has shown public demonstrations and reported production-line installation. It has not published customer deployments, supported task packages, fleet uptime, intervention rates or delivery terms. A broad general-purpose goal is harder than a controlled reception workflow, but that does not excuse the missing buyer evidence. Safety and support questions AGIBOT says the A2 uses three layers of safety monitoring and PLd-level protection, with redundant computing and safety centers. Buyers should request the underlying declaration, scope, test report and applicable standards. A marketing page is not a site safety assessment. AGIBOT also publishes a defined security-update period for the A2. The current document lists support from January 1, 2026 through January 1, 2027. One year is a concrete commitment, but it is short for industrial equipment. A procurement team should negotiate the patch period, end-of-support notice and response time in writing. Tesla does not publish an Optimus commercial safety certificate, warranty, service-level agreement or security-support term because there is no public product offer. Those items should be treated as release gates, not details to solve after a pilot begins. Which robot is better for your use case? Shortlist the base A2 only after regional confirmation The base A2 is the stronger candidate for reception, customer guidance, exhibitions and other supervised front-of-house work. There is a defined product, a contact path and a published specification set. A buyer still needs a written, variant-specific offer covering delivery territory, lead time, local support, language quality, remote access, data handling, charging workflow, battery-swap procedure and site-specific safety. Watch Optimus for general-purpose factory work Optimus is the more interesting long-term bet if Tesla turns its AI and manufacturing resources into a supported robot platform. The intended scope is broader than A2's service focus. Today, however, there is no robot a buyer can order and no current specification sheet against which to write acceptance tests. Choose neither for an open research platform Tesla does not offer Optimus as a developer product. The base A2 is a commercial service platform, not the clearest option for open robotics research. A lab that needs SDK access, low-level control and reproducible hardware should compare products designed for development rather than assuming either brand will expose the required stack. What to demand before buying For AgiBot A2, request a written quote tied to an exact variant and configuration. The acceptance plan should cover: included hands, sensors, compute and charging hardware;supported languages and noise conditions for interaction;runtime under the buyer's actual task, plus battery-swap time;autonomous operating boundaries and human intervention triggers;remote-access, data-retention and security-update terms;safety documentation, local compliance and risk-assessment support;warranty, spare parts, field service and software licensing. For Optimus, wait for the same evidence before treating a Tesla target as a product. The minimum release package is a public order path, current specification sheet, exact price, delivery territory, supported tasks, safety documentation, service terms and customer operating data. RoboZaps keeps the canonical catalog records separate from this editorial comparison. The related-robot cards below point to the live Tesla Optimus and AgiBot A2 product records for current structured data. Bottom line AgiBot A2 has the stronger procurement case for a buyer acting in 2026, conditional on a written offer for the exact variant and market. It has a defined service role, published base-A2 specifications and a commercial contact path. The buyer still needs to establish delivery, local support, compliance and performance in a real pilot. Optimus wins only as a long-term platform thesis. Tesla's goal is broader and its potential manufacturing advantage is large. Until the company publishes a product offer and customer evidence, that potential should not be scored as present-day capability. Compare on RoboZaps: Tesla Optimus Gen 2, AgiBot A2 and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Figure 03 vs Apptronik Apollo 2: Which Is Further Ahead? URL: https://blog.robozaps.com/b/figure-03-vs-apptronik-apollo-2 Author: Radica Maneva Published: 2026-07-16T13:39:34.000Z Updated: 2026-08-01T10:54:07.247Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Figure 03 has the stronger public case for manufacturing scale and a current bipedal factory workflow. Apollo 2 offers more mobility options and a broad data-collection program, but Apptronik positions it as the training platform for the future Apollo 3 fleet. Key takeaways: - Figure says it had delivered over 350 Figure 03 robots by April 2026 and demonstrated a one-unit-per-hour production cycle at BotQ. - Apollo 2 is available with legs or a wheeled base and is used for data collection across Robot Park and partner or customer sites. - Figure publishes current physical and runtime figures for Figure 03; Apptronik has not published equivalent Apollo 2 numbers, so older Apollo specs should not be copied across. - Figure uses its own Helix 02 whole-body AI stack, while Apollo 2 combines Apptronik's Artemis and Fleet Connect software with Google DeepMind's Gemini Robotics work. - Neither robot has a public list price, open order page or standard delivery terms. FAQ: Q: Which is better, Figure 03 or Apptronik Apollo 2? A: Figure 03 has the stronger public case for manufacturing scale and a current bipedal factory workflow. Apollo 2 is more flexible because it offers bipedal and wheeled configurations, but Apptronik positions it mainly as a data-collection and pilot platform for the future Apollo 3 fleet. The better choice depends on the task and the evidence each vendor can produce during a site-specific trial. Q: Is Apptronik Apollo 2 for sale? A: Apollo 2 has no public list price, checkout or standard delivery terms. Apptronik invites companies to discuss deployments and already operates units at Robot Park and customer or partner sites, so access is through enterprise pilots and direct agreements rather than an open sale. Q: Can you buy Figure 03? A: Figure 03 has no public order page or list price. Figure is building robots for internal, home-research and commercial programs. Figure and BMW have also announced a Spartanburg logistics project, but the robot is not an off-the-shelf product with public delivery terms. Q: What specifications has Apptronik published for Apollo 2? A: Apptronik confirms bipedal and wheeled configurations, swappable batteries, opportunity charging, tethered operation, patented actuators, human-facing status displays and configurable safety zones. Its current Apollo 2 product page does not publish height, weight, payload, speed, degrees of freedom or battery runtime, so older Apollo figures should not be presented as confirmed Apollo 2 specifications. Q: Which robot has better AI? A: There is not enough comparable operating data for a universal winner. Figure's Helix 02 is a vertically integrated whole-body autonomy stack demonstrated on Figure 03. Apptronik combines its Artemis and Fleet Connect software with Google DeepMind's Gemini Robotics models and a multi-site data program. Compare task completion and intervention data on the intended workflow. Figure 03 and Apptronik Apollo 2 were both put in the spotlight on June 30, 2026. Figure showed its latest robot handling a sequencing task at BMW. Apptronik unveiled Apollo 2 as the workhorse behind its expanded Robot Park and customer data-collection programs. They look like close rivals, but the public evidence points to two different products. Figure 03 is a bipedal platform that Figure is already building in the hundreds and placing into home research and commercial work. Apollo 2 is a modular training and pilot platform, offered with legs or a wheeled base, whose data is meant to improve Apptronik's next commercial generation. Verdict: Figure publishes the stronger manufacturing and product-readiness case today. Apollo 2 is the more flexible data-collection platform, but Apptronik's own announcement makes clear that its work feeds the coming Apollo 3 fleet. Neither robot has a public list price or open order page. Figure 03 vs Apollo 2 at a glance The table deliberately does not copy the familiar 2023 Apollo numbers into the Apollo 2 column. Apptronik has not confirmed that the new platform keeps the older robot's height, weight, payload, speed or battery runtime. Treating those generations as one spec sheet would make the comparison look more precise than the evidence allows. The biggest difference is product purpose Figure presents Figure 03 as its current general-purpose robot. It was redesigned for Helix, home environments and high-volume manufacturing, but Figure also says the same hand, sensor and charging changes support commercial work. Figure's June 2026 demonstration showed precise part handling, full-body movement and cart pulling during a sequencing task. Apptronik describes Apollo 2 differently. The company calls it the workhorse behind Robot Park, where robots collect data across logistics, manufacturing, retail and other customer-led tasks. The platform is already present at partner and customer sites, but its stated job is to create the experience and training data that will power Apollo 3. That does not make Apollo 2 a dead-end prototype. It makes the buying question narrower. Apollo 2 looks relevant for a company that wants to join a structured pilot or embodied-AI data program. Figure 03 looks closer to a single hardware generation that Figure intends to operate, improve and manufacture at scale. Hardware: fixed biped versus modular mobility Figure 03 is bipedal only. That gives it access to stairs, narrow passages and workstations designed around a standing person. Figure's BMW example also shows why legs can matter during a sequencing job: the robot steps, shifts its reach and repositions its whole body while handling parts and moving a cart. Apollo 2 can use legs or a wheeled base. The biped is meant for spaces where human-like mobility matters. The wheeled version gives up stairs and some terrain flexibility in exchange for stability, simpler movement and easier alignment with existing industrial mobile-robot safety practices. For a flat warehouse or factory, the wheeled Apollo 2 may be the more practical pilot platform. If the task genuinely needs a robot to move through human spaces, reach around fixtures and coordinate manipulation with foot placement, Figure 03 has published the clearer demonstration. AI: Helix 02 versus Artemis and Gemini Robotics Figure owns the full AI stack it discusses publicly. Helix 02 coordinates hands, arms, torso and feet through a pixels-to-actions model. At BMW, Figure says the system adjusted to parts and carts that were not presented in exactly the same position each time. Figure also uses its growing robot fleet to collect data and push new behaviours through its fleet-management and over-the-air update systems. Apollo 2 uses Apptronik's Artemis layer to coordinate perception, planning, controls, safety, task execution and human interaction. Fleet Connect handles monitoring and orchestration. Apptronik's research partnership with Google DeepMind adds Gemini Robotics models, while Robot Park supplies real-world data through teleoperation, autonomous execution and simulation. That partnership moved on 30 July 2026, when DeepMind released Gemini Robotics 2 and used Apollo 2 as its demonstration platform. The model drives the whole body rather than the arms alone, and it reaches 76.3% success picking an object off a shelf but only 36% screwing in a light bulb. It is limited to early-access partners for now. Our breakdown of what Gemini Robotics 2 can actually do has the published results in full. There is no honest universal winner here. Figure has shown a more vertically integrated autonomy story on its own current hardware. Apptronik has a broader modular data engine and a major external AI partner. Buyers should ask for task success rates, interventions, recovery behaviour and operating hours on their exact workflow, not choose between two polished demo reels. Manufacturing and deployment evidence Figure says it had delivered over 350 Figure 03 robots by April 29, 2026 and had demonstrated a one-robot-per-hour cycle time at BotQ. It also reported an end-of-line first-pass yield above 80%, more than 500 battery packs and more than 9,000 actuators. Those are company-reported figures, not an independent production audit, but they are unusually concrete for this market. BMW says Figure 03 will work on complex sequencing applications in logistics at its Spartanburg plant. The public evidence describes a project and demonstration rather than customer-verified throughput data, but it independently connects the current robot generation to a named factory and defined task. Apptronik says operational Apollo 2 fleets are active in Robot Park locations and at partner or customer sites worldwide. It names Google DeepMind, Mercedes-Benz and GXO, and says the robots work across customer-driven tasks. Google DeepMind separately confirms its Apptronik partnership and Gemini Robotics work on Apollo. What is missing is equally important: Apptronik has not published the number of Apollo 2 units, total operating hours, task throughput or customer-confirmed performance. On public evidence alone, Figure is ahead on disclosed manufacturing scale. Apptronik is ahead on mobility options and the breadth of its described data-collection network. Power and uptime Figure publishes a five-hour runtime for Figure 03 and says the robot can recharge at 2 kW by standing on an inductive mat. That design reduces manual battery handling and suits jobs where the robot can take planned charging breaks. Figure also pairs the dock with wireless data offload. Apptronik chose swappable batteries for Apollo 2, along with opportunity charging and tethered operation. The company says this setup is intended to support long operating windows with minimal interruption. It does not currently publish a per-pack runtime for Apollo 2, so a direct hours-versus-hours comparison would be guesswork. The trade-off is operational. Figure's dock is simpler for an autonomous fleet if the duty cycle leaves enough charging time. Apollo 2's battery swaps can keep a robot working while another pack charges, but they add battery inventory, handling and maintenance procedures. Price and availability Neither company publishes a list price for these robots. Figure 03 has no public order page. Apptronik invites companies to discuss Apollo deployments, but Apollo 2 also has no public checkout, reseller price or standard delivery promise. That makes both enterprise engagements rather than normal equipment purchases. Before treating either as available, ask for a written scope covering the robot configuration, software, support, spares, remote access, site integration, acceptance tests and responsibility for failed shifts. Our guide to humanoid robots you can actually buy separates open sales from pilots and development programs. Do not compare a rumoured hardware price in isolation. Battery packs, charging infrastructure, guarding or site changes, integration work, operator training and field service can dominate the total cost of a humanoid robot. Which robot should you choose? Choose Figure 03 for a serious bipedal pilot Figure 03 is the stronger shortlist candidate when the task requires human-space mobility, dexterous manipulation and a vendor that can point to a disclosed production ramp. The BMW sequencing example is more useful than a generic capability claim because it shows the robot combining locomotion, force and precise placement in one workflow. Choose Apollo 2 for modular pilot and data work Apollo 2 makes more sense when a wheeled base would simplify the job, or when the goal is to participate in a structured data-collection and embodied-AI program. Its Artemis, Fleet Connect and Gemini Robotics stack is built around learning across multiple sites and configurations. Choose neither for an off-the-shelf purchase If you need a published price, standard delivery date and proven customer throughput, neither robot clears that bar. Ask both vendors for the same acceptance test and compare results on your site. A successful pilot should end with measured task completion, intervention rate, uptime, safety events and total operating cost, not a promise that the next software release will close the gap. --- # Figure 03 vs Unitree H2: Which Humanoid Is Better in 2026? URL: https://blog.robozaps.com/b/figure-03-vs-unitree-h2 Author: Radica Maneva Published: 2026-07-16T13:22:08.183Z Updated: 2026-08-01T23:03:13.995Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Choose Unitree H2 if you need orderable full-size hardware or an EDU development platform. Choose Figure 03 if you are comparing autonomy, dexterous manipulation and home-oriented safety, but Figure has no public price or order program. Key takeaways: - Unitree H2 is the only robot in this matchup with a public price and order channel. - Figure 03 has the stronger evidence for autonomous, tactile, full-body manipulation. - H2 EDU is the practical developer platform; the base H2 omits dexterous hands and secondary development. - Figure 03 is designed more clearly for work around people, but neither robot is a proven consumer home product. FAQ: Q: Is Figure 03 better than Unitree H2? A: Figure 03 is ahead in demonstrated autonomy, tactile manipulation and home-oriented safety design. Unitree H2 is better for buyers and developers because it has a public base price and an EDU configuration with secondary development access. Q: How much do Figure 03 and Unitree H2 cost? A: Unitree lists the base H2 at $29,900 before tax and shipping. H2 EDU is quote-only. Figure has not published a Figure 03 list price or public order terms. Q: Can you buy Figure 03 or Unitree H2 now? A: Unitree H2 is listed on Unitree's official shop, while H2 EDU is sold through the sales team. Figure 03 has no public order page, deposit or retail delivery schedule. Q: Which robot has better hands? A: Figure 03 has the stronger standard hand system, with five fingers, palm cameras and fingertip tactile sensing. The base Unitree H2 has no dexterous hands; multiple hand options are available with H2 EDU. Q: Which robot is better for research? A: Unitree H2 EDU is the practical research choice because it supports secondary development. Figure 03 offers a more integrated autonomy stack but no public developer edition or SDK. Q: Was Figure 03 used to build 30,000 BMW cars? A: No. Figure 02 contributed to the production of more than 30,000 BMW X3 vehicles. Figure described the June 2026 Figure 03 sequencing workflow as its first Figure 03 demonstration at the plant. The short answer: buy the Unitree H2 if you need a full-size humanoid platform you can order and develop on. Figure 03 is the more convincing autonomy and manipulation system, but Figure does not offer a public order page or list price. This is a comparison of two different products and two different commercial models. Unitree sells the H2 as hardware, starting at $29,900 for the base robot. Figure is building a closed robot, AI and fleet platform around Helix, with access routed through commercial relationships rather than a public checkout. Figure 03 vs Unitree H2 at a glance The headline specifications only tell part of the story. For full background on each model, use the dedicated Figure 03 review and Unitree H2 review. This page owns the direct matchup. Price and availability: Unitree H2 wins Unitree publishes a $29,900 base price on its official store. Tax, shipping and customs are extra. The standard H2 also omits the features most research teams will care about: dexterous hands, a high-power AI module and secondary development access. Those sit with the H2 EDU configuration, which requires a sales quote. Figure publishes no MSRP, deposit, delivery schedule or public order form for Figure 03. Its website describes the robot's capabilities and its enterprise direction, but a company cannot budget around an undisclosed price and undefined service package. The H2 is therefore the only practical answer if the decision is which robot can be purchased now. For a wider view of what published humanoid prices include and leave out, see the humanoid robot cost guide. A $29,900 chassis is not a task-ready deployment once hands, compute, integration, safety work, spares and support are added. Autonomy and manipulation: Figure 03 is ahead Figure 03 is built around Helix 02, Figure's full-body vision-language-action system. In Figure's January demonstration, one neural system controlled walking, balance and manipulation through a four-minute dishwasher task with 61 ordered actions and no reset or human intervention. Figure also demonstrated tactile tasks including opening a bottle, extracting a pill and dispensing a measured syringe volume. The hardware matters here. Figure 03 has a camera in each palm and tactile sensors in its fingertips that Figure says can detect forces as small as three grams. That gives the control system close-range vision and touch when the head cameras lose sight of an object. Unitree publishes a strong mechanical specification for H2: 31 degrees of freedom, up to 360 N·m at the leg joints and about 120 N·m at the arm joints. Its product page also promises OTA improvements and support for embodied models. What Unitree does not publish is a directly comparable, long-horizon autonomous task result for the standard H2. Impressive motion is not the same thing as autonomous work. These are vendor demonstrations rather than independent benchmark results. Even with that caveat, Figure supplies substantially better evidence of integrated perception, touch and full-body task control. Developer platform: Unitree H2 EDU wins H2 EDU is the better fit for a lab that wants to write its own software, add compute and test control policies on human-scale hardware. Unitree explicitly reserves secondary development for the EDU model and offers multiple dexterous-hand configurations. The buyer still needs a quote and should confirm the exact compute, hand, warranty and support package in writing. Figure offers the more advanced integrated stack, but it is not an open developer product. There is no public Figure 03 SDK, developer edition or self-serve purchase path. That may suit an enterprise customer buying an outcome through a commercial pilot. It does not suit a university or robotics team that wants low-level access and ownership of the development environment. Home safety design: Figure 03 has the clearer case Figure 03 weighs 61 kg, nine kilograms less than the H2, and Figure designed it with soft textiles, multi-density foam around pinch points and a battery with layered protection. It also uses 2 kW inductive charging, so the robot can step onto a charging stand rather than rely on a person connecting a cable. Unitree's own H2 documentation takes a different posture. It warns users to keep a sufficient safe distance because the robot is powerful and complex. That is a sensible warning for research hardware, but it is not the pitch of a finished household appliance. Figure therefore has the stronger home-oriented design on paper. Neither vendor has published enough ordinary household operating data to call either robot a proven home product. If the requirement is a delivered, supported home helper today, choose neither. Commercial evidence: Figure 03 leads, with an important caveat Figure said in April that BotQ had produced more than 350 Figure 03 robots and demonstrated a one-robot-per-hour production rate. On June 30, Figure showed Figure 03 performing a sequencing workflow at BMW's Spartanburg plant, combining part handling with repositioning and cart movement. That BMW result was described as the first Figure 03 demonstration at the plant, not a fleet-scale production record. The widely cited 90,000 parts and 30,000 vehicles belong to the earlier Figure 02 deployment. Figure 03 inherited those engineering lessons; it did not perform that earlier work. Unitree's H2 evidence is more commercial in a different sense: the robot has a public price and sales channel. Unitree has not published comparable customer workflow metrics for H2. Figure has the stronger named-use-case evidence, while Unitree has the stronger purchasing evidence. Which should you choose? Choose Unitree H2 if: You need full-size humanoid hardware that can be ordered now.You are a research lab or developer prepared to buy the H2 EDU configuration.You value published mechanical specifications and direct hardware access over a polished autonomy stack.You have budgeted for hands, compute, integration, freight, customs, safety work and support. Watch or pilot Figure 03 if: Your use case depends on autonomous, dexterous manipulation rather than motion research.You are evaluating an enterprise deployment and can work through a commercial agreement.You care about home-oriented physical design, wireless charging and integrated fleet learning.You can wait for clear pricing, service terms and customer operating data. Choose neither if: You need a turnkey worker with published uptime, intervention and service-level data.You expect a proven consumer home robot.You cannot run a controlled pilot and complete a task-specific safety assessment. If you are comparing other systems that can actually be procured, the most advanced humanoids you can buy guide separates public sales channels from demos and future programs. Final verdict The Unitree H2 is the better purchase. It has a public base price, an official sales channel and an EDU route for secondary development. Just do not mistake the $29,900 headline for the price of a complete autonomous worker. Figure 03 is the better robot system to watch for autonomous work. Helix 02, tactile hands, palm cameras, wireless charging and a growing fleet give Figure the stronger integrated story. Until Figure publishes commercial terms and independent operating data, that lead is technical and strategic rather than a buying recommendation. Compare on RoboZaps: Figure 03, Unitree H2 and Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs Unitree H2: Which Humanoid Wins in 2026? URL: https://blog.robozaps.com/b/tesla-optimus-vs-unitree-h2 Author: Radica Maneva Published: 2026-07-16T12:34:34.469Z Updated: 2026-08-01T23:02:54.847Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Unitree H2 wins for buyers choosing a procurable research or commercial platform in 2026: it has a $29,900 base price, published specifications and an official order page. Tesla has not released a Gen 3 spec sheet, MSRP, order channel or customer delivery terms. Key takeaways: - Unitree H2 wins 2026 research and commercial procurement because it has an official $29,900 base price, order page and published specification sheet. - Tesla has not published Optimus Gen 3 specifications, customer pricing, an order channel or customer delivery terms. - H2 EDU supports secondary development; the $29,900 base H2 does not and also omits dexterous hands. - Neither company publishes enough customer uptime, intervention or multi-month task-success data to prove dependable general-purpose work. - Optimus remains the stronger long-term manufacturing thesis, but H2 is the only robot in this matchup that can enter procurement today. FAQ: Q: Is Unitree H2 better than Tesla Optimus? A: Unitree H2 is the better choice for a research or commercial team procuring one of these robots in 2026 because it has a public price, order page and detailed specifications. Tesla Optimus may have greater long-term potential, but its Gen 3 production design, customer terms and field performance are not yet public. Q: Can you buy Tesla Optimus or Unitree H2? A: Unitree lists the base H2 for direct purchase through its official shop. Tesla has not opened an Optimus order, reservation or customer delivery program. Q: How much do Tesla Optimus and Unitree H2 cost? A: Unitree lists the base H2 at $29,900 before tax, shipping and customs. The H2 EDU is contact-sales. Tesla has not published an Optimus MSRP. Q: Which robot has better specifications? A: Unitree H2 is the only robot in this matchup with a current production specification sheet. Tesla has not released Gen 3 height, weight, joint count, payload or battery figures, so a responsible comparison cannot award Optimus those categories using older hardware. Q: Which is better for robotics research? A: Unitree H2 EDU is the practical research choice because it supports secondary development and optional high-performance compute and dexterous hands. Tesla offers no public Optimus SDK or outside developer program. Q: Is Tesla Optimus or Unitree H2 ready for the home? A: Neither has public evidence supporting a dependable general-purpose home deployment. H2 is a roughly 70 kg research and commercial platform, while Optimus is not available for purchase. The verdict: Unitree H2 wins for research and commercial buyers choosing a procurable humanoid in 2026. It has a published US$29,900 base price, an official order page and a detailed specification sheet. Tesla Optimus does not yet have a customer price, order channel, delivery agreement or public Gen 3 specification sheet. That does not prove H2 will be the better robot once both machines are mature. It means H2 is the only one that can enter a real procurement process today. The current Tesla Optimus Gen 3 proposition is still a development and manufacturing plan. Buyers cannot award it categories based on older hardware or an unrevealed design. Tesla Optimus vs Unitree H2 at a glance The table looks lopsided because the disclosure is lopsided. Tesla has said what it wants Optimus to become. Unitree has published what a buyer receives, what the base model leaves out and which capabilities require the EDU configuration. Why Unitree H2 wins research and commercial procurement in 2026 It clears the procurement threshold A robot cannot win a buying comparison without a product, price and seller. Unitree supplies all three. The base H2 is listed at US$29,900 through its official shop. The page also warns that the price excludes tax, shipping and customs, which matters when comparing the sticker price with the full cost of a humanoid deployment. Tesla supplies none of those commercial terms. Its Q1 2026 update said Optimus production lines were being installed in anticipation of volume production. That is meaningful manufacturing progress, but it is not an offer to a customer. There is no published configuration, warranty, support plan, deposit, delivery window or acceptance process. H2 has a specification sheet that can be challenged Unitree states that H2 is roughly 180 cm tall and 70 kg with its battery. It lists 31 body joints, a rated arm payload of about 7 kg, a peak arm payload of about 15 kg and a 0.972 kWh battery with about three hours of runtime. Those figures come with caveats, but they give a lab or pilot team something concrete to test. Tesla has not released equivalent Gen 3 figures. The familiar Optimus height, weight, speed and joint-count numbers describe earlier hardware. Reusing them in a Gen 3 comparison would create precision where none exists. There is a defined development route H2 EDU supports secondary development and offers higher-end compute and dexterous-hand options. The base H2 does not. This distinction prevents a common procurement mistake: buying the lower sticker-price unit and discovering that it cannot support the planned software work. Tesla has no public Optimus developer program. An outside research team cannot buy a unit, obtain an SDK or design a project around documented interfaces. For robotics research in 2026, H2 EDU wins by being accessible at all. Where Tesla Optimus could still win Tesla's advantage is not a current specification. It is the possibility of manufacturing and software scale. The company describes Optimus as a general-purpose autonomous biped intended for unsafe, repetitive or boring work. Its Q4 2025 update called Gen 3 the first Optimus design meant for mass production and described an eventual planned capacity of one million robots per year. If Tesla turns that plan into a stable product, the comparison changes. High-volume production could support lower hardware costs, faster iteration and a larger service network than a research-oriented humanoid normally receives. A unified fleet could also improve software faster than one-off lab deployments. Every sentence in that case begins with “if.” Tesla still needs to reveal the production robot, publish its specifications, establish repeatable autonomous task performance and open a customer channel. Until then, scale is a thesis rather than a buying advantage. What H2 buyers should not ignore The US$29,900 price is the beginning The base price does not include tax, shipping or customs. It also does not buy secondary development or dexterous hands. A task-ready H2 program may need the EDU model, end effectors, compute, safety equipment, integration work, spares, training and local support. Compare delivered-system cost, not one line on the shop page. Three hours is not shift uptime Unitree states about three hours of battery life. That may work for research sessions and demonstrations, especially with a quick-release battery, but it does not prove continuous industrial availability. A factory pilot should measure useful task time, battery-change time, intervention rate and recovery after faults. A 70 kg humanoid needs a safety case Unitree itself warns users to maintain a safe distance. The H2's peak joint torque and human-scale mass are useful for dynamic motion, but they raise the consequence of a fall or control error. Buyers need site-specific risk controls, emergency-stop procedures, restricted operating zones and documented training before deployment around people. Neither company publishes enough field data Unitree publishes product specifications. Tesla publishes program goals and manufacturing updates. Neither source provides the customer uptime, intervention frequency, mean time to recovery or multi-month task-success data needed to prove dependable general-purpose work. H2 wins this comparison without earning a blanket reliability endorsement. Which robot fits your use case? For a wider shortlist, compare humanoid robots that can actually be bought before narrowing around a famous but unavailable platform. Final verdict: H2 wins procurement, Optimus wins the watchlist Unitree H2 is the clear winner for a lab or commercial team procuring one of these robots in 2026. It can be priced, ordered, configured and tested. H2 EDU also provides a real development path for labs that need to write software and integrate new models. Tesla Optimus remains the higher-upside manufacturing bet. Tesla's planned scale could make Optimus far more consequential than H2, but that outcome cannot be scored as a present capability. Gen 3 specifications, customer terms and repeatable field performance still need to appear. The comparison changes when Tesla publishes a production specification, opens a legitimate order channel and provides customer deployment evidence. Until then, Optimus belongs on the watchlist. H2 belongs on the shortlist. Compare on RoboZaps: Unitree H2, Tesla Optimus Gen 3 and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs UBTECH Walker S2: Which Factory Humanoid Is Ahead? URL: https://blog.robozaps.com/b/tesla-optimus-vs-ubtech-walker-s2 Author: Radica Maneva Published: 2026-07-16T12:32:43.000Z Updated: 2026-08-01T23:03:04.003Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Walker S2 is ahead for factory deployment today. UBTECH reports production and delivery and publishes a 15 kg load and three-minute autonomous battery swap. Optimus remains a development and factory-buildout program with no public order channel or current product sheet. Key takeaways: - Walker S2 is ahead for factory deployment today. UBTECH says it entered mass production and delivery in 2025 and reached 1,000-unit-level small-scale production and delivery. - Optimus is still a production program, not a purchasable product. Tesla says production lines are being installed ahead of volume production. - Walker S2 publishes the more useful industrial specification: a 15 kg stable load, 0–1.8 m work range and three-minute autonomous battery swap. - Neither company publishes a current list price. Walker S2 requires enterprise procurement, while Tesla provides no public Optimus order channel. - The evidence is not symmetrical: UBTECH provides hardware and production claims for a named model; Tesla provides factory plans without a current Optimus product sheet. FAQ: Q: Is UBTECH Walker S2 available to buy? A: UBTECH reports that Walker S2 entered mass production and delivery in 2025, but it is not a retail product with a public checkout price. A buyer needs an enterprise sales conversation and should confirm availability, lead time, integration scope and local support. Q: Can you buy Tesla Optimus? A: No public Optimus order or reservation channel was available as of 16 July 2026. Tesla's current disclosures describe development and production-line preparation, not a commercial purchase offer. Q: Does Walker S2 run for 24 hours? A: UBTECH claims 24-hour continuous operation through a dual-battery system and autonomous swaps that take about three minutes. That is not a 24-hour single-charge runtime claim. Buyers should request measured availability, swap success rate and maintenance assumptions. Q: Which robot can carry more? A: Walker S2 has a published stable-load figure of up to 15 kg. Tesla does not publish a comparable current payload specification for the intended production Optimus, so a fair strength comparison cannot be made from official current data. Q: Is Walker S2 better than Tesla Optimus for factory work? A: For a factory pilot in 2026, yes. Walker S2 has the stronger current evidence because UBTECH reports production and delivery and publishes industrial task specifications. Optimus may become a serious alternative, but Tesla first needs to show the production design, commercial availability and customer operating results. The short answer If a manufacturer is choosing between these robots for a 2026 pilot, UBTECH Walker S2 is the credible shortlist candidate. UBTECH has published a production claim, an industrial hardware specification and a concrete uptime strategy. Tesla Optimus remains the more speculative option because its current public evidence is about factories being prepared for future volume production. That does not make Walker S2 a low-risk, off-the-shelf purchase. UBTECH's production figures are company-reported, pricing is quote-based and a buyer still needs to validate task success, intervention rate, safety and service support at the intended site. Our detailed UBTECH Walker review covers the broader Walker program, while the Tesla Optimus review tracks Tesla's price, availability and demonstrated capabilities. Tesla Optimus vs UBTECH Walker S2 at a glance The key caveat is that planned factory capacity and delivered robots are different measures. Tesla's proposed line capacity should not be compared with UBTECH's reported production as though both describe current output. Production and deployment: Walker S2 has the lead UBTECH's 2025 annual report says the Walker S2 series officially started mass production and delivery during 2025. It also describes 1,000-unit-level small-scale production and delivery. That is the strongest evidence in this matchup, but it remains a company statement rather than an independently audited installed-base count. UBTECH also says the wider Walker S series has progressed through industrial field trials and production-line tasks including material handling, sorting, assembly and quality inspection. That history supports the product's factory focus. It does not prove that every named Walker partnership used the S2 model, so older Walker S and Walker S1 deployments should not be relabeled as Walker S2 deployments. Tesla is at a different stage. Its Q1 2026 update lists Optimus manufacturing in California and Texas as under construction. Tesla says preparations for its first large-scale Optimus factory were due to begin in Q2 and that first-generation production lines were being installed in anticipation of volume production. Those are significant manufacturing plans, not evidence of commercial customer deliveries. Tesla's designed annual capacity figures describe the intended lines. They do not establish present production volume, availability or field reliability. Hardware and uptime: Walker S2 provides the usable specification The official Walker S2 product page and annual report describe a 15 kg stable load, a 0–1.8 m work range and waist rotation of plus or minus 162 degrees. UBTECH also describes 52 degrees of freedom and binocular stereo vision. These are vendor specifications, so a buyer should test them against the exact object, reach, cycle time and fixture geometry at the target workstation. Walker S2's most distinctive hardware feature is its autonomous battery swap. UBTECH says the robot can complete a swap in three minutes and uses a dual-battery system to support continuous operation. The company's 24-hour language refers to an operating model built around battery changes. It should not be read as 24 hours on one charge. Tesla's current AI and robotics page positions Optimus as a general-purpose biped and says the company is developing balance, navigation, perception and interaction with the physical world. It does not publish a current, orderable-model specification for payload, runtime, work envelope or battery change time. That missing information matters. Older Optimus prototype figures are useful for following the program, but they should not be carried forward as guaranteed specifications for the design Tesla intends to mass-produce. AI and factory work: both claims still need site testing Tesla's long-term case rests on its AI work, vertical integration and ambition to manufacture Optimus at automotive scale. The company may eventually benefit from shared perception, inference and manufacturing infrastructure across its vehicle and robotics programs. Today, however, Tesla has not published the customer deployment evidence needed to judge Optimus on intervention rate, task recovery or sustained factory throughput. UBTECH describes Walker S2 as part of a factory system rather than a standalone showpiece. BrainNet 2.0 and Co-Agent are intended to coordinate individual robots and groups, while the Thinker-VLA model is aimed at loading, grasping and dual-arm handling. Those claims fit industrial use, but they remain vendor-reported results. For procurement, the useful test is not which robot has the more ambitious AI story. It is whether the robot can complete the buyer's task repeatedly, recover from ordinary errors and operate safely without shifting hidden labour to technicians. Price and procurement: neither has a public sticker price Neither company publishes a current acquisition price for the robot in this comparison. Walker S2 is sold through enterprise engagement, where the commercial package may include integration, training, support and site-specific work. Tesla has no public Optimus order or reservation page. A future price target is not the same thing as a quote or an MSRP. Buyers should therefore model total deployment cost rather than place an unsupported purchase price into a budget. Our guide to humanoid robot costs explains the extra items that usually matter: integration, safety work, tooling, supervision, maintenance and downtime. Which robot should a manufacturer choose? Choose Walker S2 for a controlled 2026 evaluation Walker S2 is the sensible option when a buyer has a defined industrial task, access to UBTECH or an integration partner, and a budget for a measured pilot. The published load, work range and battery-swap system give the evaluation a concrete starting point. The contract should define task acceptance criteria, intervention limits, cycle time, availability, safety responsibility, spare parts, software support and an exit path if the pilot misses its targets. Ask UBTECH to identify reference sites using the S2 specifically rather than another Walker generation. Track Optimus if the buying window is later Optimus belongs on a watchlist for companies whose decision window starts after Tesla demonstrates volume production and external customer use. The next useful evidence will be a stable production design, current technical documentation, commercial terms and named deployments with measurable performance. Tesla's scale plans are relevant, but planned line capacity is not a reason to design a 2026 automation project around an unavailable product. Choose neither if the task needs a proven service level now Neither option is a conventional catalogue purchase with a public price, broad independent performance data and a standard service package. If the task must go live on a fixed date, compare the humanoids with established industrial arms, mobile manipulators or purpose-built automation before committing to a humanoid form factor. Questions to ask before signing a pilot How many customer sites use this exact robot generation in production rather than a lab or demonstration?What task success rate and human intervention rate has the vendor measured on a comparable workflow?What does the uptime claim include and exclude, especially battery changes, planned maintenance and remote assistance?Who owns workstation integration, safety validation and changes to fixtures or tooling?What is the full first-year cost, including deployment staff, support, spares and software?What happens if the robot misses the agreed cycle time or cannot recover from common faults? Final verdict UBTECH Walker S2 wins this comparison for current factory readiness. It is the only one of the two with a company-reported production and delivery record, a named industrial specification and a published battery-swap strategy. Tesla Optimus is the higher-uncertainty bet. Tesla is preparing for manufacturing scale, but its public evidence still stops short of a purchasable product and customer deployment record. For a buyer making a decision now, Walker S2 belongs in a controlled pilot. Optimus belongs on the watchlist. Compare on RoboZaps: Tesla Optimus Gen 2, Walker S, Walker S2 and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Figure 03 vs Boston Dynamics Atlas: Which Is Better in 2026? URL: https://blog.robozaps.com/b/figure-03-vs-boston-dynamics-atlas Author: Radica Maneva Published: 2026-07-16T12:12:34.156Z Updated: 2026-08-01T23:03:08.949Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Atlas is the better fit for heavy industrial material handling. Figure 03 is explicitly designed for homes and lighter, varied work. Their published payload and runtime figures are manufacturer specifications, not results from a shared benchmark. Neither robot has a public list price or consumer checkout. Key takeaways: - Atlas publishes the stronger industrial specification: 30 kg sustained payload, 50 kg instantaneous payload, IP67 protection and autonomous battery swaps. - Figure 03 is explicitly designed for homes and lighter, varied work, with a smaller body, soft coverings, palm cameras and sensitive fingertips. - The payload figures are not directly comparable: Figure lists 20 kg without the sustained and instantaneous definitions Boston Dynamics provides for Atlas. - Neither company publishes a list price or offers consumer checkout. A buyer needs a paid pilot and a full installed-cost quote. FAQ: Q: Is Figure 03 better than Boston Dynamics Atlas? A: It depends on the job. Atlas has the stronger published specification for heavy industrial material handling. Figure 03 is smaller, lighter and explicitly designed for home and general-purpose work. Q: Which robot can lift more, Figure 03 or Atlas? A: Boston Dynamics publishes a 30 kg sustained payload and 50 kg instantaneous payload for Atlas. Figure publishes a 20 kg payload for Figure 03 but does not use the same sustained and instantaneous definitions, so the figures are not a direct head-to-head test. Q: Which robot has longer battery life? A: Figure 03 lists five hours, compared with four hours during typical use for Atlas. Atlas can swap its own battery, so buyers should compare productive uptime across a full shift rather than battery duration alone. Q: Can consumers buy Figure 03 or Boston Dynamics Atlas? A: No. Neither robot has a consumer checkout or public list price. Atlas has a contact-sales process for qualified enterprise prospects; Figure is allocating robots across internal programs and selected commercial work. Q: Which robot is better for factory work? A: Atlas is the stronger default for heavy material handling because of its published payload, IP67 rating, enterprise integrations and autonomous battery swap. Figure 03 may suit lighter sequencing and manipulation work, but buyers need task-level pilot data for either robot. Q: Which robot is better for the home? A: Figure 03 is the only one of the two explicitly designed for household environments. Its smaller body, soft textile coverings, foam around pinch points and wireless charging support that goal, but public demonstrations do not replace independent safety and reliability evidence. Figure 03 and Boston Dynamics Atlas are built for different jobs. Figure 03 is a lighter general-purpose humanoid with an explicit home focus. Atlas is a 90 kg industrial robot designed for material handling and enterprise systems. Verdict: Put Atlas first for heavy factory work. Put Figure 03 first when the brief involves household environments or lighter, varied manipulation. Neither is a normal consumer purchase, and neither company publishes a list price. This comparison uses manufacturer specifications and deployment updates checked on July 16, 2026. For release and pricing detail, read the Figure 03 review and our Boston Dynamics Atlas review. Evidence table: Figure 03 vs Atlas specifications Sources: [1] Figure 03 product specifications; [2] Atlas product specifications; [3] Figure 03 design details. All links were checked July 16, 2026. These are manufacturer specifications, not results from an independent head-to-head test. Runtime does not equal uninterrupted productive work, and maximum load figures do not show task speed, reach or reliability. Figure 03 and Atlas target different work Figure describes Figure 03 as a general-purpose robot moving into the home. Its hardware follows that brief: washable textile covers, foam around pinch points, palm cameras, sensitive fingertips, improved audio and wireless charging. Figure also shows the robot handling household tasks such as laundry, cleaning and dishes. Those are company demonstrations, not independent evidence of household safety or reliability. Boston Dynamics positions Atlas as an enterprise material-handling robot. The product page emphasizes barcode scanning, fleet-wide skill deployment, autonomous battery swaps and connections to MES and WMS systems through Orbit. The distinction is not absolute. In June 2026, Figure reported Figure 03 working on an automotive sequencing workflow at BMW Group Plant Spartanburg, combining two-handed manipulation with walking and balance. Figure is pursuing both home and commercial work. Atlas starts with industrial uptime, load handling and integration. Payload, size and duty cycle Atlas publishes the higher load figures: 30 kg sustained and 50 kg instantaneous. Figure lists a 20 kg payload without stating whether that is sustained, instantaneous or measured at a particular reach. Atlas therefore has the clearer heavy-load specification, but the numbers should not be treated as a like-for-like test. Figure 03 is 29 kg lighter and about 17 cm shorter. That smaller body may matter in rooms, narrow back areas and existing work cells. It does not prove that Figure 03 is safer or easier to deploy around people. Buyers still need task-specific safety evidence and a site review. Figure publishes five hours of runtime, while Boston Dynamics lists four hours during typical use for Atlas. The service model can outweigh the headline number. Figure 03 can recharge on a 2 kW wireless station. Atlas can navigate to its station, swap its battery and return to work. Compare productive time across a complete shift, including planned charging, swaps, resets and faults. Autonomy, sensing and manipulation Figure 03 runs Helix, Figure's vision-language-action system. Figure says Helix handles perception, movement and reasoning on board and in real time. Palm cameras provide close-range vision when the main cameras are occluded, while Figure says each fingertip sensor can detect three grams of pressure. Atlas uses tactile sensing and a 360-degree camera view, with 56 degrees of freedom and 2.3 m of reach. Boston Dynamics says a skill learned by one Atlas can be deployed across a fleet, while Orbit connects the robot to enterprise systems and tracks work and performance. There is no public benchmark showing that either AI stack is generally smarter. Figure gives more detail about language-led, varied manipulation. Atlas presents a fuller package for industrial fleet integration. Test both with the buyer's parts, lighting, floor layout, cycle-time target and recovery cases. Production and real-world readiness Figure has disclosed more manufacturing volume. On April 29, 2026, the company reported that BotQ had delivered more than 350 third-generation robots and demonstrated a one-robot-per-hour production cycle. Figure said those units were allocated across research, data collection, housework development and commercial use cases. This is a company-reported production milestone, not a count of paying customer deployments. The BMW sequencing work is a useful field signal because Figure 03 is operating in a factory workflow rather than only in a staged home demonstration. Figure's update does not publish uptime, intervention rate or cycle-time results, so it cannot yet be compared with a customer acceptance test. Boston Dynamics shows Atlas moving from a Hyundai customer pilot into a select early-adopter program. Its public sales path asks prospects to plan the application, evaluate the workflow, train task-specific skills and integrate the robot into operations. Neither record makes these robots commodity products. Figure has published more fleet and production data. Atlas has the more explicit enterprise product and integration specification. Price and availability Neither company publishes a list price on its current product page. Atlas directs qualified prospects to contact sales. Figure publishes product and manufacturing information but does not offer a public order page. A responsible quote needs more than the robot price. Include integration, end effectors or tooling, site work, support, training, spares, utilization and planned downtime. Our humanoid robot cost guide explains how those costs change the deployment budget. Individuals cannot check out with either robot. Enterprises can approach Boston Dynamics through its sales process; Figure is allocating robots across internal programs and selected commercial work. Buyers who need an orderable product should start with our guide to advanced humanoids available to buyers. Safety and serviceability Figure says Figure 03 uses soft textiles, multi-density foam around pinch points and a battery that achieved UN38.3 certification. UN38.3 covers battery transport testing. It is not a complete certification that the robot is safe for unsupervised home use. Atlas publishes IP67 protection and an operating range of -20°C to 40°C. Its autonomous battery swap can reduce manual intervention during a shift. Those specifications do not replace an application risk assessment. Before approving either robot, ask for stopping behavior, fault recovery, guarding assumptions, change-control procedures and evidence from the exact task. How to compare the robots in a paid pilot Set acceptance targets before either supplier tunes the workflow. Measure complete shifts, not the best demonstration cycle. Bottom line Atlas should lead the shortlist for heavy industrial material handling, harsh environments and enterprise fleet integration. Figure 03 should lead when the brief centers on a smaller body, varied manipulation and environments the company is explicitly designing around homes. Neither should win from brochure specifications alone. Run the same task, report the same measures and make the purchase decision from intervention rate, cycle time, productive uptime, recovery, safety and installed cost. Compare on RoboZaps: Figure 03 and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Boston Dynamics Robots: Atlas, Spot, Stretch and Real Costs URL: https://blog.robozaps.com/b/boston-dynamics-robots Author: Radica Maneva Published: 2026-07-16T05:33:11.000Z Updated: 2026-08-01T23:03:22.034Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: insights TL;DR: Boston Dynamics sells three robots in 2026 — the production Atlas humanoid (deployments committed this year), Spot (1,500+ deployed, quote-only at $150k-$375k per public records) and the Stretch warehouse robot — all six-figure enterprise systems with no public prices. Hyundai has owned the company since 2021; it is not publicly traded. Orbit software, now Gemini-powered, is the glue. Key takeaways: - No Boston Dynamics robot has a public price — Spot went quote-only years ago (the $74,500 figure is the dead 2020 launch price) and public records show configured units at $150k-$375k. - Atlas became a product in January 2026: production version unveiled, manufacturing started, customer deployments committed for this year. - Spot has 1,500+ robots deployed per Boston Dynamics' own count; the company's '2,000+ deployed' figure combines Spot and the Stretch warehouse robot. - Hyundai has owned Boston Dynamics since 2021 (after Google and SoftBank); it is not publicly traded — Hyundai stock is the closest proxy. - Orbit fleet software — now with Gemini-powered inspection AI — is increasingly the real product tying the robots together. FAQ: Q: Can you buy a Boston Dynamics robot? A: Only as a business, and only through a sales conversation. Boston Dynamics says its robots are not intended for individual, non-commercial purchase; there is no online ordering and no price list. Budget six figures per deployed robot based on public procurement records. Q: How much does Boston Dynamics Atlas cost? A: There is no published price. The production Atlas unveiled in January 2026 goes first to strategic customers, starting within Hyundai's orbit, and Boston Dynamics has not opened general sales. Any specific Atlas price you see is speculation. Q: Is Boston Dynamics publicly traded? A: No. Hyundai Motor Group has held the controlling stake since 2021, so Hyundai shares are the closest public-market exposure. The history: DARPA-funded MIT spinout, then Google (2013), SoftBank (2017), Hyundai (2021). Q: Is Spot still $74,500? A: No — that was the June 2020 online launch price and it is dead. Spot is quote-only today: public records show the LAPD's donated unit valued at $277,917, the NYPD's two at about $750,000 total, and Honolulu's at about $150,000, configured with payloads, software and support. Q: What robots does Boston Dynamics make? A: Three products in 2026: Atlas (electric humanoid, entering customer deployments), Spot (the quadruped inspection robot, 1,500+ deployed), and Stretch (warehouse case-handling). Orbit is the fleet-management software layer across them. Boston Dynamics robots are the most famous machines in robotics and the least straightforward to buy: none of them has a public price, the viral videos are a decade ahead of the product catalog, and the company sells through sales teams to enterprises only. This hub maps the current lineup — Atlas, Spot and Stretch — with what public records say they really cost, who owns the company, and how it stacks up against rivals that sell robots with checkout buttons. Every Boston Dynamics robot in 2026 Atlas, the electric humanoid, crossed from research demo to product at the start of 2026: Boston Dynamics unveiled the production version in January, said manufacturing would begin immediately, and committed to customer deployments this year, starting inside Hyundai's orbit. Our Atlas review tracks the specs and claims. Spot, the quadruped, is the commercial workhorse with 1,500+ robots in customer hands by the company's own count, walking inspection rounds for the likes of Michelin, National Grid and bp — and, as of a July 2026 prototype, trialling last-50-feet package delivery. The full buying picture is in our Boston Dynamics robot dog guide. Stretch unloads trailers and moves boxes in warehouses; Boston Dynamics' January 2026 figure of more than 2,000 deployed robots covers Spot and Stretch combined. Tying them together is Orbit, the fleet-management software whose inspection AI is now powered by Google's Gemini Robotics models — increasingly the actual product being sold. What Boston Dynamics robots cost There is no price list. Spot's dead 2020 online price of $74,500 still circulates, but the robot went quote-only years ago, and public procurement records put configured units between roughly $150,000 (Honolulu PD) and $375,000 each (NYPD); the LAPD's donated unit was valued at $277,917. Atlas has no published price at all — early units go to strategic customers, not open sales. Stretch is sold as an enterprise logistics program. The practical rule: budget six figures per deployed robot, and treat anyone quoting a firm Boston Dynamics price without a sales conversation as guessing. Who owns Boston Dynamics? Hyundai Motor Group has held the controlling stake since 2021, the fourth owner in the company's history: it was born from MIT spinout work funded by DARPA (the BigDog era), sold to Google in 2013, to SoftBank in 2017, and then to Hyundai at a roughly $1.1 billion valuation. The company is headquartered in Waltham, Massachusetts, and announced a $100 million expansion there in June 2026. It is not publicly traded — buying Hyundai is the closest proxy, a landscape we map in our humanoid robot stocks guide. Boston Dynamics vs the market The competitive story is polish versus price. Unitree sells a $2,800 robot dog and a $13,500 humanoid to anyone with a card; Boston Dynamics sells six-figure systems with enterprise software, support and the procurement acceptability that US government and utility buyers require — which is why both companies thrive without really competing for the same order. Against Tesla, the fight is Atlas versus Optimus for factory humanoid work, compared in our Optimus vs Atlas analysis. Boston Dynamics also organized the 2022 open letter against weaponizing general-purpose robots and enforces it in its terms of sale, a stance that shapes the police and military deployments its robots do and don't take. Where to go next Start with the Atlas review for the humanoid, the Spot guide for the robot dog and its real costs, and our humanoid robot rankings for how Atlas places against everything else shipping in 2026. Dated developments land on the humanoid robot news tracker. Compare on RoboZaps: Stretch, Spot and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robot Stocks: What You Can Actually Buy in 2026 URL: https://blog.robozaps.com/b/humanoid-robot-stocks Author: Radica Maneva Published: 2026-07-16T05:19:42.000Z Updated: 2026-08-01T23:03:25.772Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: insights TL;DR: Only one pure humanoid play trades today: UBTech (9880.HK). Tesla, NVIDIA and Hyundai (Boston Dynamics) are the proxies. Unitree's ~$618M Shanghai IPO — the sector's first pure-play listing — prices within weeks, and Agility Robotics arrives via SPAC (CCXI→AGLT) by end-2026. Pre-IPO 'shares' are the trap: Figure just declared unauthorized sales void. Not investment advice. Key takeaways: - Only one pure-play humanoid stock trades today: UBTech Robotics (9880.HK); Tesla, NVIDIA and Hyundai (Boston Dynamics' owner) are the usual proxies. - Unitree's ~$618M STAR Market IPO — the sector's first pure-play listing, ~$5.9B implied valuation — is approved and pricing, with a debut reported as early as late July 2026. - Agility Robotics goes public via Churchill Capital XI (trades as CCXI now, AGLT after the ~$2.5B merger closes, expected by end-2026); the draft S-4 hit the SEC July 14. - Pre-IPO 'humanoid shares' are the trap: Figure declared unauthorized transfers void in July 2026, naming four marketplaces — and Unitree has no ticker yet, so nobody can sell you real shares. - The cautionary tale is live: Vicarious Surgical (2021 robotics SPAC) votes on dissolution July 21, 2026, with common stockholders likely recovering nothing. FAQ: Q: What is the best humanoid robot stock? A: There is no clean answer because almost every maker is private. UBTech (9880.HK) is the only listed pure-play; Tesla is the highest-beta proxy via Optimus; NVIDIA sells the training stack to nearly everyone; Hyundai owns Boston Dynamics. Unitree's imminent Shanghai listing will create the first benchmark pure-play. This is reporting, not investment advice. Q: How do I buy Unitree stock? A: You can't yet. The IPO was approved July 3, 2026 and prices within weeks on Shanghai's STAR Market, which most non-Chinese retail investors cannot trade directly at debut — access typically comes later via Stock Connect eligibility or funds. Any 'Unitree shares' offered before the ticker exists are not the real thing. Q: Is Figure AI publicly traded? A: No. Figure is private at a $39B valuation from its September 2025 round, and in July 2026 it publicly declared any board-unapproved stock transfer void, naming Hiive, Forge Global, EquityZen and Unicorns Exchange. Buying Figure 'shares' on secondary marketplaces risks owning nothing. Q: When does Agility Robotics go public? A: The ~$2.5 billion merger with Churchill Capital Corp XI (Nasdaq: CCXI) is expected to close before the end of 2026, listing Agility as AGLT. The confidential draft S-4 was submitted to the SEC on July 14, 2026, with more than $620 million in expected proceeds including a Foxconn-led $200 million PIPE. Q: Are humanoid robot stocks a good investment? A: The demand story is real and the capital flows are huge — $1.2B+ in one July 2026 week — but valuations are being set in rich private rounds, timelines slip, and the sector's first SPAC generation produced wipeouts like Vicarious Surgical, which votes on dissolution July 21, 2026. Judge each listing on shipped robots and revenue, not demos. Not investment advice. Humanoid robot stocks are having their breakout month: Unitree is days from the first pure-play humanoid IPO, Agility Robotics filed its SPAC paperwork with the SEC, and more than a billion dollars of private money moved in a single July week. But the honest starting point for investors is that almost every humanoid maker is still private, the tradable options are mostly proxies, and the secondary market is full of traps that one maker just publicly disavowed. Here is what you can actually buy, what is coming, and what to avoid. This is reporting, not investment advice. Humanoid robot stocks you can buy today One pure-play already trades: UBTech Robotics listed in Hong Kong (9880.HK) in late 2023 and launched its $17,000-class U1 home humanoid this summer. Everything else is exposure through bigger businesses. Tesla is the highest-beta humanoid bet on public markets, with the market pricing Optimus years before revenue; its Gen 3 production story is the thing to actually track. NVIDIA sells the picks and shovels — the training stack behind most humanoid programs, now wired into the open-source LeRobot ecosystem. Hyundai owns Boston Dynamics outright, making it the quiet way to own Atlas and Spot. Xiaomi and other Chinese platform companies carry humanoid programs inside consumer-electronics balance sheets. In every case you are buying a large company where humanoids are a small slice; no listed stock today is a clean bet on the category except UBTech. The IPOs coming in 2026 Unitree is the one to watch: registration approved July 3 in the fastest STAR Market review ever, roughly RMB 4.2 billion (about $618 million) being raised at an implied valuation near $5.9 billion, debut reported as early as late July. It would be the first major pure-play humanoid listing, and its pricing becomes the sector's benchmark; our Unitree Robotics hub has the full picture. In the US, Agility Robotics is going public by merging with Churchill Capital Corp XI (currently trading as CCXI), a ~$2.5 billion deal expected to close before the end of 2026 under the ticker AGLT, with a $200 million PIPE led by Foxconn; the draft S-4 was confidentially submitted to the SEC on July 14. Behind them, Shenzhen's LimX Dynamics raised $200 million in pre-IPO money in July at a $2.21 billion valuation with its founder telling CNBC "listing is a must," likely in Hong Kong, and state-backed AI² Robotics closed roughly $735 million. The pipeline is real, but note what it means: the best assets are arriving at rich valuations set in private rounds. Pre-IPO "shares": the traps The hunger for humanoid exposure has created a gray market, and in July 2026 Figure did something unusual about it: the company publicly declared that any transfer of its stock not approved by its board is void, naming the marketplaces Hiive, Forge Global, EquityZen and Unicorns Exchange and one seller by name, and warning buyers they risk losing their money. Treat that as the template for the whole sector. Unitree has no ticker until its Shanghai debut prices, so anyone offering Unitree shares today is selling something else, and STAR Market listings are not directly tradable by most non-Chinese retail investors even after debut. If an offer of pre-IPO humanoid stock reaches you unsolicited, the realistic explanation is that you are the exit. The cautionary tale: Vicarious Surgical Public robotics markets destroy capital as efficiently as they mint it. Vicarious Surgical, a surgical-robot developer that went public in the 2021 SPAC wave with high-profile backers, asked its shareholders this month to approve dissolving the company, with the board warning that common stockholders will likely recover nothing. The vote is scheduled for July 21, 2026, one day before Tesla's earnings call. The lesson for humanoid IPO buyers is direct: robotics is capital-hungry, timelines slip, and a listing is not a destination. What to watch next Three dates matter right now: Unitree's pricing and subscription window (any day), the Vicarious dissolution vote on July 21, and Tesla's Q2 earnings call on July 22, where Optimus production timing gets confirmed or reset. We log every dated development in the humanoid robot news tracker and the weekly briefs, and the operating reality behind these valuations — which robots actually work and ship — is what our humanoid robot rankings measure. Compare on RoboZaps: Spot and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree Robotics: Every Robot, Real Prices and the IPO URL: https://blog.robozaps.com/b/unitree-robotics Author: Radica Maneva Published: 2026-07-16T05:15:43.000Z Updated: 2026-08-01T23:03:29.348Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: insights TL;DR: Unitree Robotics (Hangzhou, founded 2016) shipped 5,500+ humanoids in 2025 and sells the two most-bought legged robots: the $13,500 G1 humanoid and $1,600+ Go2 robot dog. Industrial models list at a $100,000 contact-sales placeholder. Its ~$618M STAR Market IPO — the first pure-play humanoid listing — is approved and pricing now, with a debut possibly late July 2026. No ticker exists yet. Key takeaways: - Unitree shipped 5,500+ humanoids in 2025 (IPO prospectus) on RMB 1.69B revenue — with an adjusted profit, rare in this sector. - Live store prices (July 16, 2026): G1 humanoid $13,500 (orderable), R1 from $4,900, Go2 robot dog $1,600–$4,500 + shipping; H1 lists $90,000 and H2 $29,900 with checkout disabled. - The flat $100,000 on B2, A2 and H2 Plus is a contact-sales placeholder, not a real price. - You cannot buy Unitree stock yet: the ~$618M STAR Market IPO (≈$5.9B implied valuation) was approved July 3, 2026 — the fastest STAR registration ever — with a debut reported as early as late July. - Unitree signed the 2022 anti-weaponization open letter alongside Boston Dynamics — the 'unbound Chinese rival' framing is wrong. FAQ: Q: Can you buy Unitree Robotics stock? A: Not yet. The IPO registration was approved on July 3, 2026 and Unitree is finalizing pricing for a Shanghai STAR Market debut reported as early as late July — roughly RMB 4.2 billion (~$618 million) raised at a ~$5.9 billion implied valuation. STAR Market shares aren't directly tradable by most non-Chinese retail investors at debut, and until a ticker exists, any 'Unitree shares' offer is not the real thing. Q: How much do Unitree robots cost? A: From the official store as of July 16, 2026: the G1 humanoid is $13,500 and orderable, the R1 starts at $4,900, and the Go2 robot dog runs $1,600 to $4,500 plus $399–$1,000 shipping. The H1 lists at $90,000 and H2 at $29,900 with ordering disabled, and industrial models (B2, A2, H2 Plus) show a $100,000 contact-sales placeholder. Q: Is Unitree a Chinese company? A: Yes — founded in 2016 in Hangzhou by Wang Xingxing. That matters for buyers two ways: its prices undercut Western rivals by roughly 10x, and its robots are effectively excluded from US and allied government procurement, which is why agencies buy Boston Dynamics and Ghost Robotics instead. Q: What is Unitree's cheapest robot? A: The Go2 Air robot dog at $1,600 (plus shipping) is the cheapest real robot; the R1 at $4,900 is the cheapest humanoid from any major maker, and the $13,500 G1 is the cheapest full-featured humanoid you can actually order. Q: Does Unitree ship to the US? A: Yes for consumer models — the Go2, G1 and R1 ship internationally from the official store, with shipping fees of roughly $399 to $1,000 depending on the robot. Industrial models go through sales conversations rather than checkout. Unitree Robotics is the company that turned humanoid and quadruped robots from lab demos into products with checkout buttons: it shipped more than 5,500 humanoids in 2025 per its IPO prospectus, its $13,500 G1 is the default answer to "cheapest real humanoid," and its Go2 robot dog starts at $1,600. It is also days from becoming the world's first publicly listed pure-play humanoid maker. This hub covers the company, every current robot with live store prices, the IPO status, and where each of our detailed reviews fits. The company behind the robots Founded in 2016 in Hangzhou by Wang Xingxing, Unitree built its business on quadrupeds that undercut Western rivals by an order of magnitude, and Chinese makers led by Unitree and Deep Robotics now dominate the commercial robot-dog market. The IPO prospectus discloses 2025 revenue of RMB 1.69 billion (about $235 million) with an adjusted profit, rare among humanoid startups that mostly burn cash. Its cultural breakout moment came when its H1 humanoids danced on China's 2026 Spring Festival Gala, which we analyzed in our Spring Festival performance breakdown. Unitree is also one of the signatories of the 2022 open letter against weaponizing general-purpose robots, alongside Boston Dynamics. Every Unitree robot, priced from the live store Prices below are from shop.unitree.com as of July 16, 2026. A pattern to know: Unitree lists several industrial models at a flat $100,000 with ordering disabled; that figure is a contact-sales placeholder, not a real price. Humanoids. The G1 is $13,500 and actually orderable, which is why it anchors every cheapest-humanoid conversation. The smaller R1 starts at $4,900, the cheapest legged humanoid from a major maker. The full-size H1 lists at $90,000 with checkout disabled, and the newer H2 is listed at $29,900 but is not currently orderable either, with the H2 Plus carrying the $100,000 placeholder. In practice: consumer money buys a G1 or R1 today; the full-size machines are sales conversations. Every one of these holds an FCC authorization granted before the July 2026 Covered List entry, so the whole current lineup is grandfathered rather than blocked. The evidence for each is in humanoid robots legal in the US. Quadrupeds. The Go2 runs $1,600 (Air) to $4,500 depending on trim, plus $399 to $1,000 shipping, and is the value benchmark of the whole robot-dog category; the wheeled Go2-W and industrial B2, B2-W and A2 sit behind contact-sales placeholders, and the legacy AlienGo still lists at $50,000. The Go1 remains the used-market budget entry. Our robot dogs guide ranks them against everything else on the market. Unitree Robotics stock and the IPO You cannot buy Unitree stock yet. China's securities regulator approved the company's registration on July 3, 2026, the fastest approval in the STAR Market's history, and Unitree is finalizing pricing to sell at least 40.45 million shares, at least a 10% stake, raising roughly RMB 4.2 billion (about $618 million) at an implied valuation near RMB 42 billion (about $5.9 billion). Reports point to a Shanghai debut as early as late July. Two practical notes: the listing is on Shanghai's STAR Market, which most non-Chinese retail investors cannot trade directly at debut, and until the ticker exists, anyone offering you "Unitree shares" is selling something else. We track the dated developments in our humanoid robot news tracker. The full tradable landscape, proxies included, is in our humanoid robot stocks guide. How Unitree compares Unitree's position is price-performance, not polish. Against Tesla's Optimus the comparison is shipping-versus-promised, which our Optimus vs G1 analysis covers in detail. Against Boston Dynamics, Unitree sells a $2,800 Go2 while a configured Spot costs six figures; the premium buys enterprise software, support and procurement acceptability that Chinese vendors cannot offer US government buyers. In the consumer and research market, that trade usually lands in Unitree's favor, which is why its robots fill our humanoid rankings. The other side of that trade is mapped in our Boston Dynamics hub. Where to go next Start with the G1 review if you are pricing a humanoid, the Go2 review if you want the robot dog, and the R1 vs G1 comparison if you are deciding between the two humanoids. The IPO story updates on the news tracker as filings land. Compare on RoboZaps: Aliengo and Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Loona Robot Review: The $499 Petbot With GPT-4o, Honestly Rated URL: https://blog.robozaps.com/b/loona-robot-review Author: Radica Maneva Published: 2026-07-16T04:30:46.000Z Updated: 2026-07-31T03:46:37.800Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The Loona robot (KEYi Petbot Premium) costs $499.00 on the official store in July 2026 (list $529), includes GPT-4o voice chat with no subscription for now, and has the best character animation in the consumer petbot class. It's a wheeled indoor companion, not a walking robot dog. Value pick for families; seniors are better served by the $179.99 Joy for All Pup. Key takeaways: - Loona Petbot Premium costs $499.00 on KEYi's official store as of July 2026 (list $529.00), with charging dock and game props included. - GPT-4o voice conversations are built in with no subscription required today — KEYi says cloud AI is free 'for the moment' and has flagged tariffs/cloud costs as future pricing pressure. - It's a wheeled indoor companion robot, not a walking robot dog — the personality and character animation (CES Innovation Award) are the product. - The 4.9-star/1,000+ review score is self-published on KEYi's own store; independent reviews are positive but cite battery life and connectivity gripes. - Cloud dependency is the structural risk: the AI features rely on KEYi's servers, the same category risk Sony's Aibo Japan wind-down highlighted this year. FAQ: Q: How much does the Loona robot cost? A: $499.00 on KEYi's official store as of July 2026, marked down from a $529.00 list price, including the charging dock and game prop kit. It's also sold on Amazon. KEYi has warned that tariffs could push prices up. Q: Does Loona require a subscription? A: Not today. GPT-4o conversations, voice commands and home monitoring work without a monthly fee — KEYi says cloud AI access is free 'for the moment,' while flagging cloud and tariff costs as future pricing pressure. Treat the free tier as current policy, not a guarantee. Q: Is Loona a real robot dog? A: It's marketed as a petbot and often called a robot dog, but Loona moves on wheels, not legs. It's an expressive indoor companion with games, face recognition and a patrol camera. If you want walking, that's the Unitree Go2 (from $1,600); if you want realistic fur, that's the Joy for All Pup ($179.99). Q: Is Loona good for kids or seniors? A: Families with kids are its best fit: games, chatty AI and durable-enough indoor play. For seniors and dementia care, a realistic non-wheeled companion like the Joy for All Pup is usually the better gift — it's designed for that exact use. Q: Loona vs Aibo — which robot pet is better? A: Aibo ($2,899.99 plus cloud plan) walks on legs and has the deepest behavior model ever shipped in a consumer robot pet. Loona delivers a surprising share of the personality experience for 15% of the price, on wheels, currently without subscription fees. Aibo also carries fresh platform risk after Sony's June 2026 Japan wind-down announcement. The Loona robot is the best-selling answer to a question most robot-dog searches secretly contain: what if you want a robot pet with real personality but you are not spending Aibo money? As of July 2026, KEYi's Loona Petbot Premium sells for $499.00 on the official store, marked down from a $529.00 list price, with a charging dock and game props included, GPT-4o conversations built in, and no subscription required. This review covers what Loona actually is, what the AI does, the caveats KEYi's marketing skips, and who should buy something else instead. For the whole category, see our best robot dogs guide. What the Loona robot actually is First, the honest classification: Loona is marketed as a petbot and often called a robot dog, but it is a wheeled desk-and-floor companion robot, not a walking quadruped. It scoots on two drive wheels with an expressive head, ears and big animated eyes, and its personality is the product: it recognizes faces, plays chase-and-fetch style games with the included props, reacts to gestures, gets visibly excited or sulky, and patrols your home with a camera you can view from the app. The design won a CES Innovation Award, and the character animation is genuinely the best in the consumer petbot class; it reads as a Pixar sidekick rather than a toy. The AI: GPT-4o conversations with no subscription (for now) Loona's headline 2026 feature is built-in GPT-4o voice conversations: you talk to it, it talks back in character, and it uses the model for open-ended chat rather than canned responses. KEYi says cloud AI access is currently free with no monthly fee, a real advantage over subscription-gated rivals, and the company has been direct that tariffs and cloud costs could pressure pricing later, so "free for the moment" is the accurate phrasing. Voice commands, home monitoring and the app's core features all work without extra charges today. What KEYi's marketing skips Three caveats from the buyer's side. One: Loona is not durable outdoor hardware; it is an indoor companion with a camera, and hard floors are its habitat. Two: the store's 4.9-star rating from over a thousand reviews is self-published on KEYi's own site, so weigh it accordingly; independent reviews are positive but more measured, with battery life and occasional connectivity complaints the recurring gripes. Three: the AI features depend on KEYi's cloud, which means the product's long-term value rests on a startup keeping servers on, the same structural risk every cloud-connected pet carries; Sony's Aibo wind-down in Japan is the category's cautionary tale this year. Loona vs the alternatives Against the $79.99 WowWee Dog-E, Loona is in a different league of intelligence and animation, and worth the gap for anyone who will actually engage with it. Against the $179.99 Joy for All Pup, it is a different product entirely: Loona is a clever gadget pet, the Pup is a realistic comfort companion for seniors, and our realistic robot dogs guide covers that split. Against Sony's $2,899.99 Aibo, Loona delivers a surprising share of the personality experience for 15% of the price, without Aibo's legged walking and with less depth to the behavior model. And against the walking Unitree Go2 at $1,600 plus shipping, the choice is personality versus locomotion. If a talking, game-playing companion is the goal, Loona at $499 is the value pick of the category in 2026; the heavily advertised Wuffy and Nicoo "robot dogs" at similar or lower prices are documented drop-ship bait, covered in our robot dog toys guide. Bottom line Loona is the best personality-per-dollar robot pet you can buy in 2026: $499 on the official store, GPT-4o chat with no subscription today, and character animation nothing else near the price matches. Buy it as a family gadget pet with eyes open about the cloud dependency; buy the Joy for All Pup for a senior instead, and check the full robot dogs ranking if you are still deciding which kind of robot dog you actually want. KEYi sells Loona direct, so RoboZaps does not carry it. If you want a quadruped that actually walks in the same money, the $289 Petoi Bittle and $649 Mini Pupper 2 are both in the RoboZaps robot dogs collection. --- # Tombot Jennie: Price, Release Date and What It Actually Does URL: https://blog.robozaps.com/b/tombot-jennie Author: Radica Maneva Published: 2026-07-16T04:23:38.000Z Updated: 2026-08-01T23:03:33.308Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: Tombot Jennie — the hyper-realistic Labrador puppy robot designed with Jim Henson's Creature Shop for seniors with dementia — remains waitlist-only as of July 2026. No final price is published (Tombot has cited ~$1,500) and first shipments are targeted for late 2026 after earlier slips. Need something now? The Joy for All Pup ($179.99) is the closest shipping product. Key takeaways: - Jennie is waitlist-only as of July 2026: tombot.com takes free reservations and will confirm pricing and availability closer to shipping. - No final price exists — Tombot has previously cited around $1,500; first customer shipments are targeted for late 2026 after earlier slips. - It's a Labrador retriever puppy designed with Jim Henson's Creature Shop, deliberately non-walking, built as an emotional-support device for dementia, MCI, autism, anxiety and PTSD. - Founder Tom Stevens created Jennie after his mother's Alzheimer's diagnosis forced her to give up her dog. - Anyone selling a 'Jennie' today isn't selling the real thing — and the closest shipping alternative is the Joy for All Pup at $179.99. FAQ: Q: How much does the Tombot Jennie cost? A: Tombot has not published a final price. The company has previously cited a target around $1,500, and its site says pricing will be confirmed with the waitlist closer to shipping. The waitlist reservation itself is free. Q: When will Tombot Jennie ship? A: First customer shipments are targeted for late 2026 as of July 2026. The date has moved several times through beta trials, so treat it as a target. Tombot contacts the waitlist — reported in the tens of thousands — when pricing and availability are set. Q: Is Tombot Jennie a real dog breed? A: Jennie is modeled on an 8-to-10-week-old yellow Labrador retriever puppy, chosen because Labs are America's most popular and most recognizable dog. The animation and proportions were designed with Jim Henson's Creature Shop. Q: Does Jennie walk? A: No, deliberately. Jennie is designed as a lap companion for seniors, including people with dementia who could not safely manage a mobile robot. It moves its head, tail and body, reacts to touch across its fur, and responds to voice commands with real recorded puppy sounds. Q: What should I buy instead of waiting for Jennie? A: For an immediate dementia-care or senior-companion need, the Joy for All Companion Pet Pup at $179.99 is the closest shipping product, with fur, a heartbeat and voice-responsive barks. For a tech-forward robot pet, Sony's Aibo ($2,899.99) leads on behavior. Avoid the social-media-advertised Wuffy and Nicoo, which are documented drop-ship scams. Tombot Jennie is the robot dog most people are actually asking about when they search for a realistic robotic pet: a Labrador retriever puppy designed with Jim Henson's Creature Shop as an emotional-support device for seniors with dementia and others who can no longer safely care for a live animal. Here is the honest status as of July 2026: Jennie is still waitlist-only, Tombot has not published a final price (it has previously cited around $1,500), and first customer shipments are targeted for late 2026. This page covers what Jennie does, who it is for, when you can realistically get one, and what to buy instead if you need something now. The wider market is ranked in our best robot dogs guide. What is the Tombot Jennie robot dog? Jennie is modeled on an 8-to-10-week-old yellow Labrador puppy, a breed chosen deliberately: Labs have been America's most popular dog for decades, and the familiar shape helps emotional attachment. The design was done with Jim Henson's Creature Shop, the studio behind decades of film puppetry, which is why Jennie's proportions and head movements read as animal rather than toy. Touch sensors across the body respond to where and how firmly it is petted, its barks and whimpers are recordings of real Labrador puppies, and it responds to dozens of voice commands. It is deliberately built not to walk: Jennie is designed to live in a lap, because its founding use case is a senior with dementia who cannot chase a wandering robot around the house. A companion smartphone app handles settings and software updates, and the battery is rechargeable. Why Jennie exists Tombot founder Tom Stevens created Jennie after his mother's Alzheimer's diagnosis forced him to rehome her dog, one of the hardest days of both their lives, and the replacement he wanted did not exist. The design brief has stayed clinical rather than gadget-driven: an emotional-support animal for people facing health adversities, including dementia, mild cognitive impairment, autism, anxiety and PTSD. That focus is also why Tombot works with research partners and pitches Jennie to senior-living facilities as well as families. Robotic pets are comfort products, not medical treatments, but the caregiving field has used them for years; the Joy for All companion pets that pioneered the category are widely deployed in dementia care. Tombot Jennie price and release date: the honest status As of July 2026, tombot.com takes free waitlist reservations and says it will contact the list with pricing and availability closer to shipping. The company has previously cited a target price around $1,500, and its store shows no purchasable product. First customer shipments have been targeted for late 2026, a date that has moved several times over the years as the company refined the hardware through beta trials, so treat it as a target rather than a promise. The waitlist is reported in the tens of thousands. Two warnings follow from that: any store claiming to sell a Jennie today is not selling the real thing, and the heavily advertised "realistic robot puppies" on social media, Wuffy and Nicoo among them, are documented drop-ship bait we break down in our robot dog toys guide. What to buy while you wait If the need is now, a dementia-care gift or a companion for a senior who cannot keep a live pet, the closest shipping product is the Joy for All Companion Pet Pup at $179.99: fur, a simulated heartbeat, and voice-responsive barks, from the Hasbro spin-out that has served this exact market for a decade. It is less lifelike than Jennie but it exists, today, for an eighth of Jennie's cited price. If the buyer wants a clever robot pet rather than a realistic one, Sony's Aibo at $2,899.99 delivers the most dog-like behavior ever shipped. We compare all of these in our realistic robot dogs guide. Bottom line Jennie is the most credible attempt yet at a robot dog that feels like an animal instead of a gadget, backed by a genuinely serious design partner and a founder story that keeps the product honest. But you cannot buy one today, the price is not final, and the shipping target has slipped before. Join the waitlist if Jennie is the right fit, buy the $179.99 Joy for All Pup if the need is immediate, and check our robot dogs ranking for everything in between. Jennie is waitlist-only and sold direct by Tombot, so RoboZaps does not carry it. For robot dogs that can be specified and quoted today, see the RoboZaps robot dogs collection. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # Police & Military Robot Dogs: Real Deployments, Costs and Rules URL: https://blog.robozaps.com/b/police-military-robot-dogs Author: Radica Maneva Published: 2026-07-16T04:17:23.000Z Updated: 2026-07-31T03:45:24.958Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: insights TL;DR: Police robot dogs are mainstream: 60+ US/Canadian bomb squads and SWAT teams use Boston Dynamics Spot (NYPD ~$375k each, LAPD $277,917). The military runs Ghost Robotics' ~$165k Vision 60 on base patrol. No armed robot dogs are deployed in US policing — vendor pledges and rules like NYC's proposed ASIMOV Act hold the line. Ukraine's 50,000-UGV order shows wheels, not legs, win battlefields. Key takeaways: - Police robot dogs are mainstream equipment: Bloomberg counted 60+ US and Canadian bomb squads and SWAT teams using Boston Dynamics Spot by late 2025. - Real public costs: NYPD paid ~$750,000 for two Spots (2023, forfeiture funds); LAPD's donated unit was valued at $277,917; Honolulu's ~$150,000 unit reportedly sits unused. - The US military's quadruped is Ghost Robotics' Vision 60 (~$165,000, reported): a few hundred walk Air Force base-perimeter patrols; DHS border use never went past 2022 testing. - No armed robot dogs are deployed in US policing: Boston Dynamics, Agility, ANYbotics and Unitree signed the 2022 anti-weaponization pledge (Ghost Robotics did not), and NYC's proposed ASIMOV Act would ban injury-capable police robots. - Ukraine's 50,000-UGV order for 2026 (IEEE Spectrum) shows militaries buy wheels for battlefields and legs for stairs, rubble and patrol routes. FAQ: Q: Do police really use robot dogs? A: Yes, routinely. More than 60 bomb squads and SWAT teams in the US and Canada use Boston Dynamics Spot, per Bloomberg's late-2025 reporting. The NYPD runs two ('Digidog'), the LAPD's SWAT unit has deployed its donated Spot on barricade and hazmat calls, and the typical jobs are approaching suspicious packages and clearing unsafe structures before officers enter. Q: How much does a police robot dog cost? A: Public records put configured units between roughly $150,000 and $375,000 each: the NYPD paid about $750,000 for two, the LAPD's donated Spot was valued at $277,917, and Honolulu's pandemic-era unit cost about $150,000. Ghost Robotics' military Vision 60 is reported at about $165,000. Payloads, software, training and support drive the totals. Q: Are robot dogs used in the military? A: Yes, mainly for base security rather than combat. A few hundred Ghost Robotics Vision 60s patrol US Air Force base perimeters, and the Department of Homeland Security tested them for border patrol in 2022 without a confirmed operational rollout. In actual war zones, Ukraine fields thousands of wheeled and tracked ground robots — legs remain a niche. Q: Are armed robot dogs legal? A: No US federal law bans them, but none are deployed in American policing. Boston Dynamics, Agility Robotics, ANYbotics and Unitree pledged in 2022 not to weaponize general-purpose robots and Boston Dynamics enforces it contractually; Ghost Robotics declined to sign. San Francisco reversed its 2022 lethal-robot authorization under protest, the LAPD's Spot is barred from carrying weapons, and New York's proposed ASIMOV Act would make the NYPD ban permanent. Q: Which robot dog do police departments buy? A: Almost always Boston Dynamics Spot, sold quote-only with bomb-squad payloads. Ghost Robotics' Vision 60 serves defense customers. Chinese quadrupeds from Unitree and Deep Robotics cost a tenth as much but are effectively excluded from US and allied government procurement, which is why Taiwan launched its own non-China robot-dog program in July 2026. Police robot dogs and military robot dogs are no longer a novelty story: more than 60 bomb squads and SWAT teams in the US and Canada now use Boston Dynamics' Spot, a few hundred Ghost Robotics Vision 60s patrol Air Force bases, and Ukraine is buying ground robots by the tens of thousands. This page tracks who actually deploys them, what public records say they cost, what they are allowed to do, and where the political lines are being drawn. For the civilian side of the market, see our robot dogs guide. Police robot dogs in the US: NYPD, LAPD and 60+ bomb squads The robot dog you have seen in police stories is almost always a Boston Dynamics Spot. The NYPD's "Digidog" is its nickname for the robot: a 2021 pilot was cancelled after public backlash, and Mayor Eric Adams brought it back in 2023 with two units costing about $750,000, paid from asset-forfeiture funds. They have since been used in hostage situations and structurally unsafe buildings. In Los Angeles, the city council voted 8 to 4 in 2023 to accept a $277,917 Spot donated by the LA Police Foundation, restricted to SWAT, hazmat and search-and-rescue, with no weapons and no facial recognition; commission reports counted at least seven SWAT deployments in its first years. Bloomberg reported in late 2025 that more than 60 bomb squads and SWAT teams across the US and Canada now use Spot. The typical jobs are unglamorous and genuinely dangerous: walking up to a suspicious package, opening a door on a barricade call, or checking a burning or collapsing structure before officers enter. Not every purchase works out. Honolulu's roughly $150,000 pandemic-era unit has reportedly sat unused since 2021, a reminder that a robot without a program behind it is an expensive paperweight. Military robot dogs: Ghost Robotics on bases and borders The US military's quadruped of choice is not Spot but the Ghost Robotics Vision 60, a ruggedized machine reported at about $165,000 per unit. A few hundred are in use with the Department of Defense, mostly walking perimeter-security patrols at Air Force bases, and the platform operates from -40°C heat to deep cold, covering around six miles per charge. Ghost Robotics, majority-owned by South Korea's LIG Nex1 since 2024, sells to defense and public-safety customers only; there is no consumer version. The Department of Homeland Security tested Vision 60s for southern-border patrol in 2022 to significant public criticism, and no operational border deployment has been confirmed since. The platform has also been shown carrying an arm and, in a 2021 trade-show demo by a partner company, a rifle, which brings us to the line the industry has drawn. Are armed robot dogs legal? There is no US federal law banning weaponized robots; the constraints are policy, procurement and public pressure. In 2022 Boston Dynamics, Agility Robotics, ANYbotics, Clearpath and, notably, Unitree signed an open letter pledging not to weaponize their general-purpose robots, and Boston Dynamics enforces it contractually, having cut off service to a group that mounted a paintball gun on a Spot in 2021. Ghost Robotics did not sign, arguing its defense customers should decide. City-level rules are emerging instead: San Francisco briefly authorized lethal robot force in 2022 before reversing under protest, and a proposed New York City bill nicknamed the ASIMOV Act would permanently bar the NYPD from deploying robots capable of causing injury. The LAPD's Spot authorization explicitly forbids weapons and facial recognition. So in practice: armed police robot dogs are not deployed in the US today, and the main quadruped vendors are contractually against it, but nothing in federal law prevents a military program from arming one. The battlefield reality: Ukraine's ground robots The largest real-world deployment of ground robots is not dog-shaped. IEEE Spectrum reported in July 2026 that thousands of uncrewed ground vehicles now operate in Ukraine's front-line gray zone, mostly hauling supplies and evacuating casualties, and that Ukraine ordered 50,000 UGVs for 2026, roughly triple its 2025 procurement. Wheeled and tracked platforms dominate because they are cheap and carry more; legged robots remain a niche for inspection and patrol rather than a battlefield staple. The lesson for the robot-dog category is that militaries buy legs where terrain and stairs matter, and wheels everywhere else. The next front: allied programs Quadruped procurement is going geopolitical. Taiwan launched a government-industry program in July 2026 to build domestically made robot dogs on a non-China supply chain, with its defense research institute targeting military mass production by 2028, an explicit response to Chinese dominance of the commercial quadruped market by Unitree and Deep Robotics. Chinese platforms are effectively excluded from US and allied government procurement, which is why the government market belongs to Boston Dynamics and Ghost Robotics despite Chinese robots costing a tenth as much. Our Unitree Go2 review covers the consumer side of that price gap. What police and military robot dogs actually cost Public records give real anchors: about $375,000 each for the NYPD's configured Spots, $277,917 for the LAPD's donated unit, about $150,000 for Honolulu's, and about $165,000 per Vision 60. Those prices include payloads, software, training and support, which is why they sit far above any list price. A department wanting a bomb-squad-capable quadruped should budget in the $150,000 to $400,000 range per deployed robot; details on what drives Spot's configuration cost are in our Boston Dynamics robot dog guide. Bottom line Police robot dogs are now routine equipment for bomb squads and SWAT teams, military robot dogs walk base perimeters rather than battlefields, and the armed versions remain demos and policy fights rather than deployed reality. The next few years will be decided less by robotics than by procurement politics: weaponization pledges, city ordinances like the ASIMOV Act, and allied programs built to avoid Chinese supply chains. The machine behind most of these deployments is Spot, and its full specification and quote route is on the Boston Dynamics Spot record. For the industrial inspection quadrupeds that do the same patrol work without the policing context, see ANYbotics ANYmal D and the Deep Robotics Jueying X20. --- # Realistic Robot Dogs in 2026: What Actually Looks and Feels Real URL: https://blog.robozaps.com/b/realistic-robot-dogs Author: Radica Maneva Published: 2026-07-16T04:09:45.000Z Updated: 2026-07-31T03:45:33.211Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: best TL;DR: No realistic robot dog does looks, behavior and movement at once. Most lifelike: Tombot Jennie (waitlist, ~$1,500 cited, late-2026 shipping target). Buyable today: Joy for All Pup, $179.99 with fur and a heartbeat. Most realistic behavior: Sony Aibo, $2,899.99. Most realistic movement: Unitree Go2, from $1,600. The heavily advertised Wuffy and Nicoo are documented drop-ship scams. Key takeaways: - Realism splits three ways — looks, behavior, movement — and no robot dog delivers all three; pick which one you actually want first. - Most lifelike: Tombot Jennie (designed with Jim Henson's Creature Shop) — waitlist-only as of July 2026, ~$1,500 previously cited, first shipments targeted late 2026. - Buyable today and genuinely realistic to the touch: Joy for All Companion Pet Pup, $179.99, with fur, a simulated heartbeat and BarkBack responses — no subscription. - Most realistic behavior: Sony Aibo ($2,899.99 + cloud plan) — but Japan sales are winding down per Sony's June 2026 announcement, making this a final-generation window. - The heavily advertised Wuffy and Nicoo 'realistic robot dogs' are documented drop-ship scams: cheap plush toys that don't match the ad footage. FAQ: Q: What is the most realistic robot dog? A: Tombot's Jennie, a Labrador retriever puppy designed with Jim Henson's Creature Shop, is the most lifelike robot dog shown to date — realistic fur, recorded puppy sounds and full-body touch sensors. It is waitlist-only as of July 2026, with pricing to be confirmed closer to shipping (the company has cited around $1,500). The most realistic one you can buy today is the Joy for All Companion Pet Pup at $179.99. Q: Do any realistic robot dogs actually walk? A: The furry, lifelike ones don't — Jennie and the Joy for All Pup are deliberately stationary lap companions. Robot dogs that genuinely walk, like the Unitree Go2 (from $1,600 plus shipping), look mechanical. In 2026 you choose between realistic looks and realistic movement; no product does both. Q: Are realistic robot dogs good for people with dementia? A: They are widely used in senior-living and dementia-care settings, and caregivers report they provide comfort, reduce agitation and give a sense of companionship. The Joy for All Pup was designed specifically for this and Tombot's Jennie was created after the founder's mother's Alzheimer's diagnosis. They are comfort products rather than medical treatments, so set expectations accordingly. Q: Is the Wuffy robot dog real? A: Wuffy's ads are not what ships. Scam-checking sites document the pattern: lifelike demo footage, an anonymous storefront, and a cheap drop-shipped plush toy on arrival. The same applies to Nicoo. If a social-media ad shows a walking, talking realistic puppy for under $100, avoid it. Q: How much does a realistic robot dog cost in 2026? A: $179.99 gets the Joy for All Companion Pet Pup, the realistic option that ships today. Tombot's Jennie has been cited around $1,500 (waitlist). Sony's Aibo, the behavior benchmark, is $2,899.99 plus a cloud subscription. Walking robot dogs like the Unitree Go2 start at $1,600 plus $399-$1,000 shipping but look robotic. A realistic robot dog means three different things depending on who is asking, and no product on the market does all three. If you want a robot dog that looks and feels real, with fur, weight and a heartbeat, the answer is a $179.99 companion pet or Tombot's upcoming Jennie. If you want one that behaves like a real dog, learning and reacting with something like personality, that is Sony's Aibo at $2,899.99. And if you want one that moves with the uncanny agility of a real animal, that is a Unitree. This guide sorts the genuinely realistic robot dogs of 2026 by which kind of realism they deliver, with the prices and caveats the ads leave out. For the full category ranking, see our best robot dogs guide. What makes a robot dog realistic? Realism splits three ways. Looks: fur, floppy ears, believable weight, a simulated heartbeat, the things that make your hands accept it as an animal. Behavior: reacting to voice and touch, developing habits, seeming to have moods. Movement: walking, balancing and recovering like a living quadruped. The engineering trade-offs mean these mostly exclude each other; a furry companion pet cannot walk, and the robots that move like animals are built from aluminium and look like it. Decide which realism you actually want before spending anything. Tombot Jennie: the most realistic robot dog ever built Jennie, a Labrador retriever puppy designed with Jim Henson's Creature Shop, is the most lifelike robot dog anyone has shown: realistic fur, puppy proportions and movements, real recorded puppy sounds, and touch sensors across the body. It responds to voice commands and is built deliberately without walking, because it is designed to sit in the lap of a senior with dementia or anxiety, a use case Tombot's founder arrived at after his mother's Alzheimer's diagnosis. The catch is availability: as of July 2026 Jennie remains waitlist-only, with Tombot saying it will contact the waitlist with pricing and availability closer to shipping. The company has previously cited a price around $1,500 and first shipments are targeted for late 2026, with a reported waitlist in the tens of thousands. You cannot buy one today, and any store claiming to sell a Jennie now is not selling the real thing. Track the price and shipping status in our Tombot Jennie guide. Joy for All Companion Pet Pup: the realistic robot dog you can buy today The $179.99 Companion Pet Pup from Joy for All (Ageless Innovation, the company Hasbro spun out) is the realistic option that actually ships. It has soft fur, a simulated heartbeat you can feel, and BarkBack technology that responds to voice and touch with puppy sounds and head movements. It does not walk. It was designed for older adults and is widely used in senior-living and dementia-care settings; caregivers report it provides comfort and companionship, though it is a comfort product, not a medical treatment. For a realistic-feeling robot dog under $200 with no subscription, nothing else on the market competes with it in 2026. Sony Aibo: realistic behavior in a robot body Aibo is the opposite trade. Its moulded body is unmistakably robotic, but its behavior is the most dog-like ever shipped in a consumer product: it recognizes faces, develops a personality shaped by how you treat it, shows affection and moods through OLED eyes and body language, and patrols your home on its own. At $2,899.99 plus a cloud subscription it is a luxury, and there is a timing caveat: Sony announced in June 2026 that Aibo sales in Japan are winding down while US sales continue, which makes this a final-generation window. The full picture, including the subscription costs and what happens if support ends, is in our Sony Aibo review. Unitree Go2: realistic movement, no fur Nobody would mistake a Unitree Go2 for a labrador, but watch one recover from a shove or trot across gravel and it is the most animal-like machine you can own. From $1,600 for the Air (plus $399 to $1,000 shipping), it is the choice when realism means locomotion: dynamic balance, climbing, and gaits that look biological. It is a robotics platform with a companion app, not a companion product, and it is the right buy for tinkerers, educators and anyone who wants the engineering rather than the illusion. The "realistic robot dog" ads to avoid: Wuffy and Nicoo The most-searched "realistic robot dogs" of 2026 are the ones you should not buy. Wuffy and Nicoo run heavy social-media ad campaigns showing lifelike walking, talking robot dogs; what arrives, according to documented reviews across scam-checking sites, is a cheap plush toy drop-shipped from generic wholesalers, sold through anonymous storefronts that rebrand and reappear. The ads' demo footage does not match the product. We break down exactly how the pattern works, and what to buy instead at each price, in our robot dog toys guide. Which realistic robot dog should you buy? For a senior or a dementia-care gift you need this year: the Joy for All Pup at $179.99, with the option to join Tombot's waitlist for Jennie if timing is flexible. For the most lifelike robot dog, full stop: wait for Jennie. For realistic behavior and a genuinely clever pet: Aibo, accepting the price and the final-generation risk. For realistic movement and hackability: the Go2. And if an ad on social media is showing you a walking, talking puppy for $60, close the tab. Our robot dogs ranking covers the whole market beyond the realistic niche, including the industrial machines. Of the robots on this page, the Unitree Go2 is the one RoboZaps carries: specifications and pricing are on the Unitree Go2 X record. If realistic movement matters more to you than fur, the robot dogs collection covers every walking quadruped we stock, from the $289 Petoi Bittle up to Boston Dynamics Spot. --- # Boston Dynamics Robot Dog: What Spot Really Costs and Who Buys It URL: https://blog.robozaps.com/b/boston-dynamics-robot-dog Author: Radica Maneva Published: 2026-07-16T04:02:00.000Z Updated: 2026-07-31T03:45:16.394Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: Spot, the Boston Dynamics robot dog, has no public price in 2026: the famous $74,500 was the dead 2020 launch price, and public records show configured units at $150k-$375k. It is an enterprise inspection tool (1,500+ deployed, per Boston Dynamics) with Orbit fleet software, now also being tested as a package-delivery robot. For consumers, Unitree's Go2 costs 2% of the price. Key takeaways: - There is no public Spot price in 2026: Boston Dynamics sells via sales quotes only, and the $74,500 figure everyone quotes was the 2020 launch price, long dead. - Public procurement records anchor real configured costs: LAPD $277,917 (donated unit), NYPD ~$750,000 for two, Honolulu ~$150,000. - Specs: 33.8 kg, 1.6 m/s, ~90 min per swappable battery, 14 kg payload, IP54, climbs stairs; the optional 6-DoF arm opens doors and turns valves. - The moat is software: Orbit fleet management (inspection AI now Gemini-powered), Autowalk autonomy and a self-charging dock — that's what the premium over a $1,600-$2,800 Unitree Go2 buys. - Boston Dynamics claims 1,500+ robots in customer hands; its "2,000+ deployed" figure from January 2026 combines Spot with the Stretch warehouse robot. FAQ: Q: How much does the Boston Dynamics robot dog cost? A: Boston Dynamics publishes no price in 2026 — Spot is sold through sales quotes to commercial buyers only. The widely quoted $74,500 was the June 2020 online launch price and is dead. Public records show configured units at $150,000 (Honolulu) to about $375,000 each (NYPD); third-party integrators estimate a base unit near $75,000 rising past $120,000 with the arm and enterprise software, plus annual software fees, though Boston Dynamics has not confirmed those figures. Q: Can a normal person buy a Spot robot dog? A: Effectively no. Boston Dynamics says its robots are not intended for purchase by individuals for non-commercial use, there is no online ordering, and shop.bostondynamics.com sells merchandise like plush Spots, not robots. Consumer alternatives start at $1,600 with the Unitree Go2 Air. Q: Why do police departments use Spot? A: For jobs where sending an officer is dangerous: bomb calls, hazmat, barricaded suspects and search-and-rescue. The NYPD runs two units (nicknamed Digidog, ~$750,000 for the pair), the LAPD accepted a $277,917 donated unit restricted to SWAT and hazmat use, and Bloomberg reported in late 2025 that more than 60 US and Canadian bomb squads and SWAT teams use Spot. Deployments remain politically contested — New York's proposed ASIMOV Act would bar injury-capable police robots. Q: How long does Spot's battery last? A: About 90 minutes on average per battery, and Boston Dynamics notes payloads and environment shorten that. Batteries are swappable, and the Spot dock recharges the robot in roughly two hours, so continuous operation requires a dock or spare batteries. Q: Is Spot being used for package delivery? A: It's being tested. In July 2026 Boston Dynamics showed a "porch gap" concept where Spot rides in a delivery van and carries two parcels the last 50 feet to the door, targeting pilots of 200 packages a day. The company says it is in talks with logistics companies, but no partner is named and it is not a commercial service. Q: Is Spot better than a Unitree Go2? A: They answer different questions. A Go2 costs $1,600 to $2,800 plus shipping and is the rational choice for hobbyists, education and research on a budget. Spot costs one to two orders of magnitude more and buys enterprise autonomy software (Orbit), a payload ecosystem, support contracts and procurement acceptability for utilities, factories and public agencies. The Boston Dynamics robot dog, Spot, is the most famous quadruped robot in the world and the most misunderstood purchase in robotics. The $74,500 price you still see quoted everywhere, including in press coverage from this month, died years ago: Boston Dynamics publishes no price today, sells only through its sales team, and public procurement records show configured units running $150,000 to $375,000. Here is what Spot actually costs, what it does, who really buys it, and how it compares with Chinese quadrupeds that cost less than a used car. For the full category, see our robot dogs buying guide. For the company behind it, Atlas and Stretch included, see our Boston Dynamics hub. How much does the Boston Dynamics robot dog cost in 2026? There is no public price. The Spot product page has only Contact Sales buttons, and the company's FAQ says its robots "are not intended for purchase by individuals for non-commercial use." The famous $74,500 figure was the June 2020 online launch price for the Explorer kit, which included the robot, two batteries, a charger and a tablet controller. Boston Dynamics quietly removed online ordering, and that number no longer appears anywhere on its site. What configured units really cost shows up in public records. The LAPD's donated unit was valued at $277,917 in 2023. The NYPD paid about $750,000 for two robots the same year. Honolulu's pandemic-era purchase was about $150,000. One third-party integrator estimated in 2026 that a base unit starts around $75,000 with the arm pushing it near $100,000 and a full enterprise stack past $120,000, plus $15,000 to $25,000 a year in software, but treat that as an unofficial estimate: Boston Dynamics has not confirmed any of it. And a warning for searchers: shop.bostondynamics.com looks like the place to buy a robot, but it sells hoodies, brick sets and plush Spots, not robots. Spot specs: what the robot dog actually does Spot weighs 33.8 kg with its battery, tops out at 1.6 m/s, and averages 90 minutes of runtime per swappable battery, with Boston Dynamics noting payloads and environment shorten that. It carries up to 14 kg of payload, is rated IP54, operates from -20°C to 55°C, handles 30-degree slopes, and climbs steps up to 300 mm, which covers standard stairs. Perception is 360 degrees with about a 4 m range. The optional arm changes what the robot is for. It has six degrees of freedom, roughly a metre of reach, lifts 11 kg and drags 25 kg, and its gripper carries a 4K camera with a time-of-flight sensor, enough to open doors, turn valves and flip breakers on inspection rounds. A Spot CAM 2 pan-tilt-zoom payload with 25x optical zoom and radiometric thermal imaging arrived in January 2026 for inspection work. Software is the real product: Orbit, Autowalk and Gemini What separates Spot from cheaper quadrupeds is less the hardware than the stack around it. Orbit, Boston Dynamics' fleet software, handles mission scheduling, maps, live views, alerts and integrations, and its visual-inspection AI is now powered by Google's Gemini Robotics models. Autowalk lets an operator drive a route once, after which Spot repeats it autonomously, replanning around obstacles, and returns to its dock to self-charge in about two hours. The company said in January 2026 that its autonomous door-opening feature had already opened 2,500 doors across 18 beta customers. For developers there is a mature Python SDK, currently at version 5.1.4. Who actually buys Spot Boston Dynamics' product page currently claims more than 1,500 robots in customer hands; in January 2026 the company cited more than 2,000 deployed robots, but that figure combines Spot with its Stretch warehouse robot, so nobody outside the company knows the exact Spot count. The named customers are industrial: Purina, Michelin, National Grid, BMW, bp, POSCO, GlobalFoundries and Dominion Energy use it for inspection rounds, and one unit surveys radiation inside the Chornobyl exclusion zone for the UK Atomic Energy Authority. In 2026 FIFA's security team is patrolling World Cup broadcast and stadium perimeters with Spots. The pattern across all of them: repetitive inspection and patrol routes where sending a human is expensive, dull or dangerous. Police robot dogs: NYPD, LAPD and the pushback The "police robot dog" that keeps making headlines is Spot. The NYPD's "Digidog" is its nickname for the robot: an early pilot was cancelled in 2021 after public backlash, then Mayor Eric Adams brought it back in 2023 with two units costing about $750,000, paid from asset-forfeiture funds. A city-council bill, nicknamed the ASIMOV Act, has since been proposed to permanently bar the NYPD from deploying robots capable of causing injury. In Los Angeles, the city council voted 8 to 4 in 2023 to accept a $277,917 unit donated by the LA Police Foundation, restricted to SWAT, hazmat and search-and-rescue with no weapons and no facial recognition; reports to the police commission counted at least seven SWAT deployments since. Bloomberg reported in late 2025 that more than 60 bomb squads and SWAT teams across the US and Canada now use Spot. The counterexample is Honolulu, whose $150,000 unit has reportedly sat unused since 2021. No city has enacted an outright Spot ban that we could verify. The full picture across every force and the military is in our police and military robot dogs guide. New in July 2026: Spot the delivery dog On July 14 Boston Dynamics unveiled a "porch gap" concept: Spot rides in a delivery van with a conveyor payload holding two parcels, then walks them the last 50 feet to the doorstep. The company claims parcels that size cover at least 60% of an average van's packages and is targeting pilots of 200 packages a day alongside a human driver, with routes driven manually first and then repeated autonomously. It says it is in talks with major logistics companies; no partner has been named and there is no timeline. We covered the announcement in this week's news brief. Boston Dynamics robot dog vs Unitree: the honest comparison A Unitree Go2 starts at $1,600 for the Air and $2,800 for the Pro before $399 to $1,000 in shipping; Unitree's industrial B2 lists around $85,900 through resellers; Deep Robotics' Lite3 runs from roughly $2,900. A configured Spot costs one to two orders of magnitude more than a Go2, and for a hobbyist or a lab on a budget the Chinese quadrupeds are simply the rational choice. What the Spot premium buys is the surrounding system: Orbit fleet management, hardened autonomy that survives enterprise audits, a payload ecosystem, support contracts, and a vendor US institutions are allowed to procure from. One common framing is wrong, though: Unitree signed the same 2022 open letter against weaponizing general-purpose robots that Boston Dynamics organized, alongside Agility Robotics and ANYbotics. What Spot is not Spot is not a pet and not a home robot: Boston Dynamics says plainly that it is not intended for use in the home or by individuals. It cannot be ridden; the payload cap is 14 kg. It is contractually barred from being weaponized, a term the company enforced publicly when it cut off service to a group that mounted a paintball gun on one in 2021, and the 90-minute average battery means continuous operation requires the dock or spare batteries, not marketing's version of 24/7. If you want a robot dog at home, our robot dogs guide covers what you can actually buy, and our Atlas page covers Boston Dynamics' humanoid. Bottom line The Boston Dynamics robot dog is real, deployed and genuinely useful, but it is an enterprise inspection tool with an enterprise price that Boston Dynamics no longer publishes. If your job involves gas leaks, thermal anomalies, radiation surveys or bomb calls, Spot has a decade of deployments behind it and the software moat to justify a quote. If you just want a robot dog, you want a different page: the interesting consumer story is happening at one-fiftieth of the price. Spot is in the RoboZaps database with full specifications and a quote route: see the Boston Dynamics Spot record. If you are weighing it against the cheaper end, the Unitree Go2 X is the closest thing to a Spot alternative anyone sells with a published price, and the full robot dogs collection lists every quadruped we carry. --- # This Week in Humanoid Robot News: July 9–16, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-july-9-16-2026 Author: Radica Maneva Published: 2026-07-16T03:45:32.000Z Updated: 2026-08-01T23:03:37.116Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: news TL;DR: Unitree's IPO is in its final days with a late-July debut possible, one week brought $1.2B+ in new humanoid capital (Walden, LimX, AI²), 1X showed 25-DoF hands for NEO, Tesla cleared Fremont for Optimus with production still not started, and Boston Dynamics is testing Spot as a package-delivery robot. Key takeaways: - Unitree is finalizing pricing for its STAR Market IPO — a debut as early as late July 2026 is reported, raising ~$618M at a ~$5.9B implied valuation. - One week brought $1.2B+ in new humanoid capital: Walden Robotics ($300M seed at $1.1B), LimX Dynamics ($200M pre-IPO at ~$2.21B) and AI² Robotics (~$735M; valuation reports conflict). - 1X unveiled 25-DoF tendon-driven hands for NEO; order terms are unchanged at $20,000 or $499/month, with first home deliveries still promised for 2026. - Tesla tore down the Fremont Model S/X line in 46 days for the first Optimus line — but per Musk, production had not started as of mid-July. Q2 call: July 22. - Boston Dynamics is testing Spot as a last-50-feet package delivery robot — a prototype pilot concept, not a product. FAQ: Q: When is the Unitree IPO? A: No date is confirmed yet. Unitree is finalizing pricing and subscriptions after China's regulator approved its registration on July 3, 2026, and CNBC and SCMP report a STAR Market debut as early as late July. The plan: raise about RMB 4.2 billion (~$618 million) at a roughly $5.9 billion implied valuation. Q: Has Tesla started Optimus production? A: No. As of mid-July 2026 the Fremont Model S/X line has been cleared (a 46-day teardown) and the first Optimus line is being installed, but Elon Musk said on July 1 that production had not begun and initial output will be extremely slow. Tesla's July 22 Q2 earnings call is the next checkpoint. Q: Can you buy a 1X NEO right now? A: You can order one: 1X's order page lists $20,000 to own or $499 per month with a $200 refundable deposit, and the company says the first home deliveries — with its new 25-degree-of-freedom hands — happen later in 2026. Those delivery and performance figures are company claims. Q: What is Boston Dynamics doing with Spot and package delivery? A: Boston Dynamics published a prototype concept on July 14, 2026: Spot rides in a delivery van and walks packages the last 50 feet to the door, two at a time. The company says it is in pilot talks with logistics firms targeting 200 packages a day. It is a test, not a commercial service, with no timeline. Q: What was the biggest humanoid robot news this week? A: The capital wave: Unitree entering the final stage of its IPO, plus more than $1.2 billion raised across Walden Robotics, LimX Dynamics and AI² Robotics in a single week — against the counterpoint of Vicarious Surgical asking shareholders to dissolve the company. The humanoid robot news from July 9–16, 2026 was dominated by money: Unitree moved to the last step before its Shanghai listing, three more makers raised a combined $1.2 billion, and Agility Robotics took its SPAC a step closer to a ticker. Away from the term sheets, 1X showed NEO's new hands, Tesla finished clearing the Fremont line that will build Optimus, and Boston Dynamics tried something unexpected with Spot. As always, the buyer-relevant status changes land on our humanoid robot news tracker; this post is the dated record of the week. Unitree's IPO enters its final days Unitree is finalizing underwriting, pricing and share subscriptions for its STAR Market debut, which CNBC and the South China Morning Post report could come as early as late July. China's securities regulator approved the registration on July 3, the fastest approval in the STAR Market's history. The plan filed with the exchange: sell at least 40.45 million shares, at least a 10% stake, to raise roughly RMB 4.2 billion (about $618 million), implying a valuation near RMB 42 billion (about $5.9 billion). The prospectus discloses 2025 revenue of RMB 1.69 billion and more than 5,500 humanoids shipped last year. No price or subscription date had been confirmed by July 16. When it lists, Unitree becomes the first pure-play humanoid robot company on any public market, and its pricing will become the benchmark every other maker in this section gets measured against. Our Unitree G1 and H2 pages track what the company actually sells today. One week, $1.2 billion: Walden, LimX and AI² Robotics Three raises landed in seven days. The most surprising was Walden Robotics, a Toyota Research Institute spin-out led by Russ Tedrake, which emerged from stealth on July 15 with roughly $300 million in seed funding at a $1.1 billion valuation; backers include Toyota, Nvidia, Boeing, Samsung Ventures and CoreWeave. Walden says its robots have been doing production work at a Toyota plant in North America since February, a company claim with no audited numbers behind it yet. In Shenzhen, LimX Dynamics closed a $200 million pre-IPO round on July 14 at about $2.21 billion post-money, bringing its six-month total to roughly $400 million; founder Will Zhang told CNBC that "listing is a must." And The Robot Report covered AI² Robotics' roughly $735 million in fresh capital for its wheeled AlphaBot humanoids, with state-backed funds and Moutai Group on the cap table. Reported valuations for AI² conflict, ranging from $2 billion to $3 billion depending on the outlet, so treat any single figure with caution. What is actually tradable across the sector is mapped in our humanoid robot stocks guide. Agility Robotics, meanwhile, confidentially submitted its draft S-4 with the SEC on July 14, advancing the $2.5 billion Churchill Capital XI merger that would list it as AGLT with more than $620 million in expected proceeds, including a $200 million PIPE led by Foxconn. Agility says Digit has logged over 65,000 hours across nine customer facilities. The same week supplied the cautionary tale: Vicarious Surgical, a robotics SPAC of the 2021 vintage, asked shareholders to approve dissolving the company on July 21, with the board warning common stockholders are unlikely to recover anything. 1X shows NEO's hands On July 9, 1X published a deep look at NEO's new tendon-driven hands: 25 degrees of freedom including the wrist, force-transparent quasi-direct-drive tendons, an IP68 food-safe build, and claims of ±0.2 mm positioning and 45 N fingertip force. 1X says hundreds of the hands have come off its in-house line, with capacity for 10,000 this year, and that they will ship on the first home-delivery NEOs later in 2026. Every performance number is a vendor claim demonstrated on video, tea pouring and jacket zipping included, rather than third-party testing. The commercial terms have not moved: $20,000 to own or $499 per month, with a $200 refundable deposit. Details and caveats are in our 1X NEO price and availability guide. Tesla clears Fremont for Optimus, production still not started Tesla's manufacturing team posted a timelapse around July 10 showing the Fremont Model S and X line torn down in 46 days, clearing the floor for the first Optimus Gen 3 production line. Tesla frames the line as the first step toward a design target of one million robots a year; Elon Musk, walking the site on July 1, pushed back on reports that production had quietly begun and said initial output will be "extremely slow," pointing to roughly 10,000 unique parts. So the confirmed facts are: the old line is gone, the new one is being installed, and no Optimus production had started as of mid-July. The July 22 Q2 earnings call is where Tesla either confirms the late-July-to-August output target or moves it again; our Optimus production timeline tracker follows every dated claim. Boston Dynamics tests Spot as a delivery robot Boston Dynamics published a "porch gap" concept on July 14: Spot rides in a delivery van with a conveyor payload, then walks packages the last 50 feet to the doorstep, two parcels at a time. The company estimates that covers about 60% of the packages in a typical van and says it is in pilot talks with major logistics companies, targeting 200 packages a day on routes that are driven manually first, then repeated autonomously. This is a prototype test, not a product, and there is no timeline. It is still the first serious residential use case for a robot that has lived in industrial inspection, and a notable widening of the quadruped market we map in our robot dogs guide. Robot surgeons pass a live-animal test A UC San Diego team reported in Nature on July 8 that teleoperated humanoid robots completed live gallbladder removals on pigs, one procedure with a human surgeon assisting and one performed by two humanoids working together. It is the first use of general-purpose humanoid hardware in surgery on living subjects. The researchers are direct about the limits: the procedures were slow, with recalibration and latency problems, and this is a proof of concept, not a clinical pathway. The signal is that general-purpose humanoid platforms are starting to reach domains that previously required million-dollar specialized systems. More humanoid robot news in brief Figure published a notice declaring unauthorized secondary sales of its stock void, naming four trading marketplaces and one seller. A signal of how hot, and how murky, pre-IPO humanoid paper has become; Figure 03 remains deployment-first, not sales-first.NVIDIA and Hugging Face integrated Isaac GR00T 1.7 and the Isaac Teleop data framework into the open-source LeRobot stack, consolidating the default open toolchain for humanoid developers.UMA, the Paris startup founded by ex-Tesla Optimus scientist Rémi Cadene, unveiled plans for a European-built humanoid called Northstar. It is a plan, not a robot: nothing has shipped.Apptronik unveiled Apollo 2 and opened a 90,000-square-foot training facility in Austin feeding its Google DeepMind partnership, and says Apollo 3, due next year, will be its first mature commercial product.China's industry ministry projected national humanoid output will exceed 100,000 units in 2026. A government projection via state media, not an audited count, but it is the number every China-versus-US market story will now anchor on.Taiwan launched a government-industry program to build domestically made quadrupeds on a non-China supply chain, with military mass production targeted for 2028.IEEE Spectrum documented Ukraine's order of 50,000 ground robots for 2026, the largest real-world ground-robot procurement anywhere.Agility CEO Peggy Johnson published six US policy recommendations for humanoids, from domestic component manufacturing to government procurement pathways. What to watch next week July 21: Vicarious Surgical shareholders vote on dissolution. July 22: Tesla's Q2 earnings call, the single most important date on the humanoid calendar this month, where Optimus production timing gets confirmed or reset. Any day: Unitree's price and subscription dates for the STAR Market debut. And 1X's first home NEO deliveries remain promised for later this year. For where every major robot actually stands, start with our humanoid robot rankings. Compare on RoboZaps: Figure 03, Tesla Optimus, Digit, Spot and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Sony Aibo Review 2026: Price ($2,899), Specs & the Final-Generation Window URL: https://blog.robozaps.com/b/sony-aibo-review Author: Radica Maneva Published: 2026-07-16T03:22:25.000Z Updated: 2026-07-31T03:46:21.288Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The Sony Aibo costs $2,899.99 in the US including a 3-year AI Cloud Plan ($300/year after that), and it remains the most lifelike robot dog ever sold: 22 axes, OLED eyes, personality that develops per owner, and recognition of up to 100 faces. Sony announced in June 2026 that Japan sales are ending as stock runs out — the US continues, but this generation is in its final window. Key takeaways: - Price: $2,899.99 in the US including a 3-year AI Cloud Plan; renewal is $300/year through Sony after that. - June 2026: Sony announced Japan sales end as stock runs out — the US market continues, with support, parts and cloud maintained and no end date announced. This generation is in its final window. - Hardware: 22 axes, OLED eyes, nose camera plus a back-mounted SLAM camera, touch on head/chin/back, about 2 hours of play per 3-hour charge. - Behavior: personality develops per owner, recognizes up to 100 faces, learns tricks, takes photos (500 stored in the cloud) and runs the entertainment-only aibo Patrol feature. - Honest downsides: a sluggish companion app, indoor-floor-only use per Sony's own guidance, Sony-dependent repairs, and a real cloud tether for memories and updates. FAQ: Q: How much does a Sony Aibo cost? A: $2,899.99 in the US, including three years of the AI Cloud Plan, the charging station and its toys. After year three the cloud plan renews at $300 per year through Sony. Q: Is Sony Aibo still being sold in 2026? A: In the US, yes for now. Sony announced in June 2026 that Japan sales are ending as remaining stock sells out, while saying support, parts and cloud subscriptions will be maintained and that the aibo business will continue. Q: What happens if the Aibo cloud plan expires? A: Sony says the plan is necessary for Aibo's full functionality and learning. Owners report that a lapsed plan freezes personality development and app features, and that cloud-stored memories and photos can be deleted — so treat the $300/year as part of the ownership cost. Q: Can Aibo be used as a security camera? A: Not really. The aibo Patrol feature registers up to 10 faces and reports sightings with photos through the app, but Sony explicitly states it is for entertainment purposes and not a security system. Q: How long does Aibo's battery last? A: About 2 hours of activity per charge, with roughly 3 hours to recharge. Aibo returns to its charging station on its own. Q: Is the Aibo worth it in 2026? A: If you want the most lifelike robot dog ever sold and accept the cloud subscription and wind-down risk, yes — nothing else behaves like it. If the Sony-server dependence bothers you, the honest alternatives are the $179.99 Joy for All Pup or a real quadruped like the Unitree Go2 from $1,600. The Sony Aibo price is $2,899.99 in the US, including three years of the AI Cloud Plan it needs to learn and grow. That has been true since 2018 — what changed in June 2026 is the clock: Sony announced it is ending Aibo sales in Japan as remaining stock sells out, leaving the US as the last market for the most lifelike robot dog ever sold. This review covers what the ERS-1000 actually does, what the cloud subscription really costs, and whether buying into a platform in its final generation window makes sense. What is the Sony Aibo? Aibo (ERS-1000) is Sony's flagship robot dog: a 2.2 kg, 30 cm companion with 22 axes of motion, OLED eyes, and a personality that develops from how you treat it. Sony revived the Aibo line in January 2018 in Japan after a 12-year hiatus, brought it to the US later that year, and has sold the same hardware generation since, in a handful of color editions. It recognizes up to 100 faces and treats people differently based on your history together — the closest thing to a pet that ships in a box. Sony Aibo price: what $2,899.99 buys The US bundle includes the robot, a charging station, its pink ball and aibone toys, paw pads, and a three-year AI Cloud Plan. The cloud plan is not optional garnish — it is where Aibo's memories, personality development, photo storage and software updates live. After year three, renewal is $300 per year through Sony. Owners report that letting the plan lapse freezes personality development and app features and can delete cloud-stored memories; Sony's own wording is simply that the plan is "necessary to take advantage of aibo's full functionality." Budget for it as part of the price of ownership. The 2026 situation: a final-generation window In late June 2026 Sony announced it will stop selling Aibo in Japan once inventory runs out, with no further units produced for that market. US sales continue for now, and Sony says technical support, replacement parts and cloud subscriptions will be maintained, with no end date announced, adding that "the aibo business will continue" with new products planned. Read that how you like: existing owners are supported, but the ERS-1000 generation is clearly winding down. If you have wanted one, this is the window; if you are risk-averse about long-term cloud services, that is the risk to weigh. Sony Aibo specs Size and weight: roughly 18 × 29 × 31 cm standing; about 2.2 kg.Movement: 22 axes — head (3), mouth, neck, waist, three per leg, one per ear, and a two-axis tail.Eyes and cameras: two OLED eyes; a front camera in the nose plus a dedicated SLAM camera on its back for mapping.Sensors: time-of-flight and two ranging sensors, touch on the head, chin and back, four paw pads, two 6-axis motion sensors, four microphones.Battery: about 2 hours of play per charge, roughly 3 hours to recharge; it walks itself back to the charger.Connectivity: Wi-Fi (2.4 GHz) and included LTE via the cloud plan. What Aibo actually does Aibo's core trick is becoming yours. It develops a personality from daily interaction, learns tricks (with more added through the My aibo app), takes photos of its life that sync to the cloud (up to 500 stored), and runs "aibo Patrol" — registering up to 10 household faces and reporting sightings with photos. Sony is explicit that Patrol is for entertainment, not a security system. There is also a free developer API with visual programming, still live in 2026, which has made Aibo a favorite in robotics classrooms. The honest downsides The app is the weak link. Reviewers have long found the My aibo app sluggish, and setup over 2.4 GHz Wi-Fi can be painful.It is an indoor-floor animal. Sony's own guidance rules out slippery floors, shag carpet, and anywhere near stairs. No jumping, no outdoor walks.Two hours of play, three of charging sets the rhythm of ownership.Repairs run through Sony. Owner communities report common hip failures and Sony-only servicing; there is no published repair menu, so factor in the ecosystem dependence.The cloud tether is real: $300 a year after year three, and the personality you spent years shaping lives on Sony's servers. Verdict: who should buy an Aibo in 2026 Nothing else sold today behaves like an Aibo — the Tombot Jennie (about $1,500, waitlist) is a lap companion, and Unitree's Go2 (from $1,600) is a capable quadruped rather than a pet. If you want a robot that feels like it knows you, and you accept the cloud subscription and the wind-down risk, the Aibo remains the best lifelike robot dog money can buy, and this generation will not be sold forever. If the $2,899.99 price plus $300 a year is too much dependence on Sony's servers, the honest alternatives are cheaper and dumber: see our best robot dog toys for the $80-$430 tier, or the full best robot dogs guide for the whole market. RoboZaps does not carry aibo, which Sony sells direct. If you want a walking quadruped we can quote, the Unitree Go2 X is the closest comparison at a published price, and the full robot dogs collection lists the rest. --- # Best Robot Dog Toys 2026: Real Picks From $80, and the Scam Ads to Avoid URL: https://blog.robozaps.com/b/robot-dog-toys Author: Radica Maneva Published: 2026-07-16T03:10:51.000Z Updated: 2026-07-31T03:45:41.282Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: best TL;DR: The best robot dog toy in 2026 is the $79.99 WowWee Dog-E while stock lasts; the most lifelike is the $179.99 Joy for All Companion Pet Pup, and the $499 KEYi Loona is the smart upgrade. The viral Wuffy and Nicoo "AI robot puppies" are documented drop-ship plush toys. A real robot dog that walks and responds starts above $1,500. Key takeaways: - Best real robot dog toy: WowWee Dog-E at $79.99 — app-connected with a personality-minting system, though 2026 retail stock is thinning. - Most lifelike plush: Joy for All Companion Pet Pup at $179.99 (a Hasbro spin-out), widely used with seniors — the honest version of what viral ads promise. - Smart upgrade: KEYi Loona at $499 — face recognition, games and GPT-4o voice on a wheeled base; a robot pet rather than a dog shape. - Wuffy and Nicoo are documented drop-ship products: anonymous storefronts shipping a few-dollar plush at a 10x markup, with overwhelmingly negative independent reviews. - The 30-second scam check: no company name, impossible ad behavior at a $40 price, permanent countdown discounts, and reviews that exist only on the seller's own site. FAQ: Q: What is the best robot dog toy in 2026? A: The WowWee Dog-E at $79.99 — an app-connected robot dog from a real toy company, with 200+ reactions and a unique personality per unit. Stock is thinning in 2026, so check availability at major retailers. Q: Is the Wuffy robot dog toy legit? A: No. Consumer-protection researchers document Wuffy as a drop-ship operation: an anonymous storefront shipping a basic battery-powered plush worth a few dollars wholesale. Independent buyer reviews are overwhelmingly negative, with about half giving one star. Q: Is the Nicoo robot dog toy real? A: It follows the same drop-ship pattern as Wuffy and has been traced to a product previously sold as 'Milow.' One buyer found the identical item on Temu for under $6. It has no app or AI. Q: What robot dog toy is best for kids? A: The WowWee Dog-E ($79.99) for app-era kids, or a named-brand RC dog under $60 for younger children. For a soft, lifelike companion, the Joy for All Pup ($179.99) is the best-built option. Q: Is there a robot dog toy that acts like a real dog? A: At toy prices, the Joy for All Companion Pet Pup ($179.99) comes closest — it barks back, responds to touch and has a simulated heartbeat, though it doesn't walk. Genuinely dog-like behavior with walking starts above $1,500 (Tombot Jennie) and peaks with Sony's $2,899 Aibo. Q: How much do robot dog toys cost? A: Real ones run from about $30 (generic RC toys) through $79.99 (WowWee Dog-E) to $179.99 (Joy for All Pup), with the $499 Loona as the smart-pet upgrade. Anything advertised as a lifelike AI puppy for $40 is a plush toy with a battery box. The robot dog toy market in 2026 splits cleanly in two: real products from named toy companies, starting around $80, and a flood of viral ads selling a "lifelike AI robot puppy" for $40 that arrives as a battery-powered plush. This guide covers the best robot dog toys you can actually buy, what each one really does, and exactly how to recognize the scam listings that dominate social media feeds. The short version Best robot dog toy overall: WowWee Dog-E, $79.99 list — app-connected, expressive, from a real toy company. Stock is thinning in 2026, so buy while it's on shelves.Best lifelike plush companion: Joy for All Companion Pet Pup, $179.99 — barks, responds to touch, has a heartbeat you can feel. The honest version of what the viral ads promise.Best upgrade: KEYi Loona, $499 — a wheeled robot pet with face recognition and voice AI. Not dog-shaped, but the most pet-like behavior under $500.Avoid: Wuffy, Nicoo, and lookalike ads. They are documented drop-ship products, not robots. Wuffy robot dog review: what you actually get Wuffy is one of the most-searched robot dog toys in America, and nearly all of that search volume is people checking whether the ad they saw is real. The short answer: consumer-protection researchers have documented Wuffy as a drop-ship operation. The storefront names no company and no address, the "70% off" countdown is permanent, and what ships is a basic battery-powered plush that walks a few steps and squeaks — a product available wholesale for a few dollars. Independent review platforms show overwhelmingly negative buyer feedback, with around half of reviewers giving one star and describing refund obstruction. There is no app, no AI, and nothing resembling the robot in the advertisement. Nicoo robot dog review: the same playbook Nicoo (also spelled Nico or Nicco) follows the identical pattern, and researchers have traced it to the same product previously sold under the name "Milow." It is a rebranded plush toy marked up several hundred percent through rotating storefronts and third-party marketplace listings. One Australian buyer paid over $40 and later found the identical item on Temu for under $6. Buyer reviews on independent platforms are uniformly negative. If you want a toy at this price, buy a named-brand RC toy from a retailer with a returns desk. How to spot a robot dog toy scam in 30 seconds No company name. Real toys name their maker; scam sites have a brand logo and nothing else — no address, no legal entity.The ad shows behavior no toy can do. Fluid walking, emotional reactions, and obeying speech at a $40 price point does not exist. Real robots that do this start above $1,500.Permanent countdown discounts. "70% off, today only" that never ends is a funnel, not a sale.You cannot find it at a real retailer. Genuine robot toys are on Amazon, Target or Walmart's own listings — not just marketplace sellers and a standalone Shopify site.Reviews live only on the seller's site. Check independent platforms; the ratings gap between the storefront ("4.7 stars") and independent reviews (1-2 stars) is the tell. WowWee Dog-E ($79.99): the best real robot dog toy WowWee has made robot toys since the Robosapien era, and Dog-E is its modern robot dog: an app-connected pet with a light-up POV tail display, over 200 sounds and reactions, and a "one in a million" personality minting system, so each unit develops its own character. It listed at $79.99 and has sold for less at major retailers. The caveat for 2026: retail stock is getting patchy, which usually signals end of line. If it is in stock at a real retailer, it is the best robot dog toy for kids at this price. Joy for All Companion Pet Pup ($179.99): the lifelike one Ageless Innovation's Companion Pet Pup — a Hasbro spin-out — is the toy the scam ads imitate: a soft, believable puppy that barks back when you speak, responds to touch, and has a simulated heartbeat. There is no app and it does not walk; it is built as a comfort companion and is widely used with seniors and people living with dementia. At $179.99, it is the genuine article for anyone who wants a pet-like presence rather than a gadget. KEYi Loona ($499): the smart upgrade If the budget stretches past toys, the Loona petbot is the most interesting robot pet under $500: face recognition, gesture games, a home-camera patrol mode, and GPT-4o-based conversation, all on a self-balancing wheeled base. It is not dog-shaped — think expressive robot pet rather than robot dog — but it behaves more like an animal than anything else at this price. KEYi sells it directly for $499 with no subscription currently required. See our full Loona review. What about the $30-$60 tier? Below roughly $60 the honest options are generic remote-control flip-and-walk dogs sold under rotating brand names on Amazon. They are real products and fine as party toys, but none has the app features, personality, or build quality of the Dog-E, and none behaves like the dogs in viral ads. Set expectations accordingly: at this price you are buying an RC toy, not a companion. When a toy isn't enough Real robotic dogs exist on a spectrum. The best robot dogs guide covers the whole range: Sony's $2,899 Aibo (the most lifelike robot dog ever sold, now in its final generation window), the Tombot Jennie therapy puppy (about $1,500, waitlist), and Unitree's Go2 — a real walking quadruped from $1,600. For desk-sized companions beyond dogs, see the best companion robots roundup. One tier this guide has not covered: programmable kits. The $289 Petoi Bittle and the $649 Mini Pupper 2 are real walking quadrupeds you assemble and then code, and they sit in the gap between a $180 toy and a $499 petbot. For a child who wants to learn rather than just play, they beat anything else at the price. Both are in the RoboZaps robot dogs collection. --- # Best Robot Dogs 2026: Real Prices From $80 Toys to Industrial Quadrupeds URL: https://blog.robozaps.com/b/best-robot-dogs Author: Radica Maneva Published: 2026-07-16T02:57:44.801Z Updated: 2026-08-01T10:54:26.698Z Last fact-checked: 2026-07-30T00:00:00.000Z Category: best TL;DR: The best robot dog for most people in 2026 is the Unitree Go2 (from $1,600); the best toy is the $79.99 WowWee Dog-E, the most lifelike is Sony's $2,899 Aibo (now in its final generation window), and seniors are best served by the $179.99 Joy for All Pup. The viral $40 robot dogs in social ads (Wuffy, Nicoo) are documented drop-ship plush toys, not robots. Key takeaways: - Best robot dog for most people: the Unitree Go2, from $1,600 (Air) and $2,800 (Pro) on Unitree's store, plus $399-$1,000 shipping to the US. - The viral $40 robot dogs in social ads (Wuffy, Nicoo) are documented drop-ship plush toys, not robots. Real robot pets start around $80 from named companies. - Most lifelike: Sony's Aibo at $2,899.99 with a 3-year cloud plan — and Sony announced in June 2026 that Japanese sales are ending, so this generation is in its final window. - Credible toy tier: WowWee Dog-E ($79.99, stock thinning) and the Joy for All Companion Pet Pup ($179.99) for seniors. - Industrial quadrupeds run from about $15,199 (Unitree Go2-W, reseller) to $106,000+ (B2 family); Boston Dynamics Spot no longer has a public price — the famous $74,500 was the 2020 launch price. FAQ: Q: What is the best robot dog in 2026? A: For most buyers it's the Unitree Go2: a real walking quadruped from $1,600. The best toy is the $79.99 WowWee Dog-E, the most lifelike is Sony's $2,899.99 Aibo, and for seniors the $179.99 Joy for All Companion Pet Pup ships today. Q: How much does a robot dog cost? A: Real robot dog toys run $80-$180 (WowWee Dog-E, Joy for All). Companion tier runs $499 (KEYi Loona) to $2,899.99 (Sony Aibo). Walking quadrupeds start at $1,600 with the Unitree Go2, and industrial models run from about $15,000 to over $100,000 via resellers. Q: Is the Wuffy robot dog legit? A: No. Consumer-protection researchers document Wuffy as a drop-ship pattern: anonymous storefronts, AI-generated ads, and a basic battery-powered plush worth a few dollars wholesale. Independent buyer reviews are overwhelmingly negative. Q: Is the Nicoo robot dog real? A: It follows the same drop-ship pattern as Wuffy — a rebranded plush toy marked up from a few dollars wholesale; one buyer found the identical item on Temu for under $6. It has no app or AI. Q: Is Sony Aibo still being sold? A: In the US, yes for now, at $2,899.99 including three years of the AI cloud plan. Sony announced in June 2026 that Aibo sales in Japan are ending as stock runs out, while support and cloud services continue — this generation is in its closing window. Q: What robot dog do police use? A: Boston Dynamics Spot. The NYPD's "Digidog" is a Spot (two units bought for $750,000 in 2023) and the LAPD fields a donated Spot valued at about $280,000. The military tier is Ghost Robotics' Vision 60, which is not sold to civilians. Q: How much does Boston Dynamics Spot cost? A: There is no public price anymore. The famous $74,500 was the 2020 online launch price; Boston Dynamics now sells Spot through enterprise sales only. Real deployments suggest configured units are a six-figure purchase. Q: Where can I buy a robot dog? A: The consumer tier is bought direct: Unitree sells the Go2 on unitree.com, Sony sells aibo in the US, KEYi sells Loona at $499, and Joy for All sells the $179.99 Pup on its own store. WowWee does not sell the Dog-E direct — it points buyers to retailers such as Best Buy and Macy's. Everything industrial (Boston Dynamics Spot, Deep Robotics X30, Unitree B2) is quote-only through enterprise sales or a named reseller, not a checkout page. There are more robot dogs for sale in 2026 than ever, and the price range is absurd: viral $40 toys from social media ads, an $80 app-connected pet from WowWee, Sony's $2,899 Aibo, Unitree's $1,600 Go2, and industrial quadrupeds that cost as much as a house. This guide ranks the best robot dog options at every real price point, using verified prices from official stores and named resellers. It also covers the one thing most robot dog lists skip: the ads you should not trust. How we picked: every price below came from the manufacturer's own store or a named US reseller, checked on 30 July 2026 — never from an aggregator or a marketplace listing. Where a maker sells only by quote, as Boston Dynamics and Deep Robotics do at the industrial end, we say so rather than repeat a retail figure that has been dead for years. Anything we could not price from a primary source is not ranked. The best robot dogs at a glance Best robot dog overall (consumer): Unitree Go2 Air, $1,600 from Unitree's store, a real walking, self-balancing quadruped.Best budget robot dog toy: WowWee Dog-E, $79.99 list, an app-connected robot pet from a real toy company. Retail stock is getting thin in 2026, so check availability.Best lifelike companion: Sony Aibo, $2,899.99 including a 3-year cloud plan. Sony announced in June 2026 that Japanese sales are ending, so this generation is in its final window.Best for seniors and dementia care: Joy for All Companion Pet Pup, $179.99, or the Tombot Jennie (about $1,500, waitlist, first shipments targeted for late 2026).Best developer platform: Unitree Go2 Pro ($2,800) or Deep Robotics' Jueying Lite ($2,890 to $18,900 by configuration).Industrial tier: Unitree B2 and Go2-W (reseller-listed from about $15,000 to $106,000) and Boston Dynamics Spot, which no longer has a public price at all. Working from a budget rather than a shortlist? Our robot dog cost guide breaks down every tier, plus the freight, duty and subscription costs the sticker price leaves out. Before you buy: the viral "robot dog" ads are not robot dogs Two of the most-searched robot dog names in America right now are not real robotics products. Wuffy and Nicoo, promoted through AI-generated TikTok and Facebook ads, follow a pattern that consumer-protection researchers have documented in detail: an anonymous storefront with no company name or address, a countdown discount, and a product that arrives as a basic battery-powered plush worth a few dollars wholesale. Buyer reviews on independent platforms are overwhelmingly negative; one Australian reviewer found the identical "Nicoo" item on Temu for under $6. Neither has an app, AI, or anything resembling the robot in the advertisement. The rule of thumb: a walking, sensing robot dog cannot be built for $40. Real robot pets start around $80 from named companies with support and retail distribution, and genuinely lifelike behavior starts in the four figures. If an ad will not tell you who makes the product, it has answered your question. Robot dog toys: the $80 to $180 tier Full breakdown with the scam checklist: our best robot dog toys guide. WowWee Dog-E ($79.99) WowWee has been making robot toys since the Robosapien era, and the Dog-E is the credible pick at toy prices: an app-connected robot dog with a light-up tail display, 200+ sounds and reactions, and a "one in a million" personality setup for each unit. It listed at $79.99 and has sold below that at major retailers, though 2026 stock is patchy, which suggests the line is winding down. If you want a real robot dog toy from a real company, buy this while it is still on shelves. Joy for All Companion Pet Pup ($179.99) Ageless Innovation's Companion Pet Pup, a Hasbro spin-out, is what the scam ads pretend to be: a soft, believable puppy that barks, responds to touch and sound, and has a heartbeat you can feel. There is no app and it does not walk; it is designed as a comfort companion, and it is widely used with seniors. At $179.99 on the official store (often via Amazon), it is the honest version of the lifelike-puppy promise. Programmable robot dog kits: the $289 to $649 tier The gap between a $180 toy and a $499 petbot is filled by build-it-yourself kits, and they are the most overlooked robot dogs on the market. You assemble them, then program them. For a child who wants to learn rather than just play, this tier beats anything else at the price. Petoi Bittle ($289) Bittle is a palm-sized open-source quadruped that genuinely walks, trots and recovers from being pushed over. It is programmable in block coding, Python or Arduino, which is why it turns up in classrooms and on IEEE's robot lists. The standard kit is $289 and the voice-controlled Bittle X is $319. Full specs on our Petoi Bittle record. Mini Pupper 2 ($649) MangDang's Mini Pupper 2 is the step up: a Raspberry Pi-powered quadruped built for ROS2, with SLAM and navigation working out of the box. It is aimed at university robotics courses and hobbyists who want a real robotics stack rather than a toy. Specs on our Mini Pupper 2 record. Companion robot dogs: $499 to $2,899 KEYi Loona ($499) Loona is not shaped like a dog, and it rolls on wheels rather than walking, but it behaves more like a pet than almost anything else at this price: it recognizes faces, plays fetch-style games, patrols as a home camera, and holds conversations through a GPT-4o-based voice mode. KEYi sells it for $499 (list $529) with no subscription currently required — up from $429 earlier in 2026, so older round-ups still quote the lower figure. Think of it as a robot pet rather than a robot dog in the strict sense. Our full Loona review covers the GPT-4o features and the caveats. Tombot Jennie (about $1,500, waitlist) Jennie is a therapy robot Labrador puppy designed with Jim Henson's Creature Shop for seniors and people living with dementia. It sits on a lap rather than walking, with touch sensors and an all-day battery. Tombot has been developing it for nine years, holds well over 18,000 pre-orders, and is targeting its first paid shipments for late 2026 at a price around $1,500. You cannot buy one today; you can join the waitlist. Our realistic robot dogs guide compares Jennie with everything lifelike that actually ships today. Full waitlist and price status lives in our Tombot Jennie guide. Sony Aibo ($2,899.99) Full breakdown: our Sony Aibo review. Aibo remains the most lifelike robot dog ever sold to consumers: it walks, learns, develops a personality, and recognizes its people. The US price is $2,899.99 including three years of Sony's AI cloud plan. The important 2026 news: Sony announced in June that Aibo sales in Japan are ending as stock runs out, while US sales and support continue for now. If you have been waiting on an Aibo, this generation is in its closing window. Developer and research quadrupeds: $1,600 to $19,000 Unitree Go2 ($1,600 to $2,800, plus shipping) The Unitree Go2 is the reason robot dogs became affordable: the Go2 Air is $1,600 and the Pro is $2,800 on Unitree's official store. Two caveats from our full review: Unitree charges $399 to $1,000 for shipping and the buyer handles customs, so US landed cost runs meaningfully higher, and the developer EDU tier is quote-only. The older Go1 is effectively discontinued; buy the Go2 instead. Deep Robotics Jueying Lite ($2,890 to $18,900) Hangzhou-based Deep Robotics is the most serious Chinese rival to Unitree in quadrupeds. Its Lite3 research dog is sold through US resellers in configurations from roughly $5,000 (basic remote-control) to $19,000 (LiDAR-equipped), and it competes directly with the Go2 EDU tier in university labs. Industrial and professional quadrupeds Unitree Go2-W and B2 (reseller-listed from about $15,000) Unitree's wheeled Go2-W lists at US resellers from about $15,199, and the industrial B2 family runs from roughly $85,000 to $106,000 reseller-listed, with Unitree itself quoting by contact. The B2 is the inspection and logistics flagship: IP67-rated, around 20 kg of continuous payload, and the platform behind the viral wheeled-leg off-road videos. Boston Dynamics Spot (no public price) Spot is the most famous robot dog in the world and also the hardest to price. The widely quoted $74,500 was the 2020 launch price for the online Explorer kit; Boston Dynamics no longer sells Spot online or publishes a price, and its store now sells only merchandise. Real deployments give the honest picture: the NYPD paid $750,000 for two units in 2023, and the LAPD's donated Spot was valued at about $280,000. Configured with an arm and sensors, Spot is a six-figure enterprise purchase with over 1,500 units deployed in industrial inspection. Our full Boston Dynamics robot dog guide covers the real procurement costs, specs and police deployments. Deep Robotics X30 (quote-only) The X30 is Deep Robotics' industrial flagship: IP67, rated from -20°C to 55°C, and built for power-grid and tunnel inspection. Officially it is quote-only; resellers estimate around $65,000 and up. Police and military robot dogs The robot dog you have seen with police departments is Spot: the NYPD's "Digidog" is a Boston Dynamics Spot, reintroduced in 2023 and used in hostage and standoff situations, and the LAPD fields one under SWAT-only restrictions. The military tier is different hardware entirely: Ghost Robotics' Vision 60, press-reported at roughly $150,000 to $165,000 per unit, patrols US Air Force bases and has been deployed by other governments. It is not sold to civilians, and Boston Dynamics prohibits weaponizing Spot. Our police and military robot dogs guide covers every documented deployment, cost and rule. Which robot dog should you buy? For a child or a fun desk pet, the WowWee Dog-E at $79.99 is the safe pick while stock lasts. For a senior or someone living with dementia, the Joy for All Pup at $179.99 ships today and the Tombot Jennie is worth the waitlist if you can wait until late 2026. If you want the most dog-like robot money can currently buy, the Aibo at $2,899.99 is in its final generation window. For makers, developers and anyone who wants a real walking quadruped, the Unitree Go2 from $1,600 is the best robot dog for the money in 2026, and it is not close. And if an ad on your feed offers a miraculous robot puppy for $40, it is a plush toy with a battery box. Spend the $80 on the real thing. And to compare live listings, specs and deployment evidence side by side, browse the robot dogs collection on the RoboZaps marketplace. Robot dogs are one branch of the family: see our ranking of the best humanoid robots, the guide to what humanoid robots cost, and every humanoid robot for sale right now. --- # Tesla Optimus Review 2026: Price, Gen 2 & Is It For Sale Yet? URL: https://blog.robozaps.com/b/tesla-optimus-gen-2-review Author: Radica Maneva Published: 2026-05-13T00:00:00.000Z Updated: 2026-08-01T11:16:28.220Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: reviews TL;DR: Tesla Optimus Gen 2 is not publicly orderable and has no MSRP as of July 23, 2026. Musk’s US$20,000–$30,000 figure is a long-term at-scale expectation, not a Gen 2 price. Tesla has demonstrated Gen 2 hardware and reported factory tasks by unspecified Optimus revisions, but publishes no complete Gen 2 datasheet, fleet, uptime or task-success data. Public sales remain a conditional end-2027 target. Key takeaways: - Tesla's reviewed public properties provide no Optimus order, reservation or deposit route, and Tesla has not announced an MSRP. - Musk's US$20,000–$30,000 statement is a long-term at-scale price expectation from October 2024, not the price of a Gen 2 robot. - The sub-US$20,000 figure has two dated meanings: an early 2022 selling-price guess and a 2025 production-cost target conditional on annual output above one million units. - Tesla's Gen 2 release gives relative improvements and hand, neck, foot and sensing details, but no final absolute Gen 2 height, mass, payload, battery capacity or top speed. - Tesla has demonstrated Gen 2 hardware and reported autonomous battery handling and simple factory work, but it has not published the revision, fleet scale, uptime, intervention rate or productivity behind those factory claims. - Musk said public sales might begin by the end of 2027 only if reliability, safety and functionality were high enough; Tesla has not committed to that date. FAQ: Q: How much is a Tesla Optimus Gen 2? A: Tesla has not announced an MSRP for Optimus Gen 2 or any customer configuration. Musk's US$20,000–$30,000 statement from October 2024 is a long-term expectation for Optimus at scale, not a Gen 2 list price, preorder price or quote. Q: Can you buy or preorder a Tesla Optimus in 2026? A: No public Tesla sales route existed in the official properties reviewed on July 23, 2026. Tesla offered no robot configurator, reservation, deposit or delivery contract; a legitimate future order should originate on an official Tesla domain and disclose its terms. Q: When will Tesla Optimus be available to the public? A: At Davos on January 22, 2026, Musk said public sales might begin by the end of 2027 if the robot reached high reliability, safety and functionality. That is a conditional target, not a committed release date. Q: What are the official Tesla Optimus Gen 2 specifications? A: Tesla's December 2023 release specifies Tesla-designed actuators and sensors, a 2-DoF neck, foot force and torque sensing, articulated toes, faster 11-DoF tactile hands, 30% faster walking and a 10 kg weight reduction. The last two are relative improvements, not absolute speed or mass figures. Q: What can Tesla Optimus Gen 2 do? A: Tesla's Gen 2 video demonstrates walking, balance movements, squats and egg handling. Tesla separately reports autonomous battery handling and later simple factory tasks by Optimus units, but it does not identify the exact hardware revision or publish task-success, intervention or uptime data. Q: Is Tesla Optimus autonomous? A: Tesla has reported a narrow autonomous factory task, but it has not published enough data to call Gen 2 a generally autonomous worker. Some demonstrations were edited or not labelled autonomous; the We, Robot interactions were later disclosed as human-assisted. Tesla has not published customer-grade reliability metrics. Q: What is the difference between Optimus Gen 2, version 2.5 and Gen 3? A: Gen 2 is the hardware generation Tesla unveiled in December 2023. Tesla later named version 2.5 as a separate development revision but has not published a full 2.5 datasheet. Gen 3 is a major upgrade that Tesla calls its first Optimus design intended for mass production. Q: Does Optimus work in Tesla factories, and is it in mass production? A: Tesla reports autonomous battery handling and later simple factory tasks, without publishing an audited fleet, exact revision, uptime or productivity. It also reports factory construction and line installation, but designed capacity is not actual output or customer delivery. Q: Should buyers wait for Tesla Optimus? A: Wait if Optimus is the product you specifically want, because there is no public order to evaluate. Buyers with a current deployment need should compare vendors that publish an order or quote route, the exact configuration, support terms and task-level acceptance criteria. Can you buy Tesla Optimus Gen 2, and how much does it cost? No public Tesla Optimus Gen 2 sales route exists as of July 23, 2026, and Tesla has not announced an MSRP. Tesla's public Optimus material explains the program and recruits engineers; it does not provide a robot configurator, reservation, deposit or delivery contract. That finding does not rule out a private arrangement that Tesla has not disclosed, but there is no public way for a home or business buyer to order one. The price most often attached to Optimus—US$20,000 to $30,000—is Elon Musk's long-term expectation for a robot produced at scale. He gave that range at Tesla's We, Robot event in October 2024. It is not a Gen 2 list price, a preorder price or a quote for the first commercial configuration. The most recent public-sales statement we found is conditional too. At the World Economic Forum on January 22, 2026, Musk said Tesla might sell humanoid robots to the public by the end of 2027, once the robot reached high reliability, safety and functionality. Tesla has not committed to that date or opened orders. This Tesla Optimus Gen 2 review is source-led rather than hands-on. The useful test is whether a claim is a published specification, a demonstration, a company-reported deployment or a future target. Those categories are kept separate below. The Tesla Optimus price record: target, cost and MSRP are different numbers Tesla has discussed Optimus pricing in three different ways. The distinction is more useful than any single headline number. The missing commercial details are substantial. Tesla has not published the included hand configuration, onboard compute, software entitlement, warranty, service response, spare-parts policy, training, installation, freight, tax treatment or eligible delivery markets. A factory robot may also be sold, leased or bundled with software and support; the reviewed Tesla sources do not say which model it will use. So a serious budget should not enter US$20,000 or US$30,000 as an Optimus quote. Treat the range as a long-term price expectation and keep the deployment line blank until Tesla publishes a configuration and terms. Our humanoid robot cost guide explains why purchase price is only one part of a working robot's cost. Is Tesla Optimus available to preorder? No public preorder, reservation or waitlist appears on the official Tesla properties reviewed on July 23, 2026. Tesla's AI and Robotics page is a program and recruitment page. Tesla Shop sells a 1:10-scale Gen 2 action figure; it is merchandise, not a robot reservation. The practical verification rule is simple: an Optimus order should originate on an official Tesla domain and disclose the seller, configuration, payment terms and refund policy. A third-party page using Tesla branding or collecting an “Optimus deposit” is not evidence of an authorized sale. Because Tesla has not disclosed a public program, this article cannot rule out private enterprise discussions; it can say that buyers using public channels have no official route. What is Tesla Optimus? Optimus, originally announced as Tesla Bot, is Tesla's program to build a general-purpose bipedal humanoid. Tesla describes the intended product as autonomous and aimed at unsafe, repetitive or boring work. Its current AI page says reaching that goal requires software for balance, navigation, perception and interaction with the physical world. That description is an end state, not a current capability sheet. Gen 2 is the hardware Tesla unveiled in December 2023. Version 2.5 is a later development revision that Tesla has named but not documented with a full specification. Gen 3 is the separate design Tesla says is intended for mass production. Treating those labels as interchangeable is the source of many bad Optimus price and specification claims. Tesla Bot to Gen 2: the chronology that explains today's robot The chronology matters because “Optimus specs” copied around the web often combine the 2021 concept, the 2022 design unit and the 2023 Gen 2 reveal as though they described one machine. They do not. Detailed Gen 3 reveal and production tracking belongs in our Tesla Optimus Gen 3 tracker; this page keeps the record for Gen 2, its later 2.5 revision and the buyer questions attached to them. What specifications has Tesla actually published for Optimus Gen 2? Tesla's official December 2023 Gen 2 video contains specification title cards, but it mostly reports changes relative to the previous robot. It does not provide a complete product datasheet. The video shows walking, squats, balance movements and handling an egg. It does not label those sequences autonomous, publish repeatability data or invite independent testing. The correct wording is “demonstrated in Tesla's video,” not “proven autonomous capability.” Which widely quoted Optimus specifications belong to older robots? Tesla has published absolute numbers for earlier concepts and designs. The mistake is carrying them into a Gen 2 table without their date and hardware boundary. As a result, no checked Tesla source gives a final absolute Gen 2 height, total mass, battery capacity, payload or top speed. “Not published for Gen 2” is more accurate than filling those fields with numbers from another generation. Gen 2 vs Optimus 2.5 vs Gen 3 When Tesla does not identify a public robot's hardware revision, describe it only as Optimus and state that the exact revision is unknown. Do not declare every appearance “2.5,” or attach announced Gen 3 features to an older body. Exact revision matters because a hand demonstration does not automatically establish the configuration of factory units. What can Optimus Gen 2 actually do? The most useful capability ledger records the evidence type, not just the task. An edited Tesla video, a company claim of autonomous work and an independently observed customer deployment are different grades of evidence. This record shows real hardware progress. It also shows why the phrase “Optimus can do X” needs a qualifier. Some actions were physical demonstrations, some involved disclosed human assistance, and the strongest factory-autonomy statement comes from Tesla's own shareholder material. None is a customer acceptance test. Is Tesla Optimus autonomous? Optimus has company-claimed autonomous behavior, but Tesla has not published enough data to call Gen 2 a generally autonomous worker. The Q2 2024 shareholder deck is the clearest official claim: Optimus began performing tasks autonomously in a Tesla facility, with battery handling identified. Musk later said some robots were doing simple factory tasks. Those statements are meaningful, but narrow. Tesla does not publish task-completion rate, human interventions per hour, mean time between failures, uptime, cycle time against a human baseline or performance on a previously unseen task. It also does not identify the hardware revision behind every factory statement. Teleoperation is not evidence that the hardware is fake; it can collect training data and demonstrate mechanics. The problem begins when a remotely assisted event is read as autonomous product performance. A buyer should therefore ask four questions of every Optimus clip: Was the behavior autonomous? Was it continuous or edited? Was the task repeated under variation? Were success and intervention rates published? The compute story needs the same generation boundary as the physical specifications. At AI Day 2, Tesla described the 2022 design unit as using a central computer in its torso, adapting Tesla's Autopilot hardware and software, and supporting wireless and audio communications. Those are 2022 design details, not a confirmed Gen 2 or version 2.5 compute specification. Tesla's later material describes vision and neural-network work and reports autonomous battery handling, but it does not identify the exact onboard compute revision or publish a Gen 2 autonomy benchmark. What does Optimus do inside Tesla factories? Tesla has not published an audited fleet count in the sources used here. Conflicting supply-chain estimates do not solve that gap, so this review does not turn them into a false range. “Some robots doing simple tasks,” company-reported, is the latest factory-use statement in the supplied evidence. Anything more precise needs a dated source and a definition of whether “built,” “deployed” and “doing useful work” mean the same thing. Is Optimus in mass production? Not on the evidence Tesla had published by July 23, 2026. Tesla's July 22 Q2 update says first-generation Optimus lines are still being installed at Fremont, with production "anticipated later this year," and the initial builds are designated for Tesla's internal Optimus Academy, not customers or factory output. The Q2 earnings call itself contained no Gen 2 content at all; Gen 3 is the model Tesla is preparing to produce. It described a Fremont line designed for one million robots per year and a later Texas line designed for long-term annual capacity of ten million. Those are design capacities, not installed capacity, actual annual output or delivery guidance. A line can be under construction, installed, commissioned, ramping and producing saleable units at very different times. Tesla's own update classifies the locations as construction and says the lines are being installed. The buyer-facing release gate is stricter than a factory headline. Before Optimus is commercially available, Tesla still needs to publish a customer configuration, price or quote process, order terms, supported tasks, service model and delivery schedule. A production line does not answer those questions. How reliable have Tesla's Optimus targets been? Forward-looking statements belong in the article, but they should be calibrated against later evidence rather than repeated as dates. This is not a prediction that Tesla will fail. It is a rule for handling uncertain claims: state the speaker, date and condition, then wait for measurable evidence. What can buyers do while Optimus is unavailable? None of these alternatives is a drop-in equivalent to Tesla's intended product. This is a route-and-fit check, not a full market ranking. Price disclosure is only the first filter. Buyers still need the exact configuration, support terms and task-level acceptance criteria. Our humanoid robot buying guide covers the procurement field in detail. Should you wait for Tesla Optimus? Home buyers Wait if Optimus is specifically what you want. There is no public order to join and no confirmed home configuration to evaluate. Do not pay a third party to “hold” a Tesla robot. If evaluating another home-robot offer, independently verify its current order status, warranty, remote-assistance terms, privacy terms and delivery geography. Enterprise buyers Do not base a 2026 deployment deadline on Optimus. Start with the task, environment, throughput and safety case, then compare vendors that will provide a quote and acceptance test. Keep Tesla on the watchlist and reopen the evaluation when it publishes a commercial program, service terms and performance data. Developers and researchers Optimus is not a public development platform. There is no documented SDK, developer configuration or hardware order. A purchasable research robot may be less ambitious but far more useful if it provides the interfaces, spare parts and support your project needs. Investors and observers Separate the scale thesis from reported execution. The Fremont and Texas capacity plans are significant, but designed capacity should not be booked as output. The evidence to watch is repeatable factory work, intervention and uptime data, saleable-unit yield, customer terms and actual deliveries. What would change this review? Five disclosures would materially change the buyer verdict: An official Tesla order or enterprise-quote route with eligible markets.A named customer configuration with absolute specifications and included hand, compute and software.Warranty, service, parts, training, privacy and remote-assistance terms.Task-level success, intervention, uptime and cycle-time data from continuous operation.Reported production and customer deliveries, clearly separated from designed factory capacity. Until then, new videos can show engineering progress but cannot answer the commercial questions on their own. Bottom line Tesla Optimus Gen 2 is a development robot physically demonstrated in Tesla-produced footage, with documented features and improvements involving actuators, sensing, hands, feet, neck, walking and control. Tesla has also reported autonomous battery handling and later simple factory tasks by Optimus units, although it has not published the exact revision, scale or reliability data behind those statements. It is not available through a public Tesla order route. Tesla has no public Optimus MSRP, reservation or delivery contract. US$20,000–$30,000 is a long-term at-scale expectation; sub-US$20,000 has been used both as an early price guess and, later, as a production-cost target conditional on more than one million units a year. Those are not interchangeable. The best way to follow the program is to keep the generations and evidence types separate. Use the December 2023 release for Gen 2's official hardware claims, treat version 2.5 as incompletely documented, leave unrevealed Gen 3 details to Tesla's future disclosures, and judge availability only when a real configuration and sales route appear. The RoboZaps Gen 2 database record keeps the underlying specification gaps visible instead of filling them with older concept numbers. --- # This Week in Humanoid Robots: March 16-22, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-march-16-22-2026 Author: Radica Maneva Published: 2026-03-22T00:00:00.000Z Category: news The race to mass-produce humanoid robots accelerated dramatically this week as Unitree filed for a $610 million IPO, UBTech locked in Siemens to hit 10,000 units, and Tesla confirmed Optimus Gen 3 production starts this summer. China's dominance is solidifying—and the West is scrambling to respond. Here's everything that mattered in humanoid robotics from March 16-22, 2026. Unitree Files $610M Shanghai IPO—The First Major Humanoid Robotics IPO In what could be a defining moment for the industry, Unitree Robotics filed an IPO application with the Shanghai Stock Exchange on Friday, seeking to raise 4.2 billion yuan (approximately $610 million). This marks the first major public offering from a company primarily focused on humanoid robots. The Hangzhou-based startup, best known for its Unitree G1 and Unitree H1 humanoid robots, has grown explosively since launching its first quadruped robots in 2016. The company went viral during the 2025 Spring Festival Gala when its robots performed autonomous martial arts, reaching 679 million viewers. We covered that breakthrough moment extensively. Why it matters: A successful IPO would validate humanoid robotics as a standalone investment category and unlock significant capital for R&D and manufacturing scale. It also sets a benchmark valuation for competitors like Figure AI, 1X, and others still in private funding rounds. Watch this one closely. UBTech + Siemens: 10,000 Robots by Year-End Chinese robotics firm UBTech signed a strategic cooperation agreement with Siemens Digital Industries Software on March 16, targeting annual production capacity of 10,000 humanoid robots in 2026. The partnership integrates UBTech's full-stack robotics capabilities with Siemens' expertise in industrial digitalization. Siemens will provide its Xcelerator platform for end-to-end digital workflow—from design and simulation to manufacturing management. "Mass production of tens of thousands of units has become a goal that we must achieve," said UBTech CEO Zhou Jian, noting the company has seen a surge in orders this year. UBTech's Walker S2 industrial humanoid is already being deployed in manufacturing, with total orders exceeding 1.4 billion yuan in 2025. The company also recently signed an Airbus deal to expand into aviation manufacturing. Why it matters: This is the clearest signal yet that humanoid robots are transitioning from prototype demonstrations to industrial-scale production. When a company partners with Siemens specifically for manufacturing scale, they're serious. UBTech is positioning itself as China's answer to Tesla's Optimus program. Tesla Optimus Gen 3: Production Starts This Summer At the 2026 Abundance Summit, Elon Musk provided the most detailed Optimus roadmap yet. Key revelations: Optimus Gen 3 is in final development stages and will begin initial production this summerHigh-volume production targeted for summer 202710 million square feet of factory space dedicated to robot manufacturingTesla will not lay off humans—robots will boost per-worker output insteadThe Gen 3 will feature 50 actuators per hand, representing a significant dexterity improvement Musk also predicted that recursive AI self-improvement (models training better models without human intervention) will be fully automated by end of this year or early next. His vision: robots will create massive deflation and eventually "Universal High Income," where governments issue money because AI-driven output exceeds human desire. Tesla's Fremont factory is being repositioned to mass-produce Optimus, with the Gen 3 expected to be the "most advanced robot in the world." For deeper analysis on Tesla's robotics ambitions, see our coverage of Tesla's Model S sunset. Why it matters: Tesla's scale advantage could make it the most significant player in humanoid robotics—if they can execute. Summer production start puts them behind UBTech and Unitree in timeline, but Tesla's manufacturing capabilities and capital reserves make them a formidable long-term competitor. NVIDIA Partners with European Chipmakers for Humanoid Hardware On March 16, NVIDIA announced partnerships with three major European semiconductor firms—Infineon, NXP, and STMicroelectronics—to develop and sell hardware specifically for humanoid robots. The partnerships position NVIDIA as the central AI brain while European chipmakers provide the motion control, power management, and sensor integration hardware that humanoid robots require. This vertical integration strategy mirrors NVIDIA's success in data centers. Why it matters: Hardware has been a bottleneck for humanoid robot commercialization. Better chips mean better battery efficiency, more responsive motor control, and lower costs. NVIDIA is betting that owning the full stack—from AI training to edge inference to motor control—will make them indispensable to every humanoid manufacturer. Boao Forum 2026: China Sets Global Standards The Boao Forum for Asia (BFA) Annual Conference 2026 kicks off next week in Hainan Province, with humanoid robots as both service providers and a central discussion topic. Ahead of the forum, China's Ministry of Industry and Information Technology released its first standard system covering the entire industrial chain and lifecycle of humanoid robots and embodied AI. Market forecasts presented at pre-forum briefings suggest China's embodied AI market could reach: 400 billion yuan (~$55 billion) by 20301 trillion yuan (~$138 billion) by 2035 The Guardian's in-depth feature this week, "Inside China's Robotics Revolution," revealed that China now accounts for over half of the world's new factory robot installations annually, with some estimates suggesting Chinese companies control 90% of humanoid robot shipments. Why it matters: Standards define markets. If China sets the global standards for humanoid robots—just as they did for solar panels and batteries—Western manufacturers will be forced to comply or face market exclusion. This is industrial policy playing out in real time. The Contrarian View: Mark Cuban Says Humanoids Will Fail Not everyone is bullish. At a TBPN live stream Thursday, Mark Cuban predicted humanoid robots will "fail miserably" within 5-10 years. "Everybody's making this push for humanoid robots. I think they might have a 5-year lifespan, and then they'll fail miserably. Maybe 10," Cuban said. His alternative vision: robots and spaces will be co-designed. Instead of making robots that fit human environments, we'll redesign environments to optimize for robot efficiency. "The robots aren't going to be full-form humanoids. They're going to be whatever the optimal shape is." He pointed to Amazon's 1 million+ warehouse robots—none of which are humanoid—as evidence that form follows function. Our take: Cuban isn't entirely wrong about industrial applications. Purpose-built robots often outperform general-purpose humanoids in controlled environments. But his argument misses the core value proposition: humanoids are designed for unstructured environments built for humans—homes, construction sites, disaster zones. You can't redesign a grandmother's house or a hospital corridor. The real question is whether the economics work, and this week's news suggests major players believe they will. Market Pulse Robotics funding 2026 YTD: On pace for $20B+, with February's Figure AI momentum and SkildAI's $1.4B round leading the chargeHumanoid market projection: $35 billion by 2030 (Business Research Company)China market share: 90%+ of humanoid robot shipments (Commonwealth Union)Production targets: UBTech (10K), Tesla (ramping), Unitree (post-IPO scale) What to Watch Next Week Boao Forum sessions on humanoid robotics—expect policy announcements and new partnershipsNVIDIA GTC 2026 aftermath—more hardware partnerships likely in the pipelineUnitree IPO pricing details—valuation benchmarks for the entire sector The humanoid robot industry just entered its industrial era. Mass production isn't coming—it's here. See our complete guide to the best humanoid robots to understand who's leading and why, or explore the full marketplace to see what's actually available today. --- # Tesla Optimus vs Unitree G1: Full 2026 Comparison (Price, Specs, Availability) URL: https://blog.robozaps.com/b/tesla-optimus-vs-unitree-g1 Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T11:19:37.980Z Last fact-checked: 2026-07-12T00:00:00.000Z Category: comparisons TL;DR: The Unitree G1 is buyable now from $13,500; Tesla Optimus is not for sale, with a $20,000-$30,000 target and production starting slowly from late July or August 2026. The G1 is a compact 35 kg research platform with up to 43 DoF; Optimus is a full-size 57 kg factory humanoid backed by Tesla's FSD AI stack. Key takeaways: - Unitree G1 is available now starting at ~$13,500. Tesla Optimus is not yet commercially available — Musk has estimated a $20,000–$30,000 price point when it eventually ships. - Tesla Optimus Gen 2 stands 5'8" (173 cm) and weighs 57 kg. The Unitree G1 is significantly smaller at 127 cm and 35 kg, designed as a compact research and light-duty platform. - Unitree G1 features up to 43 degrees of freedom with force-controlled three-finger hands. Tesla Optimus Gen 3 hands (2024) have 22 degrees of freedom. - Tesla Optimus is slightly faster at ~8 km/h (5 mph) versus Unitree G1 at 7.2 km/h (4.5 mph). - Tesla Optimus leverages Tesla's FSD AI stack and massive data pipeline. Unitree G1 uses reinforcement learning and imitation learning but lacks a comparable AI ecosystem. - Security concerns have emerged around Unitree products, with backdoor vulnerabilities discovered in 2025 and a U.S. congressional investigation into alleged PLA ties. - Tesla's 2025 internal-deployment target (Musk's projected "thousands") was missed; on the Q4 2025 / Q1 2026 earnings calls Musk admitted Optimus was "not in use in our factories in a material way." Fremont volume production is planned for late July/August 2026 autonomously. External sales may begin in 2026–2027. Musk announced in March 2025 that an Optimus would be sent to Mars in 2026 on a SpaceX Starship — but SpaceX shifted focus to lunar missions on February 9, 2026, ending the 2026 Mars-Optimus plan. FAQ: Q: Can you buy the Tesla Optimus robot right now? A: No. As of July 2026, Tesla Optimus is not available for commercial purchase. Tesla is using Optimus internally in its own factories and has discussed potential external sales beginning in 2026, but no firm date or ordering process has been announced. Elon Musk estimated a price of $20,000–$30,000 at the October 2024 "We, Robot" event. Q: How much does the Unitree G1 cost? A: The Unitree G1 starts at approximately $13,500 for the base model. The advanced version — which includes force-controlled three-finger hands, additional sensors, and enhanced software — costs more. You can check current pricing on Robozaps. Optional accessories, international shipping, and taxes may add to the total cost. Q: Is the Unitree G1 safe to use? What about the security issues? A: In 2025, security researchers discovered multiple vulnerabilities in Unitree products, including potential backdoors, unauthorized data collection, and a wormable Bluetooth vulnerability. Unitree patched the initial backdoor issue but had not addressed all concerns as of late 2025. If you're deploying a G1 in a sensitive environment, isolate it from your network and monitor for firmware updates. Government and defense buyers should exercise particular caution given the ongoing U.S. congressional investigation. Q: What is the difference between Unitree G1 and Unitree H1? A: The Unitree H1 is a full-size humanoid (180 cm tall) that sold for over $90,000. The G1 is a smaller (127 cm), lighter (35 kg), and vastly cheaper ($13,500) humanoid designed for research and light-duty applications. The G1 retains many of the H1's capabilities in a more accessible package. In 2025, Unitree also introduced the H2, a next-generation full-size humanoid. Q: How does Tesla Optimus compare to Boston Dynamics Atlas? A: Boston Dynamics' new electric Atlas (2024) is arguably the most physically capable humanoid robot in terms of athleticism and dynamic movement. Unlike Tesla Optimus, Atlas is designed by a company with decades of bipedal robotics experience. However, Atlas is not being marketed as a consumer or mass-market product — it targets commercial and industrial applications at a premium price point. For a detailed breakdown, see our Tesla Optimus vs Boston Dynamics Atlas comparison. Q: Will Tesla Optimus be sent to Mars? A: In March 2025, Elon Musk announced plans to send an Optimus robot to Mars in 2026 aboard a SpaceX Starship rocket. If this happens, Optimus would be the first humanoid robot deployed on another planet. However, given Tesla's history of ambitious timeline announcements that often face delays, many observers remain skeptical about the 2026 target. Q: Which robot is better for university research? A: The Unitree G1 is the clear choice for university research as of July 2026. It's available, affordable ($13,500), compact, and designed as an open development platform. Tesla Optimus isn't available for purchase and isn't designed as an open research platform. Many universities worldwide already use Unitree robots for locomotion research, AI development, and human-robot interaction studies. Q: What new Unitree robots were released in 2025? A: Unitree expanded its product line in 2025 with three new robots: the R1, A2, and H2. The H2 is particularly notable as a next-generation full-size humanoid that succeeds the H1. Unitree also continued development of its popular Go2 quadruped platform and showcased its full lineup at CES 2025. ‍ Related: Tesla Optimus Gen 2 Review · Unitree G1 Review: Pros, Cons and How it Compares · Figure 01 vs Unitree G1 · Tesla Optimus Alternatives Ready to buy? Browse humanoid robots for sale on Robozaps. The Tesla Optimus and Unitree G1 represent two fundamentally different approaches to building humanoid robots. One is backed by the world's most valuable automaker and its ambitious CEO; the other comes from a scrappy Chinese robotics startup that has already shipped thousands of units. As of July 2026, the gap between these two robots — in availability, pricing, and real-world deployment — tells an important story about the state of humanoid robotics. This comprehensive comparison covers every meaningful difference between the Tesla Optimus and Unitree G1: specifications, pricing, AI capabilities, real-world applications, and the latest 2025–2026 developments. Whether you're a researcher, business buyer, or robotics enthusiast, this guide will help you understand exactly where each robot stands today. Tesla Optimus vs Unitree G1: Complete Specs Comparison Before diving into the details, here's a side-by-side comparison of every key specification: Design Philosophy: Two Very Different Robots Tesla Optimus and Unitree G1 are built around fundamentally different design philosophies, and understanding this context is crucial for any comparison. Tesla Optimus: Full-Size Factory Worker Tesla Optimus is designed to be a full-size humanoid that can operate in human-scale environments — factories, warehouses, and eventually homes. At 173 cm tall, it's roughly the height of an average adult human. Tesla's approach is to leverage its existing AI stack from Full Self-Driving (FSD) and its massive compute infrastructure (Dojo supercomputer) to create a general-purpose robot that can learn new tasks through neural network training. The Gen 2 Optimus (December 2023) featured a slimmer, lighter build with improved actuators and Tesla-designed hands with 11 degrees of freedom. By late 2024, Tesla showcased Gen 3 hands with 22 degrees of freedom — a significant leap in dexterity that brings Optimus closer to human-like manipulation capabilities. However, it's critical to note: as of July 2026, Tesla Optimus is not commercially available. Tesla has been deploying small numbers internally in its factories, and Musk announced plans for 1,000+ units to be used at Tesla facilities. But you cannot buy one. Unitree G1: Compact, Affordable, Available Now Unitree G1 takes a completely different approach. At just 127 cm tall and 35 kg, it's a compact humanoid designed primarily for research institutions, universities, and developers who need a capable, affordable bipedal platform. It sacrifices the full human-scale form factor for portability, safety, and — critically — a price point that actually makes purchase feasible. Released in August 2024 at a starting price of $16,000, the G1 was a breakthrough moment for accessible humanoid robotics. Its predecessor, the Unitree H1, sold for over $90,000. The G1 brought the cost down by more than 80% while retaining impressive capabilities: up to 43 degrees of freedom, force-controlled three-finger hands (optional), and advanced locomotion that can handle uneven terrain. The G1 is an evolution of the H1 platform, optimized for mass production. Unitree has deep experience in volume manufacturing from its popular Go1 and Go2 quadruped robots, and that expertise shows in the G1's polished, production-ready design. Limb Articulation and Movement The number of degrees of freedom (DOF) directly determines how fluidly and precisely a humanoid robot can move. This is one area where the Unitree G1 has a clear numerical advantage. The Unitree G1, in its advanced configuration, features 43 joint motors across its body. This gives it exceptional range of motion — it can perform complex movements like crossing uneven terrain, stepping over 40 cm obstacles, and recovering from physical disturbances (being pushed or kicked). Unitree has demonstrated the G1 performing backflips (on the H1 platform) and navigating messy woodpiles, showcasing impressive agility for a humanoid. Tesla Optimus Gen 2 has approximately 28 degrees of freedom in its body, with the newer Gen 3 hands adding 22 DOF for a total hand dexterity that's approaching the G1's three-finger system. Tesla's approach focuses on smooth, deliberate movements optimized for factory tasks — picking up items, sorting, carrying loads — rather than the dynamic acrobatics Unitree demonstrates. In practical terms: the G1 is more agile; the Optimus is designed for heavier-duty, sustained work. Hand Capabilities and Dexterity Hands are arguably the most important component of a humanoid robot for real-world usefulness. Both companies have invested heavily here. The Unitree G1's optional three-finger hands are force-controlled, meaning they can sense and adjust grip pressure in real time. Unitree has demonstrated the G1 performing tasks like cracking walnuts, opening bottles, flipping food in a pan, and handling delicate objects. The three-finger design provides a good balance between simplicity and capability — fewer actuators than a five-finger hand, but enough articulation for most manipulation tasks. Tesla's Gen 3 Optimus hands (showcased in late 2024) represent a major upgrade: 22 degrees of freedom with improved tactile sensing. Tesla demonstrated egg handling without breaking — a classic robotics benchmark. Earlier demos showed wire soldering and shirt folding (though the latter was later revealed to involve teleoperation). The five-finger design theoretically allows for more human-like grasping, but it also introduces greater mechanical complexity. The key difference: Tesla's hands are more ambitious in design, but the Optimus isn't available for purchase. Unitree's G1 hands are less complex but are shipping in a real product today. Speed and Locomotion Speed matters for practical applications, especially in dynamic environments like warehouses, emergency response, or outdoor settings. Tesla Optimus is slightly faster at ~8 km/h versus the G1's 7.2 km/h max speed. It also handles rough terrain more capably, thanks to its lighter weight and lower center of gravity. The G1's compact size (127 cm) and 35 kg weight make it inherently more stable and easier to balance dynamically. Tesla Optimus, being taller and heavier, is designed for a more deliberate pace suited to factory environments where speed is less important than precision and payload capacity. AI and Learning Capabilities This is where Tesla's vision diverges most dramatically from Unitree's current offering. Tesla's AI Advantage Tesla's single greatest asset in robotics is its AI infrastructure. The Optimus robot is controlled by the same neural network architecture that powers Tesla's Full Self-Driving (FSD) system — the same end-to-end AI that processes camera inputs and makes driving decisions in millions of Tesla vehicles on the road. This gives Tesla several unique advantages: Massive training data: Tesla's fleet of vehicles generates enormous amounts of real-world visual and spatial data that can be repurposed for robot training.Dojo supercomputer: Tesla's custom AI training hardware can process robot training workloads at scale.End-to-end neural networks: Rather than relying on hand-coded behaviors, Tesla aims to train Optimus to learn tasks through observation, similar to how FSD learns driving.Simulation-to-real transfer: Tesla has invested heavily in simulated environments for training Optimus before deploying in the real world. The promise is enormous: a robot that can learn almost any task by watching humans do it. But as of July 2026, this vision is largely unrealized. Many of Tesla's impressive Optimus demos have relied on teleoperation (human remote control) rather than autonomous behavior — a fact competitors and critics have repeatedly highlighted. Unitree G1's AI Approach Unitree G1 uses a combination of reinforcement learning (RL) and imitation learning for locomotion and basic task execution. The robot ships with a capable sensor suite including 3D LiDAR, depth cameras, and IMU sensors that enable spatial awareness and navigation. The G1 is designed as an open platform for researchers and developers. This means users can program custom behaviors, integrate their own AI models, and build on top of Unitree's base capabilities. For the academic and research market, this flexibility is extremely valuable — it's why you see Unitree robots in universities worldwide. However, Unitree doesn't have anything approaching Tesla's AI infrastructure. The G1's learning capabilities are more conventional: effective for specific tasks, but lacking the grand vision of a general-purpose AI-driven humanoid. Pricing and Availability (July 2026 Update) This is the single most important practical difference between these two robots. Tesla Optimus Pricing History Elon Musk's price estimates for Optimus have shifted over time: 2022: Musk initially suggested $20,000–$25,000, positioning it as "less than half the price of a car."October 2024 ("We, Robot" event): Musk revised the estimate upward to $20,000–$30,000.2025: Tesla planned to deploy 1,000+ units internally but made no commercial sales. In January 2026, Musk admitted no Optimus robots were yet doing "useful work" autonomously. The head of the Optimus program, Milan Kovac, resigned in June 2025 and was replaced by Ashok Elluswamy from the Autopilot team.July 2026: No commercial availability. External sales may begin in 2026 or later. Unitree G1 Pricing The Unitree G1 has been available since August 2024 at a starting price of approximately $13,500. The base model includes the standard body and basic hands. The advanced version with force-controlled three-finger hands, additional sensors, and enhanced software commands a premium but remains far below what Tesla Optimus is expected to cost. For context, the G1's predecessor — the Unitree H1 — sold for over $90,000, making the G1's pricing a genuine industry disruption. You can check current G1 pricing and availability on Robozaps. Latest Developments (2025–2026) Both robots have seen significant developments since their initial launches. Here's what's happened through early 2026: Tesla Optimus: 2025–2026 Timeline March 2025: Elon Musk announced that an Optimus robot would be sent to Mars in 2026 aboard a SpaceX Starship. That timeline was effectively ended when SpaceX shifted focus to lunar missions on February 9, 2026.May 2025: Tesla shared footage of Optimus performing factory tasks including battery cell sorting and parts transport.June 2025: Milan Kovac, who had led the Optimus program since its inception in 2022, left Tesla. He was replaced by Ashok Elluswamy, the head of Tesla's Autopilot team — a signal that Tesla is more tightly integrating its vehicle AI and robotics efforts.Late 2024: Gen 3 Optimus hands with 22 degrees of freedom were demonstrated, a major improvement in manipulation capability.October 2024: At the "We, Robot" event, Optimus robots interacted with crowds but critics noted heavy reliance on teleoperation. Unitree G1 and Unitree Robotics: 2025–2026 Timeline January 2025: Unitree showcased the G1, H1, Go2, and B2-W at CES 2025 in Las Vegas, signaling expanded international ambitions.2025 product line expansion: Unitree introduced three new robots in 2025: the R1, A2, and H2 — expanding beyond the G1 into additional form factors.April 2025: Security researchers discovered backdoors in Unitree products allowing potential remote access to devices on the same network. Unitree claimed it was an unintentional vulnerability and patched it.May 2025: The U.S. House Select Committee on Strategic Competition requested investigations into Unitree's alleged connections to the People's Liberation Army and China's military-civil fusion programs. Unitree has denied selling products to the PLA.July 2025: Unitree began IPO tutoring with CITIC Securities, signaling preparations for a potential Hong Kong stock listing.September 2025: Security researchers published findings that the G1 collects and transmits multi-modal sensor data without notifying operators. A separate wormable BLE vulnerability was discovered that could allow nearby attackers to gain full control of Unitree robots.July 2025: Unitree won a WIPO Global Award for information and communications technology. Security and Privacy Concerns The security issues around Unitree products deserve their own section, as they may significantly impact purchasing decisions — particularly for government, military, or sensitive commercial buyers. In 2025, multiple security research teams found serious vulnerabilities in Unitree robots: Backdoor access (April 2025): Researchers alleged Unitree had embedded backdoors that allowed the company to remotely access devices — and potentially other devices on the same network. Unitree claimed this was unintentional.Data exfiltration (September 2025): The G1 humanoid was found to collect and transmit multi-modal sensor data without operator notification — a serious concern for any deployment in sensitive environments.Wormable BLE vulnerability (September 2025): A critical vulnerability allowed attackers in physical proximity to gain full control of Unitree Go2, B2, G1, and H1 robots via Bluetooth Low Energy. Compromised robots could then infect other nearby robots. As of late September 2025, Unitree had not commented on this vulnerability. Additionally, the U.S. Congress has raised concerns about Unitree's potential ties to the Chinese military, requesting federal investigations in May 2025. Tesla Optimus, being manufactured in the United States and not yet commercially available, has not faced comparable security scrutiny — though any networked robot will eventually face similar concerns. Real-World Applications Where Tesla Optimus Fits Tesla's vision for Optimus is ambitious: Factory automation: The primary near-term use case. Optimus is being tested in Tesla's own factories for tasks like battery cell sorting, parts transport, and assembly assistance.Warehouse logistics: Picking, packing, and organizing goods.Home assistance: Musk's long-term vision includes household tasks — cooking, cleaning, caregiving — though this is years away from practical reality.Space exploration: Musk has announced plans to send Optimus to Mars, which would make it the first humanoid robot deployed on another planet (if it happens). Where Unitree G1 Fits The G1's current market is more defined and practical: Academic research: Universities use the G1 for locomotion research, AI development, human-robot interaction studies, and computer vision experiments.Robotics education: The $13,500 price point makes it accessible for educational institutions that previously couldn't afford humanoid platforms.Light commercial tasks: Reception, guided tours, customer assistance, and demonstrations.Developer platform: The G1 serves as a foundation for developers building custom humanoid applications.Entertainment and events: Trade shows, exhibitions, and technology demonstrations. Competitors to Consider The Tesla Optimus and Unitree G1 don't exist in isolation. The humanoid robotics market is rapidly expanding. Here are the key competitors: Boston Dynamics Atlas: The most advanced humanoid robot in terms of athletic capability. The new electric Atlas (2024) is designed for commercial deployment but carries a premium price far above the G1.Figure 02: Figure AI's second-generation humanoid integrates OpenAI models for conversational AI and task planning. Recently secured massive funding and BMW factory deployments.Agility Robotics Digit: Purpose-built for warehouse logistics, Digit is already in pilot deployments at Amazon facilities.Astribot S1: A Chinese humanoid focused on upper-body dexterity and manipulation tasks.Unitree H2: Unitree's own next-generation full-size humanoid, launched in 2025, offering a larger form factor than the G1. For a complete overview, see our guide to Tesla Optimus alternatives and competitors. Which Should You Choose? The honest answer depends on what you need — and when you need it. The Bottom Line Bottom Line (July 2026): The Unitree G1 ($13,500) is the only humanoid robot you can actually buy today. Tesla Optimus ($20K–$30K estimated) remains unavailable for commercial purchase, with no confirmed release date. For immediate deployment needs, G1 wins by default — but evaluate security concerns carefully. As of July 2026, the Tesla Optimus vs Unitree G1 comparison is less about which robot is "better" and more about which robot exists as a purchasable product. The Unitree G1 is real, shipping, and affordable at $13,500. The Tesla Optimus, but it remains an internal prototype with no commercial availability and a history of missed timelines. For researchers, educators, and developers who need a humanoid robot today, the Unitree G1 is the clear practical choice — provided you're comfortable with the security concerns that emerged in 2025. For those waiting for the most capable general-purpose humanoid, Tesla Optimus represents the highest ceiling, but requires patience and a willingness to bet on Elon Musk's timeline promises. The humanoid robotics race is accelerating rapidly. By the end of 2026, we'll likely see commercial availability from multiple players — Tesla, Figure AI, Agility Robotics, and Unitree's own H2. The best time to enter this market is now; the best robot to buy depends entirely on your timeline and use case. --- # Unitree G1 Price 2026: From $13,500, EDU Cost & Full Specs URL: https://blog.robozaps.com/b/unitree-g1-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T10:54:43.687Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: The Unitree G1 costs $13,500 direct from Unitree (backordered) or $17,990 from US dealers, but that base model cannot be programmed and has no advertised upgrade path. Developers need the G1 EDU: quote-only from Unitree, listed at $43,900-$73,900 by US resellers. Strong open-source locomotion platform; no autonomous manipulation out of the box, and its security record deserves a hard look. Key takeaways: - $13,500 buys the base G1 direct from Unitree (backordered); US dealers list it at $17,990. The $16,000 price still quoted across the web is the 2024 launch price. - The base G1 cannot be programmed, with no advertised upgrade path to EDU. Real development needs the G1 EDU: quote-only from Unitree, with US resellers listing $43,900-$73,900. - The viral flips are real RL motion policies and the boxing is human-teleoperated; nothing a G1 does out of the box is autonomous manipulation. - Security deserves scrutiny: the UniPwn root exploit lists the G1, researchers dispute Unitree's patch claims, and the DoD added Unitree to its 1260H list in June 2026 (a procurement restriction, not a purchase ban). - It is still the default research humanoid: one of the broadest open SDK ecosystems, NVIDIA and LeRobot support, and real supervised deployments from Haneda Airport to a North Carolina sheriff's office. FAQ: Q: How much does the Unitree G1 cost in 2026? A: $13,500 direct from Unitree's store (backordered, excluding tax and shipping) or $17,990 from US dealers for the base model. The developer-grade G1 EDU is quote-only from Unitree; US resellers list its configurations at $43,900 to $73,900. The $16,000 figure still circulating is the 2024 launch price. Q: Is the $13,500 Unitree G1 programmable? A: No. Unitree's own store states the base G1 "does not support secondary development" - no SDK, no code deployment - and neither Unitree nor its dealers advertise any upgrade path to EDU. If you plan to write any code, your real entry price is a $43,900 EDU U1, or Unitree's smaller R1 EDU at $10,500. Q: How much is the Unitree G1 EDU? A: Unitree publishes no EDU price - it is quote-only. US resellers list eight configurations from $43,900 (U1 Standard) to $73,900 (U6 with tactile five-finger hands). No configuration sells at the commonly quoted flat "$40,000". Q: Is the Unitree G1 real, or are the videos fake? A: The robot and its acrobatics are real: the side-flip and kip-up are reinforcement-learning policies Unitree ships as motion releases. But the boxing matches are human-teleoperated, the viral clip of a G1 attacking Indonesian office workers was a staged production, and the "robot soldiers" video was AI-generated. Q: What can the Unitree G1 do out of the box? A: Walk, balance, recover from pushes and perform built-in motions from the remote and app - and on the base model, that is everything. A university lab's published deployment report calls flat-ground locomotion the platform's strongest feature, stairs unreliable, and manipulation immature: useful grasping requires an EDU plus custom teleoperation or trained policies. Q: How long does the Unitree G1 battery last? A: Unitree's official figure is about 2 hours from the 9,000 mAh quick-release battery. Reviewers report materially less under continuous manipulation or dynamic locomotion. Spare batteries are $700-$800. Q: How tall is the Unitree G1? A: 1,320 mm (1.32 m) standing, about 35 kg, per Unitree's official spec table. The 127 cm figure on several well-known sites conflicts with that; the origin of the discrepancy is unclear. Q: Is the Unitree G1 banned in the US? A: No. The US Department of Defense added Unitree to its Section 1260H "Chinese military companies" list in June 2026, which restricts DoD procurement from June 30, 2026. It is not a sanction, an import ban, or a bar on private purchase - though organizations in defense supply chains should review procurement eligibility. Q: Is the Unitree G1 secure? A: Its record deserves scrutiny. The UniPwn disclosure (September 2025) demonstrated root access over Bluetooth using hardcoded keys, lists the G1 as affected, and is wormable; patch status is disputed between Unitree and researchers. Separate research reported telemetry sent to China-based servers every five minutes; we could not locate a public Unitree statement addressing that finding specifically. Labs typically isolate the robot on a filtered network. Q: Should I buy a Unitree G1 or an R1? A: If you want to program a humanoid on a budget, the R1 EDU at $10,500 now undercuts the unprogrammable $13,500 base G1. The G1's advantages are size, payload, the LiDAR-grade sensor suite, dexterous-hand options and one of the broadest open-source ecosystems of any research humanoid - but they only unlock at EDU prices. Q: Where can I buy a Unitree G1 in the US? A: Direct from shop.unitree.com ($13,500, backordered, ships from Hangzhou with $300-$1,200 freight and import duties on top) or from authorized US dealers such as RoboStore ($17,990 with a US service center; stock fluctuates). Unitree's shop accepts no returns on any item. Q: What warranty does the Unitree G1 come with? A: 8 months on the base G1 and 18 months on the G1 EDU, per Unitree's official spec table. How much is a Unitree G1, and can you buy one? The Unitree G1 costs $13,500 direct from Unitree's own store, and yes, anyone can order one — but that base model cannot be programmed, and as of July 20, 2026 it ships on backorder. The $16,000 you'll see across most of the internet is the 2024 launch price; Unitree cut it in late 2025. US dealers list the same base robot at $17,990 with domestic support, though availability fluctuates week to week. The version researchers actually build on, the G1 EDU, has no public price from Unitree at all: it's quote-only, and US resellers list its configurations from $43,900 to $73,900. We verified every number on this page against Unitree's live store data and dealer listings on July 20, 2026. On the legal question the G1 is in the clear: its FCC authorization was granted on 3 March 2025, well before the July 2026 Covered List entry, so it is grandfathered. We track that for every model in humanoid robots legal in the US. The full price ladder, and how it got here The G1 launched in May 2024 at $16,000, a number The Robot Report carried at the time and much of the web never updated. Archived store snapshots show the price held through October 2025 and dropped to $13,500 by late December 2025. That cut has barely propagated: IEEE's robot guide, TechRadar's URL and the top Reddit thread all still say $16,000. On Unitree's own store, the base G1 is the only G1 you can put in a cart. There is no G1 EDU listing, no "G1 Air" (that's the R1 Air, a different robot), and a caution for anyone scraping prices: Unitree's shop uses $100,000 placeholder prices on quote-only items. Several aggregator sites have reprinted those placeholders as real prices. The EDU ladder is defined by resellers rather than Unitree. RoboStore, which describes itself as Unitree's official North American partner, lists eight EDU configurations (prices checked July 20, 2026; availability fluctuates): Two honesty notes on that table. Tier names above U6 are reseller inventions and differ between dealers; RobotShop's archived catalog runs to a "U10" that costs less than RoboStore's U6. Treat U-numbers as SKUs, not official grades. And the widely repeated "$43,900–$73,900" range for the EDU is real: it matches these live listings exactly, though Unitree itself will only say "contact sales." 2026 also brought new family members: the G1-D, a dual-arm data-acquisition platform in fixed-base and wheeled forms (reseller-priced roughly $38,800–$66,300), a $48,900 G1-Comp sports-competition build, and an unpriced G1 boxing edition from the "Iron Fist King" events. The $13,500 catch: what the base G1 can't do This is the single most important thing to know before buying, and the source of most G1 buyer's remorse. Unitree's store page says it plainly: the base G1 "does not support secondary development." No SDK. No code deployment. Unitree's own Amazon listing goes as far as putting it in the product title: "Unitree G1 Humanoid Robot（No Secondary Development）". It walks, balances, recovers from shoves and performs its built-in motions from the remote and the app, and that is the entire product. Neither Unitree nor its dealers advertise any upgrade path from base to EDU: treat the two tiers as separate products. The two tiers also carry different warranties: 8 months on the base robot, 18 months on the EDU, per Unitree's official spec table. So the real buying decision is binary. If you want a walking, flipping demonstration robot for events, lobbies or content, the $13,500/$17,990 base G1 is exactly that. If you intend to write a single line of code, your actual entry price is a $43,900 U1. The hobbyist alternative is Unitree's smaller R1, whose EDU version is shop-listed at $10,500. The Reddit thread that ranks near this page asking why "everyone says the G1 is only 16K" is this exact confusion, twice over. What owning one actually costs On import costs, honesty requires saying the picture is moving. The Supreme Court struck down the 2025 IEEPA tariffs in February 2026 (USTR); the replacement 10% import surcharge sunsets on July 24, 2026, and longstanding Section 301 tariffs on many Chinese robotics lines persist. A direct import from Hangzhou may owe additional duties on top of the sticker, depending on classification, exclusions and whichever regime is in effect at entry. The practical dodge: buy US dealer stock, where importation is already priced into the $17,990. Unitree G1 specs Specs below are Unitree's official figures from unitree.com/g1; flags note where circulating numbers disagree. Full verified record: Unitree G1 in the RoboZaps robot database. What a G1 actually does out of the box The most useful account of day-one G1 reality is a Slovak university lab's published deployment report on two G1 EDU U6 units (May 2026). Its verdict: the G1 is "not a turnkey humanoid worker, but a valuable research platform." Flat-ground walking and balancing are "immediately among the strongest parts of the platform": a delivered G1 walks reliably on day one from the handheld controller. Stairs, by contrast, were "unreliable" and "high-risk" despite the demo footage, the depth camera's field of view can't see the faces of people standing in front of it, the speaker is too weak for noisy rooms, and both units took about two months from order to delivery. The famous acrobatics deserve their own decoder. The side-flip and kip-up are real: reinforcement-learning policies trained in simulation on motion-capture data and shipped as versioned motion releases (Unitree labeled them V11.0 and V12.0 in March 2025), and firmware 1.4.5 added a Training Mode that records and replays guided motions. What they are not is decision-making. The boxing matches, Hangzhou's "Iron Fist King" in May 2025 and the CES 2026 bouts, were driven in real time by human operator teams; researchers quoted on the events call them "robot theater," with the robot's contribution being balance and recovery, not fight decisions. And manipulation is the honest gap: the lab report describes grasping tools as "currently rudimentary," with useful manipulation requiring custom teleoperation rigs or trained policies. Nothing a G1 does out of the box is autonomous manipulation. One more layer a 2026 reader arrives with: fakes. The 100-million-view July 2026 clip of a G1 kung-fu-kicking office workers in Indonesia was a staged entertainment production, and the February 2026 "robot soldiers live-fire" video was AI-generated outright: its original upload carried an AI disclaimer that resharers stripped. The real robot is impressive enough; a good chunk of what goes viral under its name isn't the real robot. Software: what you can build on it The reason the G1 became the default research humanoid is the stack around it. Unitree publishes open repositories for low-level control (unitree_sdk2, with Python bindings and ROS 2 support), reinforcement learning (unitree_rl_gym with pretrained G1 policies, plus Isaac Lab and MuJoCo variants), official MuJoCo models, VR teleoperation (xr_teleoperate), and a Hugging Face LeRobot integration. NVIDIA's GR00T foundation-model work supports the G1 explicitly, and in January 2026 Unitree open-sourced UnifoLM-VLA-0, a vision-language-action model for G1 manipulation, weights included. The catch remains the tier gate: all of this assumes an EDU. The Slovak lab also notes the openness is hybrid: the researcher-facing compute is open while a vendor-managed control side stays closed, and an open issue on Unitree's own GitHub reports a shoulder motor overheating and the arm losing response after a short period of dexterous-hand use. Security, telemetry and the Pentagon list No honest 2026 G1 review can skip this section, and most competitors do. UniPwn. In September 2025, researchers Andreas Makris and Kevin Finisterre published a working exploit of Unitree's Bluetooth Wi-Fi-provisioning service: hardcoded encryption keys and an auth check that amounts to encrypting the string "unitree," letting an attacker in BLE range inject commands and gain root. The affected list includes the G1 (with the Go2, H1 and B2), it spawned four CVEs including CVE-2025-35027, and it is wormable: an infected robot can compromise nearby Unitree robots. Patch status is disputed: Unitree said in late September 2025 that it had completed a majority of fixes and would roll out updates, while IEEE Spectrum's coverage reported researcher skepticism, and the NVD's June 2026 update still lists G1 firmware through 1.4.4 as affected. Public patch status remains disputed. A separate February 2026 disclosure of unauthenticated root RCE over the DDS bus on the Go2 (patched that month) suggests the attack surface is a family trait. Telemetry. Security firm Alias Robotics reported that G1 units transmit telemetry to China-based servers every five minutes, characterizing the data as including sensor streams; their analysis is on arXiv. We could not locate a public Unitree statement addressing the five-minute telemetry finding specifically, and no independent audit has settled it. The Slovak lab's practical stance is the useful takeaway for buyers: treat the G1 as a managed endpoint: dedicated VLAN, egress filtering, Bluetooth off when unused, packet capture after every firmware update. The 1260H listing. In June 2026 the US Department of Defense added Unitree to its Section 1260H list of "Chinese military companies" (official PDF). This is not a sanction, an import ban, or a bar on private purchase. What it does: prohibit direct DoD contracting with listed entities from June 30, 2026, extending to their products via subcontractors in 2027 (WilmerHale's alert is the clean explainer). If your organization sells into US defense supply chains, a G1 purchase now carries procurement-eligibility questions. If you're a lab or a private company, it changes nothing legally. But pair it with the telemetry section above and draw your own risk conclusions. Where G1s are actually working The pattern across that ledger: real pilots exist and are growing, humans supervise everything, and the robot's public failures mostly trace to third-party operation. That's a fair picture of where a $13,500–$74,000 humanoid genuinely is in 2026. Unitree in 2026: the company behind the robot Unitree is set to become the A-share market's first humanoid-robot stock: China's regulator approved its Shanghai STAR Market IPO on July 2, 2026 (Caixin), a ~$618 million raise at an implied valuation near $5.9 billion. As of July 22 it had not yet priced or begun trading. The prospectus is the first real look inside: revenue of roughly ¥1.7 billion in 2025, up from ¥392 million in 2024; statutory net profit of ¥278 million (roughly ¥590–600 million on the adjusted basis Unitree prefers); and more than 5,500 humanoid robots shipped in 2025, which the company says made it the world's largest humanoid maker by units. Two details matter for G1 buyers. Average humanoid selling price collapsed from ¥593,400 in 2023 to ¥167,600 in 2025: the G1-and-R1-driven strategy of winning on volume and price. And the whole operation is startlingly small: 480 employees as of late 2025. Founder Wang Xingxing, 36, who built his first quadruped as a master's student and was the youngest entrepreneur at Xi Jinping's February 2025 private-sector symposium, controls about 69% of voting rights. One credibility note: figures Wang cited at Davos in mid-2025 (2024 revenue "over ¥1 billion," five straight profitable years) don't match his own prospectus, which shows ¥392 million and a 2023 loss. Cite the prospectus, not the stage. Misinformation audit: ten G1 claims, checked G1 vs the field in 2026 The G1's competitive position shifted in the last year, mostly because of Unitree itself. The honest summary: at the base tier, Unitree's own R1 EDU at $10,500 now undercuts the unprogrammable $13,500 G1 for anyone whose goal is writing code on a budget. The G1's moat is at the EDU level: the sensor suite, the payload, the hand options, and one of the broadest open-source ecosystems of any commercially available humanoid. For the wider market view, see our evidence-ranked humanoid guide and cost guide, or the Optimus comparison if you're wondering about the robot that isn't for sale. Who should buy a G1, and who shouldn't Buy the base G1 if you need a reliable walking humanoid for events, demos, education showcases or content, you're clear-eyed that it won't run your code as shipped, and $13,500–$17,990 is acceptable for that. Buy a G1 EDU if you're a lab, university or R&D team that wants the most-supported research humanoid on the market and can budget $43,900+ plus hands. Look elsewhere if you want programmability on a hobbyist budget (R1 EDU, Booster K1), if you expect autonomous usefulness around the house (nothing at any of these prices delivers that in 2026; our buyable-humanoids guide is the reality check), or if your organization touches US defense contracting and the 1260H listing makes procurement review a problem. What to watch next Three dated things will change this page. Unitree's IPO is expected to price and trade in the coming weeks: the A-share market's first humanoid pure-play listing, with a stated plan to scale toward 20,000 humanoids a year. The US import-surcharge regime hits its statutory sunset on July 24, 2026, which will move landed cost for direct imports one way or the other. And the UniPwn dispute is unresolved: either Unitree ships a verifiable architecture fix, or the security case against connected deployment keeps hardening. We re-verify prices and availability on this page against Unitree's live store data; the numbers above carry their check date. The bottom line The Unitree G1 remains the most consequential humanoid robot you can actually buy: the machine that dropped a real, walking, backflipping biped under $15,000 and built one of the broadest open research ecosystems in the field. It is also a product with a trap at the bottom of its price ladder, a security record that deserves scrutiny before any connected deployment, and a manufacturer whose own R1 now undercuts it for budget developers. Know which of those three sentences applies to you, and the G1 is either the obvious buy in its class — or a $13,500 lesson in reading the spec sheet first. --- # Tesla Optimus vs Agility Robotics Digit: Full 2026 Comparison URL: https://blog.robozaps.com/b/tesla-optimus-vs-agility-robotics-digit Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T11:20:36.645Z Category: comparisons TL;DR: Agility's Digit is the proven one: the first commercially deployed humanoid, working in Amazon and GXO warehouses on RaaS contracts. Tesla Optimus targets ~$30,000 at mass production but hasn't started volume production, with only a reported few hundred built. Digit wins on commercial traction today; Optimus on AI ambition and scale potential. Key takeaways: - Tesla Optimus targets mass production at ~$30,000 per unit, leveraging Tesla's AI stack and vertical integration — but only a reported few hundred units exist, none in material factory use per Musk, with production slated from late 2026. - Agility Digit is the first commercially deployed humanoid robot, already operating in Amazon and GXO warehouses. Its RoboFab facility can produce 10,000+ units per year. - Optimus excels in AI-driven general-purpose autonomy; Digit excels in purpose-built logistics manipulation with proven commercial traction. - Both robots are expected to see significant production scaling in 2026, making this the pivotal year for the humanoid robotics industry. FAQ: Q: How much does Tesla Optimus cost vs Agility Digit? A: Tesla aims to sell Optimus at approximately $30,000 per unit (revised from an earlier $20,000 target). Agility Robotics uses a Robot-as-a-Service (RaaS) leasing model for Digit, with estimated monthly costs of $2,000-4,000 per robot (estimated). The total cost of ownership depends on deployment duration and utilization rates. Q: Is Agility Digit available for purchase in 2026? A: Yes — Agility Robotics has been commercially deploying Digit since 2023-2024 through its RaaS program. Enterprise customers like Amazon and GXO Logistics are actively using Digit fleets. Contact Agility Robotics directly for fleet pricing and availability. Q: When will Tesla Optimus be available for sale to the public? A: Tesla plans to begin external sales of Optimus in 2026, initially to manufacturers and enterprise customers. Consumer availability is further out — likely 2027 or later, pending regulatory approvals and further autonomous capability development. Q: Can Tesla Optimus or Digit work in my warehouse? A: Digit is already proven in warehouse environments and is the better near-term option for logistics automation. Optimus may become an option for warehouse work in 2026-2027 once Tesla begins external sales, but it currently lacks Digit's warehouse-specific optimizations and track record. Q: Which robot is more advanced — Tesla Optimus or Agility Digit? A: "Advanced" depends on the metric. Optimus has superior hand dexterity (22 DOF per hand vs specialized grippers), faster walking speed (8 km/h vs 5.5 km/h), and more sophisticated AI. Digit has superior commercial readiness, energy-efficient locomotion, and proven reliability in real-world deployments. For general-purpose capability, Optimus leads. For deployable, reliable logistics automation, Digit leads. Q: What happened to the original $20,000 price for Tesla Optimus? A: At the October 2024 "We, Robot" event, Elon Musk revised the Optimus price estimate upward to approximately $30,000. The original $20,000 figure was an aspirational target from early development stages. The revised price reflects more realistic manufacturing costs, though it remains aggressive compared to other humanoid robots on the market. Q: Will humanoid robots like Optimus and Digit replace warehouse workers? A: In the near term (2026-2028), these robots will supplement rather than replace human workers. They handle specific repetitive tasks — tote movement, sorting, basic inspection — freeing human workers for higher-value activities. Full replacement of warehouse labor is unlikely before 2030+, and will depend on advances in autonomous decision-making, dexterity, and cost reduction. Related: Tesla Optimus Gen 2 Review · Agility Robotics Digit Review · Tesla Optimus vs Boston Dynamics Atlas · Tesla Optimus Alternatives & Competitors Ready to buy? Browse humanoid robots for sale on Robozaps. Bottom Line: For enterprise buyers in 2026, Agility Digit wins — it's commercially deployed at Amazon and GXO right now with proven ROI. Tesla Optimus wins on raw capability (AI, dexterity, speed) but external sales haven't begun yet. Digit for today; Optimus for 2027+. Last updated: March 20, 2026  |  11 min read Tesla Optimus and Agility Robotics Digit represent two fundamentally different visions for the future of humanoid robots. As of March 2026, both have made dramatic progress — Tesla has built a reported few hundred Optimus units, which Musk says are not in material factory use, with production slated to start at Fremont from late July or August 2026, while Agility Robotics is shipping Digit fleets to Amazon, GXO Logistics, and other enterprise customers from its dedicated RoboFab manufacturing facility. This comprehensive comparison covers specs, design philosophy, real-world performance, pricing, and which robot is best positioned to dominate the next decade of commercial robotics. Tesla Optimus vs Agility Digit: Full Specs Comparison The following table breaks down the core technical specifications of both robots side by side: Design and Engineering: Two Philosophies Tesla Optimus: Vertical Integration and AI-First Design Tesla's approach to Optimus mirrors its vehicle strategy: vertical integration, aggressive cost reduction, and AI as the core differentiator. Every major component — actuators, battery pack, control electronics, and the neural network inference chip — is designed in-house. This gives Tesla complete control over the bill of materials and, critically, the ability to push over-the-air software updates to deployed units. The evolution has been rapid. The 2022 prototype ("Bumble C") could barely walk. The Gen 2 (December 2023) showcased smooth bipedal locomotion, dancing, and egg handling with 11-DOF hands. By 2024, the Gen 3 hands doubled to 22 degrees of freedom, enabling significantly more dexterous manipulation — picking up thin objects, turning knobs, and handling irregular shapes. Structurally, Optimus uses a combination of cast aluminum and engineering plastics, with Tesla-designed rotary and linear actuators featuring integrated strain wave gear reduction. The robot's 2.3 kWh battery pack (similar chemistry to Tesla vehicle cells) provides roughly 4-6 hours of moderate activity on a single charge. Agility Digit: Purpose-Built for Logistics Agility Robotics took a fundamentally different path. Digit evolved from Cassie, a bipedal research robot developed at Oregon State University in 2017. Rather than trying to build a general-purpose humanoid, Agility focused on making a robot that could do one category of tasks exceptionally well: material handling in warehouses and distribution centers. Digit's design reflects this focus. Its legs use a unique four-bar linkage mechanism that provides exceptional energy efficiency during walking — critical for all-day warehouse shifts. The robot's "head" houses a dense sensor array (LiDAR, stereo depth cameras, IMUs) designed for reliable perception in cluttered, dynamic warehouse environments where lighting conditions vary and obstacles appear unpredictably. Rather than trying to replicate human hand dexterity, Digit uses purpose-built end effectors optimized for grasping standard warehouse containers (totes, boxes, bins). This trade-off means less versatility but significantly higher reliability for the target use case. AI and Software: The Real Battleground Tesla's FSD-Derived Intelligence Tesla's biggest competitive advantage is its AI infrastructure. Optimus runs on neural networks derived from Tesla's Full Self-Driving (FSD) system, trained on data from millions of Tesla vehicles. The Dojo supercomputer — purpose-built for training AI models on video data — gives Tesla enormous compute capacity for improving Optimus's perception and decision-making. In 2025, Tesla demonstrated Optimus performing end-to-end autonomous tasks in its Fremont and Giga Texas factories: sorting battery cells, moving parts between stations, and performing quality inspections. The key breakthrough was the shift from teleoperation (which critics had noted in earlier demos) to genuine autonomous task execution using learned behaviors. Ashok Elluswamy, who took over the Optimus program in June 2025 after Milan Kovac's departure, has accelerated the integration of FSD perception models into Optimus. This leadership change aligned the robot and vehicle AI teams more closely, enabling faster transfer learning between domains. Agility Arc: Enterprise-Grade Robot Management Agility's software platform, Agility Arc, takes an enterprise-first approach. Rather than pushing the boundaries of general AI, Arc provides: Facility Mapping: Automated 3D mapping of warehouse environmentsWorkflow Definition: Drag-and-drop task programming without codingFleet Management: Coordinated control of multiple Digit unitsSafety Monitoring: Real-time compliance with ISO 13482 (robots for personal care) and warehouse safety standardsAnalytics Dashboard: Performance metrics, uptime tracking, and predictive maintenance This pragmatic approach means Digit may lack Optimus's raw AI potential, but it offers something enterprise customers value more: predictable, reliable, auditable automation that integrates with existing warehouse management systems (WMS). Real-World Deployment: Who's Actually Working? Optimus in Tesla Factories (2025-2026) Tesla has never published Optimus production numbers; reporting from The Information put cumulative builds at a few hundred, and on the January 2026 earnings call Musk said Optimus was not in use in Tesla factories in a material way. These robots are being trained on: Battery cell sorting and inspectionParts transfer between production stationsQuality control visual inspectionsLight assembly tasks Musk stated at the October 2024 "We, Robot" event that external sales could begin in 2026, with initial customers likely being other manufacturers and logistics companies. The estimated retail price of ~$30,000 (revised upward from the earlier $20,000 target) would still make Optimus significantly cheaper than most industrial robotic systems. In a bold move, Musk announced in March 2025 that an Optimus unit would be sent to Mars by end of 2026 aboard a SpaceX Starship — a largely symbolic demonstration, but one that highlights Tesla's ambitions beyond terrestrial applications. Digit in Amazon and Beyond (2023-2026) Agility Robotics has a significant head start in commercial deployment. Key milestones include: September 2023: Opened RoboFab in Salem, Oregon — the world's first dedicated humanoid robot manufacturing facility (6,500 m², capacity 10,000+ units/year)2023-2024: Amazon began testing Digit in its fulfillment centers, initially for moving empty totes2024-2025: Expanded deployment to GXO Logistics and other enterprise customers2025: Agility shifted to a Robot-as-a-Service (RaaS) model, where customers lease Digit fleets rather than purchasing outright The RaaS model is strategically important. Rather than asking warehouse operators to make a large capital expenditure on unproven technology, Agility lets them pay per-robot-per-month, lowering the adoption barrier. This approach mirrors how companies like Locus Robotics scaled their AMR (autonomous mobile robot) fleets. Mobility, Balance, and Physical Performance Both robots use bipedal locomotion, but their approaches differ significantly: Digit's advantage in energy-efficient locomotion is meaningful for warehouse applications, where robots need to operate for full 8-12 hour shifts. Optimus's faster top speed matters less in structured indoor environments but could be relevant for future outdoor applications. Hand Dexterity and Manipulation This is where the robots diverge most sharply. Tesla has invested heavily in making Optimus's hands as dexterous as possible: Gen 2 (2023): 11 degrees of freedom per hand, metallic cable-driven system with 4 fingers + thumbGen 3 (2024-2025): 22 degrees of freedom per hand, enabling precision grasps, tool use, and delicate object handling Digit, by contrast, uses specialized grippers that are optimized for grasping warehouse totes and boxes. These grippers are less versatile than Optimus's hands but achieve higher grasp success rates (>99%) on their target objects. For warehouse operators, this reliability trade-off is overwhelmingly positive. Industry Applications: Where Each Robot Fits Manufacturing Optimus has the edge in manufacturing flexibility. Its dexterous hands and general-purpose AI make it adaptable to diverse assembly, inspection, and material handling tasks. Tesla is proving this internally — Optimus units in Tesla factories handle multiple task types and can be retrained for new tasks via software updates. Digit's manufacturing applications focus on material transport — moving parts, bins, and finished goods between stations. It's less suited for fine assembly work but excels at the repetitive movement tasks that account for a significant portion of manufacturing labor. Logistics and Warehousing This is Digit's home turf. With proven deployments at Amazon and GXO, Digit has demonstrated: Autonomous tote movement and sortingPackage handling and transportIntegration with conveyor systems and shelvingMulti-shift operation with fleet coordination Optimus could enter this space, but Tesla has not yet announced logistics-specific partnerships. The robot's higher payload (20 kg vs 16 kg) and faster speed are advantages, but Digit's established customer relationships and proven reliability give it a multi-year head start. Retail and eCommerce Agility has explicitly targeted retail and eCommerce fulfillment as a growth market. Digit's ability to navigate cluttered backrooms, pick items from shelves, and transport them to packing stations aligns well with the labor challenges facing retail warehouses, especially during peak seasons. Future Applications: Healthcare, Construction, Space Optimus's general-purpose design gives it the broader long-term application potential. Tesla has hinted at home assistance, eldercare, and construction applications. The Mars mission (planned 2026) could open entirely new markets in space exploration and off-world construction. However, these remain aspirational — no concrete deployment timelines exist outside of manufacturing. Pricing and Business Model Comparison If Tesla hits the $30,000 price point, the economics become compelling compared to human labor costs. A $30,000 robot operating 16 hours/day costs roughly $3-5/hour (amortized over 3 years with maintenance), compared to $15-25/hour for human warehouse workers in the US. Digit's RaaS model has different economics — higher monthly cost but lower risk, no capital expenditure, and guaranteed uptime via service agreements. Investment and Market Context The humanoid robot market is projected to reach $38 billion by 2035 (Goldman Sachs, 2024). Both Tesla and Agility are positioned to capture significant market share, but their paths differ: Tesla: Backed by Tesla's $800B+ market cap, Dojo supercomputer investment ($1B+), and FSD data moat. Musk has suggested Optimus could eventually become more valuable than Tesla's vehicle business.Agility Robotics: Has raised over $641M in total funding ($400M Series C in March 2025), with investors including Amazon, SoftBank, and DCVC. The RoboFab facility represents a significant bet on near-term production scale. Other competitors in the humanoid robot space include Boston Dynamics Atlas, Figure AI's Figure 02, Sanctuary AI Phoenix, and Unitree H1/G1. The market is rapidly expanding, but as of March 2026, only Agility has achieved meaningful commercial deployment of a humanoid robot. Safety and Human-Robot Interaction Both robots incorporate multiple safety systems, but their approaches reflect their different deployment contexts: Tesla Optimus: Relies heavily on its vision-based perception system (derived from FSD) to detect and avoid humans. Safety features include force-limited actuators, emergency stop buttons, and speed reduction zones when humans are detected nearby. However, experts including Rodney Brooks (iRobot co-founder) have expressed skepticism about Optimus's readiness for unstructured household environments, calling the vision of humanoid robots as "catchall assistants" as "pure fantasy thinking" due to coordination challenges. Agility Digit: Designed from the ground up for shared workspace operation. Digit complies with ISO 13482 safety standards and includes proximity sensors, force/torque feedback, and automatic speed reduction when approaching humans. Its warehouse deployments at Amazon demonstrate practical coexistence with human workers — a critical validation that Optimus has yet to achieve outside Tesla's own facilities. Development Timeline: Key Milestones Category-by-Category Breakdown: Optimus vs Digit To determine which humanoid robot wins across key performance areas, we break down the comparison into seven categories. Each category declares a winner based on current specifications, deployment data, and real-world performance as of March 2026. 1. Agility & Mobility Winner: Tesla Optimus Tesla Optimus achieves a top walking speed of 8 km/h (5 mph), compared to Digit's 5.5 km/h (3.4 mph). This 45% speed advantage translates to faster task completion in environments where travel time matters — factory floors, distribution centers, and multi-zone operations. Optimus also demonstrated running capability (early tests showed bursts up to 8 km/h sustained), while Digit is optimized for efficient walking only. That said, Digit's four-bar linkage leg design offers superior energy efficiency per step — critical for all-day warehouse shifts where battery life matters more than peak speed. For pure locomotion speed and versatility, Optimus wins. For energy-efficient marathon walking, Digit holds an edge. 2. Dexterity & Manipulation Winner: Tesla Optimus This category isn't close. Tesla Optimus Gen 3 features 22 degrees of freedom per hand — each finger has multiple articulating joints, enabling precision grasps, tool use, and manipulation of irregular objects. Demonstrations include handling raw eggs, turning small knobs, and picking up thin metal sheets. Agility Digit uses purpose-built grippers optimized for warehouse totes and boxes. While these achieve >99% grasp success rates on standard containers, they cannot handle diverse object types, perform fine assembly, or use tools. For general-purpose manipulation, Optimus is substantially more capable. Digit's grippers are a deliberate trade-off: less versatility for higher reliability on target tasks. 3. AI & Software Winner: Tesla Optimus Tesla's AI infrastructure is unmatched in the humanoid space. Optimus runs on neural networks derived from Full Self-Driving (FSD), trained on data from millions of Tesla vehicles. The Dojo supercomputer provides massive compute capacity for training — purpose-built for video data at a scale no other humanoid company can match. By contrast, Agility's Arc software platform is enterprise-grade but narrower in scope. Arc excels at facility mapping, workflow definition, fleet management, and safety monitoring for warehouse operations. It's reliable, auditable, and integrates with existing warehouse management systems — but it's not designed for general-purpose AI or learning new tasks autonomously. For raw AI capability and learning potential, Optimus leads decisively. For enterprise-ready, predictable automation software, Agility Arc is more mature. 4. Sensors & Perception Winner: Agility Robotics Digit Digit's sensor suite is purpose-built for warehouse environments: LiDAR for precise mapping, stereo depth cameras for obstacle detection, IMUs for balance, and force/torque sensors for grasp feedback. This combination delivers reliable perception in cluttered, dynamic environments where lighting varies and obstacles appear unpredictably. Tesla Optimus relies primarily on cameras (derived from Tesla Vision), with force/torque sensors and IMUs. While Tesla's neural network approach can extract impressive information from camera data alone, the lack of LiDAR limits precision in certain scenarios — especially depth estimation in low-light or high-glare conditions common in industrial settings. For industrial/warehouse perception reliability, Digit's multi-modal sensor fusion wins. For consumer environments with good lighting, Tesla's camera-based approach may prove sufficient. 5. Price & Value Winner: Tie (different models suit different buyers) Tesla targets a ~$30,000 purchase price for Optimus at mass production — dramatically cheaper than any comparable humanoid if achieved. Amortized over 3-5 years, this translates to roughly $500-800/month, competitive with warehouse labor costs in many markets. Agility uses a Robot-as-a-Service (RaaS) model with estimated monthly costs of $2,000-4,000 per Digit. Higher monthly cost, but zero upfront capital expenditure, included maintenance, and guaranteed uptime — lower risk for enterprises testing humanoid automation. For buyers who want to own and can afford upfront investment, Optimus offers better long-term economics. For enterprise operators who prefer OpEx over CapEx with lower adoption risk, Digit's RaaS model wins. Different value propositions for different buyer profiles. 6. Commercial Readiness Winner: Agility Robotics Digit This isn't a debate. Digit is commercially deployed — RIGHT NOW. Amazon uses Digit fleets in fulfillment centers. GXO Logistics operates Digit in distribution hubs. Multiple enterprise customers are paying for Digit's services through Agility's RaaS program. The RoboFab facility in Salem, Oregon, can produce 10,000+ units per year. Tesla Optimus is not deployed in material numbers anywhere: Tesla has built a reported few hundred units for internal iteration, and external sales haven't begun. Elon Musk has indicated external availability in 2026, with consumer sales likely in 2027 or later. Optimus is proven internally but unproven with external customers. For a humanoid robot you can deploy TODAY, Digit is the only real option. For a humanoid robot you can deploy in 2-3 years, Optimus becomes a contender. 7. Real-World Deployment Track Record Winner: Agility Robotics Digit Digit has years of real-world warehouse deployment data. Agility can demonstrate uptime metrics, task completion rates, safety records, and ROI data from actual customer deployments. This track record is invaluable for enterprise sales — procurement teams can see proven results, not just demos. Tesla Optimus deployments are internal and limited to controlled factory environments. While early Optimus units have been shown doing tasks like battery sorting and parts transfer, Musk himself said in January 2026 they were not in material factory use, there's no third-party verification or customer testimonials. External deployments will need to prove Optimus can perform reliably outside Tesla's own walls. For proven, auditable deployment history, Digit wins decisively. Optimus needs 12-24 months of external deployments to compete on this metric. Which Should You Choose? Choose Agility Robotics Digit if you: Need proven humanoid automation TODAY — Digit is commercially available now with active deployments at Amazon, GXO, and other enterprise customers. No waiting for product launches or external sales.Operate warehouses or logistics centers — Digit is purpose-built for tote movement, package handling, and material transport. It excels at exactly these tasks.Prefer Robot-as-a-Service (RaaS) — Zero upfront CapEx, included maintenance, guaranteed uptime. Lower risk for testing humanoid automation before committing to ownership.Value reliability over versatility — Digit's specialized grippers achieve >99% grasp success on warehouse containers. It does fewer things but does them with proven reliability.Need enterprise-grade software integration — Agility Arc integrates with existing WMS systems, provides fleet management, and meets ISO 13482 safety compliance. Choose Tesla Optimus if you: Need general-purpose manipulation — Optimus's 22 DOF hands can handle diverse objects, use tools, and adapt to new tasks. Essential for manufacturing environments with varied workloads.Want lower long-term total cost of ownership — At ~$30,000 purchase price, Optimus costs $500-800/month amortized vs Digit's $2,000-4,000/month RaaS. If you can wait and want to own, the economics favor Optimus.Value AI learning and software updates — Tesla's FSD-derived neural networks and Dojo training infrastructure mean Optimus will likely improve faster via OTA updates than purpose-built alternatives.Are planning for 2026+ deployment — If your timeline allows waiting for external availability (mid-to-late 2026), Optimus becomes a viable option with potentially better specs and price.Want a single robot for multiple task types — Optimus's general-purpose design suits environments where the robot needs to assembly, inspect, sort, AND transport — not just one of these. Final Verdict Overall Winner: Agility Robotics Digit — for buyers who need humanoid automation in 2026. Digit wins on the metric that matters most for commercial adoption: it's available, proven, and deployed. While Tesla Optimus leads on raw capability (AI, dexterity, speed), those advantages don't matter if you can't actually buy and deploy the robot. Agility's head start in commercialization — the RoboFab facility, Amazon partnership, RaaS model, and years of warehouse deployment data — makes Digit the pragmatic choice for enterprises adopting humanoid robots today. However, Tesla Optimus has the higher long-term ceiling. If Tesla hits the $30,000 price point at scale, Optimus becomes the most economically compelling humanoid robot ever produced. Tesla's AI infrastructure and manufacturing scale could enable rapid improvement and cost reduction that Agility cannot match. For buyers planning deployments in 2027 and beyond, Optimus is worth serious consideration. The humanoid robot market is large enough for both to succeed. Digit owns logistics/warehouse; Optimus aims for general-purpose manufacturing and (eventually) consumer markets. This rivalry will drive innovation across the entire industry. --- # Apptronik Apollo Review 2026: Price, Apollo 2 & Real-World Deployments URL: https://blog.robozaps.com/b/apptronik-apollo-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T10:47:54.406Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: Apollo has no public price and cannot be bought today; it runs in enterprise pilots at Mercedes-Benz, GXO and Jabil. The $50,000 figure is a 2023 at-scale target, not a cost. Apollo 2 was unveiled in June 2026 as a data-collection platform, and Apptronik calls the upcoming Apollo 3 its first true commercial product. Key takeaways: - No price, no sales channel: Apollo is acquired only through enterprise pilot agreements, arranged with Apptronik on a quote basis. - The famous $50,000 figure is CEO Jeff Cardenas’s 2023 at-scale target, not a current price; the newer figure of roughly $80,000 a year (CNBC, February 2026) is investor Howard Morgan’s projection for high-volume delivery from 2027, not company pricing. - Every known deployment, at Mercedes-Benz, GXO and Jabil, is a pilot or data-collection program, and no customer has published Apollo performance metrics. - Apollo 2, unveiled June 2026, is a training and data platform offered bipedal or on a wheeled base; Apptronik calls Apollo 3 its first true commercial product, pointed at 2027. - Apptronik has raised roughly $1 billion at a reported $5 billion valuation, with Google DeepMind as both AI partner and investor. FAQ: Q: How much does the Apptronik Apollo cost? A: Apptronik has never published a price. The $50,000 figure is a 2023 at-scale target CEO Jeff Cardenas gave IEEE Spectrum. The newer figure, roughly $80,000 a year for high-volume delivery from 2027, is investor Howard Morgan’s projection reported by CNBC, a yearly service-style figure rather than a sticker price. Today Apollo is available only through quote-based enterprise pilot agreements. Q: Can you buy an Apptronik Apollo? A: No. There is no order page, no reseller network and no consumer version. Companies engage Apptronik directly for pilot programs. Apptronik’s February 2026 release said its next robot, Apollo 3, debuts in 2026, while Cardenas’s June Forbes interview pointed to "within the next year"; no month or quarter has been given. Q: What is Apollo 2? A: Apollo 2 is the generation Apptronik unveiled in June 2026. It comes in bipedal or wheeled-base configurations, runs the Artemis control stack with Fleet Connect fleet software, and is primarily a data-collection and training platform. Apptronik calls it a prototype for scaled pilots. Q: Who is using the Apptronik Apollo today? A: Mercedes-Benz runs Apollo trials at its Berlin-Marienfelde Digital Factory Campus and in Kecskemet, Hungary. GXO Logistics operates Apollo 2 units under a multi-phase R&D program, and Jabil both deploys Apollo in its plants and serves as its manufacturing partner. Google DeepMind is a research partner and investor. All of these are pilots, not paid production deployments. Q: What are the Apptronik Apollo’s specifications? A: Apptronik’s published figures come from the original 2023 Apollo and have not been reconfirmed for Apollo 2: 5 feet 8 inches (1.73 m), 160 lbs (72.6 kg), a 55 lb (25 kg) payload and about 4 hours per hot-swappable battery pack. We could not locate a manufacturer-published walking speed or degrees-of-freedom count; figures circulating online for those have no official source. Q: Is the Apptronik Apollo better than Tesla Optimus? A: They are at different stages. Apollo has named customer pilots at Mercedes-Benz, GXO and Jabil, while Optimus is used internally at Tesla with mass production slated to begin at Fremont in mid-2026. Neither robot is purchasable, and neither has a public price. Q: When will the Apptronik Apollo be widely available? A: Apptronik points to Apollo 3, which it calls its first true commercial product, with a debut expected within a year of mid-2026 and commercial quantities pointed at 2027. The timeline has slipped before: at its 2023 unveiling, Apptronik targeted full commercial release around the end of 2024. Q: What AI does the Apptronik Apollo use? A: The flagship partnership is Google DeepMind’s Gemini Robotics, which has demonstrated cross-embodiment skill transfer and natural-language task execution on Apollo. Apptronik’s own Artemis stack handles perception, planning, control and safety, with Fleet Connect managing fleets. An NVIDIA GR00T integration was announced in 2024, and Apollo’s disclosed compute is NVIDIA Jetson AGX Orin plus Orin NX. Q: What is Apptronik’s connection to NASA? A: Apptronik spun out of the Human Centered Robotics Lab at UT Austin in 2016, and its founding team worked on NASA’s Valkyrie humanoid at Johnson Space Center. NASA SBIR contracts supported parts of Apollo’s development, and NASA has discussed adapting Apollo for astronaut assistance. How much does the Apptronik Apollo cost, and can you buy one? Apptronik has never published a price for Apollo, and you cannot buy one today. Not as a consumer, and not as a business placing a normal order. The robot is deployed through enterprise pilot agreements with partners such as Mercedes-Benz, GXO Logistics and Jabil, arranged directly with Apptronik on a quote basis. The "$50,000" you will see quoted across the web is not a price. It comes from a single August 2023 interview in which CEO Jeff Cardenas told IEEE Spectrum that "long term, a humanoid needs to cost less than US $50,000." That was a goal for what Apollo should cost at scale, stated before a single unit had shipped. Nothing Apptronik has said since turns it into a sticker price. If a site shows $50,000 as Apollo's price, it is repeating a three-year-old aspiration as fact. The short answer in July 2026: there is no price, pilots are the only way in, and Apptronik itself says the commercial product is the next robot, not this one. The detail on all of that is below. One disclosure up front: this is desk research built on the primary sources linked throughout, not a hands-on test; RoboZaps has not operated an Apollo. Apptronik Apollo specifications Apptronik published one set of numbers when Apollo was unveiled in 2023 and has not updated them since, even for the new Apollo 2. These figures come from Apptronik's own materials and are tracked with per-spec verification status in the Apptronik Apollo record in our robot database. Two of those rows deserve a flag. We could not locate a manufacturer-published walking speed or degrees-of-freedom count anywhere in Apptronik's materials as of July 16, 2026. Directory sites print figures like "3 mph" or "71 DoF" anyway; those numbers have no official source, which is why our database marks them unverified rather than repeating them. Where Apptronik has stayed silent, we say so. Apollo's price history: every number, with receipts Because there is no list price, the public record on Apollo money is really a funding and projection trail. Here is each number that matters, in order. August 2023, the $50,000 target. Cardenas's IEEE Spectrum quote at Apollo's unveiling, covered above. A long-term cost goal, framed as comparable to a car. February 2025, $350 million. Apptronik raised a Series A co-led by B Capital and Capital Factory, with Google participating, to scale Apollo production (TechCrunch). A further close in March 2025 took the round to $403 million and added Mercedes-Benz, Japan Post Capital and ARK Invest as investors. Apptronik's later releases state the initial round at $415 million, implying a further $12 million or so closed quietly after March 2025. February 11, 2026, the $520 million extension. A "Series A-X" extension took the Series A past $935 million and total capital to roughly $1 billion. New investors included AT&T Ventures, John Deere and the Qatar Investment Authority, alongside returning backers Google, Mercedes-Benz, B Capital and PEAK6. The valuation was reported at $5 billion by CNBC and Reuters, around $5.3 billion post-money by TechCrunch, and over $5.5 billion by Bloomberg; Apptronik itself only said the round came at three times its 2025 valuation. February 2026, the $80,000 projection. In the same CNBC coverage, B Capital chair Howard Morgan said he expects orders reaching $1 billion worth of robots starting in 2027, with the company "delivering Apollo in high volumes for roughly $80,000 a year, about the price of a luxury car," in CNBC's words. "Think about a factory worker doing three or four shifts, and on any weekend. Eighty-thousand is cheap!" Note the framing before you treat that as a price: it is a yearly, service-style figure, and it is an investor's expectation, not company pricing. Cardenas, in the same article, declined to predict ship dates at all. That is the entire public record. No list price, one dated cost target, and one investor projection. For how this pattern compares across the industry, our humanoid robot cost guide tracks every published price in the category. How you would actually get one Apptronik's site has no order page, no configurator and no reseller network. The only direct route is Apptronik's enterprise contact form, which feeds a conversation about a pilot engagement. The three public engagements, a commercial pilot agreement with Mercedes-Benz in March 2024, a multi-phase R&D program with GXO in June 2024, and a dual manufacturing-and-deployment partnership with Jabil in February 2025, were all announced as programs of exactly that kind, not purchases. Be careful with third-party sites that present Apollo as something you can order. "Shop" listings for Apollo are quote-capture pages, not stock. There is no pre-order program, and no publicly documented outright sale or secondary market exists. On timing, Apptronik has become unusually candid. Cardenas now describes Apollo 2 as "really a prototype for scaled pilots" and says the next robot, Apollo 3, will be the real commercial product: "a mature, early product," in his words to Forbes, expected within the next year. Apptronik's February 2026 release said the new robot debuts in 2026, Cardenas's June Forbes interview pointed to "within the next year," and no month or quarter has been given; coverage and investor expectations cluster around 2027 for commercial quantities. If your plan is to buy rather than pilot, Apollo 3 is the product you are waiting for. What is the Apptronik Apollo? Apollo is a general-purpose industrial humanoid built for warehouse and manufacturing work: moving totes, delivering parts to a line, inspection, kitting. It stands at human scale so it can work in spaces built for people, and it is deliberately approachable, with a friendly face designed around a pair of stereo-camera eyes and status displays in the head and chest. The company behind it has one of the deepest pedigrees in the field. Apptronik spun out of the Human Centered Robotics Lab at the University of Texas at Austin in 2016, founded by Jeff Cardenas and Nick Paine. Its founding team worked on NASA's Valkyrie humanoid at Johnson Space Center, and NASA itself credits that lineage in Apollo's design; the company built more than a dozen robot platforms before this one. NASA SBIR contracts supported parts of Apollo's development, and NASA has discussed adapting Apollo for astronaut-assistance work. Apptronik employed around 300 people as of February 2026 reporting, has raised roughly $1 billion, and in April 2026 hired a commercial bench that says a lot about intent: a chief product officer from Waymo, a services lead from Boston Dynamics, and a software VP from Amazon. It also runs a non-humanoid industrial-automation subsidiary, Elevate Robotics, spun up in June 2025. Apollo 2: what changed in June 2026 Apptronik unveiled Apollo 2 quietly, through a website refresh in late June 2026 rather than a launch event, then confirmed it in the Robot Park announcement on June 30. The company had already been using it internally for over a year. The changes worth knowing. Apollo 2 comes in two configurations, bipedal or a wheeled base, with the wheeled version designed to conform to existing safety standards for industrial mobile robots; the stationary-pedestal option from the original Apollo is no longer marketed. The software stack got names: Artemis, which coordinates perception, planning, controls, safety and task execution, and Fleet Connect for fleet monitoring and orchestration. Apptronik claims its patented actuators run at over 90% energy efficiency and that swappable batteries enable "7x22" operation, meaning 22 hours a day, seven days a week, with short swap breaks; that is a manufacturer design claim, not verified customer uptime (official Apollo 2 page). Just as telling is what Apollo 2 is for. Robot Park is a roughly 90,000-square-foot data factory in Austin where Apollo 2 fleets run tasks, teleoperated and autonomous, to generate training data for AI models, with satellite data-collection sites at Google DeepMind, Mercedes-Benz and GXO. Apollo 2 is a learning platform at heart. Reported Apollo 2 hardware changes go further: a June interview with the company describes a larger battery with task-dependent runtime, a safety-rated pack, revised sensors and camera placement, and a wheeled configuration whose vertical axis reaches about seven feet (A3). The commercial machine, Apollo 3, was announced in the Robot Park release with no specs; its dating is contradictory, covered below. Hardware: what's confirmed and what isn't The numeric spec sheet has not moved since 2023: 5'8", 160 lbs, a 55 lb payload, and four hours per battery pack. The 25 kg payload is a headline number in a category where many rivals publish nothing, though Boston Dynamics advertises a higher lift for Atlas. The battery design is still a genuine differentiator. Packs hot-swap in place of a plug-in charge (unveiling release), and the current Apollo 2 page adds tethering and opportunity charging as supported modes. Actuation is Apptronik's core engineering bet. Apollo uses proprietary electric actuators, a mix of linear and rotary units, in a force-controlled architecture the company compares to a collaborative robot. Then-CTO Nick Paine told IEEE Spectrum the actuators used about a third fewer components than the company's previous designs, and that a failed unit swaps out with two bolts. No torque figures or actuator counts appear in the public disclosures we reviewed. Hands are the open gap. The original Apollo was shown with simplified task grippers suited to boxes and totes rather than five-fingered hands, and Apptronik has never named an Apollo 2 end-effector; the company says only that it has "tested a variety of end effectors" (A3 interview). Demos pairing Apollo with PSYONIC's Ability Hand have circulated since 2023, including a laundry-sorting demo in late 2025, but no dexterous hand is a confirmed shipping configuration. Compute is dated but official: NVIDIA Jetson AGX Orin plus Jetson Orin NX, per the March 2024 NVIDIA collaboration (The Robot Report). What runs inside Apollo 2 is undisclosed, and Apptronik was notably absent from NVIDIA's Jetson Thor early-adopter list in August 2025, which included Agility, Boston Dynamics and Figure. Safety is zone-based and built into the actuation. An impact zone halts movement when something enters close range; a configurable perimeter zone adjusts behavior further out. Independent reporting adds useful context here: CNBC's February 2026 visit found Apollo units at customer sites still working inside designated areas fenced by external sensors and light curtains, with true side-by-side collaborative operation described as engineering work still ahead. No ISO or ANSI certification has been publicly claimed. The AI stack: Gemini Robotics, GR00T and Artemis Apollo's most strategically important relationship is with Google DeepMind, announced in December 2024 alongside Google's investment. When DeepMind launched Gemini Robotics in March 2025, Apollo was named as a target platform. The September 2025 Gemini Robotics 1.5 release demonstrated cross-embodiment transfer, where skills trained on other robots ran on Apollo without retraining. By December 2025, DeepMind was publishing lab videos of Apollo executing spoken commands on objects it had never seen, bagging bread, sorting laundry by color, potting plants (Forbes). Lab demos, not deployments, but they are among the strongest public evidence of general-purpose manipulation on any commercial humanoid. The timeline extended again on 30 July 2026, when DeepMind released Gemini Robotics 2 with Apollo 2 as its headline platform. It controls the whole body rather than the arms alone, and the published task numbers run from 92% unscrewing a light bulb down to 32% sweeping with a dustpan. It remains a research release: the whole-body model goes only to early-access partners. See the full Gemini Robotics 2 results for what those demos do and do not prove. The NVIDIA GR00T integration announced at GTC in March 2024 got Apollo early attention, though there has been little public follow-up since, and no evidence GR00T is Apollo's production brain today. Apptronik's own layer is the Artemis control stack and Fleet Connect fleet software, which replaced the "point-and-click" task software the company described in the original Apollo era. The Robot Park loop ties it together: fleets generate data, Gemini Robotics models train on it, updated models come back to the fleet. Where Apollo is actually working Manufacturers list partner logos; what follows is the deployment ledger as the public record actually supports it, customer by customer. Mercedes-Benz. The pilot agreement came in March 2024: Apollo delivering parts and assembly kits to the line and inspecting components. Trials ran at the Digital Factory Campus in Berlin-Marienfelde and in Kecskemét, Hungary (Reuters). In March 2025 Mercedes put money in, a low-double-digit-million-euro investment, and expanded the Berlin program. Read the fine print of that release and the maturity is clear: Apollo has been "collecting data in a production environment," Mercedes employees transfer task knowledge through teleoperation and augmented reality, and enabling autonomous operations is the stated next step. Mercedes also simulates Apollo fleets in NVIDIA Omniverse digital twins of its plants before anything touches a real floor. GXO Logistics. Announced in June 2024 as an explicitly early-stage, multi-phase R&D program, starting in a lab before any distribution center. By mid-2026, GXO was operating Apollo 2 units and hosting Robot Park data collection. One correction worth making because it is repeated everywhere: the humanoid working at the GXO-run Spanx facility in Georgia is Agility Robotics' Digit, not Apollo. GXO pilots humanoids from at least three vendors, and its automation chief put the whole category plainly in April 2025: "We're not at wide-scale deployment and commercial viability yet, but we're not 10 years away." Jabil. The February 2025 partnership runs on two tracks. Apollo units work inside Jabil plants on inspection, sorting, kitting, lineside delivery and sub-assembly, as validation before customer shipment. And Jabil is Apollo's manufacturing partner, with the ability to scale across its 100-plus sites, which sets up the "Apollo building Apollo" loop the company likes to describe. Jabil was notably absent from the June 2026 Robot Park release, and neither company has published production volumes. Google DeepMind is a research partner and investor rather than a customer, and hosts its own Robot Park data site. The plain read of that ledger: each engagement on the public record is a pilot, R&D program or data-collection deployment. Commercial terms are not public, and no customer has published throughput, uptime or task-count figures for Apollo. Contrast Agility, which publicizes Digit's tote counts at GXO. Asked about fleet size in June 2026, Cardenas said only: "We haven't released the specific numbers of the Apollo 2s, but lots of robots." For a company this well funded, the scoreboard is still remarkably private. Strengths and weak points What genuinely stands out: the pedigree (NASA Valkyrie lineage, a decade of platforms before Apollo); funding that pushes financing risk out for years; the battery architecture, still one of the most operations-minded designs in the category; a clearly published 25 kg payload where several rivals publish nothing; the most extensively documented AI partnership in the category, with Gemini Robotics naming Apollo as a target embodiment; and a credible route to manufacturing scale through Jabil rather than a promise to build factories from scratch. What holds it back: nothing is purchasable, and the company itself now frames Apollo 2 as a prototype; no customer-site throughput, uptime or reliability data has been published; autonomy in customer sites still lives behind light curtains; the numeric spec sheet is three years old and silent on speed and dexterity, where rivals like Boston Dynamics publish full numbers; and the commercial timeline has slipped from "full release around end of 2024" at unveiling to Apollo 3 in 2027. None of these are unusual for the category. They are exactly the things a buyer needs on the table before a pilot conversation. Apollo vs Optimus, Figure 03, Digit and Atlas The question behind most Apollo searches is really "which humanoid should I be watching?" Here is the July 2026 state of Apollo's four nearest rivals. Row sources, all July 2026: the Fremont timing per Electrek; Figure's 350-plus production count per Figure and the BMW Spartanburg logistics project, announced June 25 with Figure 03 work now starting, per BMW Group (Figure's reported $20,000 volume target is press reporting, not a company figure); Digit has no public list price and is offered through robots-as-a-service; Atlas's production launch per Engadget's CES 2026 coverage, with first units going to Hyundai and Google DeepMind and The Register reporting 2026 output fully committed. The pattern in that table: of the five, only Agility's Digit is commercially acquirable today, through a robots-as-a-service model. Atlas became a production product at CES 2026 but its first-year output is committed to Hyundai and Google DeepMind. Optimus and Figure 03 have no order path at all. Apollo's pilot-only, quote-based reality is the industry norm, not a weakness peculiar to Apptronik. For the direct head-to-head, see our Tesla Optimus vs Apptronik Apollo comparison, and the best humanoid robots guide ranks the whole field. Who Apollo makes sense for Apollo is worth an enterprise conversation if you run manufacturing or logistics operations at serious scale, have repetitive gross-manipulation work (totes, kits, lineside delivery), and can fund a pilot whose payoff is learning and positioning rather than immediate ROI. Mercedes, GXO and Jabil are the profile: companies buying a seat at the table for 2027. Look elsewhere if you need a robot you can actually acquire this year (that is Digit, via RaaS), if you are a researcher or developer (a Unitree G1 EDU costs a fraction of an enterprise pilot), or if you want something for your home; no Apollo home version exists or has been announced, and the consumer-facing players are a different list. What to watch next Four dated things will tell you whether Apollo converts pedigree into product. Apollo 3's debut, which carries its own contradiction: Apptronik's February 2026 release said the new robot debuts in 2026, while Cardenas's June 30 Forbes interview pointed to "within the next year," an apparent slip toward 2027. Either way it is billed as the first true product. Jabil production numbers, which neither company has yet published. Mercedes moving Apollo from teleoperated data collection to autonomous operations, the explicitly stated next step. And any first public throughput metric at a customer site, which would end Apollo's scoreboard silence. We re-verify this page against those events. The bottom line By funding, partner quality and disclosed pilots, Apollo is among the most credible industrial humanoid programs in the world, and it is still not a product. That is not a contradiction; it is the state of the industry, and Apptronik is now unusually honest about it. The pedigree is real, the money is real, the Mercedes and Jabil engagements are real. What does not exist yet is a price, a purchasable unit, or a single customer-published performance number. If you are an enterprise evaluating pilots now, go in with the deployment ledger above and ask each vendor for the numbers they have not published. The Apollo record in our robot database tracks the verified facts as they change. --- # Boston Dynamics Atlas Robot: Price, Specs, and Can You Buy One? URL: https://blog.robozaps.com/b/boston-dynamics-atlas-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T00:02:22.151Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: Boston Dynamics has never priced Atlas, and no outsider can buy one until early 2027: 2026 output is committed to Hyundai's training center and Google DeepMind. The '$320,000' online is an anonymous-source ceiling, not a price; analyst estimates cluster at $130-145k. July 2026: a World Cup ball delivery, the first humanoid-robot strike, and a move to full Hyundai ownership. Key takeaways: - No Atlas price has ever existed. The '$320,000' is an anonymous-source ceiling from one Korean report; analyst and industry estimates cluster around $130,000-$145,000 per unit. - You cannot buy one: 2026 production is committed to Hyundai's RMAC training center and Google DeepMind, with external customers starting early 2027. - The backflip and parkour videos show the hydraulic Atlas, retired April 2024 and never sold; the production Atlas (CES 2026) is a different machine with the first official spec sheet: 1.9m, 90kg, 56 DoF, 4-hour self-swapping battery. - 'Atlas works in factories' is ahead of schedule: today's Metaplant work is a pilot; production-line deployment starts 2028, and the 30,000/year figure is planned factory capacity, not output. - July 2026 changed the story: Atlas delivered a World Cup match ball, ~35,000 Hyundai workers staged the first strike over humanoid deployment, and Hyundai announced the buyout of SoftBank's remaining stake, pending completion. FAQ: Q: How much does the Boston Dynamics Atlas cost? A: There is no price: Boston Dynamics has never published one for any Atlas generation. The '$320,000' circulating is an anonymous-source report that pricing would stay below that ceiling; analyst and industry estimates cluster around ~200 million won ($130,000-$145,000). Treat every figure as an estimate until Boston Dynamics speaks. Q: Can you buy a Boston Dynamics Atlas? A: No. Boston Dynamics' own FAQ says Atlas is 'still in the early stages of the commercial journey.' The 2026 production run is fully committed to Hyundai's robotics training center and Google DeepMind, with additional enterprise customers planned from early 2027. There is no consumer channel. Q: Is the Atlas robot for sale to the public? A: No, and it likely never will be in consumer form: Atlas is an industrial robot sold enterprise-to-enterprise. Spot and Stretch are Boston Dynamics' commercially available products; Atlas deployments outside Hyundai and Google begin in 2027 at the earliest. Q: Is the Boston Dynamics Atlas real? A: Yes. The famous videos are real robots; manufacturing of the production Atlas began after CES 2026, and electric Atlas prototypes pilot sequencing work at Hyundai's Georgia plant. Old CGI-parody videos ('Bosstown Dynamics') and a 2023 TikTok fake fed lingering doubts, but every verified Atlas video is genuine hardware. Q: What happened to the backflipping Atlas robot? A: That was the hydraulic Atlas, a DARPA-born research platform. Boston Dynamics retired it on April 16, 2024 with a farewell video and unveiled its all-electric successor the next day. The parkour robot was never for sale. Q: What are the Atlas robot's specs? A: The production Atlas's official sheet: 1.9 m tall, 90 kg, 2.3 m reach, 56 degrees of freedom with continuous 360-degree joints, 50 kg instant payload (30 kg sustained), a 4-hour self-swappable battery with autonomous recharge, IP67 sealing, and -20 to 40 degrees C operation. Price, top speed and joint torque remain undisclosed. Q: What does the Atlas robot actually do? A: Today: parts-sequencing pilot work at Hyundai's Georgia Metaplant, training-center duty at Hyundai's RMAC, and research at Google DeepMind. Production-line deployment starts in 2028, expanding toward component assembly around 2030. It also delivered the ceremonial match ball at a 2026 World Cup match. Q: Who owns Boston Dynamics? A: Hyundai Motor Group. It took 80% control in 2021 and in July 2026 announced the purchase of SoftBank's remaining ~10% stake (reported at roughly $325-335 million), which would make Boston Dynamics wholly Hyundai-owned pending completion. An eventual IPO has been floated. Q: Why did Hyundai workers strike over Atlas? A: After Hyundai disclosed plans for 25,000+ Atlas units across its plants, the Korean Metal Workers' Union demanded a labor-management agreement before any robot enters a workplace, staging rolling walkouts of ~35,000 workers in July 2026 - widely described as the first strike triggered by humanoid robots. No Korean deployment date actually exists yet. Q: Is Atlas better than Tesla Optimus? A: Atlas is further along commercially: a production product, an official spec sheet, committed deployments and a 2028 factory date, while Optimus has a bigger volume ambition and a $20-30k target price but no external customers. Neither can be bought. Our full comparison covers the matchup in depth. Q: Does Atlas use Google's AI? A: It's starting to: the January 2026 partnership aims to integrate Google DeepMind's Gemini Robotics foundation models into Atlas, and DeepMind receives its own Atlas fleet. Boston Dynamics also built a whole-robot large behavior model with Toyota Research Institute in 2025, and Atlas runs NVIDIA Jetson Thor compute onboard. Q: How much can the Atlas robot lift? A: 50 kg (110 lbs) instantaneous and 30 kg (66 lbs) sustained, per Boston Dynamics' official production spec sheet, with a 2.3 m reach and continuously rotating joints that let it lift in postures humans can't. How much does the Boston Dynamics Atlas cost, and can you buy one? Boston Dynamics has never published a price for Atlas, and no outside company can buy one in 2026: the entire first production run is committed to Hyundai's robotics training center and Google DeepMind, with additional customers planned for early 2027. The "$320,000" you'll see quoted as Atlas's price is an anonymous-source ceiling from a single Korean report, not a Boston Dynamics announcement, and the "$120,000" on some spec sites traces to nothing at all. What exists instead is an unusually rich evidence ladder about what Atlas is likely to cost, laid out below and checked against primary sources on July 22, 2026, at the end of three weeks in which Atlas delivered a World Cup match ball, triggered the first strike over humanoid robots, and moved to full Hyundai ownership. The price evidence ladder Since no list price exists, the honest way to answer the price question is to rank the evidence. One more figure to retire on sight: Spot's famous $74,500. That was the 2020 online launch price of a different, smaller robot, has been quote-only for years, and has never had anything to do with Atlas. Can you actually get one? Boston Dynamics' own FAQ answers this in one line: Spot and Stretch "are available commercially, while Atlas is still in the early stages of the commercial journey." Concretely: the production run that began after CES 2026 is fully committed to Hyundai's new Robotics Metaplant Application Center and to Google DeepMind's research fleet, and the company says additional customers come aboard in early 2027. There is no configurator, no lease page, and no waitlist a member of the public can join; prospective industrial buyers can contact Atlas sales; the company says additional customers are planned from early 2027. Anyone selling you an Atlas today is selling you a story. Three Atlases: get the generations straight Most wrong Atlas content on the internet comes from mixing three different machines that share a name. The production machine's official numbers, from Boston Dynamics' live spec page: 1.9 m tall, 90 kg, 2.3 m reach, 56 degrees of freedom with continuous 360-degree range of motion, 50 kg instant payload (30 kg sustained), a 4-hour self-swappable battery with autonomous return-to-charge, IP67 sealing, and operation from -20°C to 40°C. The official record also lists tactile sensing and 360° vision, autonomous, teleoperated and tablet control modes, human detection with fenceless guarding, and barcode/RFID plus Orbit/MES/WMS integration. Still undisclosed: price, top speed, and joint torque. Full record: Atlas in the RoboZaps robot database. A naming note: there is no "Atlas 4.0". That's YouTube clickbait; Boston Dynamics uses no public version numbers. The deployment ladder, honestly tensed Headlines routinely compress this schedule into "Atlas is working in factories." Here is the committed sequence, with the tense each step deserves. That last row is the most misquoted: the 30,000-a-year figure is the capacity of a robotics facility in Hyundai's $26 billion US investment package, not something Boston Dynamics builds today, and 2026 volumes remain undisclosed. The "Atlas beat Tesla to the factory floor" framing is likewise half-true: Atlas has committed deployments and a production product, which Optimus lacks, but its 2026 units go to a training center and a research lab; real line work starts in 2028. Atlas's July 2026: the biggest month it's ever had The first humanoid-robot strike The July walkouts deserve their own entry in the record, because they mark the moment humanoid deployment stopped being a demo-stage abstraction and became a bargaining-table issue. After Hyundai disclosed plans for 25,000-plus Atlas units across its plants, the Korean Metal Workers' Union made its position a slogan: no robot enters a Hyundai workplace without a labor-management agreement first. The rolling four-hour walkouts of July 13–15 touched roughly half of Hyundai's global output. Two facts keep the story honest in both directions: no Atlas deployment date exists for Korean plants at all (Georgia 2028 is the only committed site), and the ~₩200 million per-unit estimate cited in that debate is now one of the more grounded numbers in the entire Atlas price debate. However it resolves, every future humanoid rollout negotiation will cite this one. The AI stack: who's actually training Atlas Three partnerships, each real, each different. The Google DeepMind deal announced at CES aims to integrate Gemini Robotics foundation models into Atlas and sends DeepMind its own Atlas fleet — a homecoming, since Google owned Boston Dynamics until 2017. The Toyota Research Institute partnership produced last August's milestone: a single large behavior model controlling the whole robot, hands and feet, through long multi-task sequences; its status after the DeepMind deal is publicly ambiguous, but it has not been ended. And Atlas runs NVIDIA's Jetson Thor onboard under a collaboration expanded at GTC 2025. On autonomy, Boston Dynamics states the famous sequencing demos involve "no prescribed or teleoperated movements"; the Boston Dynamics-TRI pipeline is equally candid that the underlying skills are learned from teleoperated human demonstrations. Autonomous execution, human-taught skills: that's the honest summary of where Atlas's brain is in 2026. The DeepMind side of that produced its first public result on 30 July 2026, with Gemini Robotics 2 controlling a humanoid from feet to fingertips. Apptronik's Apollo 2 was the demonstration platform, not Atlas, and Boston Dynamics appears only in the acknowledgements. Published success rates run from 92% unscrewing a light bulb down to 32% sweeping with a dustpan. See Gemini Robotics 2 and whole-body control for what the release does and does not show. Atlas vs Optimus, in one paragraph The matchup gets a full treatment in our Atlas vs Optimus comparison; the one-paragraph version is this. Atlas has the first production humanoid with committed deployments, an official spec sheet, and a 2028 factory date; Optimus has bigger volume ambitions, a $20–30k price target (also not a price), and a production start of its own scheduled at Fremont around the time you read this. Neither can be purchased. For machines you can actually order today, our evidence-ranked guide is the honest place to start. Misinformation audit: eight Atlas claims, checked What to watch next Four dated things will move this page. The Korean labor negotiation, which now sets precedent for every humanoid rollout. The early-2027 external-customer onboarding, which produces the first Atlas deployments outside the Hyundai-Google orbit and probably the first real price signals. The 2028 Metaplant production-line date, the schedule's first hard test. And across town, Optimus's Fremont production start, which decides whether the "beat Tesla to the factory floor" framing survives the year. We re-verify this page against the primary sources above; every figure carries its check date. The bottom line Atlas in 2026 is the inverse of most humanoid stories: the robot is further along than the discourse about it. While the internet argues over a price that has never existed and shares videos of a machine that retired two years ago, the production Atlas has an official spec sheet, a committed customer, a factory date, a World Cup cameo, and a labor movement organizing against its arrival, which is the surest sign anyone has ever taken a humanoid robot seriously. You cannot buy one, probably until 2027 at the earliest, and when a price finally exists it will come from Boston Dynamics, not from the spec farms. Until then, the evidence ladder above is the closest thing to the truth that money can't yet buy. --- # 38 Best Humanoid Robots in 2026 (Evidence-Ranked) URL: https://blog.robozaps.com/b/best-humanoid-robots Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T10:54:56.545Z Last fact-checked: 2026-07-13T00:00:00.000Z Category: best TL;DR: Agility Robotics Digit is the best humanoid robot in 2026 for proven commercial work. Atlas leads on industrial hardware, Figure 03 on AI-first manufacturing, Unitree G1 on entry-hardware value, and 1X NEO among home-focused candidates. Researchers need the quote-only G1 EDU for secondary development. Key takeaways: - Digit ranks first on sustained commercial evidence, including more than 100,000 totes moved at GXO and a Toyota Canada agreement. - Atlas has the strongest published industrial hardware: 56 degrees of freedom, IP67 protection and up to 50 kg instantaneous payload. - Unitree G1 is the best-value entry hardware at a published $13,500 base price; researchers need the quote-only EDU version for secondary development. - 1X NEO is the leading home candidate, but it remains a preorder product and may use scheduled remote Expert Mode for unknown tasks. - Tesla Optimus is not commercialized and has no current official public price or complete product specification sheet. FAQ: Q: What is the best humanoid robot in 2026? A: Agility Robotics Digit is the best overall choice when verified commercial work carries the most weight: it has sustained customer deployments, more than 100,000 totes moved at GXO, and a commercial agreement with Toyota Canada. Atlas leads on industrial hardware, Figure 03 on AI-first manufacturing, Unitree G1 on entry-hardware value, and 1X NEO among home-focused candidates. Researchers need the quote-only G1 EDU for secondary development. Q: Which humanoid robot has the longest battery life? A: Most bipedal humanoids publish roughly two to five hours per charge. Digit publishes about four hours and Atlas about four hours. Walker S2 addresses uptime differently by swapping its own battery in about three minutes, while other makers publish longer runtime claims that have less independent field evidence. Q: What can humanoid robots actually do in 2026? A: Verified work includes moving warehouse totes, handling parts in automotive production, factory inspection, logistics picking, guided tours and reception tasks. No current humanoid reliably performs open-ended housework without supervision. Many systems still mix autonomy with teleoperation, controlled environments or tightly scoped workflows. Q: Which humanoid robots have real commercial deployments? A: Digit has the clearest sustained external deployment record, including GXO and Toyota Canada. Figure 02 completed a documented BMW program; Figure 03 completed its first BMW parts-sequencing demonstration in June 2026. Walker S-series robots have industrial deployments reported by UBTECH. Atlas, Apollo 2 and several Chinese platforms are in earlier production, training or customer-pilot stages. Q: What is the difference between bipedal and wheeled humanoid robots? A: Bipedal robots can use stairs and human-designed spaces but consume more energy and add balance complexity. Wheeled or hybrid humanoids are more stable and efficient on flat floors. The right choice depends on the environment and task, not on whether the robot looks more human. Q: Are humanoid robots safe? A: Safety remains use-case specific. Industrial robots usually operate under controlled procedures, supervision and task limits. Home-focused machines use lighter or compliant designs, but remote assistance and restricted operating zones are still common. Buyers should require documented safety functions, service procedures and evidence from the intended environment. Agility Robotics Digit is the best humanoid robot in 2026 when verified commercial work matters most. Boston Dynamics Atlas leads on industrial hardware, Figure 03 is the strongest AI-first manufacturing contender, Unitree G1 is the best-value entry hardware, and 1X NEO is the leading home candidate. Researchers should note that secondary development requires the quote-only G1 EDU. We rank 38 robots by deployment proof, commercial readiness, autonomy, hardware capability, uptime, safety and source transparency. July 2026 Updates Agility Robotics is going public via SPAC at ~$2.5 billion (announced July 5) and reports $300M+ in contracted Digit v5 orders1X announced a 25-DOF NEO hand revision on July 9, while the order page still lists 22 DOF per hand; its Hayward factory opened April 30 with customer shipments promised by end of 2026Unitree's ~$618M Shanghai STAR Market IPO cleared final registration in early July after a record 104-day review; the G1 base price has dropped to $13,500AgiBot rolled out its 10,000th humanoid on March 30; Omdia ranks it #1 for 2025 shipments (5,168 units, 39% share), a ranking Unitree disputes with its own 5,500+ figureNEURA Robotics raised up to $1.4 billion led by Tether (June 10, ~$7B valuation), Europe's biggest humanoid bet, with 4NE1 deliveries targeted for late 2026Sanctuary AI pivoted to hardware-agnostic robot software (June 17), and Amazon quietly acquired home-robot startup Fauna Robotics (March), the market shakeout has begun: K-Scale Labs shut down in November 2025, Cartwheel Robotics in February 2026Tesla's Optimus 3 reveal is expected around production start (late July–August) at Fremont; Musk walked the new line on July 1 and says output will be "extremely slow at first"Boston Dynamics began Atlas production in January; all 2026 deployments are committed to Hyundai and Google DeepMind, with additional customers planned from 2027 The humanoid robot industry crossed real thresholds in the first half of 2026: roughly 13,000 humanoids shipped in 2025 (per Omdia), dedicated factories are running in the US and China, and the first consumer robots are heading to homes. But hype still outruns hardware, and much of what circulates about specs and prices is wrong. In this guide, updated and re-verified July 13, 2026, we rank 38 humanoid robots that are shipping, piloting, or credibly imminent. We checked the priority facts against manufacturer or primary sources and every estimate labeled. Where a figure is an estimate, we say so. Where a claim is only the maker's, we attribute it. Quick-Glance: Best Humanoid Robots of 2026 * Reported figures; Figure publishes neither. † Press-reported, not an official datasheet. "—" means the maker has not published the figure; we don't fill gaps with guesses. Our Ranking Methodology We score every humanoid robot across seven evidence-weighted criteria: Verified external deployment (30%), sustained work for a named customer beats an internal demo.Commercial availability and production (20%), delivered or orderable machines beat announced targets.Task autonomy (15%), independent task completion beats teleoperation presented as autonomy.Manipulation and payload (10%), useful hands, reach and load capacity matter more than showreel motion.Uptime and reliability (10%), battery strategy, docking, duty cycle and field hours count.Safety and serviceability (10%), human-safe operation, maintainability and support count.Transparent specifications and pricing (5%), published evidence earns credit; missing data stays missing. Robots working in real factories, warehouses, and hospitals rank above robots that only exist in demo videos. We verify specs against manufacturer pages and primary sources, label estimates as estimates, and attribute manufacturer claims rather than repeating them as fact. Unrevealed robots (like Tesla's Optimus 3) are ranked on verified progress, not on leaked spec sheets. Last full re-verification: July 13, 2026. AI Platforms Powering Humanoid Robots Behind every capable humanoid is an AI stack for perception, reasoning, and action. Three platform camps matter in 2026: NVIDIA Isaac GR00T NVIDIA's Isaac GR00T N1, released at GTC in March 2025, was the first open foundation model for humanoid robots; the current release is GR00T 1.7 (GTC, March 2026, early access), and N1.6 shipped at CES 2026 with Cosmos Reason integration. Isaac Lab's GPU-parallel simulation compresses months of robot training into hours. Current adopters include Agility Robotics, Boston Dynamics, NEURA, Mentee and 1X, with N1.6 validated on the Unitree G1 and AgiBot Genie-1. Google DeepMind Gemini Robotics At CES 2026, Google DeepMind and Boston Dynamics announced a partnership to put Gemini Robotics foundation models on the electric Atlas (and Spot), with testing planned at Hyundai plants. Apptronik's Apollo also runs Gemini Robotics through a separate partnership, and the two companies opened a shared 90,000 sq ft training facility, "Robot Park," in June 2026. DeepMind extended that on 30 July 2026 with Gemini Robotics 2, its first model to control a humanoid's legs, torso, arms and hands under a single policy. Apollo 2 was the demonstration platform. Published task success runs from 92% unscrewing a light bulb down to 32% sweeping with a dustpan, and the whole-body model goes only to early-access partners, so it moves no ranking on this page yet. Read our Gemini Robotics 2 breakdown for the full numbers. In-House Models: Helix and the Vertical Stack Figure AI ended its OpenAI collaboration in February 2025 and built Helix, an in-house vision-language-action model that now runs Figure 03, a bet that owning the whole stack beats renting a brain. Tesla takes the same vertical approach with FSD-derived networks and its AI5 chip, XPENG with its Turing chips and VLA/VLT models, and Unitree open-sourced its UnifoLM-VLA-0 research model in January 2026. Hardware is commoditizing faster than intelligence; the AI layer is where the next rankings will be won. The 38 Best Humanoid Robots of 2026 — Full Reviews The short version, if you skip the rest: Digit is the strongest evidence-backed industrial buy; Atlas leads on hardware and Figure 03 on AI-first manufacturing. Unitree G1 is the best-value entry hardware at $13,500, but researchers need the quote-only G1 EDU for secondary development. A home robot in 2026 means pre-ordering a 1X NEO. And if you want a full-size humanoid shipping today for less than a car, compare the Unitree H2 against the Kepler K2. The other entries earn their places for narrower reasons, including speed, price, open source and weatherproofing. Each verdict line says which. Before shortlisting any of them, check it can actually be sold to you: humanoid robots legal in the US has the FCC position for every robot in our database. 1. Agility Robotics Digit — Best Proven Commercial Humanoid Manufacturer: Agility Robotics (Corvallis, OR) | ~$640M raised | Going public via SPAC at ~$2.5B (announced July 5, 2026) Digit has the most real commercial mileage of any humanoid: 65,000+ operating hours across nine facilities, per the company's SEC filing. It passed 100,000 totes moved at GXO's Flowery Branch site in November 2025 (a first-of-its-kind disclosed milestone), and its customer list (Schaeffler, GXO, Toyota Motor Manufacturing Canada, Mercado Libre) reads like a logistics who's-who. Amazon is an investor and has piloted Digit, but Digit is not deployed across Amazon fulfillment centers. The 2026 story is scale: Agility announced on July 5 it is going public via the Churchill Capital Corp XI SPAC at roughly $2.5 billion, and reports $300M+ in contracted orders for Digit v5, its cooperatively safe commercial model with swappable hands and a 50 lb lift designed to work fence-free alongside people. The RoboFab factory in Salem, Oregon can scale past 10,000 units a year. Best for: Warehouse tote handling, truck loading, logistics at scale Pros: Most disclosed real-world hours of any humanoid; blue-chip customers; dedicated factory; public listing brings capital and scrutiny Cons: No public price; narrower task range than manipulation-first rivals; battery is 4 hours, not the 8 sometimes claimed View Digit on Robozaps 2. Boston Dynamics Atlas (Electric) — Best Industrial Hardware Manufacturer: Boston Dynamics (Waltham, MA; Hyundai subsidiary) | 30+ years of bipedal R&D Production Atlas entered manufacturing in January 2026. Boston Dynamics says all 2026 deployments are committed to Hyundai's Robotics Metaplant Application Center and Google DeepMind, with additional customers planned for 2027. It has not announced open ordering. Official specifications are 1.9 m, 90 kg, 56 degrees of freedom, a 50 kg instantaneous payload and a 30 kg sustained payload. At CES 2026, Atlas won CNET's Best of CES "Best Robot" award and Boston Dynamics announced a partnership with Google DeepMind to run Gemini Robotics models on it. Hyundai says production parts-sequencing at HMGMA begins in 2028. Boston Dynamics has not published a price or announced open ordering. Best for: Automotive manufacturing, heavy industrial work, frontier R&D Pros: Most mechanically capable humanoid with the highest verified payload; in production; Gemini Robotics brain incoming; Hyundai's manufacturing machine behind it Cons: No open sales process; additional customers are planned only from 2027; no public price 3. Figure 03 — Best AI-First Manufacturing Contender Manufacturer: Figure AI (San Jose, CA) | Founded 2022 | Over $1.7B raised | $39B valuation (Sept 2025 Series C led by Parkway, with Brookfield, NVIDIA, Intel Capital, LG, Salesforce, Qualcomm) Figure AI's third-generation humanoid, announced October 9, 2025, was purpose-built around Helix, the company's in-house vision-language-action AI (Figure ended its OpenAI collaboration in February 2025 to build Helix). Each hand carries an embedded palm camera, the 2.3 kWh battery runs about 5 hours at peak and recharges wirelessly at 2 kW through coils in the feet, and the whole robot is 9% lighter than Figure 02. The deployment record matters more than the spec sheet. Figure 02 completed an 11-month pilot at BMW Spartanburg (January–November 2025), supporting production of 30,000+ X3s with about 90,000 parts loaded at 99% accuracy, and was retired in November 2025. Figure 03 completed its first BMW demonstration in June 2026 for parts-sequencing work. That demo is not the same as Figure 02's 11-month production record. Figure signed its first retail-logistics customer, Catalyst Brands (JCPenney, Aéropostale), in May. The BotQ factory is tooled for up to 12,000 units a year, targeting 100,000 robots over four years. In March 2026, Figure 03 joined Melania Trump's White House AI education summit, which CEO Brett Adcock called the first humanoid robot in the White House. Home use is in alpha testing only, no public program exists. Best for: Manufacturing, logistics, and the strongest overall AI-plus-hardware bet Pros: Figure 02 lineage at BMW; Figure 03 completed its first BMW workflow demonstration; in-house Helix AI; BotQ manufacturing Cons: No published price or open sales; home use is alpha-only; specs partly unpublished, so comparisons lean on reported figures View Figure 03 on Robozaps 4. UBTECH Walker S2 — Best Continuous-Operation Factory Humanoid Manufacturer: UBTECH Robotics (Shenzhen, China) | HKEX: 9880 | Walker S2 orders exceed ¥800M (~$112M) Walker S2, launched July 2025, is the industrial humanoid you can actually order in volume. Its signature trick is a world-first: fully autonomous hot-swap batteries. The robot changes its own pack in about three minutes, enabling round-the-clock shifts. Mass production and deliveries began in November 2025 with orders above ¥800 million, including a single ¥250M purchase, and UBTECH plans capacity of 5,000 units a year in 2026. It stands on the deployment record of its predecessor Walker S1 (172 cm, ~76 kg), which worked quality inspection at Audi's FAW plant in Changchun (the first humanoid in Audi's global production system), completed logistics validation at Foxconn's Longhua campus, and ran the world's first multi-humanoid collaborative swarm at a Zeekr factory in March 2025. Press specs for the S2: 176 cm, ~73 kg, 52 degrees of freedom, 15 kg lift. Best for: Factory inspection, assembly support, 24/7 industrial shifts Pros: Self-swapping batteries enable continuous operation; real orders at nine figures; proven S1 lineage at Audi, Foxconn, Zeekr; publicly listed parent Cons: Enterprise pricing is opaque; China-centric deployments; per-unit specs are press-sourced, not a published datasheet View Walker S1 on Robozaps 5. Apptronik Apollo 2 — Best Training Platform for Industrial AI Manufacturer: Apptronik (Austin, TX) | $935M Series A closed Feb 11, 2026 (~$5.3B valuation) | B Capital + Google co-led; Mercedes-Benz, ARK, John Deere, QIA The original Apollo published a 1.7 m height, 73 kg weight, 25 kg payload and roughly four hours per hot-swappable battery. Apptronik has not confirmed that every one of those figures carries over to Apollo 2, the current bipedal and wheeled data-collection platform. It runs pilots with Mercedes-Benz (an investor since 2024), GXO Logistics and Jabil, where the long-term goal is Apollo building Apollo. The program moved fast in 2026: Apptronik closed a $935M Series A in February, and on June 30 unveiled Apollo 2, a modular flagship offered in bipedal and wheeled-base configurations, alongside "Robot Park," a 90,000 sq ft training facility run with Google DeepMind, whose Gemini Robotics models power Apollo. Apollo 3, teased as the first true production product, is expected in 2027. On price, the old sub-$50K-at-scale target has given way to a ~$80K/year high-volume delivery framing from 2027; treat both as targets, not list prices. Best for: Training fleets, embodied-AI data collection and industrial workflow development Pros: Bipedal and wheeled configurations; Mercedes/GXO/Jabil pilot lineage; Gemini Robotics integration; nearly $1B raised Cons: No list price; volume product (Apollo 3) is a 2027 story; DOF and finer specs unpublished View Apollo on Robozaps 6. 1X NEO — Best Home Humanoid Candidate Manufacturer: 1X Technologies (San Francisco Bay Area / Oslo) | OpenAI Startup Fund + EQT + Samsung NEXT backed | first 10,000 units of planned capacity booked in five days NEO is the first credible consumer humanoid, and 2026 is its make-or-break year. To be precise about status: NEO is still pre-order only. The NEO Factory in Hayward, CA opened April 30, 2026 with first units going to R&D and internal home testing; CEO Bernt Børnich says customer shipments start before the end of 2026. 1X says the first 10,000 units of planned capacity were booked in five days. The design is genuinely home-first: 30 kg, tendon-driven soft body, 4-hour battery, and a privacy model built around teleoperation: human operators assist through "Expert Mode" with people-blurring, no-go zones, and owner approval, which is how NEO learns. On July 9, 1X announced a 25-DOF hand revision. Its order page still lists 22 DOF per hand, so we treat 25 DOF as an announced revision until 1X reconciles the shipping configuration. At $20,000 (or $499/month with a $200 refundable deposit), it's the cheapest path to a full-size home humanoid. The purchase still rests on promised improvements in autonomy. Best for: Early-adopter households, elder assistance, the home-robot bet Pros: Purpose-built for homes; real factory + 10,000 planned production slots booked; subscription lowers entry; new human-level hands announced July 2026 Cons: Not yet delivered to a single customer; autonomy today leans on teleoperation; US-only at launch View NEO on Robozaps · Home humanoid guide 7. AgiBot A2 — Highest Reported Manufacturing Scale Manufacturer: AgiBot / Zhiyuan Robotics (Shanghai) | Omdia: #1 in 2025 humanoid shipments, 5,168 units, 39% share By analyst count, nobody shipped more humanoids in 2025 than AgiBot: Omdia credits it with 5,168 units and a 39% global share (Unitree self-reports 5,500+ and disputes the ranking, read shipment claims with both numbers in mind). The scale is accelerating: AgiBot reported its 10,000th robot on March 30, 2026, three months after reporting 5,000 in December 2025. An A2 also set a Guinness World Record in January 2026 for the longest humanoid walk: 106 km from Suzhou to Shanghai. The A2 itself is a service-first humanoid (169 cm, 69 kg, 40+ degrees of freedom, a swappable 700 Wh battery good for about two hours) that threads needles in demos and handles reception, guidance, and interaction work. The A2 Ultra variant was the first humanoid certified concurrently for China (CR), the EU (CE), and the US (FCC) in May 2025, and the family spans wheeled (A2-W) and industrial variants. Best for: Customer service, exhibitions, guided tours, front-of-house work Pros: Analyst-ranked #1 shipper of 2025; tri-region certification; 106 km Guinness walk; broad product family Cons: No public pricing; service focus over heavy manipulation; volume crown contested by Unitree View AgiBot A2 on Robozaps 8. Unitree H2 — Best Full-Size Humanoid Value Manufacturer: Unitree Robotics (Hangzhou, China) | Unveiled October 20, 2025; shipping since April 2026 The H2 is the cheapest full-size humanoid actually shipping: $29,900, about 180 cm and 70 kg, with 31 degrees of freedom, a quick-release 0.972 kWh battery (~3 hours), 2070 TOPS of compute, and a static bionic face. Unveiled October 20, 2025 in Unitree's "Destiny Awakening" video and demoed at CES 2026, it started customer deliveries at the end of April 2026 and supersedes the legacy H1/H1-2 line ($90,000, now quote-only). In June 2026 Unitree stretched the family upward with the H2 Plus, a $100,000 NVIDIA Isaac GR00T reference platform with 75 DOF, Jetson AGX Thor compute, and tactile five-finger hands, due late 2026. For most commercial and education buyers, the standard H2 is the price-to-size ratio to beat. Every Unitree model and the IPO status live in our Unitree Robotics hub. Best for: Commercial service, education, full-size research on a budget Pros: Cheapest shipping full-size humanoid; quick-swap battery; Unitree ecosystem and OTA cadence; H2 Plus path for GR00T developers Cons: New platform with a thin deployment record; the "expressive face" some coverage claims is a static bionic head; EDU tiers are quote-only View Unitree H2 on Robozaps 9. Unitree G1 — Best-Value Entry Hardware Manufacturer: Unitree Robotics (Hangzhou, China) | ~$618M STAR Market IPO approved July 2026 (~$5.9B valuation) | 2025 revenue ¥1.71B, up 335% The G1 made humanoid robotics affordable, and it keeps getting cheaper: the official price is now $13,500 (it launched at $16,000 in May 2024), currently backordered. Know what the base model is: it does not support secondary development, researchers who want to program it need the G1 EDU (contact sales), which also unlocks up to 43 degrees of freedom versus 23 on the base unit and roughly 3 kg of payload versus 2. Unitree launch demos showed bottle-opening, nut-cracking, and soldering; the laundry-folding clips that circulate are research-community work, not official capability. Unitree self-reports 5,500+ humanoids shipped in 2025 and claims the volume crown, though analyst firm Omdia counts ~4,200 and ranks AgiBot first, a dispute worth knowing about when you read shipment headlines. Either way, the developer ecosystem around the G1 is the largest in consumer humanoids, and its ~$618M Shanghai IPO cleared final approval in early July 2026. Best for: Research, education, AI development, serious hobbyists Pros: Low base price; huge developer community; ROS2-compatible EDU tier; continuous OTA updates Cons: Base model can't be programmed (EDU is quote-only); small stature and ~2 kg payload rule out industrial work; ~2-hour battery View Unitree G1 on Robozaps · Full G1 review 10. Fourier GR-3 — Best Care-Focused Humanoid Manufacturer: Fourier (Shanghai, China) | Rehab-robotics heritage: company reports 2,000+ institutions in 40+ countries Fourier's current flagship is the GR-3 "Care-bot," launched August 6, 2025 and given its US debut at CES 2026: 165 cm, 71 kg, 55 degrees of freedom, 31 body pressure sensors for safe physical contact, dual hot-swap batteries (~3 hours), and a full-sensory interaction system aimed at eldercare, rehabilitation, and reception. Pricing starts above ¥200,000 (~$27,500+), sold B2B. The earlier GR-2 platform (September 2024: 175 cm, 63 kg, 53 DOF, 3 kg single-arm load, ~2-hour swappable battery) remains Fourier's research and industrial base, but note the widely-copied "50 kg payload" figure is wrong; that number belonged to GR-1 marketing and refers to total burden, not arm payload. Fourier's edge is distribution: it already sells rehabilitation robots into thousands of clinics, a channel no humanoid rival has in healthcare. Best for: Eldercare, rehabilitation, patient-facing and reception roles Pros: Only humanoid maker with a real healthcare install base; contact-safe design with pressure skin; hot-swap endurance; CES-vetted Cons: B2B only; GR-2-era spec sheets circulating online are badly wrong (payload); healthcare deployments are pilots, not yet fleets View Fourier GR-2 on Robozaps 11. Tesla Optimus Gen 3 — Most Anticipated, Not Yet Commercial Manufacturer: Tesla (Austin, TX) | Optimus 3 reveal expected around production start, late July–August 2026 Be careful with anything you read about "Optimus Gen 3": as of July 13, 2026, Tesla has not revealed Optimus 3 and has published no specifications, Musk told analysts Tesla is withholding the design because competitors "analyze it frame by frame and copy everything we do." What is verified: Optimus 3 is a ground-up redesign (not an upgrade of the current robot, which Musk calls "Optimus 2.5"), the design is in its "final stages of completion," and production is targeted for late July or August 2026 on the converted Model S/X line at Fremont. Musk walked that line on July 1 and warned output will be "extremely slow at first because everything is new." Syndicated stories claiming mass production already started (January or March 2026) trace to a single unsourced press-wire piece and are contradicted by Tesla's own timeline. On price, the last consumer figure Musk gave is $20,000–$30,000 (October 2024, "less than a car is my prediction, long term"); the sub-$20,000 numbers refer to production cost at a million-plus units a year, and Musk says the selling price "will be set by the market demand." Public sales are "likely" by end of 2027 (Davos, January 2026). We rank Optimus 3 on verified progress (a dedicated line at Fremont, the deepest manufacturing base in the industry, and Tesla's AI5 chip earmarked for Optimus), not on leaked spec sheets, which are fabrications. Full analysis: our Tesla Optimus Gen 3 guide. Best for: Watching, not buying, the highest-ceiling humanoid program in the world Pros: Dedicated ~1M unit/year design-capacity line; vertical AI stack (FSD + AI5); unmatched manufacturing muscle; reveal imminent Cons: Unrevealed, unbuyable, and every circulating spec sheet is invented; Tesla robot timelines have slipped repeatedly View Tesla Optimus on Robozaps 12. Kepler K2 Bumblebee — Best Published Industrial Price Manufacturer: Kepler Robotics (Shanghai, China) | 5th-generation platform; mass production since September 2025 Kepler is the rare industrial player with a real list price: the K2 "Bumblebee" is priced at RMB 248,000 (about US$34,000 at July 13 exchange rates) in Kepler's release. The 5th-generation robot (175 cm, 75 kg, 52 DOF) uses Tesla-style hybrid architecture with planetary roller-screw actuators, lifts 15 kg per arm (30 kg dual), and claims 8 hours of runtime on a one-hour charge. Mass production and customer shipping began September 26, 2025, with framework agreements covering several thousand units. If you see this robot listed as the "Kepler Forerunner" at similar prices, that's the 2023 K1 generation, spec sheets citing it are two generations stale. K2 worked a Chunmi appliance factory over Lunar New Year 2026 and Kepler expanded production capacity at Jiangyin in March. Best for: Light manufacturing, inspection, cost-sensitive industrial pilots Pros: Published RMB 248,000 price, unusual among industrial humanoids; hybrid-architecture efficiency; shipping at multi-thousand-unit agreement scale Cons: Younger deployment record than Digit or Walker; runtime figure is the maker's claim; brand still trades under the stale "Forerunner" name online View Kepler K2 on Robozaps 13. XPENG IRON — Most Advanced Humanoid in Development Manufacturer: XPENG (Guangzhou, China) | Next-gen IRON unveiled November 5, 2025; mass production targeted end of 2026 XPENG's next-generation IRON moved so smoothly at its November 2025 unveiling that CEO He Xiaopeng cut one open on stage to prove there was no person inside. The verified spec set, from XPENG's SEC-filed 2025 ESG report: 82 degrees of freedom total including 22 in each hand (spec sheets that add those numbers together are wrong), the industry's first all-solid-state battery in a humanoid (XPENG's claim), and three of XPENG's own Turing chips delivering about 2,250 TOPS for its VLA/VLM/VLT model stack. A partnership with Baosteel covers complex industrial inspection. Mass production is targeted for the end of 2026, with first roles in guided tours and shopping assistance; the first prototype built to automotive standards came off the line in January. One execution caveat: XPENG's robotics product head departed in June 2026 and the CEO now runs the unit directly. Not purchasable, no price, but on paper this is the most biomechanically ambitious humanoid China has shown. Best for: Watching China's most ambitious hardware program; early industrial inspection Pros: Highest credible DOF count; solid-state battery first; automaker-grade manufacturing and in-house chips Cons: Not for sale; 2024-generation specs still circulate as if current; June 2026 leadership churn adds execution risk View XPENG IRON on Robozaps 14. Galbot G1 — Best Real-World Retail Deployment Manufacturer: Galbot (Beijing, China) | ~$800M raised; $3B valuation (Dec 2025), further ¥2.5B round reported Mar 2026 | CATL is investor and deployment partner No humanoid company runs more unmanned commerce than Galbot: its wheeled, dual-armed G1 semi-humanoid operates fully autonomous "Galbot Stores" in more than 30 Chinese cities (the company's own count), picking and dispensing goods around the clock, and has done factory stints at Mercedes-Benz and Zeekr plants. Its heavy-load sibling, the Galbot S1, has worked CATL's Ningde battery lines around the clock since spring 2026 under a scale-up agreement signed in June. Galbot now publishes a G1 hardware sheet: 173 cm, 92.5 kg, 24 driven joints, 5 kg rated payload per arm, up to eight hours of battery life and IP54 protection. It still publishes no list price, though a JD.com listing during the Spring Festival sales event showed roughly ¥630,000 (~$91,000). What is well documented is the money: a CATL-led ¥1.1B round in June 2025, $300M+ at a $3B valuation in December, and a reported ¥2.5B follow-on in March 2026. Best for: Unmanned retail, warehouse picking, pharmacy and store automation Pros: Most real revenue-generating retail deployments of any humanoid maker; CATL's money and factories behind it; wheeled base means all-day uptime Cons: No public list price; wheeled semi-humanoid rather than a walker; deployment counts are the company's own figures 15. RobotEra L7 — Fastest Humanoid Robot Manufacturer: RobotEra (Beijing, China) | ~¥3B raised across two 2026 rounds (incl. SF Group); valuation above ¥10B (~$1.4B) RobotEra's L7, unveiled July 2025, holds the unofficial humanoid speed record: 4 m/s (14.4 km/h), barefoot. Speed is the headline, but logistics is the business: L7 units are being delivered at thousands-of-units scale into 10+ logistics centers with China Post and SF Express, backed by an SF Group-led ~¥2B round in May 2026. Its predecessor STAR1 (171 cm, 65 kg, 55 DOF with 12-DOF XHAND1 hands) set its own speed mark first: 3.6 m/s across the Gobi Desert, in sneakers, in October 2024. Spec sheets that call STAR1 the "fastest walking humanoid" conflate the two robots and two different gaits. RobotEra publishes no prices; figures you see online ($50K–$120K) are guesses. Best for: Logistics, dynamic environments where speed matters Pros: Fastest humanoid on record; real logistics fleet deployments; strong 2026 funding Cons: No published pricing or datasheet; China-centric; STAR1-era specs circulate as if current 16. Astribot S1 — Most Dexterous Upper Body Manufacturer: Astribot / Stardust Intelligence (Shenzhen, China) | Series B above ¥1B closed June 2026 at ~$1.4B valuation Astribot's S1 earned its reputation with demos most robots can't touch: pouring, ironing, calligraphy, and object flips at up to 10 m/s end-effector speed with a 5 kg rated payload per arm, both company claims, but ones the demo footage supports. The cable-driven design and DuoCore fast/slow architecture trade legs for arm performance (S1 rides a stand or wheeled base). The commercial story caught up in 2025-26: thousand-unit-level deliveries by end of 2025, an alliance with Seer Robotics to put 1,000+ humanoids into industrial use, a thousand-unit Thundersoft deal that includes overseas markets, and a June 2026 Series B above ¥1 billion. A cheaper sibling, the T1, starts at ¥89,900 (~$12,600). Astribot now publishes the core S1 body figures: 170 cm, 80 kg, 5 kg rated payload per arm and a four-to-six-hour battery range. Deployment-volume claims still come from the company and its partners. Best for: Dexterous manipulation, food prep, service tasks, light manufacturing Pros: Best-in-class arm speed and dexterity demos; real volume deals; well funded Cons: Deployment-volume claims are company-reported; upper-body-first design isn't a walking humanoid in most configurations View Astribot S1 on Robozaps 17. Unitree R1 — Cheapest Humanoid From a Major Manufacturer Manufacturer: Unitree Robotics (Hangzhou, China) | Launched July 25, 2025 at ¥39,900 The R1 is the cheapest confirmed-orderable bipedal humanoid from a major manufacturer, and it is shipping rather than waiting on a presale. Official tiers: R1 AIR at $4,900 and R1 at $5,900 (an R1 EDU is quote-only, and the R1-D dual-arm developer platform starts at $4,290, though that one is a manipulation platform, not a walker). At 123 cm and roughly 27-29 kg, with 20-26 degrees of freedom (up to 40 with options) and about an hour of battery, it's a serious entry point for schools and developers. Bumi is announced lower at ¥9,998, but it remains a presale with customer delivery unconfirmed. The R1's stronger claim is: full Unitree ecosystem support, worldwide shipping, and a hardware pedigree the ultra-budget robots can't match. Best for: Education, hobbyists, entry-level research, entertainment Pros: Under $5K from a major maker; shipping today; Unitree SDK and community; genuinely bipedal Cons: ~1-hour battery; limited payload and autonomy; the cheapest tier (AIR) trims sensors View Unitree R1 on Robozaps · Cheapest humanoids guide 18. LimX Dynamics Oli — Full-Size Price Breaker Manufacturer: LimX Dynamics (Shenzhen, China) | $200M Series B (Feb 2026: JD, NIO Capital, SAIC's Shangqi, CoStone) Oli reset the price bar for full-size humanoids: from ¥158,000 (~$21,800) at its July 2025 launch, in three trims (Lite, EDU, Super). The official spec sheet is refreshingly concrete: 165 cm, up to 55 kg, 31 degrees of freedom (excluding end-effectors), 5 km/h top speed, 3 kg per arm, roughly 2 hours per 9,500 mAh pack, with a larger 175 cm / 43-DOF configuration alongside, and 2-finger grippers or 5-finger hands on the end. The SDK is fully open, from low-level joint control to task scheduling, which is why labs order it through international distributors like RobotShop. A $200M Series B landed in February 2026; the concrete multi-thousand-unit delivery plan is aimed at the Middle East, while US expansion remains at the talks stage. Best for: Research labs and developers who want a full-size biped without a six-figure invoice Pros: Published specs and a real starting price; fully open SDK; fresh $200M to deliver on orders Cons: 3 kg arms are research-grade; ~2-hour battery; US availability still informal 19. Dobot Atom — Precision at a Buyable Price Manufacturer: Dobot Robotics (Shenzhen, China) | Publicly listed collaborative-robot maker Atom is what happens when an established cobot manufacturer builds a humanoid: 28 degrees of freedom, five-fingered hands, a claimed ±0.05 mm repeatability inherited from Dobot's arm business, and a "straight-knee" gait the company says cuts walking energy by 42%. It debuted in China at ¥199,000 (~$27,500) in March 2025 and, unusually for this list, has an official US store price: $79,000, add-to-cart. Dobot has run three "mass delivery" batches since its first global handover event in Nagoya in June 2025 (unit counts undisclosed), with deployments spanning industrial assembly validation and a Shenzhen cinema's service floor. Note the height: Dobot hasn't published a definitive figure in English, press reports range 1.53-1.65 m, and the "41 DOF" number circulating online contradicts the official 28. Best for: Precision assembly pilots, integrators who already trust Dobot arms Pros: Actually buyable outside China with a posted price; cobot-grade precision claims; listed-company stability Cons: US price is 3x the China price; batch quantities undisclosed; runtime and payload unpublished 20. Leju Robotics Kuavo 5 — Fastest Humanoid Factory Manufacturer: Leju Robotics (Shenzhen, China) | Huawei ecosystem: Pangu embodied AI (official), KaihongOS (reported) | ¥1.5B raised in 2025 Leju's headline isn't the robot, it's the production line. The Foshan plant it opened with contract manufacturer Dongfang Precision on March 29, 2026 turns out roughly one humanoid every 30 minutes, with 10,000 units/year of capacity across 24 assembly stages and 77 inspection points, the first automated humanoid line at that cadence anywhere. Customers already taking Kuavo units include NIO, FAW Group, and Haier. The current generation is the Kuavo 5 (and wheeled 5-W), launched October 2025 with modular limbs that convert between bipedal and wheeled configurations, plus the earlier 4Pro. Leju publishes no official spec sheet; press-reported figures (~360 N·m joints, 4.6 km/h, ~63.5 kg, ~10 kg dual-arm payload, international pricing around $50,000) should be read as unofficial. Its Huawei Pangu integration, shown at HDC 2024, is the deepest Huawei tie-in in humanoids. Best for: Volume industrial buyers in the Huawei ecosystem Pros: World's fastest humanoid production line; named automotive and appliance customers; Huawei AI stack Cons: No official spec sheet; China-centric; brand less known outside industrial circles 21. Tiangong 3.0 — Open-Source State Champion Manufacturer: X-Humanoid / Beijing Humanoid Robot Innovation Center (Beijing, China) | ~$100M commercial financing (early 2026) Tiangong is China's state-backed open-source flagship. Its predecessor, Tiangong Ultra, won the world's first humanoid half-marathon in April 2025 (2:40:42 over 21.1 km); Tiangong 3.0, unveiled February 10, 2026, vaulted a meter-high obstacle one-handed at its debut and claims industry-first touch-interactive whole-body motion control. In April 2026 it won the Beijing Robot Warrior Challenge running fully autonomously. The 169 cm, 62 kg, 43-DOF platform open-sources what commercial rivals guard: physical design blueprints, the Pelican-VL vision-language model, and the RoboMIND dataset, with 100+ open-source units seeding the largest embodied-AI community in China. Mass delivery is planned for the second half of 2026; there's no published price, this is a developer platform with national backing, not a consumer product. Best for: Research groups building on open humanoid hardware and Chinese-ecosystem AI Pros: Fully open blueprints and models; proven athletic milestones; state-institute resources Cons: No price or consumer channel; payload and battery unpublished; institute-driven roadmap 22. EngineAI SE01 — Affordable Full-Size Platform Manufacturer: EngineAI (Shenzhen, China) The SE01 (170 cm, ~55 kg, 32 degrees of freedom, 2 m/s walk, quick-release battery) launched in December 2024 at a unified ¥88,000 (~$12,000) for both commercial and education versions, one of the most aggressive full-size prices ever posted, though EngineAI currently lists no public price. Its bio-inspired, neural-network gait is among the most natural in the class. EngineAI's lineup has since widened: the PM01 (138 cm, ~$13,700 reported) performed the world's first humanoid standing front flip in February 2025, and the T800 (173 cm, 29 DOF excluding hands, 450 N·m joints) is now the flagship, its December 2025 martial-arts video went viral enough that the CEO took a kick from the robot on camera to rebut CGI accusations. Best for: Research, commercial development, locomotion work Pros: Full-size at a breakthrough launch price; standout natural gait; fast-moving lineup (PM01, T800) Cons: Current pricing opaque; payload unpublished; viral-video reputation outruns documented deployments 23. Booster Robotics T1 — Best for Robotics Education & Competition Manufacturer: Booster Robotics (Beijing, China) | Tsinghua spinout The T1 is the platform to beat in competitive robotics: Tsinghua's "Hephaestus" team won the RoboCup 2025 AdultSize Humanoid League on it, and Booster says 50 teams now build on the platform. The official spec sheet: 118 cm, ~30 kg, 23 degrees of freedom standard (up to 41 with dexterous hands), 130 N·m peak joints, Jetson AGX Orin at 200 TOPS, and roughly 2 hours of walking per charge, a crash-tolerant frame built to be knocked over and get up. Booster publishes no direct price; internationally it ships through distributors (Generation Robots in the EU lists it from €39,000 including tax, RobotShop carries three configs in North America). The smaller K1, whose platforms took first and second at RoboCup KidSize, is priced officially from $5,999 and is the better classroom entry point. Best for: University competition teams, robotics courses, RL research Pros: Championship pedigree; crash-and-recover durability; serious onboard compute; global distributor network Cons: No direct/official pricing; sub-full-size; payload unpublished 24. NEURA Robotics 4NE1 — Europe's Flagship Humanoid Manufacturer: NEURA Robotics (Metzingen, Germany) | Series C up to $1.4B led by Tether (June 2026), ~$7B valuation, >$1B claimed order book Europe's biggest humanoid bet got dramatically bigger in June 2026 when NEURA raised up to $1.4 billion led by Tether with NVIDIA, Amazon, Qualcomm, Bosch, Schaeffler and the EIB participating. The 4NE1 Gen 3.5 (180 cm, 80 kg, payload up to 100 kg, 5 km/h), designed in collaboration with Studio F.A. Porsche, carries NEURA's patented Sensor Skin for proximity-aware safety and runs NVIDIA Isaac GR00T under the Neuraverse operating layer. NEURA publishes estimated reservation pricing of €98,000 per unit for 1–19 units and €60,000 each at 20+, excluding tax and shipping, with a €100 refundable reservation; the smaller 4NE1 Mini (~132 cm) starts at €19,999 (Pro €29,999 with 12-DOF hands and SDK). Everything is reservation-only today, availability slipped from "as early as June" to end of 2026, against a 6,000-unit production target. Best for: European industrial buyers, fleet deployments, the highest payload ambitions Pros: Transparent price ladder; up-to-100 kg payload claim leads the class; Porsche-studio design; enormous new war chest Cons: Nothing delivered yet and the June 2026 ship date already slipped; payload figure is a range, not a constant; Tether-led financing raised eyebrows 25. MagicLab MagicBot Z1 — China's Consumer Breakout Manufacturer: MagicLab (Suzhou, China) Six MagicBot Z1s dancing beside pop stars at the 2026 Spring Festival Gala, 1.2 billion viewers, turned MagicLab into a household name overnight: its JD.com listing sold out within minutes of going live on New Year's Eve, and JD reported robot searches up 300% within two hours. Neither the price nor the units sold were ever published, so skip the numbers you may have seen. The Z1 itself (official spec) is a compact 1.37 m, ~40 kg biped that folds to 73 cm for transport: 24 DOF, 2 kg arm load, a 10,000 mAh quick-release battery good for about 2 hours, and a sensor head with 3D LiDAR, depth and fisheye cameras plus tactile sensing. Its industrial sibling, MagicBot Gen1, has worked electronics-factory lines, inspection, material handling, barcode scanning, since December 2024. Best for: Consumer early adopters in China, entertainment, light factory work (Gen1) Pros: Proven mainstream demand; portable folding design; rich sensor head; factory-tested sibling Cons: Price and availability opaque; 2 kg arms; China-only channel 26. Hexagon AEON — Industrial Wheeled-Legged Contender Manufacturer: Hexagon Robotics, a division of Hexagon AB (Sweden; robotics division based in Zurich) AEON, launched June 17, 2025 at Hexagon LIVE in Las Vegas, is a wheeled-legged "semi-humanoid" built by the measurement-technology giant Hexagon AB, not, as often misreported, a German startup. Launch pilots went to Schaeffler and Pilatus Aircraft; BMW began testing AEON at Plant Leipzig in December 2025 (battery assembly and components), with a formal pilot phase from summer 2026. The anchor deal is Schaeffler's: a fleet of at least 1,000 AEONs by 2032, with rollouts starting from the end of 2026, Schaeffler also supplies AEON's actuators. Hexagon's official material highlights a battery-swap mechanism so the robot "does not need to recharge"; press-reported specs (~165 cm, ~60 kg, 34 DOF, 8.6 km/h, 15 kg payload) haven't been formally published. Best for: Automotive and precision manufacturing, intralogistics Pros: Backed by a profitable industrial giant; the biggest committed fleet order in Europe (Schaeffler); BMW validation underway Cons: Wheeled base limits terrain; spec sheet mostly unpublished; pilot-stage, not open sales 27. Humanoid HMND 01 — UK's Industrial Challenger Manufacturer: Humanoid Ltd (London, UK) | Founded 2024 by Artem Sokolov | $100M invested (founder-funded) Humanoid built its first robot, the HMND 01 Alpha, in seven months, a 220 cm wheeled mobile manipulator with 29 active degrees of freedom, a 15 kg dual-arm payload and 7.2 km/h travel, with a bipedal variant in development. The company bills it as the UK's first industrial humanoid (its own claim) and runs it in warehouse, logistics, and retail pilots on a robots-as-a-service model. The serious signal came in May 2026: Bosch will contract-manufacture HMND 01 for Europe, with Schaeffler supplying components and a Beta generation due Q3 2026. Founder Sokolov has told Reuters the company holds 34,000 pre-orders, a company figure worth attributing rather than repeating, and has floated a 2029/30 IPO. Best for: European warehouse and retail automation pilots Pros: Bosch manufacturing deal de-risks scale-up; fast development cadence; wheeled + bipedal roadmap Cons: Pre-order figures are company-sourced; no published pricing; young track record 28. Engineered Arts Ameca — Best Human Interaction Manufacturer: Engineered Arts (Falmouth, UK) Ameca remains the robot people remember meeting: hyper-expressive face, eye contact, timing, and GPT-backed conversation running on Engineered Arts' Tritium platform. The company's own billing is "the world's most advanced social humanoid robot," and for exhibitions, brand activations, and human-robot-interaction research it has no real rival. Ameca Gen 3 is current; it still doesn't walk (legs remain prototype work). Specs from the official page: 187 cm and 62 kg, taller than most databases list. There is no published price: Engineered Arts sells and rents by quote, and the price ranges circulating online (from $100K to $500K) are guesses we don't endorse. Best for: Exhibitions, HRI research, hospitality, broadcast Pros: Unmatched expressiveness; proven at hundreds of events; mature software platform Cons: Cannot walk or do physical work; quote-only pricing; interaction specialist, not a laborer 29. Noetix Bumi — Cheapest Announced Presale Humanoid (¥9,998) Manufacturer: Noetix Robotics (Beijing, China) | Founded 2023 by Tsinghua PhD dropout Jiang Zheyuan | ~¥1B Series B (Feb 2026) led by CATL-affiliated Chen Dao Capital Bumi holds the affordability crown: ¥9,998 (about $1,400) for a genuinely walking, dancing 94 cm / 12 kg humanoid with 21 joints, drag-and-drop programming for kids and an open API for developers. Presale opened October 22, 2025, Noetix reported 100 units sold in the first hour and 500 in two days on JD.com, and Bumi danced at the CCTV Spring Festival Gala in February. Status check as of July 2026: JD listings scheduled first deliveries from March 2026, and we found no independent confirmation of units in customers' hands yet, treat it as "ordered, shipping ramp unproven." It remains China-only officially. Noetix's older N2 (¥39,900, 118 cm) finished second in the world's first humanoid half-marathon in April 2025, and the company says future models could reach ~¥5,000 (~$700). Best for: K-12 STEM, hobbyists, family entertainment, programming education Pros: Consumer-electronics pricing; real bipedal walking and running; beginner-friendly software; CATL-backed balance sheet Cons: Delivery track record unproven; 1-2 h battery; China-only; toy-scale payload Full Noetix Bumi review 30. Rainbow Robotics RB-Y1 — Research Dual-Arm Platform Manufacturer: Rainbow Robotics (Daejeon, South Korea) | KOSDAQ 277810 | Samsung is largest shareholder (35%) From the KAIST lab that won the 2015 DARPA Robotics Challenge, the RB-Y1 is a wheeled, dual-armed mobile manipulator: 24 DOF total (two 7-DOF arms, a 6-DOF articulated torso-leg, grippers and wheels), 3 kg per arm, 1.5 m/s travel and a 1,270 Wh pack, with a master-arm teaching system for programming by demonstration. Unusually for this list, it launched with published pricing: $80,000 research / $120,000 commercial at May 2024 pre-order (quote-based today). The bigger story is the owner: Samsung Electronics raised its stake to 35% in December 2024, consolidated Rainbow as a subsidiary, and runs a joint "physical AI" program targeting humanoid commercialization around 2028. A US subsidiary in Chicago serves the research market; RB-Y1 exhibited in the K-Humanoid pavilion at CES 2026. Best for: University labs, dual-arm manipulation research, Korean-ecosystem developers Pros: DARPA-winning pedigree; Samsung's backing and roadmap; teach-by-demonstration; real published pricing Cons: Wheeled only; 3 kg arms limit industrial use; 131 kg mass View RB-Y1 on Robozaps 31. AgiBot X1 — Best Open-Source Research Humanoid Manufacturer: AgiBot / Zhiyuan Robotics (Shanghai, China) The X1 is the rare humanoid you can rebuild from the ground up: full-stack open source, hardware documentation, source code, and the AimRT robotics framework, on a 130 cm, 33 kg bipedal platform with 34 active degrees of freedom and the same PowerFlow actuator lineage as AgiBot's commercial robots (official specs). It walks at 1 m/s, runs about two hours, and carries 0.5 kg per arm, numbers that make it a teaching and research tool, not a worker. That's the point: for university labs that need to modify everything from motor control to the AI stack without black boxes, the X1 and the AGIBOT World dataset around it are the most open package in humanoid robotics. Best for: Academic research, robotics courses, embodied-AI experiments Pros: Genuinely open hardware and software; commercial-grade actuators; large dataset ecosystem Cons: Tiny payload; 2-hour battery; research scale, not production 32. Promobot V.4 — Service & Concierge Veteran Manufacturer: Promobot (Perm, Russia; US office in New York) The V.4 is the incumbent of service robotics: a wheeled humanoid for reception, concierge, museum, and healthcare-intake work with facial recognition, document scanning, and payment processing. Partner materials cite 800+ units across dozens of countries (press from 2023 said 750+ in 43); treat the counts as company-side figures. Pricing runs "from $25,000" with $2,000/day rental official. Two things worth knowing: production is entirely in Perm, Russia (the company's own FAQ; the US presence is a New York office, with Philadelphia a legacy service center), relevant for procurement policies, and the current lineup is V.4 plus the Robo-C2 desk android; there is no V.5. Against 2026's AI-native service humanoids, the V.4's stack shows its age, but its install base keeps it the proven choice. Best for: Hotel and clinic front desks, museums, guided tours Pros: Largest service-robot install base in its class; modular commercial features; rentable Cons: Russian production complicates Western procurement; AI stack dated vs 2026 rivals; unit counts are company claims 33. Fauna Sprout — Amazon's Home-Robot Acquisition Manufacturer: Fauna Robotics (New York, USA), acquired by Amazon, March 2026 Sprout, the "Creator Edition" developer humanoid launched in January 2026 (107 cm, 22.7 kg, 29 DOF, Jetson AGX Orin, 3-3.5 h battery), was designed for shared human spaces and priced around $50,000 (CEO-stated). Its launch customer list, Disney, Boston Dynamics, UC San Diego, NYU, signaled unusual credibility for a startup founded in 2024 by CTRL-labs and DeepMind alumni. Then the twist: Amazon acquired Fauna in March 2026 (terms undisclosed). The site still takes purchase requests, but third-party availability from here is uncertain, the acquisition reads as Amazon buying a home-robotics team and platform. We keep Sprout listed for its technical merit and as the clearest signal yet that Big Tech wants consumer humanoids. Best for: Developers and labs building for human spaces; Amazon-watchers Pros: Elite early customers; polished developer platform; now Amazon-resourced Cons: Availability now hostage to Amazon strategy; $50K for a sub-half-size robot; young platform 34. PAL Robotics ARI — Social Humanoid for European Service Manufacturer: PAL Robotics (Barcelona, Spain) | Humanoids since REEM-C (2013) ARI is the workhorse of European social robotics: a 165 cm wheeled humanoid built for reception, wayfinding, and patient engagement, with a mature ROS stack that makes it a favorite in EU research programs, it served in the SPRING project at Broca hospital in Paris and multiple Horizon health projects. PAL publishes full documentation but no price; EU distributor listings cluster in the mid-$30Ks. PAL's two-decade humanoid lineage (REEM-C, the torque-controlled TALOS, the Kangaroo research biped) makes ARI less a gadget than the entry point to a serious platform family for labs that want support and spare parts on this continent. Best for: Healthcare reception, EU research consortia, HRI studies Pros: Proven in European hospitals; best-documented platform in its class; real vendor support Cons: No manipulation to speak of; wheeled; quote-only pricing 35. DEEP Robotics DR02 — Best for Outdoor & Extreme Environments Manufacturer: DEEP Robotics (Hangzhou, China) | Unveiled October 2025 The DR02 goes where other humanoids can't: it's the first humanoid with an IP66 rating (the maker's claim, unchallenged so far), sealed against rain and dust and rated for -20°C to +55°C. The 175 cm, 65 kg robot walks at 1.5 m/s (4 m/s max), carries 10 kg per arm, runs 275 TOPS of onboard compute, and uses modular quick-detach components for field maintenance (official page). DEEP Robotics earned its field credentials with quadrupeds patrolling power grids and construction sites; the DR02 extends that playbook to construction, energy, and mining humanoid work. Pricing is contact-sales. Best for: Outdoor inspection, power, construction, mining Pros: Only weather-sealed humanoid; proven outdoor-robotics parent; field-serviceable design Cons: No public price; indoor dexterity rivals beat it on manipulation; young platform 36. Xiaomi CyberOne (V2) — Watch-Tier: The Sleeping Giant Stirred Manufacturer: Xiaomi (Beijing, China) CyberOne looked dormant after its 2022 debut (177 cm, 52 kg, 21 DOF, emotion recognition; Lei Jun put its unit cost at ¥600,000-700,000, ~$90-105K, a production cost, not a price). Then in April 2026 Xiaomi showed CyberOne V2 at its investor day: hands shrunk 60% to true human scale with 22-27 DOF, a "bionic sweat gland" liquid-cooling system, and a reported 90.2% success rate in its own auto-assembly pilots. There is still no price and no commercial launch date, which is why this is a watch-tier entry, not a buy recommendation. But a top-three phone maker with EV factories re-entering humanoids is exactly the kind of thing that reshuffles this list. Best for: Watching Xiaomi's manufacturing-scale ambitions Pros: V2 hands and cooling are real engineering; Xiaomi's factories could scale fast; EV supply chain synergy Cons: Unbuyable; no roadmap to sale; four quiet years between generations View CyberOne on Robozaps 37. DroidUp Moya — The Biomimetic Moonshot Manufacturer: DroidUp / Shanghai Zhuoyide (Shanghai, China) | Founded 2021 Moya chases a different goal from every other robot here: feeling human. Temperature-regulated silicone skin (32-36°C), 25 facial drive units for micro-expressions on a 165 cm, 32 kg frame, and DroidUp's own claim of 92% human-gait likeness, the company bills it as the world's first fully biomimetic embodied-intelligence robot. It rides the Walker 3 skeleton, whose earlier version took bronze at the April 2025 Beijing humanoid half-marathon. First-batch pricing runs ¥1.2-1.5M (~$170-210K) for about 50 units, with presale opening Q4 2026. Until robots ship, treat Moya as an announced product with striking demos: premium hospitality, eldercare, and HRI research are the target markets, uncanny valley included. Best for: Premium hospitality, eldercare pilots, HRI research Pros: Unique warm-skin biomimicry; real locomotion pedigree; state-backed Shanghai funding Cons: Nothing shipped; luxury pricing; gait and expression metrics are vendor-audited only 38. LG CLOiD — The 2028 Home Bet Manufacturer: LG Electronics (Seoul, South Korea) CLOiD headlined LG's "Zero Labor Home" vision at CES 2026: a wheeled home humanoid with a tilting torso, dual 7-DOF arms and five-fingered hands that started laundry cycles, folded garments, and loaded ovens on the show floor, slowly, as hands-on reviews noted, but really. It runs LG's "Affectionate Intelligence" models and plugs into the ThinQ ON smart-home hub. LG is treating it as a program, not a demo: a Yangjae "robot data factory" built with NVIDIA opened in June to train it, a CEO-level Robotics Business Center stood up July 1, and the plan sends CLOiD to factories first with home commercialization around 2028. No price, no orders, it's here because a top-tier appliance giant just committed its org chart to home humanoids. Best for: Tracking where the appliance industry takes home robots Pros: Real household task demos; LG manufacturing and retail muscle; NVIDIA-backed training pipeline Cons: Not a product; wheeled and slow; 2028 is a long way off Dropped From This Ranking (and Why) Sanctuary AI Phoenix: ranked #6 here in early 2026; the company pivoted to hardware-agnostic robot software in June 2026 and Phoenix is now an internal R&D platform, not a product.1X EVE: the wheeled predecessor to NEO is no longer marketed; existing enterprise deployments are being maintained, not expanded.Unitree H1/H1-2: the $90,000 line is quote-only legacy hardware, superseded by the $29,900 H2 (covered at #8).K-Scale Labs and Cartwheel Robotics: shut down in November 2025 and February 2026 respectively, reminders that this market has a mortality rate.Foundation Phantom MK-1: reportedly the first humanoid sent to an active combat zone (Ukraine, February 2026), we exclude military deployments from a buying guide, and journalists who watched a controlled demo saw it fall repeatedly. How to Choose the Best Humanoid Robot for Your Needs By Use Case Factory & manufacturing: Figure 03 has the strongest AI-plus-dexterity package; Walker S2 and Kepler K2 are the volume plays you can order today; Atlas is the hardware benchmark, although no open-ordering date has been announced; Apollo 2 is collecting industrial training data in bipedal and wheeled configurations.Warehouse & logistics: Digit is the proven leader by operating hours; RobotEra L7 fleets are scaling fast in China; Galbot G1 owns retail-adjacent picking.Care & healthcare: Fourier GR-3 is the only humanoid built by a company with a real clinical install base; PAL ARI for EU hospital reception and research.Research & education: Unitree G1 EDU is the default lab platform; AgiBot X1 and Tiangong 3.0 for open-source depth; Booster T1 for competition; LimX Oli for a full-size biped on a budget; Unitree R1 or Booster K1 ($5,999) for classrooms.Customer service & hospitality: Ameca for wow-factor interaction; AgiBot A2 for AI-native service at volume; Promobot V.4 for proven concierge installs.Home: 1X NEO is the leading home preorder (US deliveries are promised in 2026, but no customer deliveries are confirmed); CLOiD, Figure's alpha and Optimus remain 2027-2028 stories. By Budget Use price as a shortlist filter, not as the ranking itself. The cheapest humanoid robots ranking owns the low-cost shortlist; the humanoid robot cost guide owns price ranges, total cost and cost drivers. This page ranks the strongest robots by evidence and use case. Humanoid Robot Market in 2026: What's Actually Happening Roughly 13,000 humanoid robots shipped in 2025 (Omdia), and the market that MarketsandMarkets valued at $2.03B in 2024 is projected to reach $13.25B by 2029, with the same firm's longer forecast now stretching to $50B by 2035. Five things define mid-2026: Factories, Not Demos Dedicated humanoid production lines now run on three continents: Figure's BotQ (up to 12,000/year), Agility's RoboFab (scalable past 10,000/year), Leju's Foshan line (one robot every ~30 minutes), UBTECH's 5,000/year plan for Walker S2, Tesla's converted Fremont line, and Eyou's automated joint factory in Shanghai feeding the supply chain. AgiBot reported its 10,000th robot in March. The bottleneck is shifting from "can we build them" to "what can they reliably do." The Money Got Serious Figure holds a $39B valuation; Apptronik closed a $935M Series A; NEURA raised up to $1.4B led by Tether at ~$7B; Agility is going public via SPAC at ~$2.5B; Unitree's IPO cleared approval in July with a ~$5.9B valuation; Galbot, Astribot, RobotEra, LimX and Noetix all closed nine-figure rounds inside twelve months. The shakeout is equally real: K-Scale Labs shut down in November 2025, Cartwheel Robotics in February 2026, Sanctuary AI pivoted to software in June, and Amazon absorbed Fauna Robotics in March. Geopolitics Arrived Chinese makers dominated 2025 shipments (Omdia puts AgiBot first at 39% share; Unitree self-reports higher), while US companies lead on AI and capital. Washington responded: the American Security Robotics Act (Cotton-Schumer, March 2026) would bar federal use of foreign-adversary robots, naming Unitree and AgiBot in sponsor statements, with companion House bills following. If you're a US government buyer, country of origin is now a spec-sheet line. Homes Are Close, Not Here Nobody has put an autonomous humanoid in a customer's home yet. NEO has 10,000 planned production slots booked and a promise of deliveries by end of 2026; Figure runs in-home alpha tests; LG targets ~2028; Tesla says public Optimus sales are "likely" by end of 2027. The honest label for 2026 is that home humanoids became orderable, not that they arrived. How to Use This Ranking This is an editorial ranking, not an availability catalogue. Use it to build a shortlist based on deployment evidence, hardware, autonomy and fit. Then check the dedicated humanoid robots for sale guide for verified orderability, buying channels and procurement checks. Conclusion: Shipping Beat Shouting in 2026 The 38 robots above tell a simple story: the humanoid industry graduated from demo videos to delivery schedules. Digit crossed 100,000 totes, AgiBot reported 10,000 robots by March, Walker S2 booked nine-figure orders, and you can now buy a walking humanoid for less than a used car, while the ¥9,998 Bumi remains an announced presale with customer delivery unconfirmed. The failures are part of the same story: two startups shut down, one pivoted to software, and the most famous robot on the list still hasn't been revealed. Our advice holds across budgets: buy on verified deployments and published specs, not on demo reels. Where a maker won't publish a price or a datasheet, we've said so, and where a number is the company's own claim, we've labeled it. That's also how we'd read every other ranking on the internet. We update this guide as robots launch and prices move; the core ranking facts were re-verified against manufacturer pages and primary sources on July 13, 2026. For corrections: dean@robozaps.com. For individual reviews and buying guidance, explore blog.robozaps.com or browse humanoid robots for sale on Robozaps. For the company landscape behind these machines, see our humanoid robot companies ranking; for what they do in the field, the applications of humanoid robots; and for the latest changes, the news tracker. Prefer four legs? See the best robot dogs of 2026. --- # Tesla Optimus Gen 3 (V3): Release Date, Price, Specs & Production URL: https://blog.robozaps.com/b/tesla-optimus-gen-3 Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-10T01:24:56.191Z Last fact-checked: 2026-08-10T00:00:00.000Z Category: insights TL;DR: As of August 10, 2026, Tesla has not revealed Optimus Gen 3 or announced a release date, price or preorder. The July 22 Q2 update says first-generation Optimus lines are being installed at Fremont and production is "anticipated later this year." Musk said the ramp will be "quite flat and long." His public-sale target remains the end of 2027. Key takeaways: - Gen 3, Optimus 3 and V3 name the same robot. As of August 10, 2026, Tesla has not revealed it, and online "Gen 3 spec sheets" still recycle Gen 2 figures or unsupported claims. - The reveal slipped from Q1 2026 to "middle of this year." Tesla's July 22 update and call set no new date, and the company has said it is withholding the design until production is close. - Tesla's July 22 Q2 deck says first-generation Optimus lines are being installed at Fremont and production is "anticipated later this year." As of August 10, Tesla has not announced that production has started. - There is no price: $20,000 to $30,000 is Musk’s long-term consumer target, roughly $20,000 is a production-cost goal at a million units a year, and public sales are pointed at end of 2027. - The only V3 hardware figure Musk has given is the hand, 50 actuators per robot (25 per forearm and hand); the patented hand design was publicly scrapped in April 2026, so treat even that figure as directional. - The forecast record argues for discounting every date: 10,000 robots promised for 2025 became a few hundred to roughly 1,000 by external estimates, the 50,000-unit 2026 target became "impossible to predict," and the Q1 reveal slipped to mid-year. FAQ: Q: Is Optimus Gen 3 the same as Optimus V3? A: Yes. Optimus Gen 3, Optimus 3 and Optimus V3 all refer to the same ground-up redesign of Tesla’s humanoid robot. The confusion exists because "Gen 3" was earlier used for the upgraded 22-degree-of-freedom hands shown in November 2024 on the old Gen 2 body, a platform Musk himself called Optimus 2.5. Q: When will Tesla Optimus Gen 3 be revealed? A: There is no official reveal date. Tesla missed its original Q1 2026 target, said "probably middle of this year" in April, then set no new date in the July 22 Q2 update and call. Specific late-July or August dates circulating online are rumors, not Tesla announcements. Q: What is the Tesla Optimus Gen 3 release date? A: Tesla has not announced a release date. Its July 22 Q2 deck says first-generation Optimus lines are being installed at Fremont and production is "anticipated later this year." That refers to internal production, not retail availability. Musk said at Davos in January 2026 that Optimus would likely be available to the public by the end of 2027. Q: How much will Tesla Optimus Gen 3 cost? A: There is no price. Musk’s long-term consumer target is $20,000 to $30,000 (October 2024), his production-cost goal is around $20,000 per unit at a million units a year, and he has said the actual selling price "will be set by the market demand." Any site quoting a firm Optimus 3 price is guessing. Q: Can you buy or pre-order a Tesla Optimus? A: No. As of August 10, 2026, Tesla has not opened Optimus preorders or reservations. Security researchers have documented fake Tesla websites collecting $250 "Optimus deposits" through search ads; any site taking money for an Optimus should be treated as a scam unless the offer appears on tesla.com. Q: What are the Tesla Optimus Gen 3 specs? A: Tesla has published none: no height, weight, payload, battery, speed or joint count. Musk has given one hardware figure, a 50-actuator hand architecture (25 per forearm and hand), whose mechanical design has already changed at least once, plus a rough 10,000 unique-parts count. The program lead’s April 2026 silhouette suggests, but does not confirm, a soft outer covering. Spec sheets listing $30,000, 180 cm, 45 kg or an IP68 rating are fabricated. Q: What chip does Optimus Gen 3 use? A: Undisclosed. Tesla positions its AI5 chip as the Optimus brain, but AI5 only taped out in April 2026 and reaches volume production in 2027, so the first Optimus 3 units cannot ship with volume AI5 silicon. The AI software stack derives from Tesla’s Full Self-Driving work, and Musk has confirmed Grok handles voice interaction. Q: Who leads the Tesla Optimus program? A: Ashok Elluswamy, who also heads Tesla's Autopilot AI teams, has led Optimus since Milan Kovac left in June 2025. He remained publicly active on the program in July 2026, posting that Optimus has "big shoes to fill" on the converted Model S/X line. Q: How many Optimus robots exist today? A: Tesla has never published a count. External estimates put total Optimus output through 2025 at a few hundred units or fewer, and Musk said in January 2026 that the robots were "not in usage in our factories in a material way." Claims of a 1,000-robot working fleet are unsourced aggregator content. Q: What is Optimus 2.5? A: Musk’s own label for the interim platform: the existing Gen 2 body fitted with the upgraded 22-degree-of-freedom hands demonstrated in November 2024. Much of the "Gen 3" coverage from that period actually describes Optimus 2.5, not the V3 redesign. Q: What will Optimus 3 look like? A: Tesla has shown only a silhouette, via engineering leader Konstantinos Laskaris at an April 2026 ETH Zurich keynote: human proportions, thicker forearms and no exposed mechanics, which suggests but does not confirm a soft outer covering. Musk’s description is that it "won’t even seem like a robot. It’ll seem like a person in a robot suit." Photos captioned "Optimus Gen 3" today show earlier-generation bodies. Q: Can you invest in Optimus? A: Not directly. Optimus is a Tesla program, not a separate company, so the only public exposure is Tesla stock (TSLA). Any pitch offering "Optimus shares" or pre-IPO access is a scam, as are the deposit sites offering pre-orders. What is Tesla Optimus Gen 3, and is it the same as Optimus V3? Yes: Optimus Gen 3, Optimus 3 and Optimus V3 name the same ground-up redesign of Tesla's humanoid robot. As of August 10, 2026, Tesla has not publicly shown it or announced a release date, price, preorder or official spec sheet. The latest official update says Tesla is installing first-generation Optimus lines at Fremont and anticipates production later this year, without a more specific date, in the July 22 Q2 shareholder deck. One disclosure up front: this is desk research built on the sources linked throughout, primary wherever they exist; Tesla has not publicly shown this robot. The misinformation audit: what circulates vs the record The Optimus Gen 3 search results are a case study in fabricated certainty. Before this page leans on anything, here is the audit. Optimus 3, V3, Gen 3, Optimus 2.5: the naming decoder "Gen 3" has meant two different things, and most published spec sheets mix them up. Through 2024 and early 2025, coverage used "Gen 3" for the upgraded hands Tesla demonstrated in November 2024: 22 degrees of freedom per hand, three more in the wrist and forearm, actuation moved into the forearm, mounted on the existing Gen 2 body. A Tesla engineer later confirmed that demo was teleoperated. Musk gave the hands-upgraded platform its own label on the Q2 2025 earnings call: "we are on Optimus version two right now, sort of two and a half" (transcript). Since mid-2025, Optimus 3 (or V3) has meant the full redesign, the robot this page tracks. One recurring trap: in March 2026 Tesla displayed an Optimus at the AWE expo in Shanghai that much of the press captioned "Gen 3." CNEVPost covered it as the upcoming third-generation robot while noting the final version may differ; the unit visibly used established-generation body hardware to showcase the new hands, and Tesla has never presented it as the V3 reveal, which the company still describes as unshown. If an article shows you a photo of "Optimus Gen 3," it is almost certainly showing you a Gen 2-era body. The reveal saga: every statement, dated October 22, 2025. On the Q3 earnings call, Musk set the first target: "We look forward to unveiling Optimus V3, you know, probably in Q1" (Q3 2025 call). January 28, 2026. On the Q4 call (official webcast), Musk kept it close, saying Tesla would unveil Optimus 3 "in a few months." The same call announced the end of Model S/X production to free the line that would build it. March 31, 2026. Q1 ended without a reveal. Musk on X: the robot "is walking around, but needs some finishing touches before it's ready to be shown" (Not a Tesla App). April 22, 2026. On the Q1 call, the target moved to "probably middle of this year," and Musk explained the secrecy: competitors analyze everything Tesla shows "frame by frame" and copy it, so the design stays hidden until production is close (Electrek). July 1, 2026. Musk walked the newly installed Fremont line and posted a team photo (Teslarati), and shut down a fan theory that Tesla was secretly ahead: "No, Optimus production will be extremely slow at first, as everything is new. This is not like making a car" (Electrek). Where that leaves the date: Tesla missed its Q1 reveal target, said "probably middle of this year" in April, then set no new date in the July 22 update and call. Tesla has tied the reveal to production being close without committing to a day. Reports of a specific late-July reveal trace to the Teslascope account citing "internal communications," not a Tesla announcement. What Tesla has actually said about the robot Strip away the fabricated spec sheets and the verbatim record is short but vivid. From the Q2 2025 call: you have to "design every part of it from physics first principles," because "there's nothing off the shelf that actually works," and the new design has "all the degrees of freedom that you really want or need" (transcript; the popular "designed from scratch" line is a paraphrase of this). From the Q3 2025 call: it "won't even seem like a robot. It'll seem like a person in a robot suit," so real "you'll need to like poke it" (Q3 call). From the March 2026 Abundance Summit: Tesla is "in the final stages of completion of Optimus 3, which is really going to be by far the most advanced robot in the world" (Teslarati). From the Q1 2026 call: the robot has roughly 10,000 unique parts, none previously mass-produced, and Musk's larger frame, that Optimus will be "not just Tesla's biggest product ever, but probably the biggest product ever" (Q1 2026 call transcript). The closest thing to an official look is a silhouette shown by Tesla Optimus engineering leader Konstantinos Laskaris in an April 2026 ETH Zurich keynote: human proportions, thicker forearms, no exposed mechanics, which suggests but does not confirm a soft outer covering (Not a Tesla App). That silhouette, one hand figure, and a parts count are the entire Tesla-sourced hardware picture. The hand saga: 60% of the problem Musk has said the hand and forearm represent roughly 60 percent of the engineering difficulty of the whole robot, and the V3 hand's public history shows a design still in motion. Row sources beyond those linked in the table: the Gen 2 hands per the December 2023 reveal, the November 2025 statements. Net position: 50 actuators per robot is Musk's stated figure, the mechanism behind it has changed at least once since, and the shipping V3 hand has never been shown. Any article presenting the patent diagrams as the final hand is out of date by the patent-holder's own account. The scale of the ambition is real, though: 50 actuators in two forearms is roughly triple the actuator count of the hands shown on stage in November, and hands are exactly where supply-chain reporting said the program stumbled in 2025. Chips and AI: what will actually power it Tesla positions its AI5 chip as the Optimus brain, and the dated record lets you test that against physics. AI5 taped out around April 15, 2026, per Musk's own announcement, with volume production expected in 2027 and TSMC fabricating, per Electrek; Musk showed a first packaged sample that same week, covered by Tom's Hardware. His performance claims for it have ranged from "40 times better than AI4 by some metrics" (Q3 2025 call) to the more specific 8x compute and 9x memory framing in later statements. Do the arithmetic and one popular claim collapses: robots built in summer 2026 cannot ship with volume AI5 silicon that does not exist until 2027, and Tesla has never said which computer the first V3 units carry. Musk has said early AI5 supply goes to Optimus and data centers rather than cars, which tells you about 2027 allocation, not 2026 hardware. The first-unit compute therefore remains undisclosed; the plausible candidates are the current AI4 family or an interim revision, with engineering-sample AI5 an outside case. The software story is more settled: the AI stack derives from Tesla's driving work ("the real-world intelligence we've developed for the car, most of that transfers to Optimus," Q3 2025 call), and Musk confirmed in June 2025 that V3 uses Grok for voice. The skeptical footnote belongs beside it: the transfer thesis, car vision to robot manipulation, is an argument, not a demonstrated result, and Tesla publishes no autonomy metrics of any kind against which to test it. What Gen 3 must prove: the autonomy context The reveal will land on top of a mixed record. Tesla's current fleet has one factory task the company explicitly claims runs autonomously end-to-end (battery-cell sorting, May 2024), a set of 2025 showpieces it claims were learned behavior (dancing, kung fu), and a string of demonstrations that turned out to be teleoperated, two by Tesla's own admission, including the shirt-folding video and the Tesla Diner popcorn robot. The full demo-by-demo ledger lives in our Optimus Gen 2 review. What it means here: when V3 is revealed, the credibility question will not be whether it looks amazing, it will, but whether Tesla shows unedited, verifiable autonomy and publishes any metric at all. None of the commercial humanoid makers we track has published independently audited capability benchmarks; Tesla, whose public demo record contains more confirmed teleoperation than any rival we cover, has the most to gain from being first, and the least history of doing so. The Fremont story, in full Production is the half of the Gen 3 story that happened in public. On the Q4 2025 call (January 28, 2026), Musk announced the end of Model S and Model X production, "an honorable discharge" (CNBC), freeing their Fremont line for Optimus. The last S and X rolled off in early May 2026, and Tesla's Q1 investor deck listed the Optimus facilities as under construction (SEC exhibit). The teardown took 46 days, documented in a Tesla timelapse (via Yahoo Finance), and Musk framed the four-month line swap as "an insanely fast speed." On July 1 he walked the installed line with the team. Tesla's dates now need to be read alongside three ramp warnings. On April 22, Musk said the line will move only as fast as the slowest of roughly 10,000 parts, making 2026 output "literally impossible to predict." On July 1, he said production would be "extremely slow at first." On the July 22 Q2 call, he said the initial S-curve would be "quite flat and long" because everything on the robot is new. As of August 10, Tesla has not announced that Fremont production has started. The capacity ladder Every number in that ladder is real in the sense that Tesla said it, and none of it has produced a single robot for sale. Which is why the next table exists. The credibility ledger: forecasts vs what happened This is not a gotcha exercise. It is the correct prior for reading every Optimus date in this article, including the ones Tesla gives on July 22. The price record, complete Two different numbers get conflated constantly: what a robot might sell for, and what one costs Tesla to build. The dated record of both: The July 22, 2026 Q2 call added nothing to this table: price went unmentioned. Against those targets, the independent cost analysis is sobering. Morgan Stanley estimated the Gen 2 Optimus bill of materials at roughly $46,000 leveraging China's component ecosystem, and about $131,000 if built entirely outside it (Tech Digest); the bank's separate "Humanoid Tech 25" thesis argues value accrues to component suppliers rather than robot brands. Chinese financial media report a sub-$20,000 BOM target for Optimus 3 and name Sanhua Intelligent Controls and Tuopu Group as actuator suppliers, with Tuopu's Mexico plant reportedly targeting 300,000 actuator units of capacity; all of that is reported, not company-confirmed. The gap between a $46,000 Gen 2 BOM and a sub-$20,000 target is the central economic bet of the Fremont line. The scam ecosystem Because Optimus has no ordering channel, the demand for one has been monetized by criminals instead. Security researchers have documented an active malvertising network of fake Tesla sites, offers-tesla.com, exclusive-tesla.com, prelaunch-tesla.com and rotating successors, that buy search ads against Optimus queries and collect $250 "deposits" plus card details (Cybernews; also documented by SANS ISC). The rule that outlives every domain rotation: Tesla had no Optimus ordering channel of any kind as of August 10, 2026, so any site taking Optimus money today should be treated as fraudulent. If Tesla ever opens real orders, it will be on tesla.com and it will be global news; you will not learn about it from an ad. Internal fleet reality Tesla has never published an Optimus count, and external estimates conflict: Forbes and Omdia-derived reporting put cumulative output through 2025 at a few hundred robots, while DigiTimes-derived supply-chain reporting suggested roughly 1,000 assembled by mid-2025. What is documented: a mid-2025 production pause for a hand and forearm redesign, supplier reports of joint-motor overheating, weak hand load capacity, short transmission life and limited battery runtime, and deployment limited to battery-workshop tasks at reportedly less than half the efficiency of human workers (Electrek, citing LatePost and The Information). Musk's own January 2026 words, "not in usage in our factories in a material way," remain the honest baseline. The program's leadership tells the same transition story: Milan Kovac, who led Optimus from 2022, left in June 2025 (TechCrunch); Ashok Elluswamy, who also runs Tesla's Autopilot AI, has led it since, reportedly told staff 2026 would be the program's "hardest year" (Quartz), and posted on July 13 that Optimus has "big shoes to fill" on the converted line (Benzinga). Optimus 4 and the annual cadence At the March 2026 Abundance Summit, Musk said Tesla will have the "Optimus 4 design complete next year" and intends to release a new robot design every year (Teslarati). Take it seriously as strategy and skeptically as schedule: an annual cadence would mirror how Figure and 1X are already iterating, and it also means Optimus 3 is, by Tesla's own roadmap, an interim model. For anyone imagining buying one, the cadence matters more than the reveal: by the time public sales open (end of 2027, per Davos), Optimus 3 may already be the previous generation. The field on reveal day Optimus 3 will be revealed into a market that stopped waiting for it. Figure 03 has an official spec card, 350-plus units built, published factory yields and logistics work starting at BMW Spartanburg. Boston Dynamics Atlas became a production product at CES 2026 with first-year output committed to Hyundai and Google DeepMind. 1X NEO has an open order book at published prices. Unitree's G1 has been orderable since 2024, currently listed from $13,500 and backordered, and Agility's Digit does paid warehouse work and publishes throughput metrics. Tesla's counter is unchanged and untested: nobody else is building for a million units a year, and nobody else has an automotive supply chain and factory base to lean on. The reveal is where that argument finally meets evidence: either the robot and its line justify the two-year wait, or Tesla joins the field as one credible competitor among several, a year behind on shipped proof. The reveal-day checklist: how to judge it When the unveil happens, this is the scorecard we will run, and you can run it live. We will publish the scored checklist the day of the reveal, alongside the first sourced spec sheet if Tesla releases real numbers. About that "mass production started in January" story Syndicated articles claiming Optimus mass production began on January 21, 2026 trace to a single third-party press-wire piece with no Tesla source and no named reporter. Tesla's own record contradicts it three ways: one week after that date, Musk announced Model S/X production would end to free up the very line that would build Optimus; that line was still assembling cars until May; and Tesla's Q1 2026 investor deck listed the Optimus facilities as under construction (SEC exhibit). Content farms have since compounded it with a "1,000-unit working fleet" claim that is equally unsourced. None of it survives contact with Tesla's own record, and the episode is why this page treats production-start claims as unverified until Tesla itself speaks. Latest status: no reveal, price or release date Update, August 10: Tesla's latest official disclosure remains the July 22 Q2 shareholder deck and call. The deck says first-generation Optimus lines are being installed at Fremont, production is "anticipated later this year," and both Fremont and Texas Optimus capacity are listed as "Construction." Musk called Optimus Tesla's hardest product to scale and said the initial ramp would be "quite flat and long." Tesla did not provide a reveal date, production start, unit count, price, preorder or public release date, and no robot was shown on the audio-only call. Three events would change this status. First, Tesla confirmation that production has started. Second, the Optimus 3 reveal; April guidance said "middle of this year," but the July 22 shareholder deck and call set no new date. Third, an official retail launch or preorder page; Musk's public-sale target remains the end of 2027. Until Tesla publishes real specifications, every Optimus Gen 3 spec sheet remains unverified. For the robot Tesla operates today, see the Optimus Gen 2 review; for the factory story, the Fremont conversion tracker. The bottom line Optimus Gen 3 remains unrevealed, with no announced release date, price, preorder or spec sheet. What Tesla has confirmed is narrower: it is a ground-up redesign, first-generation production lines are being installed at Fremont, and production is anticipated later in 2026. Tesla's missed targets and Musk's warning that the ramp will be flat and long argue against treating rumor dates as commitments. The next meaningful update is Tesla confirmation of a reveal, production start or retail plan, not another unsourced timetable. Compare on RoboZaps: Tesla Optimus Gen 3, Figure 03, Tesla Optimus, Digit and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 1X NEO Review: Price, Pre-Order & Release Date (2026) URL: https://blog.robozaps.com/b/1x-neo-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-07-28T05:09:27.051Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: insights TL;DR: 1X NEO is the one home humanoid you can actually order: $20,000 or $499/month, reserved with a $200 refundable deposit, US first. No customer delivery had been verified as of July 16, 2026; 1X says some buyers get theirs this year, some later. It ships with basic onboard autonomy (Redwood) plus scheduled human teleoperation for the rest, which is both its superpower and its privacy debate. Key takeaways: - NEO is the only purpose-built home humanoid with published US purchase and subscription pricing: $20,000 or $499/month with a $200 deposit, refundable until 14 days after your Order Invitation per 1X's terms. - No customer delivery had been verified as of July 16, 2026; 1X’s own July statements ("some this year, some later") point to a staggered rollout against 10,000+ pre-orders and 10,000/year factory capacity. - The July 9, 2026 hand upgrade is the big hardware news: 25 degrees of freedom, tendon-driven, tactile fingertips with slip detection, IP68, shipping on the first customer units. - Expert Mode means a scheduled, approval-gated human teleoperator can drive the robot in your home; 1X’s safeguards (blurring, no-go zones, US-based operators, session indicator) are company claims not yet independently audited. - The WSJ hands-on found every household task in its session teleoperated, and Børnich maintains the launch film’s autonomous portions were genuine; production-unit reviews will settle it. FAQ: Q: How much is the NEO robot? A: $20,000 to buy outright with Early Access perks (three-year warranty, priority delivery), or $499 per month on subscription, reserved with a $200 deposit. Per 1X's pre-order terms, the deposit is refundable until 14 days after your Order Invitation arrives, and advertised pricing can change before you accept. A six-month minimum subscription term was reported at launch but no longer appears on the order page. Q: When will the NEO robot be available? A: US deliveries are slated to start in 2026, but no customer delivery had been verified as of July 16, 2026, and 1X says "some of you will get your NEO this year, some will get them later." International orders are slated to follow in 2027. Q: What can the NEO robot do? A: On its own, basic tasks via 1X’s onboard Redwood AI: navigation, door answering, retrieving objects. For complex chores it relies on Expert Mode, where a scheduled human teleoperator drives the robot. In the only major journalist hands-on (WSJ), every household task was teleoperated; 1X says autonomy is improving via its World Model research. Q: Who makes the NEO robot? A: 1X Technologies, a Palo Alto-based company founded in Norway in 2014 as Halodi Robotics, led by CEO Bernt Børnich, with backers including the OpenAI Startup Fund and EQT Ventures. It builds NEO at its own factory in Hayward, California, rated for 10,000 robots a year. Q: How tall is the NEO robot? A: NEO stands 5'6" (168 cm) and weighs about 66 lbs (30 kg), per 1X’s official specifications. It runs about 4 hours on its 842 Wh battery and operates at a claimed 22 decibels, roughly whisper-quiet. Q: Can NEO’s operators see inside my home? A: During scheduled Expert Mode sessions, yes: a 1X teleoperator sees through the robot’s cameras to complete tasks. Per 1X’s launch briefing and published FAQ: sessions require your explicit approval, people can be blurred, no-go zones are software-enforced, operators are US-based 1X employees, and the robot’s ear light indicates an active session. None of these controls has been independently audited yet. Q: Is the NEO robot autonomous? A: Partially. It ships with foundational autonomy from the onboard Redwood model and uses human teleoperation for tasks it cannot do alone. The widely shared "60-70% autonomous" figure has no primary source. 1X’s World Model research program, launched as a dedicated lab in June 2026, is its stated path to reducing teleoperation. Q: How does NEO compare to Figure 03? A: NEO is orderable today with published pricing and a teleop backstop; Figure 03 has no price or pre-order and stays partner-only until Figure’s autonomy meets its own bar, though Figure’s published demo ledger is deeper. They are the two serious home plays, betting opposite strategies. See our Figure 03 review and NEO vs Tesla Optimus comparison. Q: Can I buy NEO outside the United States? A: Not yet. Ordering is US-only, with international deliveries slated for 2027. The order page does not currently offer non-US checkout. Q: Is NEO safe around children and pets? A: That is the design intent: 30 kg, soft knit covering, compliant tendon-drive limbs and whisper-quiet operation are choices made for shared households. But 1X's own FAQ tells owners to stay attentive around children, vulnerable people and pets, and no production units are in homes yet, so there is no real-world safety record either way. Treat the design claims as engineering intent pending the first independent reviews. How much is the 1X NEO, and can you order one? You can order one today, which no other purpose-built home humanoid offers: 1X's order page takes a $200 fully refundable deposit against either a $20,000 purchase or a $499-a-month subscription. What you cannot yet do is take delivery: as of July 16, 2026, we could find no verified customer delivery of a NEO anywhere, and 1X's own materials still describe first customer shipments in the future tense. One disclosure up front: this is desk research built on the sources linked throughout, not a hands-on test; no production NEO has yet reached any independent reviewer, us included. 1X NEO specifications: the official sheet 1X publishes a full spec sheet on its official product page, rare candor in a category where Tesla publishes nothing. These are manufacturer claims; no production unit has been independently measured. Verification status per spec is tracked in the NEO record in our robot database. The misinformation audit: what circulates vs the record What is NEO, and who is 1X? NEO is the home robot 1X Technologies has been building toward for a decade. The company started in Norway in 2014 as Halodi Robotics, rebranded as 1X in 2022, and now runs from Palo Alto with manufacturing in Hayward, California and Moss, Norway. Its CEO and founder, Bernt Børnich, is an actuator engineer by background, and the company's signature technology, low-noise tendon-drive actuation, is what makes a 30 kg whisper-quiet humanoid possible. Before NEO came EVE, a wheeled humanoid deployed in factory and security work from 2022 per 1X's company history; EVE is no longer marketed, and NEO is the company's whole bet. Reports since September 2025 have described talks to raise around $1 billion at a $10 billion-plus valuation; no close had been announced as of mid-July 2026, so treat the $10 billion as a target, not a fact. The soft-body design philosophy Every other credible humanoid is a machine you keep away from your children; NEO is designed to be safe to bump into. The choices compound: tendon-drive actuation instead of high-ratio gearboxes gives compliance and quiet (22 dB claimed, against the 60-plus of most industrial humanoids), the 3D-lattice polymer skeleton keeps mass at 30 kg, less than half a Figure 03, and the knit suit is literally machine-washable. The trade-off is honest physics: NEO lifts 154 lbs briefly but carries 55, and it will never pull a BMW parts cart. 1X chose the living room over the loading dock, and the entire hardware stack follows from that choice. The hands: the July 2026 upgrade On July 9, 2026, 1X announced a redesigned hand it says will ship "on every NEO" (official announcement): 25 degrees of freedom in total, 22 actuated in the fingers and palm plus 3 in the wrist, driven by forearm-mounted motors through tendons, with fingertip tactile sensing that detects normal force, contact location and shear, real-time slip detection, up to 45 N of fingertip flexion force, ±0.2 mm positioning, IP68 sealing and food-safe materials. Forbes described the hands as approaching human-level capability and Dezeen relayed the design detail, while WIRED added the caveat that matters: real-world performance is unverified and some demonstration clips were operator-driven. 1X says it can build about 10,000 hands in-house in 2026. Børnich's line: built "to match or surpass human capability across every dimension that matters." For pre-order holders the practical meaning is simple: the hands were the weakest part of the launch-era robot, and the version that reaches customers is meaningfully better than what the press handled in October. Ordering, decoded: what your $200 actually buys Pricing has not moved since launch. The six-month minimum subscription term reported by launch coverage no longer appears on the order page, so we list it in the misinformation audit rather than the table. Read the deposit for what it is: a cheap, refundable option on an early delivery slot, not a commitment. Two fine-print points matter more than the marketing page: the refund window closes 14 days after your Order Invitation arrives, and the $20,000/$499 figures are advertised pricing that 1X's terms allow to change before you accept. The delivery scoreboard The capacity math explains the quiet: a 10,000-unit-a-year line that started in May cannot clear a 10,000-plus order book in the calendar year, before internal, B2B and testing units take their share. 1X has notably avoided promising every deposit-holder a 2026 delivery. Our read, labeled as inference: the earliest customer units land late 2026, and a meaningful share of the book slides into 2027. If you hold a deposit, your queue position matters more than any public date. Redwood: what actually runs on the robot Redwood, announced June 2025, is NEO's onboard brain: a 160-million-parameter vision-language model that runs entirely on the robot's embedded GPU at roughly 5 Hz, working even with spotty internet. It handles navigation, door operation, object retrieval and whole-body coordination, moving arms, hands, pelvis and gait together, and it was trained on teleoperated and autonomous episodes from 1X's EVE and NEO fleets. Its small size is a deliberate privacy and reliability argument: the robot does not need to stream your home to a datacenter to walk to the kitchen. Its limitation is equally structural: a 160M-parameter model is not going to reason its way through novel multi-step chores, which is exactly the gap Expert Mode exists to fill. Expert Mode: how the teleoperation actually works When NEO cannot do a task autonomously, you schedule a session, and a 1X operator connects and pilots the robot through it. The mechanics, as 1X has described them: sessions happen only in windows you choose and each one requires your explicit approval; tasks are queued through the app in what 1X has compared to a strategy-game interface; and every teleoperated session doubles as training data for the autonomy. This is the engine of the whole product thesis: 1X sells the robot before the AI is finished and uses paying customers' homes, with consent, as the training distribution. It is candid, clever and unprecedented, and it means early NEO ownership is closer to hosting a supervised apprentice than employing a butler. The privacy ledger The safeguards, each with its provenance. From 1X's launch briefing: operator sessions require explicit user approval, people in the home can be blurred in the operator's feed, and users define no-go zones the robot respects even under teleoperation. From 1X's published FAQ: Expert operators are US-based 1X employees, the robot's ear-ring light changes whenever a session is active, limited sensor data from autonomous requests is processed but, per 1X, not stored, sharing data for improvement is optional, and 1X says it does not profile users or sell their data. Børnich's summary claim is that the model is "more secure than hiring a cleaner." The criticism record is just as concrete. The WSJ's Joanna Stern, reported that every household task in her session was driven by a teleoperator in a VR headset and framed early adoption as trading "the privacy of your inner sanctum" for a robot that is still learning. Marques Brownlee criticized the launch as a hype reel for a robot whose demonstrated actions were teleoperated, drawing Børnich's on-record rebuttal that the launch film's autonomous segments were genuinely autonomous. The New York Times' Hard Fork team ran a second hands-on with similar results: a failed floor pickup, balance trouble, and tasks completed under full teleoperation (episode transcript). Privacy-alarm threads dominate NEO discussion on Reddit, and outlets from CNET to TechRadar ran versions of the same warning. Every safeguard above is a company claim; none has been independently audited, because the robots are not in customers' homes yet. That is the honest state of the debate. The autonomy roadmap: World Models and a reshuffled AI bench 1X's answer to "when does the teleoperator go away" is world models: teaching the robot to predict the visual consequences of its own actions so it can learn tasks without a human demonstrating each one. The company shipped a self-learning World Model for NEO in January 2026, and in June stood up a dedicated World Model Lab, hiring Luma AI's Samarth Sinha to lead it, with Børnich framing the thesis to Forbes as "you can't fine-tune your way to AGI." The bench turned over around the same push: Eric Jang, the VP of AI who shaped 1X's learning stack, left in January 2026, with Mohi Khansari promoted over robot learning. 1X has promised early World Model results before the end of 2026; Børnich's standing claim is a "mostly fully autonomous" robot in 2026. Score those against the delivery scoreboard above, and judge in December. What NEO can and can't do Three buckets, honestly sorted. Claimed autonomous, per 1X: door answering, object retrieval and handoffs, navigation, the launch film's unassisted segments, all Redwood-driven and defended by Børnich as genuine. Teleoperated by design: most complex chores today, including essentially everything in the WSJ session, laundry handling, tidying, kitchen tasks. Aspirational: the full unattended chore list, which awaits the World Model program. And 1X's own disclosed limitations deserve equal billing: cooking will initially be restricted, only the hands are waterproof (moisture can damage the body, and outdoor use is limited), and maintenance is recommended through 1X. A NEO in 2026 is an indoor apprentice with excellent hands, not a housekeeper. The factory, and the B2B tension The Hayward plant is 1X's answer to the question every pre-order holder should ask: can they actually build these? It is 58,000 sq ft, billed as America's first vertically integrated humanoid factory, producing motors, batteries, structures, transmissions and the knit suits in-house, rated at 10,000 NEOs a year now with a target above 100,000 a year by the end of 2027. The tension is the EQT deal: up to 10,000 humanoids across EQT's portfolio companies between 2026 and 2030, in logistics, warehousing, manufacturing and facilities. The release describes shared intent to make robots available, with each portfolio company deciding independently whether to deploy, so it is a pipeline rather than booked revenue, and still a second claimant on the same production line consumers are queued for. When you read "consumer shipments planned for 2026," remember the line is also building robots for warehouses. NEO vs Figure 03, Tesla Optimus, and the robots you can already buy Precision matters here: NEO is not the only humanoid an individual can put money on, a Unitree G1 has a literal add-to-cart button (currently backordered) and Neura takes 4NE1 Mini reservations, but it is the only purpose-built home humanoid with published US purchase and subscription pricing. Against Figure 03, the contrast is strategy, not hardware: 1X monetizes the learning phase with customers inside it; Figure keeps the learning phase in-house and sells nothing yet. Against Optimus, NEO simply has more verifiable existence: published specs, published prices, a running factory you can tour on video. Deeper matchups: NEO vs Tesla Optimus and the best humanoid robots guide. Who should pre-order, and who should wait Pre-order if the $200 is disposable, an early slot has option value to you, and you are genuinely comfortable being part of the training loop, operators, scheduled sessions, novelty failures and all. Wait if you want a robot that already does chores by itself (none exists at any price), if a remote human in your home is disqualifying no matter the safeguards, or if you would rather let the first production units absorb the early-adopter risk and read the reviews that follow. And if what you actually want is any capable humanoid to experiment with this year, a G1 costs less and is orderable today (backordered at this writing), it just will not fold your laundry either. What to watch next Four dated threads decide NEO's next chapter: the first verified customer delivery, which converts 1X's entire story from promise to product; the first production-unit reviews that follow it, which will test the autonomy, runtime and noise claims in real homes; the World Model results 1X has promised before the end of 2026; and any movement on the reported $1 billion raise, which would fund the 100,000-a-year ambition. We re-verify this page at each. The bottom line NEO is the most honest dishonest product in robotics: honest because 1X publishes its prices, specs and even its teleoperation dependence in plain sight, and dishonest only in the sense that every home robot sold today is selling a future its software has not reached. As a piece of hardware and a company bet, it is credible, a real factory, a real order book, the best hands yet announced, and a privacy architecture that at least takes the question seriously. As a purchase, it is a $200 refundable option on being an early participant in that bet, with delivery timing 1X itself will not fully commit to. Put the deposit down if that is exciting; keep your $200 if you want proof first. The NEO record in our robot database tracks the verified state as it changes. --- # 1X NEO vs Tesla Optimus: Full Comparison (2026) URL: https://blog.robozaps.com/b/neo-robot-vs-tesla-optimus Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T23:06:52.454Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: 1X NEO is the more purchase-ready home robot because it has a public pre-order, price framework and delivery target. It is not shipping off the shelf: final terms may change and early units will have basic autonomy. Tesla Optimus has no consumer price, order page or confirmed delivery date, so there is not enough current evidence to name it the better performer. Key takeaways: - 1X accepts a $200 NEO pre-order deposit and advertises $20,000 ownership or a $499 monthly subscription; its terms say final pricing and subscription details may change before acceptance. - Tesla Optimus is not available to consumers. Tesla was installing initial production lines as of April 2026 but had not announced a retail price or consumer delivery date. - NEO has a clearer home-use proposition and publishes more current specifications, including a 66 lb body, four-hour runtime and new 25-DoF hands including the wrist. - Neither robot has enough independent consumer deployment evidence to support broad claims about real-world reliability, child safety or household autonomy. FAQ: Q: Can I buy Tesla Optimus right now? A: No. Tesla has not opened consumer sales, reservations or a public Optimus waitlist. Its April 2026 update said production lines were being installed, but it did not provide a retail price or consumer delivery date. Q: How much does the 1X NEO robot cost? A: 1X currently advertises $20,000 for Early Access ownership, $499 per month for a Standard subscription and a $200 pre-order deposit. Its terms say final pricing and detailed subscription conditions may change before an order is accepted. Q: Is 1X NEO available now? A: NEO is available to pre-order, not for immediate delivery. 1X says US deliveries start in 2026 and expansion to other markets starts in 2027, but a pre-order does not guarantee an individual delivery date. Q: Is 1X NEO safe around children? A: 1X designed NEO with a soft body, covered joints and low-inertia tendon drives for home use. Those features do not guarantee safe unsupervised use around children. Early owners should follow 1X's instructions and supervise operation. Q: Which robot is better for household chores? A: NEO has the clearer household focus and the only consumer offer, but early units are expected to arrive with basic autonomy and may use scheduled Expert Mode for unfamiliar chores. Optimus has no consumer task list or home-use benchmark, so a reliable performance winner cannot yet be measured. Q: Can NEO learn new tasks? A: 1X says Redwood AI learns and repeats tasks, while scheduled Expert Mode lets a 1X operator supervise chores NEO does not know. That is different from claiming every new task will be learned or completed fully autonomously. Q: When will Tesla Optimus be available to consumers? A: Tesla has not confirmed a consumer sales date. Its official 2026 updates discuss production preparation and large-scale manufacturing plans, not a retail launch schedule. Q: What are the privacy concerns with 1X NEO? A: NEO uses cameras and microphones, offers app-based monitoring and can use a scheduled remote Expert Mode. Buyers should review 1X's current privacy policy and understand how remote access, recordings and room restrictions are handled. 1X NEO and Tesla Optimus are not two finished robots sitting on the same store shelf. NEO is accepting refundable pre-orders for a planned consumer rollout, while Tesla has no public Optimus order page, price or consumer delivery date. That difference matters more than a speculative spec-sheet winner. This comparison uses the latest first-party information available on July 16, 2026. For a deeper look at each product, read our 1X NEO review and Tesla Optimus price and availability guide. 1X NEO vs Tesla Optimus: quick answer NEO is the more concrete option for a home buyer, but it is still an early-access pre-order, not an off-the-shelf product. 1X lists a $20,000 ownership option, a $499-per-month subscription and a $200 deposit. The company says US deliveries start in 2026. Its pre-order terms also say final pricing and subscription details may change before an order is accepted. Optimus is further from a consumer purchase. Tesla's April 2026 update said first-generation production lines were being installed in anticipation of volume production. It did not publish a retail price, open reservations or give consumers a delivery date. Several widely repeated Optimus numbers belong to older prototypes or comments rather than a current Gen 3 product sheet. We have left those cells blank instead of presenting estimates as confirmed specifications. Price and availability 1X NEO: a deposit secures a place in the queue The 1X order page currently advertises three headline numbers: $20,000 for Early Access ownership, $499 per month for the Standard subscription and a $200 deposit due at pre-order. It also says US deliveries start in 2026. That is not the same as buying a shipping robot. Under the 1X pre-order terms, the company later sends an Order Invitation. The final price, add-ons and detailed subscription terms can change before acceptance, and a subscription may require a minimum number of months. The deposit can be refunded until 14 days after the invitation is sent; after that point, the terms say it becomes non-refundable. 1X says its Hayward factory has begun full-scale NEO production and that current units are being prioritized for internal development and home testing while it works toward customer deliveries. That is meaningful manufacturing progress, but it is not evidence that consumer orders are already arriving. Tesla Optimus: production preparation, not consumer sales Tesla's January 2026 update described Gen 3 as its first design intended for mass production and said production was planned to begin before the end of 2026. The company's April 2026 update then said Optimus production lines were being installed. Those statements do not establish a consumer launch. Tesla has not opened a public Optimus checkout, reservation list or subscription. It also has not published an official retail price. Any $20,000 to $30,000 figure should be treated as an aspiration or estimate, not a current offer. Hardware: NEO publishes more usable specifications NEO is 5 ft 6 in tall and weighs 66 lb. 1X lists a 154 lb lift rating, 55 lb carry rating and 18 lb arm payload. Those are different measurements, so the 154 lb lift number should not be compared directly with another robot's carry capacity. The company also claims a four-hour runtime, a 1.4 m/s walking speed and a maximum 6.2 m/s running speed. A July 2026 1X hand update says the version intended to ship has 25 degrees of freedom: 22 actuated degrees in the fingers and palm plus three at the wrist. That is newer than the 22-per-hand figure still shown on parts of 1X's product page. Tesla's older Optimus prototypes have public dimensions and demonstration footage, but Tesla has not released an equivalent current Gen 3 spec sheet. Using older Gen 2 height, weight, battery and hand figures as if they describe the production-intent machine would create false precision. Autonomy and household usefulness 1X markets NEO as a household robot. Its order page says early units arrive with basic autonomy powered by Redwood AI and gain capabilities over time. Owners can schedule chores, speak to the robot and use an app for monitoring or remote control. When NEO does not know a task, the owner can schedule Expert Mode so a 1X operator supervises the robot. That combination is more honest than claiming fully solved home autonomy, but it changes the value proposition. An early NEO may depend on scheduled human help for complex chores. Buyers should judge it as an evolving service and hardware package, not as a fully autonomous housekeeper on day one. Tesla describes Optimus as a general-purpose autonomous humanoid for unsafe, repetitive or boring tasks. Its AI and robotics page highlights vision, planning, balance, navigation and physical-world interaction. Tesla has not, however, published a consumer task list, household autonomy benchmark or service model that can be compared directly with NEO's current offer. Safety and privacy NEO has the stronger home-oriented design story. 1X lists a soft lattice-covered body, covered joints, low-inertia tendon drives and a 66 lb frame. Those features are relevant when a robot will operate around furniture, pets and people. They are not a guarantee that NEO is safe for unsupervised use around children. Manufacturer claims and component specifications cannot replace independent testing, careful setup and adult supervision during an early-access rollout. Privacy is also part of the purchase decision. NEO uses cameras and microphones, its app supports remote monitoring, and Expert Mode lets a scheduled operator supervise unfamiliar chores. Prospective buyers should review 1X's current privacy policy, decide which rooms are off limits and understand how remote sessions and recordings are handled before placing a robot in the home. Optimus cannot yet be assessed as a consumer privacy product because Tesla has not published a home-user offer or operating policy for it. Which robot should you choose? Consider a 1X NEO pre-order if you: want the only one of these two robots with a public consumer pre-order;accept that early units may have limited autonomy and use scheduled human assistance;are comfortable with the deposit rules and the possibility that final terms will change;value a lighter, soft-bodied design made specifically for homes. Wait for Tesla Optimus if you: want verified Gen 3 specifications before comparing performance;do not want to commit money to an early-access robot;prefer to wait for a retail price, warranty, service plan and consumer delivery date;are more interested in Tesla's eventual scale than being early to home robotics. Verdict NEO wins on purchase readiness, not proven real-world performance. It has a public price framework, a refundable pre-order process, published hardware details and a product designed for homes. Optimus has no consumer offer and too few current specifications for a fair performance verdict. The important caveat is that NEO is still a pre-order. The most accurate decision in 2026 is therefore not “buy NEO instead of Optimus,” but “consider joining NEO's queue if you accept early-access risk; otherwise wait for delivered units and independent testing.” If neither option is ready enough, compare robots with stronger delivery evidence in our best humanoid robots guide, or browse the current humanoid robots for sale. Compare on RoboZaps: NEO and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robot Industry Report 2026: Prices, Funding & Market Data URL: https://blog.robozaps.com/b/humanoid-robot-industry-report-2026 Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T23:07:28.678Z Category: reviews TL;DR: This report tracks 26 humanoid robots across 7 countries — 14 commercially available — against more than $5 billion of industry investment. Consumer units now price below $25,000 while industrial platforms run about $150,000-$320,000, with Figure AI's $39 billion valuation the sector's high-water mark. Key takeaways: - Tracks 26 humanoid robots across 7 countries, 14 of them commercially available. - Total industry investment exceeds $5 billion, including $2.7 billion+ in venture capital since 2020. - Consumer humanoids now price below $25,000, while industrial units run roughly $150,000-$320,000. - Marquee funding: Figure AI's $39 billion valuation and Apptronik's $403M Series A with Mercedes-Benz. - The demand drivers: aging labor forces, manufacturing reshoring, and large-model AI advances. This is the most comprehensive public database of humanoid robot specifications, pricing, funding, and deployment data available. Updated monthly, this report tracks 26 humanoid robots across 7 countries, with over $5 billion in total industry investment (including acquisitions). Last Updated: March 2026 | Next Update: March 2026 Executive Summary: The State of Humanoid Robotics in 2026 26 humanoid robots currently tracked in our database$4 billion+ in venture capital raised by humanoid-focused startups since 2020$5 billion+ total industry investment including acquisitions14 robots commercially available today (available for purchase or pre-order with confirmed pricing)Consumer prices range from $4,900 to $25,000 — humanoids are becoming affordableAverage robot price: $94,359 across all categories (skewed by industrial units)China leads production with 10 robots (38%), USA follows with 9 (35%) Humanoid Robot Price Index Consumer humanoid robots are now priced below $25,000 — comparable to a new car. Industrial units range from $150,000 to $150K-$320K. Here's every robot with confirmed or estimated pricing: Price Tier Summary Humanoid Robot Funding Tracker Humanoid robot startups have raised over $2.7 billion in venture capital since 2020. Including acquisitions and corporate investments, total industry investment exceeds $5 billion. Corporate Investments & Acquisitions Production & Deployment Tracker Commercial humanoid robot deployments began in earnest in 2024-2025, with pilots at Amazon, BMW, Mercedes-Benz, and others. Here's the current state of production and deployment: Announced Production Targets Confirmed Commercial Deployments Global Humanoid Robot Production by Country Industry Timeline: Key Milestones (2020-2026) *Events marked with asterisk could not be independently verified from primary sources Complete Specifications Database The most comprehensive public database of humanoid robot specifications. All measurements verified from official manufacturer sources where available. Performance Rankings Fastest Humanoid Robots (Running Speed) *1X NEO speed unverified from official manufacturer source Longest Battery Life Highest Payload Capacity *Some payload claims unverified from official sources Most Degrees of Freedom (DOF) Note: Human body has approximately 244 degrees of freedom. Most humanoid robots prioritize key joints for practical tasks rather than matching human DOF count. Market Size Projections Note: Market projections for humanoid robots vary significantly across research firms. The following estimates are commonly cited in industry coverage but should be independently verified. Robozaps does not endorse specific projections. Key Market Drivers Labor shortages: Aging populations in developed economies creating demand for automationManufacturing reshoring: Companies bringing production back onshore need automation to be cost-competitiveAI advancements: Large language models and vision AI enabling more capable robotsCost reduction: Consumer-tier robots now available under $25,000 (vs. $100K+ in 2023)Proven deployments: Amazon, BMW, Mercedes pilots demonstrating real-world viability Investment Milestones 2024: Figure AI raised $675M Series B, later $1B+ Series C at $39B valuation2024: Total VC investment in humanoid startups exceeded $1B for the first time2025: Apptronik closed $403M Series A with Mercedes-Benz, ARK Invest2026: Industry tracking toward $500M+ annually in new investment Methodology Data Sources Verification Note: Data verified from official manufacturer sources where accessible. Some specifications, prices, and deployment claims could not be independently verified due to limited public disclosure. Unverified data is marked with * or —. Official manufacturer websites and press releasesSEC filings and investor presentations (where applicable)Verified news reports from TechCrunch, Reuters, Bloomberg, IEEEIndustry databases (Crunchbase, PitchBook)Direct communication with manufacturers Pricing Methodology Confirmed prices: Official pricing from manufacturer websites or press releasesEstimated prices: Based on target pricing statements, production cost disclosures, or industry analysis. Clearly labeled as estimates.All prices in USD. Currency conversions at time of data collection. Update Schedule This report is updated monthly. Price changes, new funding rounds, and deployment announcements are added as they occur. Major updates are announced via the Robozaps newsletter. For Journalists & Researchers This data is free to cite with attribution. Suggested citation: For media inquiries, high-resolution graphics, or interview requests: dean@robozaps.com Download the data: CSV export (coming soon) Explore more: Browse all humanoid robots | Humanoid Robot Price Guide | Robozaps Blog Sources & References All data verified from primary sources where accessible. Last verified: February 11, 2026. Manufacturer Specifications Boston Dynamics Atlas - Official product page with full specifications Unitree H1 - Official specifications Unitree G1 - Official specifications and pricing Apptronik Apollo - Official product page 1X NEO - Official product page Figure AI - Official company page NEURA Robotics - Official company page Fourier GR-1 - Official specifications Verified Pricing Unitree R1 - $4,900 - Official shop Unitree G1 - $13,500 - Official shop 1X NEO - $20,000 - Official pre-order NEURA 4NE1 Mini - €19,999 - Official reservation Deployment Announcements Agility Digit at GXO Logistics - November 2025, Flowery Branch, GA Agility Digit at Amazon - Testing announcement, October 2023 Agility RoboFab - 10,000 robots/year capacity Funding & Investment Figure AI Series B - TechCrunch, February 2024 Apptronik Series A - Official announcement, March 2025 Market Research Markets and Markets - Humanoid Robot Market Report (Code: 99567653) Precedence Research - Humanoid Robot Market Size Report Allied Market Research - Humanoid Robot Market Analysis Note: Some manufacturer specifications, funding amounts, and deployment details could not be independently verified due to limited public disclosure. Unverified claims are marked with asterisks (*) throughout this report. Last updated: February 11, 2026. Data compiled by the Robozaps research team. Robozaps is the world's largest humanoid robot marketplace. We track 26 robots across 7 countries and maintain comprehensive specifications, pricing, and availability data. Compare on RoboZaps: LimX Oli, AgiBot X1, Xiaomi Cyberone, Unitree R1, Figure 03 and Ameca. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Figure AI Review 2026: Team, Investors, History, Robots & $39B Valuation URL: https://blog.robozaps.com/b/figure-ai-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: Figure is a top-tier humanoid robotics contender with real BMW production evidence, an in-house Helix AI stack and a credible manufacturing ramp. Its $39 billion valuation is much harder to support: revenue, paid fleet size, unit economics, intervention rates and current customer concentration remain largely undisclosed. Figure is technically credible, commercially early and highly speculative. Key takeaways: - Figure was founded by Brett Adcock in 2022 and has moved through three robot generations in four years. - The best reported team-size point is 360 people in October 2025; Figure does not publish a current executive roster. - Jerry Pratt was Figure's CTO but left in 2024; he should not be listed as current leadership. - Figure has disclosed more than $1.745 billion across Series A, B and C rounds, with Parkway as its repeat lead investor. - The latest round valued Figure at $39 billion post-money, but the company does not publish revenue, margins or backlog. - BMW is the strongest customer evidence: the underlying production work is independently confirmed and Figure reports more than 1,250 operating hours. - Figure says BotQ produced more than 350 Figure 03 robots by April 2026, but it has not disclosed how many were sold to customers. - Helix 02 unifies language, vision, manipulation, walking and balance, but most performance evidence remains company-produced. - Figure 03 has no public consumer order page, list price or confirmed broad home release date. - An active safety-related employment lawsuit and limited leadership disclosure add governance risk; the allegations have not been adjudicated. - Figure's 2024 securities filing disclosed six directors, four executive officers and 45 Series B investors, but it is not a current governance roster. - Public fund filings provide private-share reference marks, but the positions are restricted, illiquid and valued with significant unobservable inputs. - Figure had at least 13 listed U.S. patent grants by July 2026 across hardware, kinematics and embodied-AI control. - BotQ's 12,000-unit nameplate goal is not current output; Figure reported more than 350 completed Figure 03 units and a demonstrated one-per-hour cycle. - The active safety-related employment case remains in discovery, with no ruling on the truth of either side's claims. FAQ: Q: What is Figure AI? A: Figure AI is a privately held U.S. robotics company founded in 2022. It develops general-purpose humanoid robots, the Helix embodied-AI system and the BotQ manufacturing operation. Q: Who founded Figure AI? A: Brett Adcock founded Figure AI and remains its CEO. He previously co-founded recruiting marketplace Vettery and electric-aircraft company Archer Aviation. Q: Who is on Figure AI's leadership team? A: Figure publicly identifies Brett Adcock as founder and CEO but does not maintain a complete current executive roster. TIME identified Corey Lynch as head of AI, Logan Berkowitz as vice president of business operations and Lee Randaccio as vice president of growth in October 2025. Q: Is Jerry Pratt still Figure AI's CTO? A: No. Figure named Jerry Pratt as CTO in its 2023 launch materials, but TechCrunch reported in June 2024 that he left Figure to co-found Persona AI. Q: How many employees does Figure AI have? A: TIME reported about 360 employees in October 2025, while LinkedIn currently places Figure in its broad 201-500 employee band. Figure has not published a precise July 2026 headcount. Q: Who are Figure AI's investors? A: Named investors across Figure's rounds include Parkway Venture Capital, Microsoft, OpenAI Startup Fund, NVIDIA, Bezos Expeditions, Intel Capital, Align Ventures, ARK Invest, Brookfield, Macquarie, LG Technology Ventures, Salesforce, T-Mobile Ventures, Qualcomm Ventures, Tamarack and others. Q: How much funding has Figure AI raised? A: Figure has disclosed more than $1.745 billion across a $70 million Series A, $675 million Series B and more than $1 billion in committed Series C capital. Q: Why is Figure AI valued at $39 billion? A: The $39 billion figure is the post-money valuation from Figure's September 2025 Series C. It reflects investor expectations for humanoid robotics, Helix and manufacturing scale, not publicly disclosed revenue or profit. Q: Is Figure AI publicly traded? A: No. Figure AI is private and has no public ticker. The company says private share transfers require board approval and has warned about unauthorized secondary-sale attempts. Q: What robots has Figure AI built? A: Figure's public generations are Figure 01, Figure 02 and Figure 03. Figure 01 was the first working platform, Figure 02 produced the strongest BMW deployment evidence and Figure 03 is the current production-focused model. Q: What is Helix 02? A: Helix 02 is Figure's full-body vision-language-action system. It combines semantic reasoning, visuomotor control and a fast whole-body controller for walking, manipulation and balance. Q: What has Figure AI proved at BMW? A: BMW independently confirms that Figure robots worked in production at Plant Spartanburg. Figure reports that Figure 02 logged more than 1,250 hours, handled over 90,000 parts and contributed to more than 30,000 BMW X3 vehicles. Q: How many Figure 03 robots have been made? A: Figure reported that BotQ had produced more than 350 Figure 03 robots by April 2026. It has not disclosed how many were sold to paying customers versus kept for research, data collection and deployment development. Q: Can consumers buy Figure 03? A: No public consumer ordering channel exists. Figure has not published an official list price, broad home release date, warranty or consumer service plan for Figure 03. Q: What are Figure AI's biggest risks? A: Major risks include a valuation far ahead of public financial evidence, safety and privacy in homes, undisclosed intervention and reliability rates, reliance on BMW for the strongest customer data, manufacturing support at scale, leadership opacity and an active employment lawsuit involving safety allegations that have not been adjudicated. Q: Who was on Figure AI's board in 2024? A: Figure's March 2024 Series B Form D listed Brett Adcock, Colby Adcock, Jesse Coors-Blankenship, Grant Hosking, Peter Welinder and Gregg Hill as directors. The filing is a dated offering disclosure, not proof that the same board remains in place in July 2026. Q: Was Figure AI founded in 2022 or incorporated in 2023? A: Figure describes the business as founded in 2022. Its Series B Form D lists the Delaware issuer Figure AI Inc. as incorporated in 2023. The operating start and legal incorporation date can differ. Q: How many investors joined Figure AI's Series B? A: Figure's SEC Form D reported 45 investors and $674,981,967 sold in the $675 million Series B offering. The public announcement named only a subset of those investors. Q: Does Figure AI have patents? A: Yes. Justia listed 13 U.S. grants assigned to Figure AI as of July 16, 2026, covering hands, head and neck assemblies, torso and leg kinematics, robot control and a bipedal action model. Applications and non-U.S. records may extend beyond that snapshot. Q: What are Figure AI private shares worth? A: There is no public market price. Fund filings show different restricted-share fair values, including $135 per Series B share at ARK in April 2026 and about $194.93 per Series C share in an Alger filing. These are Level 3 or good-faith accounting estimates, not executable quotes, and share-class rights differ. Q: What is Figure AI's work culture? A: Figure publicly describes a high-intensity, in-person, mission-first culture built around moving fast, technical fearlessness, product focus, aggressive optimism, long-term impact and a championship mindset. The company explicitly says the culture is not for everyone. Q: What is the current status of the Figure AI safety lawsuit? A: As of July 16, 2026, Gruendel v. Figure AI remains in discovery. A July 6 order set document-production and discovery-planning steps. It did not decide whether the former employee's allegations or Figure's counterclaims are true. Figure AI has earned its place in the top group of humanoid robotics companies. Since 2022 it has built three robot generations, developed an in-house vision-language-action system, opened a manufacturing line and put robots into live production work at BMW. That is a serious record for a four-year-old company. It is also nowhere near enough to validate a $39 billion valuation on operating results alone. Our verdict: Figure is one of the most technically credible and commercially ambitious humanoid startups, but the public evidence is lopsided. The BMW deployment is real and measurable. The BotQ production ramp is meaningful but self-reported. Home demonstrations show breadth, not consumer readiness. Revenue, unit economics, fleet uptime, intervention rates and current customer concentration remain largely undisclosed. Investors are paying for the possibility that Figure becomes a full-stack robotics platform, not for a mature business that already exists. This review covers the company behind the robots: its history, founder, team, funding, investors, AI strategy, manufacturing plan, customers and risks. Detailed Figure 03 specifications, price rumors and release-status updates stay in the separate product review. For the catalogue view, see the Figure AI manufacturer record. Figure AI quick facts What Figure AI actually does Figure is building general-purpose humanoid robots for physical work. The basic thesis is straightforward: homes, warehouses and factories were designed around the human body, so a machine with legs, arms and hands can use existing doors, shelves, tools and workstations without every site being rebuilt for automation. Figure initially emphasized structured jobs in manufacturing and logistics. Figure 03 and Helix have widened the story to household work. The company is unusually vertically integrated. It designs the robot body, hands, actuators, battery, sensing and charging system; trains the Helix AI models that interpret instructions and control movement; operates data-collection programs; and runs BotQ manufacturing. A defect found on a factory shift can therefore inform a mechanical redesign, a new test at the end of the assembly line and the next training cycle. In theory, more deployed robots create more data, which improves the models, which makes the robots more useful and easier to sell. What Figure has not explained publicly is the commercial wrapper. It has not published a hardware price, lease rate, robot-as-a-service fee, maintenance contract, gross margin or expected payback period. BMW has not disclosed the financial terms of its relationship. The business may eventually combine hardware, recurring software and service revenue, but that is a plausible model rather than a disclosed one. Today, the most defensible description is a private robotics developer moving from paid industrial trials toward repeatable commercial deployments. That distinction matters. A robot can complete a useful task and still be a poor product if it needs too much human supervision, fails often, takes too long to repair or costs more per productive hour than the alternatives. Figure publishes more engineering detail than many startups, but the numbers needed to judge the business are still thin. Figure AI history: from basement idea to BotQ The speed is the striking part of this history. Figure launched publicly with 40 employees in March 2023, reported 50 by its Series A that May and 80 by the Series B in February 2024. TIME reported a 360-person company in October 2025. Within the same period, Figure moved through three complete robot designs and shifted from an external AI collaboration to its own embodied-AI stack. The sequence also shows how the strategy changed. Figure 01 was pitched primarily at warehouses, manufacturing and retail. Figure 02 became the industrial proof point at BMW. Figure 03 was redesigned with softer materials, wireless charging, quieter operation and a more domestic appearance, even though the strongest evidence still comes from factories. Figure is trying to preserve the near-term industrial route while aiming at the much larger, much harder home market. Founder Brett Adcock Brett Adcock is Figure's founder and chief executive. Before humanoid robots, he co-founded the recruiting marketplace Vettery, which was sold to Adecco in 2018 for about $110 million. He then co-founded electric-aircraft company Archer Aviation, which went public through a SPAC transaction in 2021. Those companies gave him experience raising capital, recruiting technical teams, selling an enterprise platform and taking a hardware story to public markets. Figure's master plan, dated May 20, 2022, frames humanoids as a response to labor shortages and unsafe or undesirable work. Adcock's argument is that a general-purpose machine should eventually spread its development cost across many tasks rather than remain trapped in one narrow application. He has since expanded the claim to homes and has predicted that humanoids could become one of the world's largest markets. Those forecasts explain the scale of Figure's fundraising, but they should be read as founder ambition, not demand forecasts backed by disclosed orders. Adcock is a high-velocity operator and an unusually effective fundraiser. Figure's product cadence supports that reputation. It also creates key-person risk. TIME reported that a group of colleagues followed him from Vettery and Archer into Figure, including senior operations and growth leaders. Business Insider reported in May 2026 that he had also launched and become CEO of a new AI hardware company, Hark, after raising $700 million. Running another heavily funded company raises a fair question about attention, governance and how responsibilities at Figure are divided. There is also a disputed point in his prior history. Adcock told TIME that he chose to leave Archer in 2022 to start Figure; an Archer spokesperson said his departure followed a board decision, while Figure said the resignation was voluntary. The disagreement does not determine Figure's prospects, but it belongs in a complete founder profile rather than a sanitized biography. Leadership, team and hiring Figure does not publish a current executive-team page. Its company site names Adcock and describes the collective team's experience, while the careers page lists open roles. That opacity makes it easy for third-party profiles to preserve old titles. The clearest example is Jerry Pratt: Figure identified the veteran roboticist as CTO in 2023, but TechCrunch reported in June 2024 that he had left to co-found Persona AI. He should not be presented as Figure's current CTO. The team has grown quickly, but the best point estimate is dated. TIME counted 360 people around the Figure 03 launch in October 2025. LinkedIn currently places Figure in its broad 201-500 employee band. Neither figure tells us how many people are permanent employees, contractors or hourly robot operators, and Figure has not published a July 2026 headcount. A precise current number would be false confidence. The hiring mix is more revealing than the total. Figure's current openings span reinforcement learning, whole-body control, AI training infrastructure, electrical and actuator engineering, manufacturing, quality, health and safety, deployment logistics, field service, robot operation and commercial site leadership. Roles in Reno and Los Angeles, alongside San Jose, suggest the company is building the operating layer needed to stage, deploy and support robots outside headquarters. Figure's early releases emphasized recruits from Boston Dynamics, Tesla, IHMC, Google X, DeepMind, Cruise, Apple and Archer. That pedigree is useful, but brand-name résumés do not replace current accountability. For a company preparing machines to work near people, a published leadership roster covering engineering, AI, manufacturing, safety and commercial operations would materially improve transparency. Corporate structure and the board disclosed in 2024 Figure describes itself as founded in 2022, which is when Brett Adcock published the master plan and began assembling the company. The legal issuer used for the Series B was Figure AI Inc., a Delaware corporation whose Form D lists 2023 as its year of incorporation. Those statements are compatible: a company can begin operating before the financing entity is formally incorporated. The distinction is worth recording because company histories often collapse the operating start date and legal-entity date into one number. The March 2024 Form D is also the best public snapshot of Figure's formal governance. It listed Brett Adcock as both an executive officer and director; Dana Berlin, Lee Randaccio and Logan Berkowitz as executive officers; and Colby Adcock, Jesse Coors-Blankenship, Grant Hosking, Peter Welinder and Gregg Hill as directors. The filing did not give their job titles, committee assignments, voting rights or independence status. It should not be treated as a July 2026 board roster: Form D records the related persons relevant to that offering at the time it was filed. The composition provides useful context for the Series B. It included the founder, operating executives, representatives connected to Figure's repeat lead investor and a senior figure from its then-current AI partner. That can help a young company move quickly and connect financing to technical resources. It does not tell outsiders whether the board had a lead independent director, a dedicated safety committee or formal succession arrangements. Figure remains a private company, so it does not file annual proxy statements or public 10-K governance disclosures. It also does not publish a complete current leadership or board page. For a company valued at $39 billion, building machines intended to work beside people and potentially inside homes, that is a meaningful transparency gap. A current roster covering the board, senior engineering, manufacturing, safety, legal and commercial leadership would make accountability easier to assess. Culture, hiring bar and retention risk Figure's culture page is unusually explicit. It says the company will hire, recognize, reward and fire based on its values, warns that the culture is not for everyone and says it is not seeking candidates whose priority is minimizing workload or maximizing compensation. Its five published values are Move Fast & Be Technically Fearless, Product First Mission Focused, Aggressively Optimistic, Maximize Future Impact and Championship Mindset. That operating philosophy helps explain the product cadence. Figure completed three robot generations, built an AI stack and stood up a factory system in roughly four years. The company also says it wants in-person collaboration, minimal bureaucracy and a team oriented around one mission. Clear expectations can be an advantage in a technically demanding startup because candidates know the intensity before joining. The trade-off is organizational. A culture built around speed, exceptional commitment and aggressive optimism can make it harder to surface dissent, retain people through repeated hardware cycles or distinguish a prudent safety objection from resistance to execution. That does not establish a problem at Figure; it identifies a governance question that matters more in physical robotics than in ordinary software. The healthiest version of this culture would pair speed with strong stop-work authority, documented escalation paths and leaders rewarded for finding failures early. Adcock reported in December 2025 that Figure had received 176,000 job applications over three years and hired about 425 people. If those self-reported figures use the same denominator, that is roughly a 0.24% hiring rate. It shows enormous applicant interest and a narrow funnel, but it is not an employee-retention metric. Figure has not published voluntary attrition, average tenure, offer-acceptance rates or the share of hires assigned to AI, hardware, manufacturing, deployment and robot operations. Funding history The disclosed rounds total more than $1.745 billion. That sum is based on company announcements and does not assume undisclosed seed financing. The progression is extraordinary: Figure's valuation moved from $2.6 billion at Series B to $39 billion post-money roughly 19 months later, a fifteen-fold step-up. Hardware progress and the Helix story advanced during that period, but public operating disclosures did not grow at the same rate. Parkway Venture Capital is the central financial backer. It led the Series A and Series C and participated in the Series B. The repeat support matters because it signals sustained conviction across hardware generations. It also means the investor map is not a collection of one-off celebrity checks; Figure has a lead investor willing to finance the company through multiple stages. The Series B brought a second kind of value. Microsoft offered Azure infrastructure; NVIDIA and Intel connected Figure to compute and silicon ecosystems; OpenAI supplied brand and an announced model-development partnership; Jeff Bezos invested through Bezos Expeditions. The OpenAI collaboration ended within a year, which is a useful warning: a strategic investor logo can create optionality without guaranteeing a lasting operating relationship. Who invested in Figure AI, and why the mix matters Figure's investors fall into four groups. Financial investors such as Parkway, Align, Tamarack, ARK, Aliya, Bold and FJ Labs are underwriting company value. Compute and component companies such as NVIDIA, Intel, Qualcomm, Microsoft and LG can support training infrastructure, chips, connectivity and manufacturing relationships. Brookfield can provide real estate, operating environments and potential deployment sites. Salesforce, T-Mobile and Macquarie bring enterprise distribution, communications and capital-markets reach. Brookfield is the clearest example of money tied to an operating thesis. Its 2025 agreement covers human-video data collection across varied properties, infrastructure for Helix and potential commercial deployments. That could give Figure access to homes, offices, logistics sites and other real-world environments that are difficult for a robotics startup to assemble alone. It is still a partnership plan, not a published fleet result. NVIDIA and Intel participated in more than one round, while Qualcomm and LG joined the Series C. Their presence is strategically sensible because humanoids require training compute, on-device inference, sensors, communications and power electronics. It does not mean those companies have validated Figure's unit economics or safety performance. Investors validate willingness to invest at negotiated terms; customers validate willingness to pay for useful work. What the Series B securities filing adds The Series B press release rounded the round to $675 million. Figure's SEC notice supplies the transaction mechanics: the company offered $675 million of equity under Rule 506(b), reported $674,981,967 sold, only $18,033 remaining and 45 investors in the offering. The first sale was recorded on February 27, 2024, and the notice was signed by Adcock as CEO on March 12. The Form D is a notice of an exempt offering, not an audited financial statement. The SEC explicitly warns that it has not necessarily reviewed the information for accuracy or completeness. Still, it is a primary legal filing and more precise than copying the financing announcement. It confirms the investor count, exact amount sold, legal issuer and related persons while leaving revenue, share rights and operating results undisclosed. Valuation, private stock and IPO status The $39 billion figure is a post-money private valuation from the September 2025 Series C. It is not market capitalization, revenue or a daily price set by public trading. Private valuations can reflect preferred-share rights, negotiated terms and a small amount of capital relative to the full implied company value. Without audited financials, readers cannot calculate a credible revenue multiple or compare the valuation with mature robotics businesses. The Wall Street Journal reported in April 2025 that investor materials showed no revenue in the prior year, only a few dozen robots in production and forecasts of more than 200,000 robots and $9 billion in revenue by 2029. Those figures predate the later Series C and BotQ ramp. They are useful as evidence of how aggressively the company was being marketed, not as current results or guidance that Figure has publicly reaffirmed. Figure AI is not publicly traded and has no ticker. In July 2026 the company warned that unauthorized attempts to sell its shares may be invalid and said transfers require board approval. Secondary-market listings and special-purpose vehicles do not create the same liquidity, disclosure or investor protections as a public exchange. Figure has not announced a confirmed IPO date. Private-share marks: useful evidence, not a public price A few regulated investment funds disclose their Figure holdings. Those filings offer rare per-share reference points, but none is equivalent to a quoted market price. The securities are restricted, different preferred series may carry different rights, and the funds classify the positions as Level 3 or otherwise valued with significant unobservable inputs. A reported fair value is an accounting estimate at a date, not proof that the entire company could trade there. The spread between $135 and about $157 for two Series B references illustrates why secondary headlines need care. The marks come from different holders and dates, and they may use different valuation procedures. The Series C figure is higher still, but preferred-share seniority and contractual rights can differ across rounds. Dividing a fund's value by its share count is informative; multiplying that result by an unknown fully diluted share count would not produce a defensible company valuation. The filings do support one conclusion: some regulated funds have materially marked up Figure's earlier shares, and investor demand extends beyond the companies named in press releases. They do not show liquidity. Figure's own July 2026 notice says transfers require board approval and warns that unauthorized sale attempts may be invalid. A buyer in an SPV may own an interest in the vehicle rather than Figure shares directly, with additional fees, transfer restrictions and counterparty risk. The $39 billion valuation also needs the right denominator. The 2025 investor materials reported by The Wall Street Journal projected $9 billion of revenue in 2029. Dividing $39 billion by that projection gives about 4.3 times a four-year-forward revenue target, not a current revenue multiple. The same valuation is 15 times Figure's $2.6 billion Series B value. Without current revenue, gross margin, cash burn and dilution data, neither comparison is enough to decide whether the Series C price was justified. Figure's robot portfolio The Figure 01 review covers the first platform. Its importance is historical: Figure says it took its first steps in May 2023, roughly a year after the company was founded. Figure 01 supported early proof-of-concept work, but it was not a commercial consumer product. The Figure 02 review covers the model that moved into BMW's Spartanburg plant. Figure integrated the battery into the torso, improved the hands and refined the package for deployment. The company later described forearm electronics and dynamic cabling as reliability lessons that informed the Figure 03 wrist design. Figure 03 is a ground-up redesign with lower-cost manufacturing processes, palm cameras, tactile fingertips, a lighter body, softer external materials and wireless charging. Those changes serve both the factory and home stories. For current dimensions, specifications, availability and price status, use the dedicated Figure 03 review and Figure 03 database record. Helix: Figure's in-house AI strategy Figure's strategic pivot in 2025 was to make robot intelligence a core internal product. The company had announced an OpenAI collaboration alongside the Series B in February 2024. By February 2025 Adcock said Figure would end the arrangement and rely on its own models. He argued that large language models were becoming more commoditized while high-frequency robot control remained the harder problem. Whatever the relationship's internal history, Helix made Figure responsible for the entire perception-to-action stack. The first Helix release used two layers. System 2 was a 7-billion-parameter vision-language model that interpreted images and instructions at roughly 7 to 9 Hz. System 1 was an 80-million-parameter visuomotor policy that converted that context into upper-body action at 200 Hz. Figure said it used about 500 hours of teleoperated training data and controlled 35 degrees of freedom across the head, torso, wrists, hands and fingers. Helix 02 added a third layer for whole-body autonomy. System 0 is a 10-million-parameter controller running at 1 kHz, trained in simulation using more than 1,000 hours of retargeted human-motion data and 200,000 simulated environments. System 2 selects and interprets the goal, System 1 maps perception to desired movement, and System 0 handles balance, contact and fast joint control. The design lets walking and manipulation become one continuous learned behavior rather than separate scripted modes. The demonstrations are broad. Figure has shown robots sorting packages, putting away groceries, folding towels, loading a dishwasher, making a bed and coordinating a two-robot room reset. That breadth is relevant because general-purpose economics depend on reusing the same hardware and model across many tasks. The weak point is measurement. Figure rarely publishes task-success distributions, intervention rates or performance over long unscripted periods in homes. TIME's visit before the Figure 03 launch gives useful context. Reporters saw towel-folding and laundry demonstrations fail before later succeeding, and Figure acknowledged that the robot was not ready for domestic use at launch. That does not negate the successful demonstrations. It shows why edited clips and best-case runs cannot stand in for reliability data. A home robot must handle clutter, children, pets, reflective surfaces, bad lighting, dropped objects and thousands of edge cases without a technician standing nearby. Patents and the intellectual-property moat Figure's public patent record has become substantial enough to examine. Justia's Figure AI assignee page listed 13 U.S. grants as of July 16, 2026, with grant dates from July 2025 through June 2026. The list spans industrial design and mechanical architecture, not merely a logo or outer shell. It includes hands and end effectors, head and neck assemblies, central-body and hip geometry, ankle mechanics, whole-body control and an AI action-model architecture. The bipedal-action-model grant is particularly relevant to Figure's full-stack claim. It describes a larger model that processes language and a smaller model that produces robot-control output, trained in a unified framework. The named inventors include Corey Lynch, Yevgen Chebotar, Toki Migimatsu and Michael Ahn. Other grants identify a wider bench across mechanical design, hands, locomotion, industrial design and control. The pending torso application adds details that marketing pages do not emphasize. Its claims cover a rechargeable battery in a 1.5 to 5 kWh range, internal airflow past compute or battery cooling devices, and energy-absorbing protection around an arm interface. Patent claims are drafted broadly and do not confirm the exact production configuration, but they show which engineering problems Figure considered worth protecting. A patent portfolio is evidence of original work and a potential defensive asset. It is not proof of technical superiority, freedom to operate, enforceability or customer value. Competitors can design around claims, patents can be challenged, and the hardest advantage may sit in manufacturing know-how, fleet data and model-training operations that never appear in a patent. Figure's likely moat is the combination of hardware IP, tacit production knowledge, deployed-robot data and capital, not any single grant. BotQ manufacturing BotQ is Figure's attempt to solve the hardware scale problem in-house. Announced in March 2025, the facility was designed for initial capacity of up to 12,000 humanoids per year. Figure also set a goal of producing 100,000 robots over four years and eventually using its own robots in the manufacturing process. Capacity and goals are not the same as output, orders or customer deliveries. The April 2026 production update was more concrete. Figure said it had produced more than 350 Figure 03 robots, improved final assembly from one robot per day to a one-hour cycle and reached end-of-line first-pass yield above 80%. It also reported 99.3% first-pass yield on the battery line, more than 500 battery packs, over 9,000 actuators and more than 80 functional tests on each completed robot. Those numbers show a real production system, not a lab assembling one hero prototype. They are still company-reported and do not reveal the commercial mix. Figure allocates robots to AI research, data collection, home testing and commercial development, so more than 350 produced does not mean more than 350 sold. The decisive manufacturing metrics will be field failure rates, warranty cost, repair time, component life and the percentage of output accepted by paying customers. Factory capacity, quality control and the service burden BotQ's headline capacity is 12,000 robots per year, but capacity, demonstrated line speed and completed output are different measurements. Figure reported more than 350 Figure 03 units produced by April 29, 2026 and said it had demonstrated a one-robot-per-hour cycle time. One robot per hour sustained for one eight-hour shift on 250 working days would equal 2,000 robots a year; sustained around the clock for 365 days would equal 8,760. Reaching 12,000 would require more parallel capacity, a faster effective cycle or additional shifts and lines. Figure calls 12,000 the capability of its first-generation manufacturing line, not current annual deliveries. The manufacturing architecture is more sophisticated than a prototype workshop. Figure says its custom manufacturing execution system links production data with more than 150 networked workstations, part genealogy and component test results. It uses product-lifecycle, enterprise-resource and warehouse-management software, qualifies suppliers against incoming inspection criteria and performs thousands of-cycle burn-in exercises on completed robots. Figure 03 was redesigned around injection molding, die casting, metal injection molding and stamping instead of relying heavily on slow CNC machining. Figure says parts that took more than a week to machine can be produced in under 20 seconds with tooling. That shift is necessary for cost and volume, but it moves expenditure forward into molds, supplier qualification and line equipment. The economics work only if volume becomes large enough to amortize the tooling and designs remain stable long enough to use it. Service may be as difficult as assembly. Figure says it has built field-service and fleet-management systems, over-the-air updates, alerting, failure analysis, fallback ladders and processes for fleet-wide upgrades and recalls. Those are the right systems for commercial deployment. They are also indirect evidence of the support burden: a robot in a customer's factory or home creates maintenance logistics, spare-parts inventory, software rollback, technician training and incident-response obligations that do not exist for an internal demo. The missing unit-economic bridge is labor. Figure has not disclosed assembly hours per robot, service technicians per deployed unit, average repair time, warranty reserve, cost of spare parts or the share of robot-operating hours that require remote assistance. A one-hour factory cycle does not mean one labor-hour of total cost, and an 80% first-pass yield does not reveal the cost of bringing the remaining units to acceptance. From eight hours of training data to an eight-hour shift Package handling provides a useful view of how Figure develops a task. In February 2025 the company said eight hours of carefully selected human demonstrations were enough to train a Helix policy to grasp, reorient and move varied packages on a conveyor. The task forced the model to handle boxes and deformable bags, find labels, judge depth and timing, transfer across robots and recover when a grasp was imperfect. In May 2026 Figure streamed three Figure 03 robots working through an eight-hour package-sorting shift under Helix 02. Adcock described the run as fully autonomous and at human performance levels, with packages oriented barcode-down for scanning. TechRadar reported that the robots appeared to complete the work mostly as required, noted a small number of exceptions in the footage its writer reviewed and recorded continuing skepticism from observers about speed and possible teleoperation. The livestream is stronger evidence than a short edited clip because viewers can observe extended operation and imperfect cases. It is still not a customer audit. Figure did not publish the complete run log, intervention count, throughput distribution, downtime, energy use, robot swaps, package mix or exception taxonomy. Three robots taking turns also makes it important to separate cell throughput from per-robot throughput. The right next step is not a longer promotional video. It is a dated operating table showing total packages attempted, success without human touch, recoveries, interventions, faults, mean repair time, energy per productive hour and throughput at each percentile. The same standard applies at BMW: the move from Figure 02 sheet-metal loading to Figure 03 sequencing is technically meaningful, but commercial judgment requires repeatable output, not capability alone. Customers, deployments and partnerships BMW is Figure's strongest customer reference. The companies announced a commercial agreement in January 2024 and began by identifying suitable automotive tasks. Figure later reported that Figure 02 robots worked 10-hour weekday shifts over an 11-month deployment, logged more than 1,250 operating hours, handled over 90,000 sheet-metal parts and contributed to production of more than 30,000 BMW X3 vehicles. BMW independently confirmed the underlying production work. It described Figure 02 inserting sheet-metal parts into fixtures at Plant Spartanburg and later said the robot supported production of more than 30,000 X3 vehicles. Figure's detailed hours, step count and parts totals remain company-reported, but this is much stronger evidence than an unnamed pilot or a demo inside Figure's office. Figure 03 arrived at the same plant in June 2026 for a sequencing workflow in Hall 52. The task combines part selection, dexterous handling, carrying and whole-body movement, including moving a wheeled cart. It appears more complex than Figure 02's earlier placement task. Figure has not yet published an equivalent multi-month report showing Figure 03 uptime, interventions and sustained cycle performance. Catalyst Brands is the next named deployment path. Its May 2026 agreement targets physically demanding distribution and logistics work at a Reno facility. Catalyst's retail portfolio includes JCPenney, Aéropostale and Brooks Brothers. The announcement does not specify fleet size, commercial terms, deployment dates or measured results, so it should be classified as a signed agreement rather than a proven operating fleet. Brookfield is an investor and strategic partner, not yet a comparable customer case study. Its agreement covers data collection across varied environments, AI infrastructure and potential deployments in Brookfield properties. The partnership may be strategically valuable because real homes and commercial buildings provide more diverse training data than a robotics lab. Public metrics will determine whether it becomes a deployment channel or stays at the development stage. What is independently verified? Figure's evidence is better than its critics sometimes imply and weaker than its marketing can make it appear. A named automaker allowed robots onto a live production line and later confirmed the work. That matters. The company has also published specific architecture and manufacturing numbers rather than relying only on cinematic videos. At the same time, most details originate with Figure and few have been audited. The standard should rise as the valuation and claims rise. A startup proving its first robot can move parts deserves latitude. A $39 billion company discussing homes, mass manufacturing and hundreds of thousands of robots should publish fleet uptime, intervention frequency, safety incidents, customer concentration and economics per productive hour. Business model and commercial prospects Figure has chosen a sensible first market. Factories and distribution centers offer repeatable workflows, controlled access, professional maintenance and measurable labor costs. A robot does not need human-level general intelligence to create value if it can perform one high-frequency task safely for most of a shift. BMW provides Figure with an environment where reliability, cycle time and part quality are visible. The industrial route can also finance learning. Each deployment produces failure cases, maintenance records and task data. If Figure can reuse that learning across customers without expensive site-specific engineering, margins should improve as the fleet grows. If every installation needs a large integration team and persistent teleoperation, the supposed general-purpose advantage shrinks. Homes offer a much larger theoretical market and a harsher product test. Consumers need an affordable machine that is quiet, safe, privacy-preserving, easy to insure and useful across many chores. They will not tolerate factory-style safety cages, frequent technician visits or long calibration procedures. Figure 03's softer design and wireless charging are steps toward that product, but there is no public price, order page, service plan or broad release schedule. Commercial success will be visible in a few simple metrics: repeat orders from named customers, growing paid fleet size, declining human interventions, high uptime, shorter repair cycles and a cost per useful hour below human or fixed-automation alternatives. Figure has not disclosed enough of these metrics to model the business with confidence. Figure AI competitors Figure's main advantage is integration plus deployment evidence. Tesla has far greater capital and manufacturing infrastructure, but its Optimus program still exposes limited external customer data. Agility has the strongest pure-play logistics posture and more named commercial deployments. Apptronik can use Jabil's global manufacturing network rather than build every production capability internally. Each model tests a different scaling strategy. The home race is particularly revealing. 1X accepts orders at disclosed prices and explicitly describes a remote Expert Mode for tasks the robot cannot perform alone. Figure has published stronger industrial evidence but no consumer offer. Unitree sells lower-cost hardware to developers, which can build ecosystem and data even when the product is not a turnkey household worker. Figure is asking the market to believe it can bridge enterprise reliability and home generality before either market is mature. Risks and unresolved questions Valuation risk. A $39 billion post-money price assumes Figure captures a substantial share of a future humanoid market. Public revenue, gross margin, cash burn and backlog are absent. Even strong engineering progress can coexist with a future down round if commercialization takes longer than investors expect. Safety risk. A humanoid is a heavy, powered machine operating around people. A former head of product safety sued Figure in federal court in November 2025, alleging he was fired after raising concerns about injury risks. Figure disputes the claims in active litigation; the court has not adjudicated them. The Northern District of California docket showed continuing filings through July 6, 2026. The allegations should not be treated as findings, but they make Figure's limited public safety documentation more consequential. Home privacy risk. A useful household robot sees rooms, faces, possessions, routines and conversations. Adcock told TIME that Figure intended to use data from home robots to improve future models and described plans to scrub personal information. Figure has not published a detailed consumer data-retention, consent or model-training policy tied to a retail product because that product is not yet on sale. Reliability risk. Home demos and short clips do not reveal failure distributions. BMW proves that a narrow task can run in production, but a general household worker needs far broader reliability. Figure should publish task success over hundreds of attempts, mean time between failures, remote-assistance rates and recovery behavior after dropped or misplaced objects. Customer-concentration risk. BMW is the only customer with substantial public operating metrics. Catalyst is an announced agreement and Brookfield is a strategic pathway. Until several independent customers expand from pilots to repeat orders, Figure's commercial traction remains concentrated. Manufacturing risk. BotQ's ramp is impressive, but producing hundreds of units is different from supporting tens of thousands in the field. Component durability, supplier quality, repair logistics, warranty reserves and spare-parts availability become harder as deployments spread. Leadership and governance risk. Figure is founder-led, does not publish a current leadership roster and has already seen a prominent CTO depart. Adcock's simultaneous leadership of Hark increases the need for clarity about executive responsibility, conflicts, board oversight and how technology or people move between the companies. AI-strategy risk. Bringing Helix in-house gives Figure control and differentiation, but it also concentrates model, data and safety responsibility inside one company. The ended OpenAI collaboration shows that strategic relationships can change quickly. Figure must fund frontier AI, robot hardware and manufacturing at the same time. Safety lawsuit: procedural status in July 2026 A former principal robotics safety engineer, Robert Gruendel, sued Figure in November 2025, alleging retaliation after raising safety concerns. Figure disputes the claims and has brought related counterclaims. The case is Gruendel v. Figure AI, Inc., No. 5:25-cv-10094, in the U.S. District Court for the Northern District of California. The allegations are not findings of fact, and the case has not reached a merits judgment. A July 6, 2026 discovery order addressed a collection of Figure documents retained by Gruendel. The magistrate judge directed Figure to review and produce responsive, non-privileged materials under specified procedures, ordered the collection preserved, set July 17 for production of non-privileged documents referenced in parts of the complaint and required the parties to submit a discovery plan by July 24. Fact discovery is scheduled to close November 14, 2026. That order is procedural. It does not validate the complaint's safety allegations or Figure's counterclaims. Its relevance to this review is narrower: the dispute is active, discovery is under way and public information may change as documents are produced or later motions are decided. Any article that says the lawsuit was resolved, dismissed or proved without a later docket entry would be inaccurate. The broader diligence issue remains safety governance. Figure says it created dedicated safety and reliability teams during the Figure 03 manufacturing redesign, and a granted control patent includes a safety-related limit. Those are positive signals. Public reporting still lacks incident-rate data, external safety certification, a named senior safety owner, red-team results for home use and a clear explanation of who can halt a deployment. The diligence questions Figure still does not answer Figure does not need to publish every customer contract or proprietary model detail. It could answer most of these questions with ranges, standardized fleet metrics and an annually refreshed governance page. Until then, the company deserves credit for technical velocity and real deployment evidence, while the valuation and scale claims deserve a larger discount for missing operating data. What to watch next The most important next disclosure is not another household video. It is a Figure 03 fleet report covering several months of paid work: robot count, scheduled hours, autonomous hours, interventions, task success, safety events, repair time and cost per productive hour. A comparable Catalyst report would show that BMW was not a one-customer exception. Watch BotQ for customer shipment mix rather than gross output. Watch the careers page for field-service and deployment hiring outside California. Watch for a public safety framework before any broad home program, and for consumer terms that explain price, remote assistance, data use, warranty and liability. On the financing side, a tender, new round or IPO filing would reveal whether the private valuation can withstand greater disclosure. Final verdict Figure AI is a credible robotics company with an aggressive valuation, not vaporware and not a proven mass-market business. The team has moved unusually fast, BMW validates that its robots can perform useful production work, Helix gives the company a coherent AI architecture, and BotQ shows preparation for scale. Very few humanoid startups can point to all four. The weak side of the case is commercial transparency. Figure does not disclose enough about revenue, pricing, paid fleet size, customer concentration, interventions, field reliability or unit economics. Its home story is several steps ahead of the evidence. The active safety litigation, sparse leadership disclosure and Adcock's second CEO role add governance questions that a comprehensive review cannot ignore. So the rating is conditional: top-tier technical contender, early commercial operator, speculative valuation. Figure deserves close attention because it has built and deployed real machines. It does not yet deserve the assumption that mass household adoption, attractive margins or a $39 billion outcome are inevitable. Methodology and source standard We checked Figure's company, careers, product, AI, manufacturing, funding, partnership and stock-transfer materials; BMW's independent statements; contemporaneous reporting from AP, TIME, Forbes, TechCrunch, The Wall Street Journal and Bloomberg Law; the federal court docket; and official competitor materials. Company figures are labeled as company-reported unless a customer or independent source confirms them. Funding announcements establish transaction terms, not business performance. Allegations in active litigation are identified as allegations. This page was fact-checked on July 16, 2026. Figure is private and the market moves quickly, so customer metrics, team size, financing and legal status may change. The dedicated robot pages remain the source for model-level specifications and availability updates. --- # Honor Humanoid Robot: What Honor Actually Revealed at MWC 2026 URL: https://blog.robozaps.com/b/honor-humanoid-robot Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T23:07:40.880Z Last fact-checked: 2026-07-06T00:00:00.000Z Category: news TL;DR: Honor unveiled its first humanoid robot at its March 1, 2026 MWC launch as part of a broader AI device ecosystem push. Official sources confirm the category and target scenarios, but not price, release date, full specs, or a verified top-speed ranking. Key takeaways: - Honor's humanoid robot is real, but public information remains limited. - Official Honor copy positions the robot around shopping assistance, workplace inspections, and supportive companionship. - Speed claims should be treated as reported unless Honor publishes a full public spec sheet. - Pricing, availability, payload, battery life, autonomy level, and commercial launch timing remain unknown. FAQ: Q: When did Honor reveal its humanoid robot? A: Honor unveiled it during its March 1, 2026 MWC launch event in Barcelona. Q: How fast is the Honor humanoid robot? A: Honor has not published a full public spec sheet. Some pre-launch reports cited up to 4 m/s, but this should be labeled as reported rather than treated as a confirmed ranking. Q: How much will Honor's humanoid robot cost? A: Honor has not announced pricing or availability for the humanoid robot. Do not rely on speculative consumer price ranges. The reveal has already happened This article used to read like a preview. That made it stale. Honor's MWC 2026 launch happened on March 1, 2026, and Honor says it unveiled its first humanoid robot at the event. The safest current version is a post-event recap: Honor has entered humanoid robotics, but the official public details are still thin. The company has shared broad scenarios and strategic framing, not a full spec sheet or buyer-ready launch plan. What Honor confirmed In its MWC 2026 launch release, Honor says its first humanoid robot is part of a wider AI device ecosystem push. The company positions the robot around three core scenarios: shopping assistance, workplace inspections, and supportive companionship. That is meaningful because Honor is not a traditional robot maker. It brings mobile hardware, cameras, batteries, sensors, device ecosystems, and consumer UX experience. But none of that proves a finished humanoid product by itself. What remains unknown Price and availability.Payload, runtime, speed, size, weight, and full mechanical specification.Autonomy level and whether remote human assistance is involved.Safety certification, warranty, service model, and target launch markets.Whether the robot is a commercial product, a platform preview, or an ecosystem signal. Be careful with speed and first-mover claims Some pre-launch coverage reported that Honor's robot could reach up to 4 m/s. Without a public Honor spec sheet, that should stay in the "reported" bucket, not become a headline fact or fastest-robot ranking. The same caution applies to "first smartphone brand" claims. Honor is one of the most visible smartphone companies to move into humanoid robotics, but broad first-mover language is risky without carefully defining the peer group. Robot Phone is separate Honor also previewed a Robot Phone at MWC 2026, and the company published more detail about that product concept than about the humanoid. Keep the two separate. Robot Phone is a phone concept with embodied motion; the humanoid robot is a broader robotics move with fewer public details. Bottom line Honor's humanoid robot is a serious strategy signal, not yet a buyer guide. The page should track confirmed scenarios and missing evidence, then wait for pricing, specs, safety, and launch data before making stronger claims. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # Noetix Bumi Price: World's Cheapest Humanoid Robot at $1,400 (2026) URL: https://blog.robozaps.com/b/noetix-bumi-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Category: reviews TL;DR: The Noetix Bumi costs $1,400, making it the world's cheapest humanoid robot: a 94 cm, 12 kg child-sized robot with 21 degrees of freedom that walks, dances and responds to voice. It sold out within 48 hours of launch and remains hard to get outside China. It is built for education and entertainment, not household labor. Key takeaways: - World's cheapest humanoid at $1,400 — Less than an iPhone, more than a robot - 94 cm tall, 12 kg — Child-sized, safe for home and classroom environments - 21 degrees of freedom — Can walk, run, dance, and interact via voice/touch - Sold out in 48 hours — High demand, limited availability outside China - Spring Festival Gala appearance — Demonstrated capabilities to 700+ million viewers - Best for education and entertainment — Not designed for household labor FAQ: Q: How much does the Noetix Bumi cost? A: The Noetix Bumi costs ¥9,998, which is approximately $1,400 USD. This makes it the world's cheapest humanoid robot—less expensive than an iPhone 17 Pro Max. The price reflects Noetix's vertical integration, lightweight materials, and nearly all domestic Chinese supply chain. Q: Can the Noetix Bumi do household chores? A: No. Noetix explicitly positions the Bumi for education and family entertainment, not household labor. The robot lacks the payload capacity, manipulation dexterity, and sensor suite needed for tasks like cooking, cleaning, or carrying objects. Think of it as a companion and educational tool. Q: Is the Noetix Bumi available in the US or Europe? A: Not currently. As of March 2026, the Bumi is only available in China through JD.com. Noetix has an English-language website suggesting international expansion is planned, but no specific timeline or sales channels have been announced for North America or Europe. Q: How tall is the Noetix Bumi? A: The Bumi stands 94 cm (3.1 feet or 37 inches) tall and weighs 12 kg (26.5 lbs). This child-sized form factor was intentional—designed to avoid the "uncanny valley" effect and fit comfortably in homes and classrooms without intimidating children. Q: What can the Noetix Bumi actually do? A: The Bumi can walk, run, and dance using bipedal locomotion. It responds to voice commands, supports touch interaction, and includes facial recognition via its front camera. Developers can program custom behaviors through its open interfaces. It was demonstrated live at China's Spring Festival Gala in March 2026. Q: How does the Bumi compare to the Unitree R1? A: The Bumi ($1,400) is more than 3x cheaper than the Unitree R1 ($4,900). However, the R1 is taller (123 cm vs 94 cm), heavier (25-29 kg vs 12 kg), and designed for research/developer use with more degrees of freedom. The Bumi targets consumers and educators; the R1 targets serious roboticists. Q: Is Noetix a legitimate company? A: Yes. Noetix Robotics was founded in September 2023 in Beijing by Jiang Zheyuan, a Tsinghua University researcher. The company has raised over $41 million in funding from investors including Vertex Ventures, achieved positive cash flow in 2024, and delivered real products—including the N2 robot that finished second in the world's first humanoid half-marathon. The Noetix Bumi is officially the world's cheapest humanoid robot—priced at just $1,400 (¥9,998), this child-sized bipedal robot from Beijing startup Noetix Robotics is cheaper than an iPhone 17 Pro Max. In February 2026, the Bumi made international headlines when it appeared at China's Spring Festival Gala alongside human actors, performing in front of 677 million viewers. The robot sold out within days of its October 2025 launch, with 500 units snapped up in just 48 hours on JD.com. In this comprehensive Noetix Bumi review, we'll cover everything you need to know: full specifications, real-world performance, how Noetix achieved this breakthrough price, and how it compares to competitors like the Unitree R1 ($4,900). Whether you're an educator, hobbyist developer, or simply curious about owning a humanoid robot, the Bumi represents a genuine inflection point in consumer robotics. Noetix Bumi Specifications About Noetix Robotics: The Company Behind Bumi Noetix Robotics (officially Songyan Dynamics Beijing Technology Co., Ltd.) was founded in September 2023 by 27-year-old Jiang Zheyuan, who left his doctoral studies at Tsinghua University to commercialize humanoid robotics technology. The founding team includes engineers from China's top institutions, including Tsinghua University and Zhejiang University. Unlike many robotics startups that focus on showcase demonstrations and long development timelines, Noetix prioritized rapid commercialization from day one. This pragmatic approach paid off—the company achieved positive cash flow in Q4 2024 and has completed five funding rounds, culminating in a Pre-B round of nearly $41 million led by Vertex Ventures in October 2025. The company gained international attention in April 2025 when its larger N2 robot finished second in the world's first humanoid half-marathon in Beijing, completing 21 kilometers in 3 hours and 37 minutes. This achievement generated over 2,500 pre-orders and tripled Noetix's valuation. The Bumi represents their strategic move to capture the mass-market consumer segment. How Noetix Achieved the $1,400 Price Point At $1,400, the Bumi costs roughly the same as an iPhone 17 Pro Max—a remarkable achievement for a walking, talking humanoid robot. Founder Jiang Zheyuan has been transparent about the three engineering pillars that made this possible: 1. Vertical Integration Noetix designs its own control boards and motor drivers in-house rather than purchasing standard modules. This eliminates supplier markups and allows tight hardware-software optimization. By controlling the entire stack, they can make engineering trade-offs that commercial component suppliers can't. 2. Structural Redesign The team adopted lightweight composite materials with metal reinforcement only where structurally necessary. Cutting total weight to just 12 kg created cascading cost savings: lighter frames need smaller motors, smaller motors need smaller batteries, and the reduced component count simplifies assembly. 3. 100% Domestic Supply Chain Almost every component—from motors and sensors to the Rockchip processor—is sourced within China. This localized supply chain provides faster iteration cycles, lower logistics costs, and significant price advantages over international competitors managing complex global supply chains. Spring Festival Gala: The Bumi's TV Debut In March 2026, the Noetix Bumi achieved something no consumer humanoid robot has done before: a live televised performance at China's Spring Festival Gala, the world's most-watched television broadcast with over 700 million viewers. Four Noetix humanoid robots appeared alongside human actors in a comedy skit, demonstrating their ability to walk, gesture, and interact in an unscripted entertainment environment. The performance aired on Lunar New Year's Eve (February 16, 2026) and featured robots from four Chinese companies: Unitree, MagicLab, Galbot, and Noetix. According to China Daily, orders for Chinese-made humanoid robots "surged" following the gala, with delivery dates for some models delayed until April 2026. For Noetix specifically, the exposure validated their positioning: robots that aren't just athletic, but socially aware for daily life scenarios. Performance & Capabilities Locomotion The Bumi can walk and run on flat surfaces using bipedal locomotion with dynamic balance correction. Its 21 degrees of freedom—distributed across the legs, hips, torso, and arms—enable coordinated movement patterns including dancing and basic terrain adaptation. While not as capable as full-size industrial humanoids, the locomotion is genuinely impressive for a $1,400 robot. Interaction Bumi supports voice commands via its microphone array and includes touch-based interaction. The front-mounted camera enables facial recognition and basic object detection. The robot integrates with JD.com's Joy Inside 2.0 ecosystem and offers open programming interfaces for developers who want to extend its capabilities. What It Can't Do Noetix has been clear about Bumi's intended use: education and family entertainment, not household labor. The robot lacks the payload capacity, dexterity, and sensor suite needed for tasks like cooking, cleaning, or heavy lifting. Think of it as a sophisticated companion and educational tool, not a household assistant. Who Should Buy the Noetix Bumi? Ideal For: Educators and schools — A genuine humanoid robot at a price point schools can affordHobbyist developers — Open programming interfaces for experimentationSTEM programs — Hands-on robotics learning with real hardwareFamilies — Educational entertainment for children fascinated by robotsResearch labs — Affordable platform for proof-of-concept testing Not Ideal For: Anyone expecting household task assistanceCommercial/industrial applicationsBuyers outside China (limited availability)Users requiring robust outdoor operation Noetix Bumi vs. Unitree R1: Affordable Humanoid Comparison The closest competitor to the Bumi is the Unitree R1, priced at $4,900—still making the Bumi more than 3x cheaper. The Bumi wins on price and accessibility, while the R1 offers more capability for serious developers. For educational institutions with limited budgets, the Bumi is the clear choice. For research labs needing a more capable platform, the R1 justifies its higher price. Looking for more options? Check our best humanoid robots guide for a complete market overview. Pros and Cons Pros World's cheapest humanoid — $1,400 breaks all price barriersGenuine bipedal locomotion — Walks, runs, dances on flat surfacesChild-safe size — 94 cm avoids intimidation factorOpen programming interfaces — Developer-friendly for customizationProven company — Marathon performance, significant funding, real salesVoice and touch interaction — Multiple ways to engage Cons China-only availability — No official international salesCurrently sold out — High demand, production catching upNot for household tasks — Education/entertainment focusLimited battery life — 1-2 hours depending on activityYoung company — Founded 2023, limited track record Availability: How to Buy the Noetix Bumi Current Status: Sold Out The Noetix Bumi launched for pre-order in October 2025 on JD.com (China's major e-commerce platform), timed to coincide with the Double 11 and Double 12 shopping festivals. The first 500 units sold within 48 hours, and the robot is currently sold out with deliveries delayed until April 2026. For International Buyers: Unfortunately, Noetix has not announced international sales channels. The company's website (en.noetixrobotics.com) provides English-language information, suggesting future global expansion is planned, but no timeline has been announced. Getting Notified: Follow Noetix's official website for announcementsMonitor JD.com for restocks (requires Chinese address)Check Robozaps for availability updates The Bottom Line: Should You Buy the Noetix Bumi? The Noetix Bumi represents a genuine milestone in consumer robotics. At $1,400, it demolishes the price barrier that kept humanoid robots out of reach for individuals, schools, and small organizations. The fact that it sold out within 48 hours and appeared on the world's most-watched television broadcast validates both the technology and the market demand. If you're an educator looking to bring robotics into your classroom, the Bumi offers an affordable entry point into humanoid technology. If you're a hobbyist developer, the open programming interfaces make it an interesting platform for experimentation. If you're a family with robot-obsessed kids, this could be the most educational toy you'll ever buy. The limitations are real: it's not available outside China, it's currently sold out, and it won't do your dishes. But for what it is—the world's cheapest real humanoid robot—the Noetix Bumi delivers on its promise. For buyers ready to invest more, consider the Unitree R1 at $4,900 for greater capability, or explore our full humanoid robot buyer's guide for all available options. Last updated: March 2026 --- # MagicBot Price 2026: MagicLab Specs & Factory Deployment URL: https://blog.robozaps.com/b/magiclab-magicbot-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-08-01T23:07:45.274Z Category: reviews TL;DR: MagicLab's MagicBot has no public price: it is sold enterprise-only via contact sales, and units sold out on JD.com after its Spring Festival Gala appearance. The young company (founded December 2023) already runs multi-robot collaboration on production lines and raised hundreds of millions of RMB by May 2025. Gen1 specs remain partly undisclosed. Key takeaways: - Price: Contact sales (no public pricing); estimated industrial deployment cost significantly lower than Western competitors like Figure 02 ($250,000+) - Standout Feature: Multi-robot collaboration capabilities already deployed in production lines — rare for such a young company - Best For: Industrial automation (inspection, material handling, assembly), research institutions, and robotics developers seeking modular platforms - Key Limitation: Very new company (founded Dec 2023) with limited track record; specs not fully disclosed for Gen1 model - Availability: Sold out on JD.com following Spring Festival Gala appearance; contact-sales model for enterprise customers - Funding: Raised 150 million RMB (~$21M USD) in angel round (Dec 2024) + hundreds of millions RMB in Series B (May 2025) FAQ: Q: What is the price of MagicLab MagicBot? A: MagicLab does not publicly disclose pricing for either the MagicBot Gen1 or Z1 models. Both operate on a contact-sales model typical of industrial robotics. Based on comparable Chinese humanoids, estimated pricing likely ranges from $50,000-$100,000 USD for enterprise deployments, though this is unconfirmed. During the March 2026 Spring Festival Gala, MagicBot units sold out on JD.com within minutes, indicating strong demand despite undisclosed pricing. Q: Is MagicBot actually deployed in factories or just a demo? A: Yes, MagicBot Gen1 is confirmed to be deployed in real production environments. In December 2024, MagicLab released footage of multiple Gen1 robots collaborating at an electronics factory, performing product inspection, material transport, precision assembly, and barcode scanning. However, a company researcher stated the robots are "still in the skill training and learning phase" and not yet fully autonomous, suggesting deployments currently require human oversight. Q: How does MagicBot compare to Tesla Optimus? A: MagicBot Gen1 and Tesla Optimus Gen 2 target similar industrial automation markets but differ significantly in approach. MagicBot is already factory-deployed (as of December 2024) with 42 DOF and 20 kg per-arm payload capacity, while Optimus remains in Tesla's internal testing phase with limited third-party deployments. Tesla projects Optimus pricing at $20,000-$25,000 USD (if mass-produced), while MagicBot's pricing is undisclosed but likely higher. MagicBot's multi-robot collaboration (MagicNet) is a key differentiator, while Optimus benefits from Tesla's AI and manufacturing ecosystem. Q: What is the difference between MagicBot Gen1 and Z1? A: MagicBot Gen1 is a full-size (175 cm) industrial humanoid focused on heavy-duty tasks with 42 DOF, 20 kg per-arm payload, and 4-5 hour battery life. It's designed for factory automation and features IP66 weather resistance. MagicBot Z1 is a compact (140 cm) agile model prioritizing mobility over payload, with 24-50 DOF, 3 kg per-arm payload, 2.5 m/s walking speed, and advanced acrobatic capabilities. Z1 targets research, education, and dynamic applications like entertainment, while Gen1 focuses on industrial production environments. Q: Can I buy a MagicBot for personal/home use? A: MagicLab has not announced consumer or personal versions of MagicBot. Both Gen1 and Z1 are marketed to enterprise, industrial, and research customers through a contact-sales model. The robots sold on JD.com during the Spring Festival Gala were likely enterprise units, as evidenced by the high price points (e.g., Galbot G1 at roughly $91,000). For personal humanoid robots, consider alternatives like Unitree G1 ($13,500) or wait for Tesla Optimus consumer availability (timeline unconfirmed). Q: What software/SDK does MagicBot use? A: MagicBot operates on MagicLab's proprietary Magic Atom motion control platform, which supports high-fidelity humanoid movements, imitation learning (IL), reinforcement learning (RL), and multi-robot coordination via MagicNet. However, MagicLab has not released public SDK documentation or developer tools. The Z1 "Development Version" (50 DOF) offers "secondary development" support, suggesting API access for enterprise and research customers. The platform supports OTA (over-the-air) updates via WiFi 6, but detailed software capabilities remain undisclosed. Q: Is MagicBot partnering with ByteDance/TikTok? A: MagicLab confirmed in November 2024 that it is in talks with ByteDance (TikTok's parent company) to integrate the Doubao large language model into third-generation MagicBot robots. Doubao specializes in text, image, and video generation and is significantly cheaper than OpenAI's GPT-4. The integration would enable natural language task instructions, creative problem-solving, and video-based quality control. As of March 2026, the partnership has not been formally announced, and current MagicBot deployments do not appear to use Doubao or other LLM-based decision-making systems. Q: How long does MagicBot's battery last? A: Battery life varies by model. The MagicBot Gen1 provides 4-5 hours of operation per charge, suitable for half-shift industrial work with planned recharging intervals. The compact MagicBot Z1 has a significantly shorter 2-hour runtime (10,000 mAh, 15-cell battery), limiting its use in extended applications. Both models feature quick-release battery packs and support fast charging (62V 5A). The short Z1 battery life is a notable limitation for industrial or research applications requiring extended autonomous operation. Introduction: China's Rising Humanoid Star Takes Center Stage When six MagicBot Z1 humanoid robots performed synchronized dance moves with pop stars at China's 2026 Spring Festival Gala — watched by over 1.2 billion viewers — the world took notice. Within minutes, MagicLab's robots sold out on JD.com (there is no public MagicBot price; units sell enterprise-only), marking a watershed moment for what may be China's most ambitious humanoid robotics startup. Founded just over two years ago in December 2023, MagicLab (Magic Atom Robotics Technology) has rapidly evolved from a basement startup to a production-floor pioneer with robots already deployed in industrial settings. This MagicLab MagicBot review examines the specifications, capabilities, pricing, and real-world applications of both the full-size MagicBot Gen1 and the compact MagicBot Z1. Based on our analysis of technical documentation, factory deployment videos, and third-party assessments, we evaluate whether MagicLab's aggressive commercialization strategy justifies the market hype — or if this is another overpromised robotics moonshot. MagicLab MagicBot Specifications MagicLab offers two distinct humanoid platforms: the full-size MagicBot Gen1 (industrial focus) and the compact MagicBot Z1 (research/agile applications). Below are the confirmed specifications: Note: MagicLab has not disclosed full specifications for the Gen1 model. The above data is compiled from third-party reseller listings, factory deployment videos, and official Z1 documentation. Price and Availability: Sold Out in Minutes MagicLab does not publicly disclose pricing for either the Gen1 or Z1 models, operating on a contact-sales model typical of industrial robotics suppliers. However, the market reaction to their Spring Festival Gala appearance provides revealing pricing signals. JD.com Sellout and Market Demand On February 16, 2026, during the live broadcast of China's Spring Festival Gala, online retailer JD.com listed multiple humanoid robots for sale. According to CNN and Global Times reporting, robots from MagicLab, Unitree Robotics, and Noetix sold out within minutes of being listed. JD.com data revealed that within two hours of the broadcast: Robot searches surged 300%Customer service inquiries increased 460%Order volume jumped 150% While MagicLab has not confirmed the JD.com listing prices, similar Chinese humanoid robots on the platform range from approximately 50,000 RMB (~$7,000 USD) for compact research models to over 630,000 RMB (~$87,000 USD) for industrial-grade units like the Galbot G1. Price Comparison: MagicBot vs. Competitors Value Assessment: MagicLab appears positioned in the mid-tier pricing segment for Chinese humanoids — significantly more affordable than Western alternatives like Figure 02, but likely more expensive than Unitree's consumer-focused G1. The lack of transparent pricing is a barrier to adoption for smaller enterprises and research institutions. Performance and Mobility: From Factory Floors to Festival Stages MagicLab's robots demonstrate a rare combination of industrial robustness and entertainment-grade agility. The company's dual-model strategy — Gen1 for heavy-duty tasks, Z1 for dynamic movement — addresses different market segments effectively. MagicBot Gen1: Industrial Workhorse In December 2024, MagicLab released footage of multiple Gen1 robots collaborating on a production line at an unnamed electronics factory. The deployment video showed robots performing: Product inspection: Visual quality control using onboard cameras and force-torque sensors to detect defectsMaterial transport: Carrying components weighing up to 20 kg (44 lbs) per arm between workstationsPrecision assembly: Using dexterous 6-DOF hands with sub-millimeter precision (pressure sensors enable force feedback)Barcode scanning and inventory management: Autonomous navigation with real-time location tracking The Gen1's 525 N·m peak torque per joint enables handling of automotive-grade components — a capability typically reserved for purpose-built industrial cobots. According to MagicLab researcher Albert Wei, the robots are "still in the skill training and learning phase" and not yet fully autonomous, suggesting current deployments operate with human oversight. MagicBot Z1: Compact Acrobat The Z1 model prioritizes mobility over payload capacity. Standing just 1.4 meters (4.6 feet) tall and weighing 40 kg (88 lbs), the Z1 achieved several industry firsts during its Spring Festival Gala debut: Thomas 360 spin: First humanoid robot of similar size to complete a 360-degree breakdance spinWalking speed: Up to 2.5 m/s (9 km/h / 5.6 mph) — comparable to Unitree H1's performanceObstacle climbing: Can negotiate obstacles up to 15 cm (6 inches) highJoint rotation range: Up to 320° on certain joints (exceeds human range of motion)Knee torque: 130 N·m enables explosive movements like jumps and rapid directional changes The Z1's expandable DOF architecture (24 standard, up to 50 with optional hands and wrist modules) makes it particularly appealing for research institutions. The proprietary Magic Atom motion control platform reportedly enables training of new movements in as little as 24 hours using imitation learning (IL) and reinforcement learning (RL). Sensors and Perception: 360-Degree Awareness The Z1 model features MagicLab's most comprehensive sensor suite: 3D LiDAR: Enables real-time SLAM (simultaneous localization and mapping) for autonomous navigation in unstructured environmentsDepth camera (Intel RealSense D435): Provides RGB-D data for object detection and manipulation tasksDual fisheye cameras: 360-degree visual field enables full peripheral awareness — critical for multi-robot collaboration scenariosHead tactile sensor: Pressure-sensitive contact detection for safe human-robot interactionMicrophone array: Noise reduction and echo cancellation enable voice command recognition in noisy factory environments The Gen1 model uses a simpler sensor package (cameras and force-torque sensors), reflecting its focus on structured industrial environments where navigation paths are predefined. Industry Comparison: The Z1's sensor array is comparable to Boston Dynamics' Atlas (LiDAR + stereo cameras) but falls short of Figure 02's 16-camera rig optimized for OpenAI VLM processing. For industrial applications, MagicLab's sensor selection prioritizes cost-effectiveness over data redundancy. AI and Software: ByteDance Partnership in the Works MagicLab operates its proprietary Magic Atom motion control platform, which supports "high-fidelity humanoid movements" through a combination of classical control algorithms and machine learning. According to company statements, the platform enables: Rapid skill acquisition: New movements can be trained in 24 hours using imitation learning from human demonstrationsAdaptive joint stiffness: Real-time adjustment of compliance for energy efficiency and precision tasksMulti-robot coordination: Fleet control system (MagicNet) enables synchronized operation of multiple robots ByteDance/Doubao Integration (Pending) In November 2024, MagicLab confirmed it was in talks with ByteDance (TikTok's parent company) to integrate the Doubao large language model into third-generation MagicBot robots. Doubao is significantly cheaper than OpenAI's GPT-4 and specializes in text, image, and video generation — capabilities that could enable MagicBots to handle: Natural language task instructions from human workersCreative problem-solving in unstructured environmentsVideo-based quality control and anomaly detection As of March 2026, the ByteDance partnership has not been formally announced, and current MagicBot deployments do not appear to use LLM-based decision-making. Software Availability MagicLab has not released a public SDK or developer documentation for either model. The Z1 "Development Version" (50 DOF variant) advertises "secondary development" support, suggesting API access for enterprise and research customers. The platform supports OTA (over-the-air) firmware updates via WiFi 6. Design and Build Quality: Industrial-Grade Durability Both MagicBot models use fully electric actuators — a strategic choice that differentiates them from hydraulic competitors like Boston Dynamics' Atlas. The advantages: Lower maintenance: No hydraulic fluid leaks or contamination risks (critical for food/pharmaceutical manufacturing)Energy efficiency: Electric motors generate less waste heat and enable regenerative brakingQuieter operation: Important for human-robot collaboration in shared workspaces Material and Structural Design The Z1 uses high-strength aluminum alloy and engineering plastics optimized through topology simulation and thermal analysis. The Gen1 model features: IP66 rating: Dust-tight and protected against powerful water jets — suitable for cleanroom and outdoor industrial environmentsIndustrial-grade roller bearings: Capable of withstanding 8.7 kN (1,956 lbf) impact force — enables operation in high-vibration environments like automotive plantsIntelligent air-cooled heat dissipation: Active cooling prevents thermal throttling during sustained high-torque operations Hands and Manipulation MagicBot's dexterous hands are a standout feature. The standard configuration uses 6 mini high-torque servo actuators with pressure sensors, enabling: Sub-millimeter precision for delicate assembly tasks70% coverage of human hand gestures (according to company claims)Force feedback for adaptive grip strength The optional 11-DOF MagicHand S01 (available for Z1 Development Version) expands manipulation capabilities further, though specific performance data has not been disclosed. During the Spring Festival Gala, Galbot's robots (using a competing hand design) demonstrated folding clothes and rolling walnuts — tasks that require both precision and force modulation. MagicLab has not released comparable demonstration videos for fine manipulation tasks. Real-World Use Cases: Where MagicBot Excels Based on confirmed deployments and technical specifications, MagicBot is best suited for: 1. Electronics Manufacturing and Assembly The electronics factory deployment (December 2024) demonstrates MagicBot's ability to handle circuit board inspection, component placement, and barcode scanning. The 20 kg per-arm payload capacity enables handling of heavy sub-assemblies like power supplies and metal chassis. 2. Automotive Quality Inspection The IP66 rating and force-torque sensors make Gen1 suitable for automotive quality control tasks — inspecting welds, testing door closures, and validating assembly tolerances. Multi-robot collaboration enables simultaneous inspection of multiple vehicle sections. 3. Warehouse Logistics and Material Handling The 3D LiDAR and autonomous navigation capabilities position MagicBot as a mobile manipulator alternative to traditional AGVs (automated guided vehicles). Unlike wheeled robots, bipedal locomotion enables traversing stairs and uneven surfaces in legacy warehouses not designed for automation. 4. Research and Education The Z1's modular design (24-50 DOF) and rapid learning capabilities make it attractive for university robotics labs and AI research institutions. The Magic Atom platform's 24-hour training cycle is competitive with platforms like Unitree's H1 (which uses reinforcement learning for parkour skills). 5. Entertainment and Demonstration The Spring Festival Gala performance showcased MagicBot's suitability for synchronized choreography, theme park attractions, and corporate demonstration events. The Thomas 360 spin and other acrobatic moves generate significant media attention — valuable for brand marketing. 6. Search and Rescue (Future Potential) MagicLab's stated mission includes "disrupting search and rescue" operations. The Z1's obstacle-climbing ability and compact form factor could enable deployment in disaster zones with collapsed structures, though no real-world rescue deployments have been confirmed. Pros and Cons: The MagicBot Reality Check Pros Already Factory-Deployed — Unlike many competitors still in R&D phase, MagicBot Gen1 is performing real production tasks (inspection, assembly, material handling) at electronics factories as of December 2024.Multi-Robot Collaboration — MagicNet fleet control enables coordinated operation of multiple robots, critical for scaling industrial automation beyond single-robot cells.High Payload Capacity (Gen1) — 20 kg per arm (40 kg total) rivals industrial cobots and exceeds most humanoid competitors like Unitree G1 (2 kg/arm) and Tesla Optimus (unconfirmed).Fully Electric Actuators — 525 N·m peak torque without hydraulics means lower maintenance costs, no fluid contamination risk, and suitability for cleanroom environments.Rapid Development Cycle — Founded December 2023, factory-deployed by December 2024, and national TV showcase by March 2026 demonstrates exceptional execution speed.Dexterous Hands — 6-DOF hands with sub-millimeter precision and pressure sensors enable delicate manipulation tasks (70% of human hand gestures claimed).Strong Funding Trajectory — Raised 150M RMB angel round + hundreds of millions RMB Series B, indicating investor confidence and capital runway for R&D. Cons Very New Company (Founded Dec 2023) — Limited operating history raises questions about long-term support, spare parts availability, and warranty fulfillment for enterprise customers.No Public Pricing — Contact-sales model creates friction for smaller enterprises and research institutions that need budget clarity before procurement approval.Incomplete Spec Disclosure — Gen1 model lacks confirmed specifications for height, weight, walking speed, and battery capacity, making competitive evaluation difficult.Not Fully Autonomous — Company admits robots are "still in the skill training and learning phase" and require human oversight, limiting immediate commercial viability.China-Focused Market — No confirmed international distribution, support infrastructure, or regulatory approvals (CE/UL certification) for Western markets.No Public SDK — Lack of developer documentation limits third-party application development and academic research adoption compared to open platforms like Unitree.Short Battery Life (Z1) — 2 hours runtime severely limits use cases; frequent recharging disrupts workflow in industrial settings. Competitor Comparison: How MagicBot Stacks Up Verdict: MagicBot Gen1 sits between Unitree's consumer-friendly G1 and UBTECH's enterprise-focused Walker S. For buyers prioritizing proven industrial deployment and multi-robot coordination, MagicBot offers compelling value. Buyers seeking transparent pricing and a mature support ecosystem should consider Unitree or UBTECH. The Verdict: Is MagicBot Worth the Hype? MagicLab's MagicBot represents one of the most aggressive commercialization pushes in humanoid robotics history. A company founded just 25 months ago (December 2023) has achieved factory deployments, multi-robot collaboration capabilities, and a sold-out consumer moment on national television — a trajectory that rivals far more established competitors. Who Should Buy MagicBot? Ideal for: Electronics manufacturers seeking modular automation for assembly, inspection, and material handling with 20 kg payload requirementsAutomotive quality control teams needing IP66-rated robots for harsh factory environments with multi-robot coordinationChinese enterprises prioritizing domestic suppliers and government-aligned "Made in China 2025" procurement policiesResearch institutions (Z1 model) focused on bipedal locomotion, imitation learning, and multi-agent robotics with rapid prototyping needsEarly adopters willing to accept a young company's risks in exchange for cutting-edge multi-robot collaboration technology Not recommended for: Western enterprises requiring CE/UL certification, established support networks, and transparent warranty termsBudget-constrained buyers seeking fixed pricing and public SDKs (consider Unitree G1 at $13,500 instead)Fully autonomous applications — MagicLab admits robots still require human oversight and are "in the skill training phase"Mission-critical deployments where a 2-year-old company's long-term viability poses supply chain risksPersonal/consumer use — no consumer models available; contact-sales model excludes individual buyers The Bottom Line MagicBot is a high-risk, high-reward bet on China's humanoid robotics ambitions. The technology is real (factory deployments prove it), the execution speed is extraordinary (25 months from founding to production), and the multi-robot collaboration capabilities address genuine industrial needs. But the lack of pricing transparency, incomplete spec disclosure, and minimal operating history make this a decision for bold early adopters, not conservative procurement teams. For enterprises with existing relationships in China's robotics supply chain and tolerance for emerging-vendor risk, MagicBot Gen1 offers a compelling industrial automation platform. For everyone else, wait 12-18 months for pricing clarity, international support infrastructure, and third-party validation before committing capital. Final Score: Promising technology with proven factory deployment, but buyer beware of the startup risk profile. Last updated: February 24, 2026 Sources: MagicLab official website and Z1 specifications (magiclab.top)China Daily: "Orders for robots surge after Spring Festival Gala" (February 21, 2026)CNN: "China's biggest TV event had a clear star: the robot" (February 18, 2026)Global Times: "Humanoid robots at Spring Festival Gala grab headlines" (March 2026)TechNode: "Humanoid robots take center stage at 2026 Spring Festival Gala" (February 17, 2026)Interesting Engineering: "MagicLab's humanoid army turns factory into precision powerhouse" (December 10, 2024)Mike Kalil: "MagicLab MagicBot Collaborative Industrial Humanoid Robots" (November 2, 2025)Aparobot: MagicBot Z1 specifications and use casesThird-party reseller catalogs (American Satellite, Latin Satelital, Europa Satellite) Robozaps is an online marketplace for humanoid robots. Our reviews are based on manufacturer specifications, third-party assessments, and publicly available deployment data. We do not accept payment for reviews or editorial coverage. Compare on RoboZaps: MagicBot Gen1, Tesla Optimus Gen 2, MagicBot Z1, AgiBot A2, Galbot G1 and Walker S. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # DroidUp Moya Review: $173K Biomimetic Robot Specs & Price [2026] URL: https://blog.robozaps.com/b/droidup-moya-review Author: Radica Maneva Published: 2026-03-10T00:00:00.000Z Updated: 2026-07-16T02:42:54.209Z Category: reviews TL;DR: The DroidUp Moya costs approximately $173,000, with pre-order estimates from $165,000 to $200,000+ depending on customization. Its signature is warm-skin technology, active thermal regulation at 32-36°C, plus 92% human-like walking. It is not shipping yet: deliveries are expected late 2026, with only about 50 units in the first production batch. Key takeaways: - Price: The DroidUp Moya costs approximately $173,000 USD, with pre-order estimates ranging from $165,000 to $200,000+ depending on customization — positioning it in the premium companion robot segment. - Warm Skin Technology: First commercial full-body humanoid with active thermal regulation maintaining 32-36°C body temperature, designed to trigger natural human bonding responses. - Walking Performance: 92% human-like walking accuracy at speeds up to 0.83 m/s (3 km/h / 1.9 mph) — an elegant pace, significantly ahead of most expressive humanoids that cannot walk at all. - Battery Life: Approximately 6 hours per charge using tendon-assisted actuation that minimizes energy consumption. - Best For: Healthcare facilities, eldercare, museums, high-end hospitality, and research institutions focused on human-robot interaction studies. - Key Limitation: Not yet available — expected late 2026 with only ~50 units in the first production batch. DroidUp is a 2021 startup with no consumer track record. FAQ: Q: How much does the DroidUp Moya cost? A: The DroidUp Moya costs approximately $173,000 USD, with estimates ranging from $165,000 to over $200,000 depending on customization. This positions it as a premium institutional robot rather than a consumer product. DroidUp has not announced financing options or leasing programs, though these may emerge as commercial deployments begin in late 2026. Q: When will the DroidUp Moya be available? A: DroidUp expects to begin shipping Moya units in late 2026. The first production run will be limited to approximately 50 units, likely prioritizing Chinese institutional customers in healthcare and public venues. International availability has not been announced. Q: Can the DroidUp Moya actually walk? A: Yes. Unlike many expressive humanoids that are stationary or wheeled, Moya achieves full bipedal locomotion using DroidUp's Walker 3 skeleton. The company claims 92% human-like walking accuracy at speeds up to 0.83 m/s (1.9 mph). The Walker 2 platform (predecessor to Walker 3) won bronze at the 2025 Beijing Humanoid Robot Half Marathon, demonstrating proven bipedal capability. Q: Why does the DroidUp Moya have warm skin? A: Moya maintains body temperature between 32-36°C (89.6-96.8°F) through embedded heating elements in its synthetic skin. Research shows humans subconsciously use touch temperature to assess connection and kinship. DroidUp designed the warm-skin feature specifically to trigger these bonding responses, making interactions feel more natural and emotionally comfortable than with cold-surfaced robots. Q: Is the DroidUp Moya safe to interact with? A: At 32 kg (71 lbs), Moya is significantly lighter than most full-size humanoids, reducing collision risks. The tendon-assisted actuation system enables smoother, more controlled movements than high-torque industrial actuators. However, as with any humanoid robot, institutional deployments will require safety assessments and likely some supervision. DroidUp has not published specific safety certifications. Q: How does DroidUp Moya compare to Sophia the robot? A: Sophia (by Hanson Robotics) and Moya both prioritize realistic humanlike appearance and emotional expressiveness. However, Sophia cannot walk — it's primarily a bust or wheeled platform. Moya combines full bipedal locomotion with expressiveness. Moya also adds warm skin technology that Sophia lacks. Sophia has more global brand recognition and years of public appearances, while Moya is a 2026 newcomer. Q: What is DroidUp's track record? A: DroidUp (also known as Zhuoyide) was founded in 2021 in Shanghai. The company previously demonstrated hyper-realistic android busts at events like the World Artificial Intelligence Conference (WAIC) and enrolled an android in Shanghai Theatre Academy's doctorate arts program. Their Walker biped skeleton won bronze at the 2025 Beijing robot half marathon. However, Moya is their first commercial humanoid product, and the company has no consumer track record. Q: Is the DroidUp Moya worth $173,000? A: For institutions that specifically need a mobile humanoid with emotional expressiveness and realistic human touch — healthcare, premium hospitality, human-robot interaction research — Moya offers capabilities no other robot provides. If you need general-purpose manipulation or don't require the warmth/expression features, alternatives like Unitree H1 ($90K) or upcoming Tesla Optimus (~$25-30K) offer better value. The answer depends entirely on whether Moya's unique biomimetic features align with your use case. Bottom Line: The DroidUp Moya price is about $173,000, and it is worth it for healthcare, eldercare, and premium hospitality buyers who need emotional connection and warm-skin technology. NOT recommended for manipulation tasks or budget-conscious buyers. Availability: Late 2026, ~50 units first batch. Last updated: March 20, 2026  |  12 min read The DroidUp Moya is doing something no other humanoid robot has attempted: feeling genuinely human to the touch. With synthetic skin that maintains body temperature between 32-36°C (89.6-96.8°F), micro-expressions across 25 facial degrees of freedom, and 92% human-like walking accuracy at a measured 0.83 m/s pace, Moya represents China's most ambitious push into biomimetic robotics. But at $173,000, is the what DroidUp calls the world's first "fully biomimetic" humanoid worth the investment? This comprehensive DroidUp Moya review covers everything you need to know: real-world specifications, pricing breakdown, performance analysis, and how Moya compares to competitors like Ameca and the upcoming Xpeng Iron. DroidUp Moya Specifications The DroidUp Moya — world's first fully biomimetic humanoid robot with human-like warmth and expressions. DroidUp Moya Price: What Does It Actually Cost? DroidUp has confirmed pricing of approximately $173,000 USD for the Moya, though final prices may range from $165,000 to over $200,000 depending on customization options. As a pre-production robot with limited initial availability (~50 units), pricing remains somewhat fluid. At this price point, Moya positions itself as a premium institutional robot rather than a consumer product. DroidUp is clearly targeting healthcare facilities, museums, and research institutions with budgets for cutting-edge human-robot interaction technology. Here's how the DroidUp Moya price compares to other humanoid robots on the market: For the price, Moya offers a unique value proposition: it's the only humanoid robot that combines full bipedal locomotion with realistic warmth and micro-expressions. Ameca has better facial expressions but cannot walk. Tesla Optimus can walk but has no emotional expressiveness. Moya sits at the intersection — though you pay a premium for that convergence. Performance and Mobility: Real-World Capabilities The DroidUp Moya achieves what most expressive humanoids cannot: actually walking. Built on DroidUp's Walker 3 skeleton — the successor to Walker 2, which won bronze at the 2025 Beijing Humanoid Robot Half Marathon — Moya delivers genuinely impressive bipedal performance. Key mobility specifications: Walking Speed: Up to 0.83 m/s (3 km/h / 1.9 mph), a measured pace optimized for elegant movement rather than speedWalking Accuracy: 92% human-like gait — DroidUp claims this is measured against biomechanical analysis of natural human movementWeight: Just 32 kg (71 lbs) — remarkably light for a full-size humanoid, enabled by tendon-assisted actuationBattery Life: Approximately 6 hours per charge, comparable to most bipedal humanoidsTurning: Smooth directional changes without the jerky stops common in earlier humanoids The lightweight build is notable. At 32 kg, Moya is lighter than Ameca (62 kg), roughly half the weight of Tesla Optimus (~73 kg), and comparable to 1X NEO's ~30 kg. This low mass, combined with tendon-assisted actuators similar to 1X's approach, enables longer battery life and more energy-efficient movement. However, observers at the March 2026 Shanghai debut noted that while Moya's gait is smooth, it still shows that 8% gap from fully human — some describe it as similar to walking in heels. The robot is clearly optimized for elegant, measured movement rather than dynamic athletics like running or jumping. Sensors and Perception The DroidUp Moya's sensor suite prioritizes human interaction over environmental navigation: In-Eye Cameras: Cameras positioned behind Moya's eyes capture facial data for real-time expression mirroring and eye contact maintenance. This enables Moya to respond to micro-expressions in the people it's interacting with.Thermal Sensors: Monitor and regulate the synthetic skin temperature to maintain the 32-36°C warmth that distinguishes Moya from other robots.Facial Tracking System: Reads human facial expressions and movements to inform Moya's reactive expressions. The system processes movement data in real-time to generate contextually appropriate responses. Unlike industrial humanoids that prioritize depth sensing and object detection (LiDAR, Intel RealSense, etc.), Moya focuses on social perception. The sensor array is designed to answer: "What is this person feeling, and how should I respond?" — not "What objects are in this room and how do I manipulate them?" This focus makes sense for Moya's target applications in healthcare and hospitality where emotional connection matters more than object manipulation. AI and Learning Capabilities DroidUp Moya employs what the company calls the "Zhuoyide cerebellar motor control model" — a proprietary AI system that handles real-time movement coordination and social interaction: Biomimetic AI: The AI processes facial recognition data and generates corresponding micro-expressions — joy, surprise, concern — in real-time. DroidUp claims the system can produce natural transitions between emotional states rather than discrete expression changes.Motor Control: The cerebellar model coordinates the Walker 3 skeleton's movements, managing balance and gait while maintaining the smooth, elegant motion DroidUp prioritizes. This handles the complex coordination between walking and upper-body expressiveness.Speech Integration: Moya includes speech recognition and natural language processing for conversation, though DroidUp has not disclosed which LLM backbone powers the system. The SDK situation is unclear. DroidUp has not announced public API access or ROS compatibility. Given the company's focus on institutional customers rather than research labs, developer accessibility may not be a priority. This is a notable contrast to platforms like Unitree H1 that actively court the research community with open development tools. Design and Build Quality Moya's design philosophy centers on one goal: feel less like a robot and more like a person. This drives every material and engineering choice. The synthetic skin incorporates embedded heating elements that maintain human body temperature. Studies on haptic perception show that warmth triggers subconscious bonding responses — we instinctively associate warmth with life and kinship. DroidUp is explicitly exploiting this psychological response to create stronger human-robot connections. Beneath the warm skin, Moya features a simulated rib cage and soft material layers that mimic human fat and muscle. The result is a tactile experience closer to touching a person than touching a machine — though whether this enhances comfort or deepens uncanny valley discomfort varies by individual. The 25 degrees of freedom in Moya's face enable micro-expressions: subtle eye movements, slight smiles, small nods that humans make unconsciously during conversation. These aren't programmed animations but real-time generated responses to observed human behavior. The modular platform architecture allows different gender presentations and facial configurations. DroidUp can customize appearance for specific deployment contexts — a significant differentiator for institutional customers who need robots matching specific personas. Real-World Use Cases 1. Healthcare and Eldercare DroidUp explicitly targets healthcare as Moya's primary market. China's rapidly aging population creates urgent demand for care supplements. Moya's warm touch, emotional responsiveness, and non-threatening presence could provide companionship and basic interaction for elderly patients. The 6-hour battery life supports partial shift deployment, and the lightweight build (32 kg) reduces safety concerns compared to heavier industrial robots. 2. Museums and Exhibitions Interactive museum guides benefit from Moya's combination of walking ability and emotional expressiveness. Unlike stationary systems, Moya can escort visitors through spaces while maintaining engaging conversation. The customizable appearance allows museums to create period-appropriate or thematically relevant characters. 3. Premium Hospitality High-end hotels and venues seeking differentiation could deploy Moya as a premium concierge experience. The emotional responsiveness creates more memorable interactions than typical service robots, while the warm-skin technology makes handshakes and greetings feel more natural. 4. Human-Robot Interaction Research Researchers studying uncanny valley effects, social robotics, and human-robot bonding have limited platforms that combine locomotion with realistic emotional expression. Moya provides a unique research tool — though the unclear SDK situation may limit academic applications. 5. Banking and Financial Services DroidUp mentions banks as a target deployment. Premium financial services branches increasingly use technology to differentiate customer experience. A biomimetic greeter could elevate perception of service quality — though ROI calculations at $173,000 per unit require high-value customer contexts. DroidUp Moya: Pros and Cons Pros First Commercial Full-Body Warm-Skin Humanoid — 32-36°C body temperature triggers natural bonding responses; no other humanoid offers thisCombines Walking + Expressions — Ameca can't walk; Tesla Optimus can't emote. Moya does both, uniquely positioning it for social applicationsLightweight Design (32 kg) — Tendon-assisted actuation keeps weight low, enabling safer human proximity92% Human Gait Accuracy — Walker 3 skeleton proven at robot marathon; movement is smooth and elegantCustomizable Appearance — Modular platform allows gender, face, and persona customization for institutional branding25 DOF Facial Expressions — Micro-expressions respond in real-time to human interaction, not pre-scripted animations Cons Not Available Until Late 2026 — Pre-orders only; first batch limited to ~50 units. You cannot buy one today.New Company Risk — DroidUp founded in 2021 with no consumer track record. Long-term support uncertain.Uncanny Valley Concerns — Early reactions are mixed; some find Moya engaging, others describe it as "like a living ghost." Realistic ≠ comfortable.Limited Specs Disclosed — No published payload capacity, full-body DOF count, or detailed technical documentationHigh Price Point ($173,000) — 6-8x more expensive than general-purpose alternatives like Tesla Optimus or Unitree H1China-Focused Initially — First deployments expected in China; international availability unclearSDK/Developer Access Unknown — No announced API or ROS compatibility; may limit research applications How DroidUp Moya Compares to Competitors vs. Ameca: Ameca has comparable facial expressiveness (27 facial DOF vs Moya's 25) and is available today. But Ameca cannot walk — it's a torso on a stand or wheeled base. If your application requires a mobile, walking presence with emotional expressiveness, Moya is the only option. vs. Xpeng Iron: Both are Chinese humanoids targeting 2026 launch with realistic appearances. Iron comes from a major EV manufacturer (Xpeng) with proven mass production capability, while DroidUp is an unproven startup. Iron demonstrated walking in early 2026 but also showed balance issues. Neither has disclosed full pricing. Verdict: Should You Buy the DroidUp Moya? The DroidUp Moya is attempting something genuinely new in humanoid robotics: creating a robot that doesn't just look human but feels human. The warm skin, micro-expressions, and elegant walking motion combine into an experience designed to trigger emotional connection rather than utility. At $173,000, you're not buying a tool — you're buying a presence. Buy the Moya if: You're a healthcare facility, museum, or premium hospitality venue specifically seeking a humanoid that creates emotional connections with visitors or patients. You have the budget for experimental technology and understand you're an early adopter with a 2021 startup. You need walking + expressiveness combined in one platform — no alternative offers this. Don't buy the Moya if: You need manipulation capabilities (carrying objects, opening doors, performing tasks). You want a proven platform with established support — consider Ameca for pure expressiveness or Unitree H1 for athletic bipedal research. You're price-sensitive — wait for the market to mature. Moya represents a bet on the future of social robotics. If DroidUp executes on their vision and survives as a company, early adopters will own groundbreaking technology. If not, that $173,000 becomes an expensive museum piece. Given the late 2026 timeline and ~50 unit first batch, most buyers should watch the first deployments before committing. Interested in the DroidUp Moya? View the full DroidUp Moya listing on Robozaps or browse all humanoid robots to compare alternatives. Last updated: March 8, 2026. Specs sourced from DroidUp press releases (March 2026), New Atlas, Mike Kalil, and Tekedia coverage. Pricing confirmed at ~$173,000 by multiple sources. Robozaps is a humanoid robot marketplace — we maintain hands-on product databases and may earn referral fees from qualifying purchases. See where Moya sits among the best humanoid robots of 2026. --- # XPeng Iron Price 2026: Is It For Sale, Cost & Release Date URL: https://blog.robozaps.com/b/xpeng-iron-review Author: Radica Maneva Published: 2026-03-09T00:00:00.000Z Updated: 2026-08-01T23:07:52.697Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: You cannot buy an XPeng Iron in July 2026 and no price exists: the $150,000 online is an unsourced estimate, one '$120,000 In Stock' listing is fake, and 'XPENG IRON' crypto is unrelated. XPeng's real commitments: mass production by end-2026 (1,000+/month capacity), its own Chinese showrooms Q1 2027, overseas later in 2027, households ~2028. The robot is real; it was cut open on stage to prove it. Key takeaways: - No Iron is for sale and no price exists; He Xiaopeng's only framing is that manufacturing cost could approach car costs. The $150k figure is an aggregator estimate, and one $120k 'In Stock' listing is fake. - Mass production by end-2026 is not a sale date: showroom jobs come Q1 2027, overseas 2027, households around 2028; '1 million by 2030' is a may, not a commitment. - The human-in-a-suit saga ended November 6, 2025, when XPeng unzipped the robot on video and cut its leg open on stage; the 'ear' was a microphone array. - Most pages mix two robots: gen-1 (2024, '200 DoF' marketing) and next-gen Iron (82 DoF, 22 per hand, three Turing chips at ~2,250 TOPS, claimed solid-state battery). - The factory story flipped: XPeng's own CEO says assembly-line work wasn't economic (hands wore out in a month); the fleet mostly collected training data, and He Xiaopeng personally took over the unit in June 2026. FAQ: Q: Is the XPeng Iron robot real or fake? A: Real. After viral claims that a human was inside, XPeng posted an unedited video on November 6, 2025 unzipping the robot's covering, and later that day cut open its leg on stage while it stood powered on, showing actuators and wiring. The 'ear' people spotted was a microphone array. Q: How much does the XPeng Iron cost? A: There is no price. XPeng has never announced one; He Xiaopeng has only said the manufacturing cost of humanoid robots could eventually approach that of cars, which is not a retail-price statement. The $150,000 figure circulating online is a third-party estimate, and the $120,000 'In Stock' listing on one marketplace is not authorized: no one can buy an Iron. Q: Can you buy an XPeng Iron? A: No. Not as a consumer, not as a business. Mass production is slated to start by end-2026, with the first robots deployed in XPeng's own Chinese showrooms from Q1 2027, overseas later in 2027, and households around 2028 at the earliest. Q: When will the XPeng Iron be released? A: XPeng's committed sequence: large-scale mass production by the end of 2026 with capacity above 1,000 units a month, commercial deployment in its Chinese stores from Q1 2027, global commercial deliveries in 2027, and household use around 2028. None of these is a retail on-sale date yet. Q: Why did the XPeng Iron fall? A: At its first public outing in a Shenzhen mall on February 1, 2026, Iron completed its catwalk and then fell backwards while standing; staff caught it, and it appeared the next day strapped to a support frame. He Xiaopeng compared it to a toddler learning to walk. Public-space reliability is still maturing. Q: How many degrees of freedom does the Iron have? A: The next-gen Iron claims 82 degrees of freedom with 22 per hand, per XPeng's official release. The '200 degrees of freedom' still quoted on many sites is the 2024 first-generation robot's marketing count, which used a different counting method. Q: Does the XPeng Iron use a solid-state battery? A: XPeng claims the next-gen Iron uses an all-solid-state battery, which would be an industry first. No supplier, cell specification or independent teardown has ever been disclosed, so treat it as an unverified vendor claim for now. Q: Why is the XPeng Iron designed as a female robot? A: XPeng's stated position is that body styling is customizable, with female and male variants planned, because buyers will pick body types like car trims and service robots should feel approachable. The claim that XPeng only builds a 'female robot' is wrong, though the launch showcased the female variant first. Q: Does the XPeng Iron work in factories? A: Less than the 2024 marketing implied. He Xiaopeng has admitted the screw-tightening experiment wasn't economic (the hands wore out within a month), and reporting found the factory fleet was mostly used to collect training data. The strategy is now showroom and service jobs first, with industrial inspection via partner Baosteel. Q: Is the XPeng Iron better than Tesla Optimus? A: Neither can be bought and neither has a price, so the comparison is promises: both EV makers target volume production around end-2026, Iron with the more human-like gait and a nearer commercial venue (XPeng's own showrooms), Optimus with bigger scale ambitions. For robots that actually ship today, a Unitree G1 starts at $13,500. Q: Is XPENG IRON a cryptocurrency? A: A meme token using the name trades on BNB Chain and Solana, and its price pages rank in search results for the robot. It has no connection to XPeng. The robot has no price and cannot be bought. Q: Can I invest in XPeng's robot business? A: Only via XPeng's listed EV stock (XPEV); there is no robotics spinoff or IPO, and in June 2026 He Xiaopeng consolidated the robotics unit under his personal leadership. At least one analysis argues XPEV is overvalued on the robot story, so size the bet on the EV business, not the robot. Is the XPeng Iron for sale, and what does it cost? No, you cannot buy an XPeng Iron in July 2026, at any price, from anyone, and XPeng has never announced one. What XPeng has committed to is mass production by the end of 2026, which is not the same thing as a sale date: the first Irons are slated to work in XPeng's own Chinese showrooms from early 2027, go overseas later that year, and reach households in 2028 at the earliest. Every dollar figure you've seen attached to Iron is third-party guesswork, one "listing" selling it is fake, and there's a cryptocurrency squatting on its name. We verified all of it against primary sources on July 22, 2026. The is-it-real saga, told precisely Iron is famous because people refused to believe it. On November 5, 2025, at XPeng's AI Day in Guangzhou, the next-generation Iron walked a catwalk with a gait smooth enough that Chinese and Western social media concluded a human was inside the suit, citing an "ear" outline (actually a microphone array) and contours under the back covering. XPeng's response came the next morning, November 6, not "days later" as widely retold: He Xiaopeng posted a single-take, unedited video in which a technician unzips the robot's covering to show a mechanical torso, and later that same day, on stage at a separate launch event, staff cut open Iron's leg with scissors while it stood powered on, exposing actuators and wiring. Elon Musk liked the video. The hashtag for what Iron looks like without skin topped Douyin. Two honest footnotes the viral clips skip. First, robotics analyst Scott Walter argued from frame analysis that the robot walking on stage was the older-generation chassis running new gait software, with the next-gen hardware appearing only in renders; XPeng never directly rebutted him. Second, Iron's first public outing after the reveal, at a Shenzhen mall on February 1, 2026, ended with the robot falling backwards after its catwalk; it appeared the next day strapped to a support frame, and He Xiaopeng compared it on Weibo to a toddler learning to walk. The robot is real. It is also still learning to stand around unpredictable humans. The price truth: three debunks in one section XPeng has never stated a price. He Xiaopeng's only on-record framing is directional, and it's about cost, not price: the manufacturing cost of humanoid robots could eventually approach that of cars. That is not a retail-price announcement. From there, three things pollute the search results. The "$150,000" figure repeated across robot-listing sites is an analyst-style estimate none of them source to XPeng. A marketplace called robotixmarket.com displays a "$120,000 — In Stock" Iron listing with an add-to-cart button for a robot that no one, anywhere, can purchase; treat it as decoration at best. And an unrelated "XPENG IRON" meme token trades on BNB Chain and Solana, which is why crypto price pages rank when you search the robot's cost. None of these is XPeng, and none of them knows the price, because there isn't one. Specs, by generation (this is where everyone gets it wrong) Most competitor pages mix two different robots. The 2024 Iron and the 2025 next-gen Iron share a name and little else; the figures below are XPeng's own claims, labeled, with the conflicts shown. The aggregator numbers you'll see elsewhere — "4-hour runtime," "7 km/h," "20 kg grip" — appear on spec farms only and trace to no XPeng statement; we've omitted them. What Iron actually does today Here the honest record diverges sharply from the 2024 marketing. XPeng said then that Iron was "already operational" building EVs in its Guangzhou factory. A year later, 36kr's reporting found the factory fleet was used mostly for algorithm tuning and teleoperation data collection, and He Xiaopeng himself admitted the screw-tightening experiment failed economically: the hands wore out within a month and cost more to replace than years of a worker's wages. That admission drove the strategy you see now: commercial service first (showroom guide, receptionist, tour host), selective industrial inspection with sole named partner Baosteel, and households last. It is a more credible plan precisely because it was arrived at the hard way. Iron's 2026, dated Worth noting what's absent: Iron skipped July's WAIC in Shanghai, where rivals AgiBot and Unitree filled the halls; XPeng ran its own Munich show instead. And the leadership churn behind the June takeover cuts both ways: it's the strongest possible commitment signal, and it's also three senior departures in the year before the hardest deadline the company has ever set. The female-design question, using XPeng's actual words The next-gen Iron's feminine styling generated a commentary cycle largely detached from what XPeng said. The company's stated position: the design is customizable, with female and male body variants planned, because buyers will choose body types the way they choose car trims, and service robots should feel approachable. The robotics VP put it flatly: "We see the different options because there are different kinds of humans." Whether the launch-choreography choice to debut the female variant on a catwalk was wise marketing is a fair debate; the claim that XPeng builds only a "female robot" is simply wrong. Iron vs Optimus: the only comparison that matters The honest framing: Iron and Optimus are two of the most closely watched unpurchasable robots on Earth, both promising the same year, both from EV makers betting the factory. Iron's differentiators are anthropomorphism (the skin, the gait, the body variants) and a nearer commercial venue (its maker's own showrooms). Optimus's are scale ambition and the Tesla brand. Neither has shipped a product. Our evidence-ranked guide covers the robots that actually have. The stock question There is no XPeng robotics IPO. Searchers conflate Iron with XPeng's listed EV stock (XPEV) and with rival Unitree's approved listing; the actual July 2026 signal points the other way, with He Xiaopeng consolidating the robotics unit under his personal control rather than spinning it out. XPEV's valuation debate now partly rides on Iron: at least one analysis argues the stock is substantially overvalued on the robot push. If you want listed humanoid exposure with actual shipments behind it, that conversation currently runs through Unitree's pending IPO, covered in our G1 review. Misinformation audit: eight Iron claims, checked What to watch next Iron's calendar is unusually checkable. End of 2026: does mass production actually start, at anything like 1,000 units a month? Q1 2027: do Irons appear in XPeng showrooms doing real shifts? And between now and then, whether the promised European availability gets a date and whether XPeng publishes the number everyone is searching for. Every one of those has a falsifiable deadline, which is more than most humanoid roadmaps offer, and we'll update this page as each lands or slips. The bottom line The XPeng Iron is real, remarkable, and not for sale. It owns one of the most humanlike publicly demonstrated gaits in the industry and the scars to prove it's a machine; it also owns a public fall, an unrebutted question about which chassis actually walked that catwalk, and a factory story its own CEO admits didn't work the first way. If the end-2026 production line and the Q1 2027 showroom jobs happen on schedule, Iron would give ordinary showroom visitors routine access to one of the most humanlike service robots yet deployed. Until then it is the best advertisement yet for a product that does not exist, and every price you read for it, including any future one here, deserves the same test: did XPeng say it, or did someone guess? Compare on RoboZaps: Xpeng Iron, Tesla Optimus and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # This Week in Humanoid Robots: March 1-7, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-mar-7-2026 Author: Radica Maneva Published: 2026-03-07T00:00:00.000Z Category: news This week's headlines told a clear story: the race to build humanoid AGI is intensifying, and the money is following. From Musk's bold claims about Tesla achieving artificial general intelligence in robot form first, to a €1 billion Tether-backed bet on German robotics, the industry is moving past demos and into deployment mode. Here's what shaped the humanoid robot industry from March 1-7, 2026. Musk Claims Tesla Will Be First to AGI in Humanoid Form Elon Musk posted on X that "Tesla will be one of the companies to make AGI and probably the first to make it in humanoid/atom-shaping form." The statement came as Tesla confirmed plans to convert its Fremont Model S and Model X production lines to Optimus robot manufacturing—a major strategic shift that signals robotics is becoming central to Tesla's future. Why it matters: This isn't just Musk being Musk. Tesla plans to spend over $20 billion in capital expenditure in 2026—up sharply from $8.5 billion in 2025—with a significant chunk going toward Optimus production and supporting infrastructure. The company is targeting 1 million units annually as its initial production goal. That's a concrete, measurable bet on humanoid robots becoming Tesla's next major business line. The timing aligns with reports that xAI, Musk's AI company, recently merged with SpaceX, with rumors of an even larger consolidation across all Musk companies including Tesla. An Optimus robot running xAI's Grok models would represent a significant capability leap. Our take: The AGI claim is aspirational, but the Fremont conversion is real. We covered Tesla killing the Model S for Optimus in detail. What's notable is how Tesla's robotics ambitions are now directly cannibalizing its legacy vehicle business. Ending production of the vehicles that built Tesla's brand to make room for robots? That's commitment. See our full Tesla Optimus Gen 3 guide for where the hardware stands today. Neura Robotics Raises €1 Billion Backed by Tether German humanoid robot maker Neura Robotics is raising approximately €1 billion ($1.2 billion) in a funding round led by Tether, the stablecoin issuer behind USDT. The deal values the company at €4 billion—lower than the €8-10 billion rumored last November, but still a massive valuation for a European robotics startup that most people haven't heard of. Neura develops the 4NE1 humanoid robot, a 5.9-foot system that understands natural language instructions and can carry up to 220 pounds while moving at about three miles per hour. The company also builds industrial robotic arms that can be programmed through visual interfaces rather than custom code, and logistics robots capable of moving 1.5 tons of goods. Why it matters: This is the largest single funding round we've tracked for a pure-play humanoid robotics company in 2026. It also marks Tether's biggest bet on physical AI—an interesting signal of where crypto money sees opportunity. According to reports, Neura has accumulated a €1 billion order book, suggesting strong commercial traction. Our take: Crypto money flowing into robotics is a trend worth watching closely. Tether has been diversifying beyond stablecoins, but humanoid robots represent a bold thesis on physical-world AI becoming the next major technology platform. The company also recently acquired ek Robotics, adding 300 employees and warehouse logistics expertise. Read our NEURA Robotics 4NE1 review for the full technical breakdown on their flagship humanoid. Xiaomi Begins Factory Trials of Humanoid Robots Xiaomi CEO Lei Jun announced that the company's humanoid robots have begun trial operations at Xiaomi's automobile factory in China. The machines are already performing real tasks: loading self-tapping nuts at assembly stations and transporting material boxes across the facility. Lei said large-scale deployment across all production facilities is planned within five years. The robots operate using Xiaomi's proprietary Xiaomi-Robotics-0 VLA (vision-language-action) foundation model. By integrating multimodal perception and reinforcement learning, the humanoids can perform autonomous operations without constant human guidance. Key performance indicators like mean time between failures and task success rates are "steadily improving," according to Lei. Why it matters: Xiaomi isn't just building robots for demos and trade shows—they're putting them to work in their own operations, validating capabilities in real manufacturing environments. This is exactly the approach needed to prove humanoid robots can deliver actual ROI. Our take: This is how you validate humanoid robots: deploy them in your own operations first, then sell to others. Xiaomi's approach mirrors Tesla's strategy with Optimus. It's telling that two of the world's largest technology manufacturers are both using their own factories as testing grounds. Our Xiaomi CyberOne review covers their flagship humanoid specifications. Five years to large-scale deployment sounds conservative given the pace of progress—expect it sooner. OpenAI Robotics Leader Resigns Over Pentagon AI Deal Caitlin Kalinowski, who led OpenAI's hardware and robotic engineering teams since November 2024, resigned from the company. In detailed posts on LinkedIn and X, she cited specific concerns about "surveillance of Americans without judicial oversight and lethal autonomy without human authorization" as issues that "deserved more deliberation than they got." Her departure followed OpenAI's agreement with the Pentagon to deploy AI models on classified government networks. The deal came shortly after Anthropic walked away from similar negotiations, reportedly pushing for stricter limits on domestic surveillance and autonomous weapons. The optics looked bad—OpenAI appearing to step in after its rival took a principled stand. CEO Sam Altman later acknowledged the rollout looked "opportunistic," and OpenAI clarified restrictions on military uses. But Kalinowski was already gone. Why it matters: The robotics industry's relationship with military applications is becoming a major fault line for talent. Kalinowski's resignation puts a spotlight on where companies draw ethical boundaries around autonomous systems—and whether top engineers will stay when those boundaries get tested. Our take: This won't slow OpenAI's robotics ambitions materially, but it does highlight the tension companies face as AI moves into defense applications. The best robotics engineers have options, and companies that lose talent over ethical concerns may find the cost higher than expected. Our piece on humanoid robots in military and defense explores this increasingly complex space. Canadian Startup Mirsee Prepares for Mass Production Mirsee Robotics, a small company based in Cambridge, Ontario, announced plans to move to mass production of its MH3 humanoid robot in 2027. CEO Tarek Rahim predicted the robotics revolution will be "bigger than the automotive revolution in the early 20th century, happening at ten times the speed" and that there will eventually be "more robots than cars." Unlike bipedal humanoids dominating headlines, the MH3 uses wheels for mobility—a deliberate design choice aimed at maximizing battery life and stability. It's harder to knock over and can operate longer between charges. The robot uses a Canadian-made vision system for object manipulation, and the company is adding voice command capabilities powered by AI. Why it matters: While Chinese and American companies dominate headlines, regional players like Mirsee represent the industry's global expansion. The wheeled design also shows there's still room for different form factors in what's becoming a crowded humanoid space. Not every application needs legs. Our take: North American manufacturing capability for humanoid robots remains limited outside of Tesla. Mirsee is small—planning to reach just 20 employees—but they represent important industrial base development. Toyota Canada also announced Digit deployments at their Woodstock facility this week, suggesting the Ontario region is becoming a North American robotics cluster. See our complete humanoid robot companies guide for the full competitive landscape. Chinese Robotics Companies Showcase at AW 2026 in South Korea China's leading humanoid developers—Unitree, Leju, and AgiBot—gathered at Smart Factory and Automation World 2026 (AW 2026) in South Korea. The event showcased their latest hardware and commercialization strategies for the Korean manufacturing sector. Perhaps more significant: Unitree signaled openness to technology cooperation with South Korean companies. Given South Korea's manufacturing expertise (Samsung, Hyundai, LG) and China's lead in humanoid hardware, this could open meaningful partnership opportunities. Why it matters: This represents China's robotics industry pushing aggressively into new markets. AW 2026 is a major industrial automation show, and the presence of multiple Chinese humanoid makers signals serious commercial intent beyond domestic deployments. Our take: Unitree continues its aggressive expansion strategy following their viral Spring Festival performance that reached 679 million viewers. Their willingness to partner with Korean firms suggests they're building an ecosystem, not just selling individual robots. This could accelerate adoption across Asian manufacturing. Check our Unitree G1 review and Unitree H2 review for hardware details. Market Pulse Funding: Neura Robotics' €1B round brings total 2026 humanoid robot funding past $3 billion in the first quarter alone—already exceeding most of 2025.Valuations: Apptronik is now valued at $5 billion (up from $15 million when they started), with Google DeepMind as a partner. Neura at €4 billion. The industry is minting unicorns.Production targets: Tesla targeting 1M Optimus units annually; Xiaomi planning large-scale deployment within 5 years; Mirsee moving to mass production in 2027.Market forecast: New research published this week projects the humanoid robot market reaching $243.4 billion by 2035—driven by personal assistance, healthcare, and education applications. What to Watch Next Week Tesla Fremont conversion timeline: Details on when Model S/X lines officially transition to Optimus production could come any day. This will be a major milestone for the industry.Neura Robotics deal close: The €1B round reportedly may be followed by additional fundraising—watch for closing announcements and how the capital will be deployed.OpenAI robotics restructuring: With Kalinowski's departure, OpenAI's robotics roadmap may shift. Expect updates on team leadership and whether the military deal affects hiring. Want to stay ahead of humanoid robot developments? Bookmark our humanoid robot news hub for continuous coverage, or browse the latest robots available at Robozaps. --- # This Week in Humanoid Robots: February 22-28, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-feb-28-2026 Author: Radica Maneva Published: 2026-03-01T00:00:00.000Z Updated: 2026-08-01T23:07:48.847Z Category: news The week's biggest story? Reality checks. From BMW proving humanoids can actually work production lines to a legendary roboticist declaring the whole endeavor "fantasy," this week forced the industry to confront the gap between demo videos and deployable technology. BMW Brings Hexagon's AEON Robots to Europe After U.S. Success The headline numbers are hard to ignore: 30,000 cars produced, 90,000+ parts handled, 1,250+ hours of runtime. That's what Figure AI's humanoid robots achieved at BMW's Spartanburg plant in South Carolina over 11 months. Now BMW is taking the experiment to Europe. On February 27, the automaker announced it will deploy humanoid robots at its Leipzig plant in Germany—the first time "Physical AI" of this kind has entered European automotive production. Why it matters: This isn't a demo. Figure 02 ran 10-hour shifts, Monday through Friday, on an active assembly line. The robots loaded sheet-metal parts with 5-millimeter precision in just 2 seconds per cycle. When you're building X3 SUVs, that kind of consistency matters. The deployment also generated critical data that shaped Figure 03's design. The forearm—the robot's most failure-prone component—was completely re-architected for the new model. Every intervention, every failure mode, every hour of runtime informed the next generation. Our take: This is the deployment milestone the industry needed. Flashy videos of robots folding laundry are one thing; running an automotive production line for nearly a year is another. BMW's expansion to Germany signals that the pilot exceeded expectations. For more on Figure's latest, see our Figure 03 review and Figure AI company analysis. iRobot Cofounder: Humanoid Vision Is "Pure Fantasy" Rodney Brooks, the MIT roboticist who cofounded iRobot (makers of the Roomba), unloaded on the humanoid robot industry this week. His verdict on Elon Musk's vision of humanoid assistants: "pure fantasy thinking." Brooks argues that today's humanoid robots "will not learn how to be dexterous" regardless of how many billions VCs pour into training. The problem? Touch. Human hands contain 17,000 mechanoreceptors for detecting pressure and texture. While AI has been trained on massive datasets of speech and images, "we do not have such a tradition for touch data," Brooks wrote. Training robots by filming humans performing tasks—the approach used by Tesla and Figure—won't solve this fundamental gap. Why it matters: Brooks isn't some armchair critic. He's been building robots for three decades. His claim that robots won't look like humans in 15 years—instead sporting wheels, multiple arms, and only being called humanoids—directly challenges the form factor every major player has bet on. Our take: Brooks has a point about the touch data problem, but dismissing the entire humanoid effort feels premature. BMW's deployment shows real-world value exists even with current limitations. The question isn't whether today's robots are perfect—it's whether they're useful enough to justify continued investment. Still, his critique about transparency is valid. When companies hide their teleoperation rates, the public can't properly evaluate progress. For context on what these robots actually cost, check our humanoid robot pricing guide. China Owns 90% of the Humanoid Robot Market The numbers are stark: Chinese companies shipped roughly 90% of all humanoid robots sold globally in 2025. Unitree moved 5,500 units. AgiBot shipped 5,168. Meanwhile, Figure, Agility Robotics, and Tesla each sold around 150. That's not a typo. Unitree shipped 36 times more humanoid robots than its closest American competitor. "China has a more robust hardware supply chain—much of it built up through the EV sector, from sensors to batteries—and the world's strongest manufacturing base," analyst Selina Xu told TechCrunch. Even Elon Musk acknowledged the competitive reality at Davos: "China is very good at AI, very good at manufacturing, and will definitely be the toughest competition for Tesla. To the best of our knowledge, we don't see any significant competitors outside of China." Why it matters: This is the EV playbook all over again. Early state support, industrial policy, rapid iteration, cost advantages—and before Western competitors could scale, Chinese companies owned the market. Global humanoid shipments were just 13,317 units last year. By 2035, that's projected to reach 2.6 million. The early leader often becomes the permanent leader. Our take: The U.S. still leads in AI and software. Figure's deployment at BMW demonstrates capabilities Chinese competitors haven't matched publicly. But hardware matters, and China's supply chain advantages are formidable. For the latest on Chinese robots, see our Unitree G1 review, Unitree H2 review, and analysis of the best humanoid robots on the market. MIT Technology Review: The Hidden Human Labor Behind Humanoids A worker in Shanghai recently spent a week wearing a VR headset and exoskeleton while opening and closing a microwave door hundreds of times a day—training the robot beside him. Welcome to the strange new world of humanoid robot training. MIT Technology Review published a deep investigation into the human labor powering the "autonomous" humanoid industry. Key revelations: Data harvesting at scale: Figure partnered with Brookfield, which manages 100,000 residential units, to capture "massive amounts" of real-world data in homes.Delivery workers as sensors: One company had workers wear movement-tracking sensors while moving boxes, generating training data.Teleoperation everywhere: 1X's Neo robot, shipping to homes this year, relies on remote operators to handle tricky tasks. If the robot gets stuck, a person in Palo Alto takes over. "Just as our words became training data for large language models, our movements are now poised to follow the same path," the report notes. Why it matters: If home humanoids aren't genuinely autonomous, the business model is "a form of wage arbitrage that re-creates the dynamics of gig work while, for the first time, allowing physical tasks to be performed wherever labor is cheapest." Our take: Transparency matters. When companies hide their teleoperation rates, the gap between marketing and reality becomes dangerous—literally, as Tesla's Autopilot lawsuits demonstrate. We're not saying teleoperation is bad; 1X gets customer consent. But the industry needs honest communication about what these machines can actually do today. For more on what you can actually buy, see where to buy humanoid robots. Honor Enters the Humanoid Race Chinese smartphone giant Honor will unveil its first humanoid robot at Mobile World Congress 2026 in Barcelona. The announcement, made via teaser video on X, marks another major consumer electronics company entering the humanoid space. Honor joins Xiaomi, which launched CyberOne in 2022, in the phone-maker-to-robot-maker pipeline. Why it matters: When smartphone companies start building humanoids, it signals the technology is approaching a commercial tipping point. These companies have massive manufacturing capabilities, consumer distribution networks, and experience shipping millions of complex hardware units annually. Our take: The consumer electronics crossover validates the humanoid form factor for home applications. Honor's robot remains mysterious for now, but we'll be watching MWC closely. For more on the smartphone-to-robot trend, see our Xiaomi CyberOne review. Market Pulse What to Watch Next Week Mobile World Congress (March 1-4): Honor's humanoid reveal, plus any robotics announcements from other consumer electronics players.Figure 03 deployment updates: With Figure 02 retiring, watch for announcements about where the next-generation robots will be deployed.China policy signals: The "Made in China 2025" robotics push is accelerating. Watch for new funding announcements or pilot programs targeting the 2.6 million unit 2035 goal. Stay updated with the latest humanoid robot news by visiting our humanoid robot news hub. Ready to buy a humanoid robot? Check out our marketplace. Compare on RoboZaps: Unitree H2, Xiaomi Cyberone, Figure 03, Figure 02 and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 8 Cheapest Humanoid Robots in 2026: Prices and Configurations URL: https://blog.robozaps.com/b/cheapest-humanoid-robots-2026 Author: Radica Maneva Published: 2026-02-24T00:00:00.000Z Updated: 2026-08-01T23:06:02.521Z Last fact-checked: 2026-07-20T00:00:00.000Z Category: best TL;DR: Unitree R1 AIR is the lowest-priced complete biped in this verified shortlist at $4,900 before tax and shipping. Booster K1 starts at $5,999 and clearly lists secondary-development support; higher-priced families still require careful SKU, compute and hand comparisons. Key takeaways: - Unitree R1 AIR is the lowest-priced complete biped in this verified shortlist at $4,900 before tax and shipping. - Booster K1 starts at $5,999 and is the cheapest option here whose named tiers all list secondary-development support. - The advertised Unitree G1 and H2 base prices do not include the compute, hands or secondary-development access offered on their quote-only EDU versions. - LimX Oli has the lowest published full-size starting price here at CNY 158,000, but the price is not mapped to a specific Oli tier. - Noetix Bumi's CNY 9,998 figure was a sold-out 2025 launch price, so it is no longer ranked as a current offer. FAQ: Q: What is the cheapest humanoid robot in 2026? A: Unitree R1 AIR is the cheapest complete bipedal humanoid with a current manufacturer price in this verified shortlist, at $4,900 before tax and shipping. It has 20 DoF and no dexterous hands; Unitree reserves secondary development for the quote-only R1 EDU. Q: Can you buy a humanoid robot for under $10,000? A: Yes. Unitree lists R1 AIR at $4,900 and R1 at $5,900, while Booster advertises K1 from $5,999. Confirm the exact development access, compute, battery and hand configuration before ordering. Q: What is the cheapest programmable humanoid robot? A: Booster K1 is the lowest-priced complete biped in this shortlist whose manufacturer explicitly lists secondary-development support across every named configuration, starting from $5,999. Booster does not map that starting price to Geek, Education or Professional, so request a configuration-specific quote. Q: What is the cheapest full-size humanoid robot? A: LimX Oli has the lowest manufacturer-published full-size starting price in this shortlist, from CNY 158,000. The price announcement does not map that figure to Oli Lite, EDU or Super, so buyers should confirm the current tier and included hardware in writing. Q: Can you buy a useful home humanoid for $20,000? A: 1X NEO is the only home-first option here at $20,000 ownership or $499 per month, but it remains a preorder. 1X says early units begin with basic autonomy and can use scheduled remote human supervision for unfamiliar tasks. Q: Why is Bumi not called the cheapest humanoid anymore? A: CNY 9,998 was Bumi's 2025 launch and first-batch price. Noetix's current Bumi page lists newer configurations through contact sales and does not show that price, so it is not treated as a current buying price. Q: Which number should I compare when requesting quotes? A: Compare the price of the smallest configuration that can complete your task, not the bare chassis. Put the SKU, hands, compute, sensors, software access, batteries, warranty, freight, tax and acceptance test on one quote sheet. Cheapest Humanoid Robot in 2026 Among the complete bipedal humanoids with current manufacturer prices verified for this shortlist, the lowest is the Unitree R1 AIR at $4,900 before tax and shipping. The lowest-priced option here that clearly includes secondary-development support is the Booster K1, advertised from $5,999, although Booster does not say which of its three configurations receives that starting price. Those two answers differ because a low chassis price does not guarantee the SDK access, compute, sensors or hands needed for development. This guide compares the advertised configuration behind each price, not just the number in the headline. 8 Cheapest Humanoid Robots in 2026 With Published Prices Prices and product pages were checked July 20, 2026. We included only complete bipedal humanoids with a manufacturer-published fixed or starting price and a live product, order or preorder page, or a company statement that production had begun. Prices remain in the seller's currency, so ordering across USD and CNY is approximate. A lower $4,290 Unitree price sometimes appears in searches, but it belongs to R1-D, a separate application/mobile-manipulator family rather than the complete bipedal R1 AIR. EngineAI SA01 and other lower-body platforms are excluded for the same reason. How Much Does a Cheap Humanoid Robot Cost in 2026? Current entry prices for complete bipeds start at $4,900 for R1 AIR. Development-oriented K1 starts at $5,999, while base G1 is $13,500. The next group includes NEO at $20,000, Oli from CNY 158,000, PM01 at CNY 188,000 and H2 at $29,900. A company release prices Kepler's industrial K2 at RMB 248,000 per unit. These are chassis or program entry prices, not delivered project budgets. The configuration tables below show why two robots with similar headline prices can produce very different quotes. What Each Entry Price Actually Buys “Programmable” is not a single feature. A robot may expose basic control interfaces while reserving secondary development, perception compute, simulation tooling or dexterous hands for a higher tier. Before comparing quotes, ask the seller to name the exact SKU and mark every included item. 1. Unitree R1: Lowest Current Price in This Shortlist Unitree publishes three R1 configurations: R1 AIR at $4,900, R1 at $5,900 and R1 EDU through contact sales. Prices exclude tax and shipping. AIR has 20 DoF and a monocular camera; the $1,000 step to R1 adds a 26-DoF layout, waist and head joints, and binocular vision. Both list about one hour of battery life. The product page markets open joint and sensor control interfaces, but its comparison table marks secondary development only for R1 EDU, which ranges from 26 to 40 DoF depending on options. If your project needs custom models, extra compute or hands, request an EDU quote and the relevant development manual before ordering. 2. Booster K1: Lowest Starting Price Here With Explicit Development Support Booster advertises K1 from $5,999. Its three 22-DoF configurations are Geek, Education and Professional, and all three list secondary-development support. The common body is about 95 cm and 19.5 kg, with a stereo depth camera, IMU, microphone array and speaker. Booster lists 48 dense TOPS and about 30 minutes of walking battery life for Geek, 117 TOPS and about 80 minutes for Education, and 200 TOPS with the larger battery for Professional. Geek carries a three-month warranty; Education and Professional list one year. The starting price is not tied to a named tier, so request a written quote for the compute, battery and warranty you require. 3. Unitree G1: Affordable Chassis, Quote-Only Research Configuration Unitree lists the base G1 at $13,500 before tax and shipping. It has 23 DoF, a depth camera, 3D LiDAR and about two hours of listed battery life. The G1 EDU price is not public. Base G1 does not include dexterous hands, a high-compute module or secondary-development support in Unitree's own matrix. EDU can expand the body to as many as 43 DoF and add compute and hand options. Our Unitree G1 review covers the wider platform; for a purchase decision, compare the seller's quote with the exact EDU suffix and hand model. 4. 1X NEO: Cheapest Home-First Preorder 1X NEO is offered at $20,000 for Early Access ownership or $499 a month for the Standard subscription, with a refundable $200 deposit. The ownership package lists a three-year warranty, premium support and priority delivery. 1X says US deliveries start in 2026. NEO is not comparable with a research chassis. It is being sold around household tasks, but 1X says early owners receive basic autonomy and can schedule Expert Mode, in which a remote human supervises unfamiliar work. Treat it as a preorder with a service component, and read the privacy, connectivity and cancellation terms before comparing the monthly figure with outright ownership. 5. LimX Oli: Lowest Published Full-Size Starting Price Here LimX announced Oli preorders from CNY 158,000 in July 2025. The current Oli pages route buyers to order consultation and describe a full-size family spanning 165 to 175 cm and 31 to 43 DoF, with quick-swap batteries, multiple end-effector choices and development tools that vary by tier. The current configuration matrix separates Lite, EDU and Super, while the price announcement does not say which tier starts at CNY 158,000. Oli therefore has the lowest manufacturer-published full-size starting price in this shortlist, not a confirmed current price for an EDU or hand-equipped unit. 6. EngineAI PM01: Current Price, Unclear Priced Version EngineAI's current purchase page lists PM01 at CNY 188,000. The linked product matrix separates Commercial and Education versions, both around 1.4 m tall with about two hours of listed battery life. The store card does not identify the version behind CNY 188,000. Commercial lists 23 DoF, no perception sensors, no high-compute module and no secondary development; Education lists 24 DoF, two depth cameras, an NVIDIA Jetson Orin NX and secondary-development support. Do not rely on the expired CNY 88,000 promotional price still repeated on older pages and third-party lists. 7. Unitree H2: $29,900 Base, Development Extras Quoted Unitree lists H2 at $29,900 before tax and shipping, with H2 EDU sold through contact sales. Both configurations list 31 DoF, binocular vision and about three hours of battery life. The $29,900 configuration does not include dexterous hands, a high-performance compute module or secondary-development support. H2 EDU can add those items and has a longer listed warranty. The base price is useful for chassis comparison, but it does not describe a complete manipulation or embodied-AI package. 8. Kepler K2 Bumblebee: Industrial Price With Company-Reported Delivery A company-issued Kepler release states that K2 is priced at RMB 248,000 per unit and says mass production and deliveries began in 2025. Kepler positions it for industrial work rather than consumer or classroom use. The price and delivery evidence are company claims, and the release does not connect the published figure to a complete configuration, end-effector package or service level. An industrial buyer should require an acceptance test, duty cycle, payload at reach, safety package, software access, spares and on-site support in the quote. Cheapest Humanoid Robot by Use Case Choose by required capability first. For a lab that needs secondary-development access, a $5,999 platform that includes it may cost less than a $4,900 chassis that does not. What the Starting Price Usually Leaves Out The comparison above uses seller prices, not landed project cost. A written quote should answer all of the following: Record the exact SKU, including any base, EDU, developer, commercial or professional suffix.Specify whether the arms end in covers, grippers or dexterous hands, and name the hand model.List joint control, high-level API, secondary development, simulation assets and commercial-use terms.List the onboard processor, accelerator, cameras, LiDAR, storage and usable interfaces.Include batteries, charger, spare-battery availability and realistic duty cycle for the task.Confirm stock status, freight, tax, customs, warranty, parts, remote support and repair location.Define the task, payload, speed, uptime and safety test that the robot must pass. For market-wide ranges and the costs beyond the chassis, use our humanoid robot cost guide. For order channels and seller checks, see the verified humanoid buying guide. Those pages cover different questions; this ranking stays focused on published entry prices and configurations. Why Noetix Bumi Is No Longer Ranked at CNY 9,998 Bumi's CNY 9,998 figure was a 2025 launch and first-batch price. Noetix said that initial 500-unit batch sold out, but its current product page shows newer Pro, EDU-Pro and EDU-Max configurations with “Contact Consultant” instead of a price. The current specifications also differ from the launch version. We therefore keep CNY 9,998 as a historical price, not as evidence that a buyer can order today's Bumi at that figure. What We Excluded No current manufacturer price for a purchasable configuration: Tesla Optimus, Figure, AgiBot, Xiaomi CyberOne, Clone Alpha, Fourier, UBTECH Walker and Digit.Outside this article's form-factor scope: wheeled mobile manipulators, fixed upper bodies and lower-body development platforms. Adding estimate-based rows would make the list longer and less useful. Quote-only systems still matter in our overall humanoid capability comparison, but they cannot be ranked honestly by price here. Which Low-Cost Humanoid Fits Which Buyer? R1 AIR is the price floor in this verified shortlist at $4,900 before tax and shipping. Booster K1 is the clearer development option because every named tier lists secondary-development support, although the $5,999 starting price is not tied to a tier. NEO is a home preorder; Oli, PM01 and K2 require configuration-specific quotes. Compare the configuration that does the work, preserve the seller's currency and treat every missing hand, computer, SDK and support term as an unanswered cost. Compare on RoboZaps: LimX Oli, Bumi, Unitree H2, Kepler K2, Clone Alpha and Xiaomi Cyberone. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 53 Humanoid Robot Companies Compared: The 2026 Directory URL: https://blog.robozaps.com/b/humanoid-robot-companies Author: Radica Maneva Published: 2026-02-24T00:00:00.000Z Updated: 2026-08-01T23:06:11.927Z Last fact-checked: 2026-07-30T00:00:00.000Z Category: insights TL;DR: RoboZaps tracks 95 robot brands; 53 currently qualify as humanoid robot companies across nine countries, and ten flagships have source-backed commercial availability today. Compare them all by status, country, funding and availability, from Figure and Tesla to Unitree, UBTECH, Agility and Boston Dynamics. Key takeaways: - The directory is generated from canonical company and robot records rather than a hardcoded article table. - As of 30 July 2026, 53 of 95 tracked brands qualify through at least one published humanoid robot. - Company operating status and flagship robot availability are separate fields and should not be read as the same thing. - Funding totals include disclosed, source-backed events only; an unknown amount is not an estimate of zero. - Companies are alphabetical. Leadership depends on the metric: deployments, access, maturity, capital or buyer fit. - Nine companies are US-headquartered. The FCC's 28 July 2026 Covered List addition restricts new equipment authorizations for foreign-produced devices; it does not revoke existing ones. FAQ: Q: How many humanoid robot companies are there? A: As of 30 July 2026, RoboZaps lists 53 humanoid robot companies from a broader database of 95 robot brands. A company counts only when it has at least one published robot tagged as a humanoid, so the live total can change when canonical records are added, corrected or reclassified. Q: Which companies make humanoid robots? A: The directory includes companies such as 1X, Agility Robotics, Apptronik, Boston Dynamics, Figure, MagicLab, PUDU Robotics, Tesla, UBTECH, Unitree and XPeng Robotics, plus every other qualifying record in the live table. Search by company or robot, or filter by status, country, funding stage, availability and flagship. Q: Which humanoid robot company is the best? A: There is no universal winner; it depends on the metric. By funding and valuation, Figure AI leads (about $39 billion post-money). By 2025 shipments, AGIBOT and Unitree lead, each around 5,000 units. By proven paid deployments, Agility Robotics (Digit) and Figure are furthest along. By affordability, Unitree. Decide which of those matters for your use case, then compare the companies on that dimension rather than on the loudest claim. Q: Which humanoid robot companies are publicly traded? A: The directly listed rows in this update are Dobot (HKEX: 2432), Faraday Future (NASDAQ: FFAI), Tesla (NASDAQ: TSLA), UBTECH (HKEX: 9880) and Xiaomi (HKEX: 1810). Rainbow Robotics also shows KOSDAQ: 277810 but is classified as a Samsung-controlled subsidiary. A listed parent and its robotics subsidiary are not the same exposure. Q: Which humanoid robots are commercially available? A: Ten flagships have source-backed commercial availability, including AGIBOT A3, Booster K1, Dexmate Vega, MagicBot Gen1, Robosen K1 Pro, UBTECH Walker S2 and Unitree H2. Availability belongs to that model and does not guarantee stock, immediate delivery or support in every country; confirm the configuration and lead time with the supplier. Q: How does RoboZaps verify funding and availability? A: Funding is stored as dated, source-backed events, while availability comes from sourced robot records. When current evidence is missing or does not fit the canonical taxonomy, the directory shows unknown rather than inferring a value. Q: How many US humanoid robot companies are there? A: Nine of the 53 companies in this directory are headquartered in the United States: 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Dexmate, Faraday Future, Figure AI, Kinisi Robotics and Tesla. Seventeen further rows record an unknown headquarters while sourcing is completed, so nine is a floor rather than a settled total. Q: Does the FCC rule ban Chinese humanoid robots? A: No. On 28 July 2026 the FCC added foreign-produced advanced robotic devices to its Covered List, which under section 2.903(a) blocks covered equipment from receiving new equipment authorizations. Devices authorized before that date keep their authorizations, and a same-day waiver (DA 26-789) lets them receive software and firmware updates at least until 1 January 2029. The rule also turns on where a device is produced rather than the nationality of the company that sells it. Most lists of humanoid robot companies become stale as soon as a funding round closes, a robot generation changes or a purchase page goes live. This directory is generated from the same canonical company and robot records that power RoboZaps, so the table can be corrected once at the source instead of copied into another article. As of 30 July 2026, RoboZaps tracks 95 brands in total. The 53 humanoid robot companies in this directory qualify because they have at least one published humanoid robot record. Those companies span nine headquarters countries, and ten listed flagships have source-backed commercial availability. Use the filters to compare status, country, funding stage, availability and flagship model. The default order is alphabetical, not a disguised ranking. Humanoid robot companies: live directory A company qualifies when it has at least one published humanoid robot in the RoboZaps database. The broader brand database is shown separately in the summary, so a company that only makes quadrupeds, industrial arms or service robots is not quietly counted as a humanoid company. What the current market snapshot shows The company count and the commercial count tell different stories. All 53 companies clear the database inclusion rule, while only ten current flagships have evidence strong enough for the commercially available label. The rest may be announced, offered for pre-order, restricted to selected pilots, waitlisted or genuinely unknown. A polished demonstration does not move a robot into the commercial column. Ten flagships now carry the commercially available label, including AGIBOT A3, Booster K1, Dexmate Vega, MagicBot Gen1, Robosen K1 Pro, UBTECH Walker S2 and Unitree H2. That label is deliberately model-specific. It does not mean every configuration is in stock, that delivery is immediate in every country, or that a buyer can deploy it without integration work. The linked source shows the evidence used, while a supplier quote remains the final authority on configuration, lead time and territory. The headquarters filter measures where the company is based, not where every robot is designed, manufactured or supported. Global groups can have research centers and sales offices across several countries. Treat the country view as an industry map, then verify service coverage separately if you are buying. A headquarters address says very little about spare parts, field support or regulatory readiness in your operating region. Funding needs similar restraint. The table totals disclosed events that have a date and source; it does not estimate hidden rounds, infer valuation from investment, or treat a public parent company's market capitalization as money raised by its robotics subsidiary. An undisclosed result means the evidence does not fit the canonical funding model yet. It is not a claim that the business has raised nothing. US humanoid robot companies in this directory Nine of the 53 companies are headquartered in the United States: 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Dexmate, Faraday Future, Figure AI, Kinisi Robotics and Tesla. China is the largest single cluster at 19. The United Kingdom has two, and Belgium, Canada, Israel, Italy, Japan and South Korea have one each. Seventeen rows still read unknown because headquarters sourcing is incomplete, so treat the US and China counts as floors rather than settled totals. The US group is weighted toward well-funded developers that have not yet opened general sales. The four largest disclosed raises in the directory all belong to US companies: Figure AI at a Series C of at least $1 billion, Apptronik at a Series A of at least $935 million, Agility Robotics at a $150 million Series B and 1X Technologies at a $100 million Series B. Against that, only Dexmate's Vega carries the commercially available label and 1X's NEO is pre-order, while availability for Digit, Apollo, Atlas, Figure 03 and Optimus is recorded as unknown rather than assumed. What the FCC Covered List addition changes On 28 July 2026 the FCC's Public Safety and Homeland Security Bureau added foreign-produced advanced robotic devices to the Covered List (DA 26-786). Section 2.903(a) of the Commission's rules prohibits covered equipment from receiving equipment authorizations, and section 2.911 requires every authorization application to certify that its equipment is not covered. Equipment that cannot be authorized generally cannot be lawfully imported or marketed in the United States. The action is narrower than a ban on Chinese robots, because it reaches only new authorizations. A waiver issued the same day (DA 26-789) lets every advanced robotic device already authorized for US use keep receiving the software and firmware updates that mitigate consumer harm, at least until 1 January 2029. Producers can also ask the Department of War to evaluate a device for a Conditional Approval, which exempts it from the list. None of this maps onto the country column above. The Covered List turns on where a device is produced, and the Commission adopted the interagency body's definition of foreign-produced rather than writing a company-nationality test; the parallel power-inverter entry reaches inverters produced in a foreign country regardless of the nationality of the producer. A company headquartered in the United States that manufactures abroad is not automatically outside the rule. Read the headquarters column as company data, not as a compliance signal. For per-model authorization status, including which grants predate the rule, see which humanoid robots are legal in the US. Current generations versus published robot pages Flagship generations can change faster than a full robot profile can be reviewed. Figure 03 and XPeng's Next-Gen IRON are the current manufacturer-backed generations in this update, replacing Figure 02 and PX5 as company flagships. When a current flagship's full RoboZaps detail page is still passing its publication checks, the directory shows the model name without creating a dead link. The company still qualifies only through a separate published humanoid record; an unpublished flagship cannot qualify a company by itself. Showing the current model without linking to an unfinished page avoids two problems: an obsolete generation at the top of a company row, and an incomplete robot profile published only to make the table look current. The flagship filter remains useful while the new profile completes its evidence and media checks. How to read the company data The table keeps company status separate from robot availability. An active company can have an announced flagship, while a commercially available robot can belong to a subsidiary or public issuer. Combining those concepts into one “status” label creates false certainty, so the directory does not do it. Company status describes the organization: active, acquired, pivoted, inactive, defunct or unknown.Country is the company headquarters country, not the manufacturing location or every office it operates.Funding is event-based. Totals include disclosed, source-backed announcements only; ≥ means the announcement said “at least” or “over.”Commercial availability belongs to the listed flagship robot and does not imply that every model from the company can be ordered.Flagships are explicitly curated where possible. A deterministic database selection is labelled when an explicit flagship is unavailable.Unknown is a real data state. It means the current evidence is incomplete—not that RoboZaps estimated a value. Which humanoid robot company is leading? The honest answer is that it depends on the metric, so here is the leader on each. By funding and valuation, Figure AI leads, at roughly $39 billion post-money in its 2025 round. By 2025 shipments, the Chinese makers AGIBOT and Unitree dominate, each reporting on the order of 5,000 units. By real paid deployments, Agility Robotics (Digit, working in warehouses) and Figure (on a BMW line) are furthest along. By low-cost access, Unitree undercuts almost everyone. There is no single winner because leading in funding is not the same as leading in deployments, access or serviceability, so compare the dimension that matches your use case. Deployment evidence: named customers, repeat orders, paid work and duration in operation.Commercial access: order, pre-order, waitlist, pilot-only or announcement status.Product maturity: reliability, safety documentation, autonomy, teleoperation and task fit.Capital and ownership: disclosed funding events, public listing or strategic parent support.Buyer fit: payload, runtime, environment, integration, support territory and total cost. For a robot-level decision, start with the task rather than the loudest company claim. A warehouse buyer, a university lab and an event operator need different evidence, safety controls and support. The company directory tells you who is active and what each supplier currently presents as its flagship; the robot records hold the model-level specifications. Publicly traded, private and subsidiary companies Ownership structure matters, especially for investors. A public company row refers to the listed issuer shown in the table. A subsidiary row identifies an operating company controlled by a parent; it is not automatically a separately investable stock. Private companies may disclose funding without disclosing valuation, and the directory does not convert one into the other. If your intent is investment research, use the dedicated humanoid robot stocks guide, which separates direct exposure, parent-company exposure and speculative adjacency. How buyers should compare humanoid suppliers A company list is a starting point, not a procurement shortlist. Before contacting a supplier, define the task and operating environment, then ask for evidence at the exact level you need: a purchasable configuration, a paid pilot, integration requirements, safety documentation, service coverage and a complete quote. Confirm whether “available” means a standard order, a pre-order, a waitlist or a supplier-selected pilot.Request evidence for the task, payload, runtime, floor conditions and human interaction in your own environment.Price the full deployment: robot, end effectors, software, integration, training, support, spares and downtime.Check data handling, remote access, update policy, warranty, safety process and local service response.Treat demos, targets and planned production as different evidence classes from repeatable customer operation. Use the humanoid robots for sale and buying guide for current buying routes, and the humanoid robot price guide for a more complete view of acquisition and operating cost. What counts as a humanoid robot company? For this directory, the inclusion rule is mechanical: at least one published robot record must carry the humanoids tag and link back to the company. That keeps the scope reproducible. It also means the count can rise or fall when a record is published, corrected, reclassified or retired. This is a narrower definition than lists that count component suppliers, research projects, telepresence platforms, service robots with a screen or every company that has announced an embodied-AI project. Those businesses can matter, but mixing them into a humanoid manufacturer count makes comparisons unreliable. RoboZaps keeps the broader 95-brand universe visible while reserving the 53-company figure for records that pass the stated rule. A familiar company may be absent because its humanoid page is still a draft, its robot is not yet tagged correctly, or the evidence does not support a clean manufacturer-to-product relationship. Those cases stay in the verification queue instead of padding the total. Once the canonical record clears the rule, the table updates without a hand-edited company list. Browse the wider robot brand database when you want manufacturers across all robot categories, or the robot database when you want individual models rather than companies. If you are comparing these companies for a purchase, our robotics advisory compares platforms across manufacturers against one evidence standard. Update and correction policy The initial directory is server-rendered for search engines and accessibility, then hydrated for filtering in the browser. Company and robot facts remain in Medusa; editorial explanation remains in Payload. That separation prevents a copied table in an article from drifting away from the product database. Where the evidence is incomplete, the table says unknown. Source links sit beside status, funding and availability where available, and verification dates remain attached to the underlying records. Read the full RoboZaps methodology or report a correction when a primary source has changed. Compare on RoboZaps: AGIBOT A3, Walker S2, MagicBot Gen1, Unitree H2, Vega and Booster K1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree H2 Humanoid Robot: $29,900 Price, Specs, vs H1 (2026) URL: https://blog.robozaps.com/b/unitree-h2-review Author: Radica Maneva Published: 2026-02-24T00:00:00.000Z Updated: 2026-08-01T10:48:55.428Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: reviews TL;DR: The Unitree H2 is publicly listed at $29,900 on Unitree's official store, one of the only full-size humanoids with a real list price, though at our July 17 check the listing showed out of stock (contact sales). Base units ship with placeholder hands, development needs the quote-only EDU edition, and NA pricing starts at $40,900. The $100,000 H2 Plus is NVIDIA's GR00T reference robot. Key takeaways: - The H2 has a genuine public list price, $29,900 on Unitree's official store, though the listing showed out-of-stock/contact-sales at our July 17, 2026 check; base units ship with non-functional placeholder hands and development requires the quote-only EDU edition. - North American buyers pay more: Unitree's regional partner lists the H2 from $40,900 and the EDU from $68,900, with phased deliveries announced from April 2026 and no audited delivery reports yet. - The $100,000 H2 Plus, announced at Computex 2026, is NVIDIA's first Isaac GR00T reference humanoid: Jetson Thor compute plus tactile five-finger hands for a 75-DoF total, shipping late 2026. - Unitree is the industry's commercial outlier: ~¥1.7B revenue (+335%), profitable (~¥278M statutory), ~5,500 humanoids shipped in 2025, with STAR Market IPO registration approved July 2, 2026. - US institutional buyers have homework: Unitree was added to the Pentagon's 1260H list in June 2026, and the vendor stack has a documented security history (Go1 backdoor, UniPwn), though the H2 itself is not on the affected list. FAQ: Q: How much does the Unitree H2 cost? A: $29,900 on Unitree's official store, China-direct, plus tax, shipping and customs, a price unchanged since the October 2025 launch. Through Unitree's North American partner, pricing starts at $40,900, and the developer-ready EDU edition is quote-only. Functional hands cost extra on all base configurations. Q: Can you buy the Unitree H2 now? A: It is publicly listed at $29,900 on shop.unitree.com, one of the only full-size humanoids with a real list price, but at our July 17, 2026 check the store's own product data showed the listing out of stock, routing buyers to contact sales. North American pre-orders opened in February 2026 through the regional partner with a $2,500 deposit and phased deliveries announced from April 2026. Q: Does the Unitree H2 come with hands? A: Not functional ones. The base H2 ships with non-working placeholder hands that complete the silhouette. Five-finger dexterous hands are paid add-ons, and the $100,000 H2 Plus includes tactile five-finger hands as standard. Q: What is the difference between the H2 and H2 Plus? A: Same chassis, different brain and hands: the $100,000 H2 Plus adds NVIDIA Jetson Thor compute, dual Sharpa Wave tactile five-finger hands (22 actuated DoF each), wrist cameras and a remote e-stop, totaling 75 DoF. NVIDIA made it the first Isaac GR00T reference humanoid, with availability pointed at late 2026. Note that Unitree's feature copy presents the hands as included while its spec table marks them optional; confirm at order time. Q: Is the Unitree H2 faster than the H1? A: No, and it is not trying to be. The H1 holds the humanoid speed record at 3.3 m/s; the H2 trades speed for articulation, 31 body degrees of freedom against the H1's roughly 19, with smoother, more human-like motion. Unitree has published no official H2 top speed. Q: How tall is the Unitree H2? A: About 1.8 m and roughly 70 kg with its battery. Unitree's own pages list both 180 cm and 182 cm, so treat the height as approximately 1.8 m rather than a precise figure. Q: What can the Unitree H2 do? A: Out of the box it is a research and performance platform: the famous dancing and kung-fu videos are real but choreographed motion showcases, not task autonomy, and Unitree made no autonomy claim for them. Useful work requires the EDU edition, add-on hands and your own development. It is not a home robot; Unitree's CEO says household humanoids are years away. Q: When will the Unitree H2 be delivered? A: China-direct orders ship on Unitree's standard timelines with the buyer handling customs. North American phased deliveries were announced to start in April 2026 by the regional partner; no independently audited delivery reports existed as of July 17, 2026. Q: Is the Unitree H2 safe to buy for a US institution? A: It requires review. Unitree was added to the Pentagon's Section 1260H list in June 2026 (bringing DoD contracting bans), a House committee held a hearing on the company in March 2026, and the vendor's stack has a documented security history including the Go1 backdoor and the UniPwn exploit, which did not list the H2 but affected sibling models. Private labs can and do buy them; network isolation and procurement review are the sensible baseline. How much does the Unitree H2 cost, and can you buy one? The Unitree H2 is listed at $29,900 on Unitree's official store, plus tax, shipping and your local customs duties, and the price has held steady since launch in October 2025. One live caveat: when we checked the store's own product data on July 17, 2026, the H2 listing showed out of stock, routing buyers to contact sales rather than checkout. That still makes it one of the very few full-size humanoids on earth with a public list price, just not, at this writing, an instant buy button. The number comes with fine print nobody puts in a headline: the base robot ships with non-functional placeholder hands, secondary development requires the EDU edition (quote-only), and North American buyers ordering through Unitree's regional partner pay from $40,900. All of it is unpacked below. One disclosure up front: this is desk research built on the sources linked throughout, primary wherever they exist, not a hands-on test; we verified the store listings by direct fetch on July 17, 2026. The misinformation audit: what circulates vs the record What is the Unitree H2? The H2 is Unitree's full-size flagship humanoid, unveiled October 20, 2025 in a launch video titled "H2 Destiny Awakening" that showed it dancing, throwing kung-fu combinations and walking a runway, the first Unitree humanoid with a lifelike bionic face (TechNode). Unitree's own framing at launch: "standing 180 cm tall and weighing 70 kg... born to serve everyone safely and in a friendly way." Where the older H1 chased speed, it still holds the humanoid sprint record at 3.3 m/s, the H2 chases range of motion: more joints everywhere, optional dexterous hands, and motion quality good enough that the launch clips were widely mistaken for CGI. It is the machine Unitree sells while rivals show renders, and in June 2026 NVIDIA made its Plus variant the reference humanoid for its GR00T research stack, which tells you exactly what market this robot actually serves: labs, not living rooms. Official specs, and what's still unpublished Sources: the official store listing, H2 specification page, H2 Plus page and Unitree's June 2026 announcement, all fetched directly. Our Unitree H2 database record is queued for a data refresh against this review's store-verified findings. The hands gotcha nobody explains The single most important thing a $29,900 buyer needs to know: the base H2 ships with non-functional placeholder hands. They complete the silhouette; they do not grasp. Functional five-finger hands are paid add-ons (Unitree's Dex5 family, sold through distributors, with circulating per-hand prices in the four-to-five-figure range that Unitree does not publish), and secondary development of any kind, your own code on the robot, officially requires the EDU edition, which is quote-only via Unitree's sales channel. So the honest arithmetic for a research buyer is $29,900 for the platform, plus hands, plus the EDU delta, plus tax, shipping and customs. It is still dramatically cheaper than anything comparable, and it is not $29,900 out the door for a robot that can pick things up. H2 vs H2 Plus: the $70,000 difference In June 2026 the H2 line split. The H2 Plus, announced at Computex with Jensen Huang on stage, is the same chassis upgraded into NVIDIA's first official Isaac GR00T reference humanoid for academic research: Jetson Thor T5000 compute onboard, dual Sharpa Wave tactile five-finger hands with 22 actuated DoF each and over 1,000 tactile pixels per fingertip, wrist cameras, and a 75-DoF total count, listed at $100,000 on Unitree's store with availability pointed at late 2026. One manufacturer-side contradiction worth budgeting around: Unitree's feature copy presents the Sharpa Wave hands as part of the system while its detailed spec table marks them optional (official announcement). Its launch labs, the Allen Institute for AI, ETH Zurich, Stanford's robotics center and UCSD, define the customer. The decision rule is clean: the $29,900 H2 is the body; the $100,000 H2 Plus is the body plus the research-grade nervous system. There is also a third sibling now: the H2-D, an officially listed wheeled dual-arm variant (135 kg, 1.5 m/s, built for industrial manipulation) with no published price or date yet. H2 vs H1: successor, not upgrade The pricing inversion is the story: the newer, more capable robot lists at a third of the old one's nominal price, because the H1 was a low-volume research flagship and the H2 is a volume product from a company that shipped roughly 5,500 humanoids in 2025. If you are choosing between them, the H1 only wins if you specifically need its raw locomotion pedigree; our Unitree H1 review covers it, and the $13,500 G1 remains the budget entry to the same ecosystem. Ordering, decoded China-direct: $29,900 on the official store, buyer pays local customs; EDU by sales contact. North America: Unitree's regional partner Toborlife opened pre-orders in February 2026 with a $2,500 deposit, base H2 from $40,900 and H2 EDU from $68,900, with phased deliveries announced from April 2026, figures from the partner's own release rather than audited reporting; no independently verified North American deliveries existed at our check. European distributor listings run higher still. The pattern to internalize: "$29,900" is the China list price, a landed, working, region-supported H2 runs meaningfully higher everywhere, before hands, and stock state moves; at our July 17 check the official listing itself was in contact-sales mode. A published price with a sales queue is still a different universe from rivals' no-price pages. What it can actually do Be precise about what the demos show. The launch video's dancing, kung fu and runway walk are real robotics, not CGI, and they are choreographed and RL-trained motion showcases; Unitree made no task-autonomy claim for them. The H2's biggest public moment, appearing as the Monkey King at the 2026 Spring Festival Gala astride a cloud of B2-W robot dogs, was theater, and Unitree's claim that the accompanying robot kung-fu segment ran "fully autonomously" is a company claim without independent verification (SCMP). What the H2 is genuinely for is being a body for someone else's brain: embodied-AI research, university labs, entertainment bookings, and increasingly the NVIDIA GR00T ecosystem. It is not a home robot, and Unitree's own CEO is the source: Wang Xingxing has said household humanoids are two to three years away for safety reasons and that humanoids have not had their "ChatGPT moment." Security and the Washington problem Any institutional buyer needs this section. Unitree products have a documented security history: the 2025 Go1 backdoor (CVE-2025-2894), a tunnel service that allowed remote access to quadrupeds found at MIT, Princeton and CMU (Axios), and the 2025 "UniPwn" Bluetooth exploit affecting Go2, B2, G1 and H1, per the IEEE Spectrum disclosure, whose late-2025 reporting described Unitree's patch response as incomplete; the H2 was not on the affected list, which covered its sibling models. The policy layer escalated in 2026: a House Homeland Security hearing on Unitree in March, and on June 8 the Pentagon added Unitree to its Section 1260H list of "Chinese military companies" (TechCrunch), a designation that restricts US Defense Department procurement. None of this stops a private lab from buying an H2; all of it belongs in any US university's or government-adjacent buyer's procurement review, alongside standard mitigations (network isolation, no sensitive-environment deployment). Unitree, the company Unitree is the industry's commercial outlier: it makes money. Its IPO prospectus shows roughly ¥1.7 billion of 2025 revenue, up ~335%, with about ¥278 million statutory net profit (a ¥600 million-class figure on an adjusted basis excluding share-based payments), a 60% gross margin, and roughly 5,500 humanoids shipped in 2025, which the prospectus, citing industry data, calls the global #1 (Caixin). Regulators approved its Shanghai STAR Market IPO registration on July 2, 2026, targeting a roughly ¥4.2 billion (~$618 million) raise, with Chinese financial coverage pegging the implied valuation near ¥40 billion; pricing and the listing date were still pending as of July 16. Founder Wang Xingxing targets 10,000 to 20,000 humanoid shipments in 2026. The caveat inside the prospectus: Q1 2026 adjusted profit fell by half as R&D and sales spending spiked, so the profitability story is real but not effortless. H2 vs the field The comparison collapses to one sentence: if you want a full-size humanoid you can actually order this year at a published price, the H2 is essentially the list, with its own G1 sibling below it and everything else either enterprise-gated (Walker S2) or unpurchasable (Figure 03, Optimus). The wider rankings live in our best humanoid robots guide and cost guide. Who should buy it, and who shouldn't Buy it if you are a lab, university program, integrator or entertainment operator that needs a full-size, well-supported research body at the industry's best price-to-hardware ratio, and budget honestly for the EDU edition plus hands if you intend to develop on it. Think harder if you are a US institution with federal exposure, the 1260H listing and the security history belong in your review, or if you expect a robot that does useful work out of the box: the H2 is a platform, and the intelligence is your job. Skip it entirely if you want a home robot; nothing Unitree sells is one, by its founder's own timeline, and the 1X NEO is the product actually chasing that market. What to watch, dated Four threads move this page. The IPO pricing and debut, possibly within weeks of this writing, which will publish fresh audited numbers. H2 Plus availability, pointed at late 2026 with NVIDIA reportedly citing October. Verified North American deliveries, which would upgrade the April-2026 partner claims from announced to audited. And any H2-D pricing, which would extend the line into industrial manipulation. We re-verify this page at each. The bottom line The Unitree H2 is the least hypothetical humanoid you can name: a real price, a real buy button, real shipment volume behind it, and a maker that turns a profit, all of which remain rare enough in this industry to be differentiators on their own. Its honest limits are equally clear: placeholder hands out of the box, development locked behind a quote-only EDU tier, no published speed, choreographed demos rather than task autonomy, and a security-and-sanctions file that US institutional buyers cannot ignore. As a research body at $29,900 it is the value benchmark the entire industry gets measured against; as anything more autonomous than that, it is exactly as unproven as everyone else. The H2 record in our robot database, currently queued for a refresh against this review, tracks the verified state as it changes. --- # Galbot G1 Price 2026: Cost, Can You Buy It & What It Is URL: https://blog.robozaps.com/b/galbot-g1-review Author: Radica Maneva Published: 2026-02-24T00:00:00.000Z Updated: 2026-07-27T12:18:32.108Z Last fact-checked: 2026-07-27T00:00:00.000Z Category: reviews TL;DR: You cannot buy a Galbot G1 off the shelf: it is an enterprise robot sold by quote in China. The real prices are a ~630,000 yuan (~$91k) JD.com Spring Gala flash sale and a standing JD listing around 700,000 yuan; Western 'in stock' listings are not confirmed Galbot channels. It is a wheeled dual-arm manipulator from Galaxy General, and the certified robot in China's first licensed robot pharmacy. Key takeaways: - No global list price: the G1 is an enterprise robot sold in China by quote. The clearest public price was two units on JD.com during the 2026 Spring Festival Gala at ~630,000 yuan (~$91k), which sold out in minutes; JD later showed a standing China listing around 700,000 yuan. - Enterprise deployments run about 700,000 yuan a unit, and a '$97k' tag is just the dollar version of the real JD listing. The '$119,995 in stock' and '$55k' Western listings are not confirmed Galbot channels, and low-five-figure 'G1' prices belong to the unrelated Unitree G1. - It is wheeled, not bipedal: a dual-arm torso (suction cup plus gripper) on a 360-degree omnidirectional base, with a lifting column that reaches about 2.4 m. - Its real first: it runs China's first pharmacy licensed to operate with a robot (a Beijing Nepstar store, with Meituan Medicine), as the country's first certified robot pharmacy assistant. - The autonomy is real but narrow. It won an autonomous sorting competition, but live-store robots are slow, fail some grasps, handle few products, and still need human staff; Galaxy General reportedly ships fewer than 150 robots a year. - Galaxy General is China's highest-valued embodied-AI company (~$3 billion), backed by CATL, Meituan and IDG Capital. There is a G1 Premium (Jetson Thor) but no 'G2'; the S1 is a separate heavy-industrial robot. FAQ: Q: How much does the Galbot G1 cost? A: There is no global list price; it is sold to businesses in China by quote. Two units sold on JD.com during the 2026 Spring Festival Gala at about 630,000 yuan (~$91,000), and JD later showed a standing China listing around 690,000 to 700,000 yuan, close to what enterprise deployments reportedly cost. A '$97,000' tag is just the dollar version of that; the '$119,995 in stock' and '$55,000' Western listings are not confirmed Galbot channels. Q: Can you buy a Galbot G1? A: Not as a consumer, and there is no established channel outside China, though one G1 was imported to Japan for business testing in 2026. It is an enterprise robot sold via a quoted deployment through Galbot. Western pages showing it 'in stock' with an add-to-cart button are not confirmed Galbot channels. Q: Is the Galbot G1 the same as the Unitree G1? A: No. They are different robots from different companies that share only the 'G1' name. The Unitree G1 is a roughly $16,000 bipedal humanoid; the Galbot G1 is a wheeled enterprise manipulator with no public price. Low-five-figure 'G1' prices (around $13,500 to $16,000) refer to the Unitree robot. Q: Does the Galbot G1 walk? A: No. It uses a 360-degree omnidirectional wheeled base with a lifting column that raises and lowers the torso, reaching about 2.4 m. It is a wheeled semi-humanoid, not a bipedal walking robot. Q: Who makes the Galbot G1? A: Galaxy General (Beijing Galbot Co.), founded in 2023 and co-founded by Peking University professor Wang He. By its December 2025 funding it was valued at about $3 billion, China's highest-valued embodied-AI company, backed by CATL, Meituan, IDG Capital and Sequoia China. Q: What is the Galbot G1 used for? A: Enterprise manipulation in retail, pharmacy, logistics and factories: picking, sorting, packing and restocking. Its flagship use is robot-assisted pharmacies and autonomous retail kiosks in China. Q: Is the Galbot G1 fully autonomous? A: Partly. It won an autonomous pharmaceutical-sorting competition in 2025, and its store runs use Galaxy General's own AI models rather than teleoperation. But reporters found live robots slow and error-prone, handling few products and still needing human staff for payment and prescriptions. Q: What is the difference between the G1, G1 Premium and S1? A: The G1 runs on NVIDIA Jetson Orin; the G1 Premium upgrades to Jetson Thor for more onboard compute. There is no 'G2.' The S1 is a separate heavy-load industrial robot: Galbot's launch release claimed a 50 kg-class dual-arm payload, while its current developer manual specifies 30 kg combined. It is not a G1 successor. Q: How many Galbot G1 robots have been deployed? A: Galaxy General reported over 170 autonomous retail units across 40-plus cities by mid-2026, and around seven Beijing robot-pharmacy deployments by mid-2025 growing to about ten by March 2026, one of them the licensed Nepstar store. One report put its production at fewer than 150 robots a year, so the 'thousands of units' quoted elsewhere are order backlog, not an installed base. Q: What are the Galbot G1's specifications? A: Per Galbot's developer manual: about 92.5 kg and 1.73 m tall (reaching roughly 2.4 m via a lifting torso), 21 articulated joints, a 5 kg per-arm / 10 kg combined payload, an 8-hour laboratory runtime, and NVIDIA Jetson AGX Orin 64GB compute (Jetson Thor on the G1 Premium). Its sensors include a head binocular camera, four torso RGB cameras, wrist depth and six-axis force sensors, a chassis 3D LiDAR and eight ultrasonic sensors. You cannot walk in and buy a Galbot G1, and the one time the public could, it was a stunt. The G1 is an enterprise robot sold in China by quote, not a product with a shelf price. The only retail moment came in February 2026, when two units went up on JD.com during the Spring Festival Gala broadcast for about 630,000 yuan (~$91,000) each and sold out in minutes. Real deployments run closer to 700,000 yuan a unit. JD later listed it in China at around 690,000 to 700,000 yuan too. What is not real is the Western commerce: the "$119,995 in stock" and "$55,000 prototype" listings are not confirmed Galbot channels. What makes the G1 worth understanding is not the price, though. It is that Galbot's maker, Galaxy General, is the highest-valued embodied-AI company in China, and the G1 runs a genuine first: the country's first pharmacy licensed to operate with a robot, where the G1 is the certified robot pharmacy assistant. This page covers what it costs, who makes it, what it can actually do, and where the marketing outruns the hardware. Can you buy a Galbot G1, and what does it cost? No, not as a consumer, and there is no established channel outside China, though one G1 was imported to Japan for business testing in 2026. Galbot sells the G1 to businesses through a quoted deployment, integrated per site. There is no public list price and no "buy" button on its own site. The real price signals are the flash-sale price, a standing JD listing, and enterprise pricing; treat the Western commerce listings as unofficial. Quick facts and key specifications Unlike most Chinese makers, Galbot publishes a full spec sheet in its developer manual, so the figures below are official rather than the reseller guesses that disagree with each other elsewhere online. Full verified specs also live in the Galbot G1 record in the RoboZaps robot database. The one thing nearly every aggregator gets wrong: the G1 is not bipedal. It rides a 360-degree wheeled base on a single lifting column that raises and lowers the whole torso, so it can reach a top shelf or a low bin without legs. Anyone describing it as a walking humanoid has not read Galbot's own spec sheet. Which prices are real, and which are not There is no global list price, but there are real Chinese figures and then listings that are not confirmed Galbot channels: ~630,000 yuan (~$91,000) was the flash-sale price for two units on JD.com during the Spring Festival Gala. Promotional, and only two units.~690,000 to 700,000 yuan is the standing figure JD showed for the G1 in China after the Gala, and roughly 700,000 yuan is what enterprise pharmacy deployments reportedly cost. A "$97,000" tag online is just the dollar conversion of that, not an invented number.$119,995 "in stock" and $175,000 AUD "ships within a week" appear on Western storefronts with "add to cart" buttons. Galbot sells no inventory through these Western storefronts, so treat them as unconfirmed, not channels you can actually buy through.A $55,000 "prototype" on a spec aggregator is likewise unverified and traces to no Galbot source. One more trap worth naming: the Unitree G1 is a completely different robot from a different company: a roughly $16,000 bipedal humanoid that shares only the "G1" name. Prices like "85,000 yuan" or "$13,500" belong to the Unitree machine, not this one. If a page quotes a low-five-figure price for a "G1," it is almost certainly describing Unitree's robot by mistake. Who makes it: Galaxy General Galbot is the product brand of Galaxy General (Beijing Galbot Co.), founded in May 2023 and now, by valuation, the leading embodied-AI startup in China. It was co-founded by Wang He, a Peking University professor with a Stanford robotics PhD who still runs a joint lab at PKU, alongside Yao Tengzhou, who leads hardware and commercialisation. By the company's own December 2025 announcement it had raised over $800 million at a roughly $3 billion valuation, and it followed that with a reported 2.5-billion-yuan round in March 2026. Its backer list reads like a who's-who of Chinese strategics: battery giant CATL led a $153 million round in 2025, and Meituan, IDG Capital, Qiming and Sequoia China are all in. The pitch is full-stack embodied AI: build the data, the foundation models and the robot in-house, and use unmanned retail and pharmacy as the way in. The design and hardware The G1's shape follows its job. Manipulation, not walking, is the point, so Galaxy General skipped legs and put a capable dual-arm torso on a stable wheeled base. The left arm ends in a suction cup for flat and bagged items; the right uses an adaptive parallel gripper for grasping. The lifting column gives it a working envelope from floor level to roughly 2.4 metres, which is what lets a single robot stock a shelf and pick from a bin. On compute, there are two tiers. The standard G1 runs on NVIDIA's previous-generation Jetson Orin class. The G1 Premium upgrades to NVIDIA's Jetson AGX Thor, which Galbot says made it the first company to put Thor in a humanoid, claiming roughly 7.5 times the AI compute and 3.5 times the energy efficiency of the Orin generation. Galbot's developer manual lists a fuller sensor suite than the two arms suggest: a binocular camera in the head, four torso RGB cameras, depth cameras and six-axis force sensors at both wrists, a 3D LiDAR and eight ultrasonic sensors on the chassis, and inertial units for balance. It is built to navigate and manipulate a real store, not just pose for a demo. The AI, and how autonomous it really is The interesting part is the software. Galaxy General trains its robots largely on synthetic data generated in simulation, then fine-tunes on real data. Its grasping foundation model, GraspVLA, was pre-trained on a billion-frame synthetic grasping dataset and published as a paper in 2025; a retail-specific model, GroceryVLA, handles picking varied and fragile products off cluttered shelves without being reprogrammed for each item. The company's strongest genuine autonomy proof is a competition: at the 2025 World Humanoid Robot Games, the G1 won the pharmaceutical-sorting event outright and fully autonomously, while, as founder Wang He pointed out, most rivals were teleoperated. That said, the live-deployment reality is narrower than the highlight reels. Reporters who visited Galaxy General's own stores found the robot slow and brittle: at a FamilyMart it handled fewer than ten of the store's roughly one thousand products and failed to grasp chip bags about four times in ten, and the store still ran four human staff. At the pharmacy, one reporter noted the robot "had not yet learned the locations of medicines." Humans still handle payment everywhere, and pharmacists still review every prescription. The autonomy is real, but it is a curated set of tasks done reliably, not a general worker. The deployment record Galaxy General's track record is genuinely ahead of most humanoid makers on one axis, regulated retail, and thinner than the headlines on the rest. Two caveats keep this honest. First, the impressive industrial names are orders and collaborations, not confirmed running deployments, and CATL is also an investor, which blurs "customer." Second, on scale: one on-the-ground report put Galaxy General's actual output at fewer than 150 robots a year, so the "thousands of units" you see quoted are order backlog, not an installed base. The pharmacy authorisation, though, is the real milestone: a regulator licensed a store to run a robot pharmacy assistant, which no competitor has matched. Around seven Beijing robot-pharmacy deployments were reported by mid-2025, growing to about ten by March 2026, before and around that permit. G1, G1 Premium, and the S1 Three names get confused, so here is the map. The G1 is the standard wheeled manipulator on Jetson Orin. The G1 Premium is the same platform with Jetson Thor compute for buyers who want maximum on-board autonomy. There is no "G2." The next distinct product is the S1, a heavy-load industrial robot powered by a CATL battery and reported on CATL's own production lines in 2026. Galbot's own launch release claimed a 50 kg-class dual-arm payload, though its current developer manual specifies 30 kg combined. It is a different machine for a different job, not a G1 successor. How it compares The G1 is not really competing with the walking humanoids people picture, like Tesla's Optimus, Figure's robots, or Unitree's bipeds. Those chase general mobility; the G1 is a purpose-built manipulator that trades legs for a stable base and spends its intelligence on grasping. Against other wheeled manipulators and enterprise service robots it is unusually far along commercially, precisely because Galaxy General narrowed the problem to shelves and counters. If you are comparing it to a Unitree G1, stop: that is a different, cheaper, bipedal research robot that happens to share a name. Who the Galbot G1 is for This is an enterprise robot for Chinese retail, pharmacy, logistics and, increasingly, factory operators who want a robot that can pick and place across a real catalogue without bespoke tooling for each item. It is not for consumers, has no established buyer channel outside China (beyond a single unit imported to Japan for testing), and is not for anyone who needs a robot that walks or works out of the box today. If you are a Western buyer who found a "G1 in stock" page, it is not a confirmed Galbot channel; the only real path is a quoted deployment through Galbot. What to watch Three things will tell you whether Galaxy General's lead is real. Watch whether it actually reaches mass production in 2026 rather than the sub-150-units-a-year it reportedly ships now; whether the November-2025 shift to a joint-stock structure turns into an IPO; and whether any of the marquee industrial partnerships convert into disclosed, running deployments rather than announced orders. The pharmacy authorisation gives it a genuine moat in regulated retail. The open question is whether the hardware can scale as fast as the valuation implies. The bottom line The Galbot G1 is the most commercially advanced wheeled manipulator to come out of China's embodied-AI boom, backed by its best-funded company and validated by a first-of-its-kind licensed robot pharmacy. But you cannot buy one unless you are an enterprise buyer, almost always in China, it has no global list price, and the Western "in stock" listings are not confirmed Galbot channels. Judge it on what it verifiably does, which is pick and pack a curated catalogue, slowly but in a licensed store, and on whether Galaxy General can turn a $3 billion valuation and fewer than 150 robots a year into the scale it is promising. --- # Tesla Optimus Production Timeline 2026: Status, Start Date & What's Official URL: https://blog.robozaps.com/b/tesla-model-s-optimus-robot-factory-conversion Author: Radica Maneva Published: 2026-02-20T12:00:00.000Z Updated: 2026-08-01T23:07:33.398Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: news TL;DR: Formal Optimus production at Fremont has not started as of July 23, 2026. The July 22 Q2 shareholder update confirmed the Model S/X lines are decommissioned and first-generation Optimus lines are being installed, with production 'anticipated later this year.' No start date, no unit count, no price. Initial builds go to Tesla's internal Optimus Academy, not customers. Key takeaways: - Formal production on the new Fremont line has not started: the July 22 Q2 deck says Optimus lines are being installed at Fremont with production 'anticipated later this year,' and Tesla's production reports have never carried an Optimus line item. - The Model S/X trade is now official: Q2 is Tesla's first investor document to confirm the decommissioning as complete and publish line photos; the plan itself was announced in January. - Musk spent the call lowering expectations: 'the hardest product to scale manufacturing that we've ever made,' no existing supply chain, and an initial S-curve 'quite flat and long.' - 1 million a year is the Fremont line's design capacity, not a forecast; the first builds go to Tesla's internal Optimus Academy for training, not to customers. - Ignore the fake numbers: the 'January 21 mass production' story is a sourceless hoax, and the $685M Sanhua actuator order was denied by Sanhua itself. FAQ: Q: Has Tesla Optimus production started? A: Formal production on Tesla's new Fremont line has not started. As of July 23, 2026, Tesla's own Q2 shareholder update says first-generation Optimus lines are being installed at Fremont with production 'anticipated later this year,' and Tesla's July 2 production report, like every one before it, contains no Optimus count at all. Q: When will Optimus production start? A: Tesla's July 22, 2026 guidance is 'soon' and 'later this year,' with no specific date. Tesla replaced earlier talk of a late-July or August start with the broader guidance before that window elapsed. Given the program's forecast record, treat the direction as real and the date as soft. Q: How many Optimus robots exist? A: Tesla has never published a count. Omdia told AP fewer than 500 were shipped in 2025, while DigiTimes-derived supply-chain reporting counted roughly 1,000 assembled by mid-2025; the metrics differ, and Tesla has confirmed neither. Claims of thousands already working in Tesla factories contradict Musk's January 2026 statement that they were 'not in usage in our factories in a material way.' Q: Where will Optimus be built? A: Fremont first: the converted Model S/X line will build Optimus 3, with a design capacity of 1 million robots a year. Optimus 4 is slated for Gigafactory Texas in Austin, 'aspirationally 10 million units a year' per Musk, with no date attached. Q: Are the Model S and Model X discontinued? A: Production has ended. The last cars were built in early May 2026, the lines were torn down in about 46 days, and Tesla's Q2 update confirms the manufacturing lines were decommissioned to make room for Optimus. Q: Is the 1 million robots a year figure real? A: It is the Fremont line's design capacity, confirmed by Musk on the July 22 call. It is not a production forecast: Musk says the initial ramp will be 'quite flat and long' because roughly 10,000 parts on the robot are new and much of the supply chain had to be created from scratch. Q: When can I buy a Tesla Optimus, and at what price? A: You can't yet. There is no ordering channel and no price; Musk's $20,000 to $30,000 talk is a long-term at-scale target, and he pointed public sales at the end of 2027. Tesla has no preorder or deposit program: sites posing as Tesla and taking Optimus money are scams, and third-party 'reservations' are not Tesla orders. Q: What is the Optimus Academy? A: Tesla's internal training program. Per the Q2 update, the initial Optimus builds 'will be used in our Optimus Academy for training data collection and further functionality development,' where fleets of robots practice tasks and feed the learning pipeline before any external deployment. Q: Did Optimus mass production start in January 2026? A: No. That claim traces to a single third-party press-wire article with no Tesla source and no named reporter, and no Tesla production report has ever carried an Optimus line item. Q: What is Optimus 4? A: The next-generation robot Musk says will be built in Austin on a far more vertically integrated supply chain, targeting 'an order of magnitude more production' than Optimus 3, aspirationally 10 million units a year. No production date has been given, and Tesla says scaling depends on its yet-unsited Terafab chip plant. Formal Optimus production on Tesla's new Fremont line has not started. That is the short answer as of July 23, 2026, and for once it comes from Tesla itself rather than from reading tea leaves: the Q2 2026 shareholder update, published July 22, says the Model S and X lines at Fremont have been decommissioned and that Tesla is "installing the first-generation lines for Optimus, where we expect to start production soon," with production "anticipated later this year." Tesla's production reports have never disclosed an Optimus count, no start date has been given beyond "soon," and the program's own history says to treat every date with caution. Prototype and pilot units existed before this line; what has not begun is output from the line built to make robots at scale. What makes July 22 a real milestone anyway is that the factory conversion is now official. The trade of two flagship cars for a robot line had been reported piecemeal for months; the Q2 deck is the first Tesla investor document to confirm the decommissioning as complete and to publish photos of the new line; the plan itself dates to the January 28 announcement. This page tracks that conversion and the production timeline around it, with every claim dated and sourced, because this is the corner of the Optimus story where hype and fabricated numbers circulate most freely. What Tesla has actually said, in order Note what is missing from that record: a start date, a unit count, and a price. The call accompanying the deck was audio-only, so no robot was shown either. What the July 22 deck actually shows Three things in the shareholder update matter for this timeline. First, the words: "We have decommissioned the manufacturing lines for Models S & X at our Fremont Factory and are installing the first-generation lines for Optimus, where we expect to start production soon." A separate outlook line hedges that to "in anticipation of production in 2026." Second, the pictures: two slides captioned "Optimus – first generation production line in Fremont," the first official images of the line in an investor document. Third, the installed-capacity table now has a Robotics category listing Optimus at both the California and Texas factories, each with status "Construction" and no capacity figure. One reading note: "first-generation lines" refers to the first generation of production lines, which will build the third-generation robot, Optimus 3. And the deck is explicit about where the first robots go: "The initial Optimus builds will be used in our Optimus Academy for training data collection and further functionality development." Not to customers, but to an internal training program where fleets of robots practice tasks and feed the learning pipeline. Tesla has not disclosed where the Academy runs or whether training overlaps with its factory floors. Why the ramp will be slow, in Musk's own words Tesla's CEO spent much of the July 22 call lowering expectations, which is worth taking seriously precisely because it cuts against his usual direction. Optimus, he said, is "the hardest product to scale manufacturing that we've ever made at Tesla, because everything on the robot is new," and unlike electric cars, "there is no supply chain. We've had to build up a supply chain in its entirety, or in-house the production." The production ramp "will follow the normal S-curve of a manufacturing ramp, but the initial portion of the S-curve will be quite flat and long." He has previously put the robot's part count around 10,000 and said the line "will move as fast as the least lucky, slowest, dumbest part" of them all. The Fremont line's design capacity is 1 million robots per year, a figure Musk repeated plainly on the call. Design capacity is not output: a line built for a million units a year can and, by Tesla's own guidance, will produce a trickle for a long time first. Anyone quoting "1 million Optimus robots" as a production number rather than design capacity is misreading the same sentence everyone else read. The forecast record: why dates get discounted Sources for the misses: Electrek's April 2026 timeline review, AP's reporting of the Omdia shipment estimate, the DigiTimes-derived assembly estimate, and TechTimes on the undisclosed production record. Note the estimates measure different things over different windows, shipped in 2025 versus assembled by mid-2025, and Tesla has confirmed neither. This is not a gotcha exercise; it is the correct prior. When Tesla says production is anticipated "later this year," the record says to believe the direction and discount the date. Austin, Optimus 4, and the chip constraint The July 22 call also extended the roadmap past Fremont. Optimus 4, Musk said, "will be built in Austin" on a much more vertically integrated supply system, aiming for "an order of magnitude more production... sort of aspirationally 10 million units a year versus 1 million units a year of Optimus 3." He attached no date, and when an analyst asked directly when Optimus 4 production would start, none was given. Site development at Gigafactory Texas is underway per the deck, with building construction "in full swing." Two supporting details are worth tracking. Tesla's AI5 chip "will initially go into Optimus," ahead of the cars. And Musk said Tesla expects "to announce a location soon" for its Terafab chip plant, "without which we will be constrained in our ability to scale Optimus production, because we simply won't have enough AI chips." In other words, by Tesla's own account, the ten-million-a-year vision depends on a chip factory that does not yet have an announced site. The fake numbers, debunked This page's search results are polluted by fabricated production claims, so here is the register. The story that "Optimus mass production began January 21, 2026" traces to a single third-party press-wire piece with no Tesla source and no named reporter; Tesla's own record contradicts it, and Tesla's production reports have never carried an Optimus line item. The syndicated item is still live; note the absence of any Tesla source or named reporter. Content-farm claims of "1,000+ Optimus already working in Tesla factories" collide with Musk's January 2026 statement that the robots were "not in usage in our factories in a material way." And the widely repeated report that Tesla placed a $685 million actuator order with China's Sanhua was denied by Sanhua itself, which called the rumor untrue; treat any story built on that figure accordingly. Tesla has never published an Optimus production count. Until a number appears in a Tesla document, there is no number. What to watch next Three things move this timeline now. A production-start announcement from Tesla, which the company has tied to "later this year" and which its own deck photos suggest is genuinely close. The public reveal of Optimus 3, which Tesla has said it is holding until near production start so competitors cannot copy the design. And the Q3 earnings call around late October, where "later this year" either becomes a date and a number or slips again. For what the robot itself is and how it compares to the field, see our Optimus review and the Optimus vs Atlas comparison. The bottom line The factory conversion is real, official, and photographed: Tesla traded the Model S and X for a robot line and now says so in its own investor materials. Formal production on the new line, as of July 23, 2026, has not started. The honest summary of the timeline is Tesla's own hedged language plus the program's forecast record: lines installing now, production "anticipated later this year," output that will be deliberately small at first even if the date holds, and a history that says dates usually don't. This page will update when either a start announcement or a unit count exists in a Tesla document. Compare on RoboZaps: Tesla Optimus and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # 1X NEO vs Unitree H2: Which $20-30k Home Humanoid Should You Pre-Order? [2026] URL: https://blog.robozaps.com/b/1x-neo-vs-unitree-h2 Author: Radica Maneva Published: 2026-02-20T00:00:00.000Z Category: news TL;DR: The 1X NEO ($20,000 or $499/month) is the safer, home-optimized pick with its soft-bodied design; the Unitree H2 ($29,900) is the larger, more versatile full-size humanoid. NEO wins on price, safety and home fit, the H2 on size and physical capability. Both ship in 2026 and take refundable deposits. Key takeaways: - 1X NEO is the safer, more affordable option at $20,000 (or $499/month)—built specifically for home environments with a revolutionary soft-bodied design. - Unitree H2 is the more versatile, full-size humanoid at $29,900—capable of both home tasks and light commercial applications. - NEO wins on price, safety, and home optimization. H2 wins on size, versatility, and physical capability. - Both robots ship in 2026, accept refundable deposits, and promise OTA software updates to expand capabilities over time. FAQ: Q: Is the 1X NEO better than Unitree H2? A: Neither is objectively "better"—they serve different needs. NEO is safer and more affordable for home-only use. H2 is more capable and versatile for mixed home/commercial applications. If you have children, NEO's soft-bodied design is the responsible choice. Q: How much does NEO cost vs H2? A: NEO costs $20,000 (or $499/month subscription) with a $200 deposit. H2 costs $29,900 with a $2,500 deposit. NEO is 33% cheaper with a 12x lower deposit requirement. Q: When will these robots ship? A: Both target 2026. H2 has a specific April 2026 target for North America. NEO says "2026 for US orders" without specifying a month. Q: Can I cancel my pre-order? A: Yes. NEO's $200 deposit is fully refundable. H2's $2,500 deposit terms vary—check with ToborLife for current cancellation policy. Q: Which robot is safer around children? A: NEO, by a significant margin. Its soft-bodied, tendon-driven design was specifically engineered for safe human coexistence. H2 uses traditional rigid construction that requires more caution around vulnerable family members. Q: Will these robots improve over time? A: Yes. Both companies commit to OTA (over-the-air) software updates that will expand capabilities post-purchase. This is the new standard for humanoid robots—you're buying a platform, not a fixed product. Two humanoid robots. Two very different philosophies. One shared goal: bringing automation into your home. The 1X NEO ($20,000) and Unitree H2 ($29,900) represent the first generation of humanoid robots actually priced for consumer pre-order—and both are shipping in 2026. If you've been waiting for the "buy now" button on a home humanoid, it's finally here. But which robot deserves your deposit? This comprehensive 1X NEO vs Unitree H2 comparison breaks down everything: price, size, capabilities, design philosophy, and which robot fits your specific needs. 1X NEO vs Unitree H2: Head-to-Head Comparison 1X NEO: The Safety-First Home Robot ($20,000) 1X Technologies (formerly Halodi Robotics) designed the NEO from the ground up for one thing: living safely with humans. Unlike every other humanoid robot on the market, NEO uses a soft-bodied, tendon-driven design inspired by human muscle and ligament structure. Design Philosophy Where most humanoids are rigid metal skeletons with exposed joints and hard surfaces, NEO is soft. Its "muscles" are artificial tendons that give it compliant, human-like movement. If you bump into NEO—or if it bumps into you, your furniture, or your kids—it yields like a person would, not like a metal cabinet falling over. This isn't just marketing. It's the fundamental engineering decision that makes NEO viable for home use. Traditional industrial robots require safety cages. NEO can hand you a cup of coffee without risk of cracking your ribs. Pricing Options 1X offers two ways to own NEO: Purchase: $20,000 one-time paymentSubscription: $499/month ($5,988/year) The subscription model is interesting. At $499/month, you'd pay off the $20,000 purchase price in 40 months (3.3 years). If you're uncertain about the technology or expect rapid hardware improvements, subscribing makes sense. If you're committed long-term, buying saves money. The deposit is only $200 and fully refundable—the lowest commitment of any humanoid robot on the market. What NEO Can Do (2026 Launch) NEO ships with "basic autonomy"—1X's term for foundational home navigation and simple task completion. The company has been transparent that capabilities will expand significantly through over-the-air (OTA) software updates. Expected launch capabilities include: Home navigation and mappingBasic object manipulation (picking up, carrying, placing)Following simple commandsSafe human interaction Unitree H2: The Full-Size Workhorse ($29,900) Unitree—the Chinese robotics company famous for the $13,500 G1 humanoid and ultra-affordable quadruped robots—launched the H2 at CES 2026 with a clear message: this is the cheapest full-size humanoid robot ever sold. Design Philosophy At 180cm (5'11") and 70kg (154 lbs), the H2 is built to human adult specifications. It can reach the same shelves you can, use the same tools you use, and operate in environments designed for people without modification. The trade-off: H2 uses traditional rigid robotics construction. It's stronger and more capable than NEO, but also harder and less forgiving in collisions. This isn't a robot you want falling on your toddler. Technical Specifications The H2 packs serious hardware into its frame: 31 degrees of freedom: Full-body articulation for complex movements180cm height: Full adult human scale70kg weight: Substantial but mobileIndustrial-grade actuators: Higher payload and strength than consumer robots Pricing and Availability The H2 is priced at $29,900—approximately $10,000 more than NEO, but dramatically cheaper than any comparable full-size humanoid. The Unitree G1 (smaller model) starts at $13,500, making the H2's per-capability value remarkable. Pre-orders require a $2,500 deposit through ToborLife, the exclusive North American distributor. Use code TOBORBOTINFO200 for $200 off. Delivery is targeted for April 2026 in North America. Category-by-Category Comparison 1. Price and Value Winner: 1X NEO At $20,000 vs $29,900, NEO costs roughly 33% less than H2. The subscription option ($499/month) drops the entry barrier even further. NEO's $200 refundable deposit vs H2's $2,500 makes testing the waters dramatically easier. However, value depends on your use case. If you need a full-size humanoid's capabilities, the H2's $29,900 is still historic—full-size humanoids from Figure, Boston Dynamics, and others cost $50,000-$250,000+. 2. Safety and Home Integration Winner: 1X NEO NEO's soft-bodied design is purpose-built for homes with children, pets, and fragile furniture. The tendon-driven actuators provide inherent compliance—if something goes wrong, the robot yields rather than pushes through. H2 is a traditional industrial robot. It's safe by industrial standards (which require extensive risk assessment), but it's not designed with the assumption that a 3-year-old might grab its leg mid-stride. 3. Physical Capabilities Winner: Unitree H2 The H2's larger frame, 31 DOF, and industrial actuators give it strictly superior physical capabilities. It can reach higher, carry more, and perform more complex movements than NEO's compact form allows. If you need a robot that can help with garage work, reach top shelves, or handle heavier loads, H2 is the clear choice. 4. Versatility Winner: Unitree H2 NEO is designed exclusively for home consumers. H2 targets both home and light commercial applications. If you're a small business owner considering a robot for both home and shop, H2 can do double duty. Unitree also offers an "H2 EDU" variant for educational institutions, signaling broader SDK and developer support. 5. Software Ecosystem Winner: Tie Both robots promise OTA updates to expand capabilities over time. Unitree has a longer track record shipping consumer robotics (their Go1/Go2 quadrupeds have active developer communities), while 1X has focused more on commercial deployments with their EVE robot. Neither company has published detailed SDK documentation for their consumer humanoids yet. This is a "wait and see" category. 6. Delivery Timeline Winner: Tie (slight edge to H2) H2 has a specific target: April 2026 for North America. NEO's timeline is "2026 for US orders" without a specific month. Unitree's track record with the G1 launch suggests they can hit aggressive timelines; 1X is newer to consumer delivery. Which Should You Choose? Buy the 1X NEO if you: Have children or pets: NEO's soft-bodied design makes it the only safe choice for homes with small kids or animals that might collide with or grab the robot.Want the lowest financial risk: $200 refundable deposit and $499/month subscription option let you test the technology without major commitment.Prioritize aesthetics: NEO's soft, rounded design and color options (Tan, Gray, Dark Brown) look more like furniture than factory equipment.Need a dedicated home robot: If your use case is purely domestic—laundry, dishes, tidying—NEO is optimized for exactly this. Buy the Unitree H2 if you: Need full-size reach and strength: At 180cm, H2 can access everything in your home designed for adult humans—no compromises.Want commercial versatility: Small business, workshop, or retail applications alongside home use.Plan to develop custom applications: Unitree's developer ecosystem and SDK support are more mature.Value proven hardware execution: Unitree has shipped tens of thousands of consumer robots; their manufacturing is battle-tested. Final Verdict: 1X NEO vs Unitree H2 For most home buyers, the 1X NEO is the better choice. Its safety-first design, lower price, minimal deposit, and subscription option make it the lower-risk entry point into home humanoid robotics. If you're buying a robot to help around the house and you have family members who might interact with it unpredictably, NEO is the responsible pick. However, the Unitree H2 wins if you need serious capability. Its full adult-size frame, 31 DOF, and industrial-grade construction mean it can tackle tasks NEO simply cannot. If you're a power user, developer, or small business owner who needs maximum versatility, H2 delivers more robot for the premium. Both robots represent a historic moment: the first humanoids priced for real consumers, shipping in 2026. The wait is almost over. Ready to pre-order? See 1X NEO on Robozaps | See Unitree H2 on Robozaps | Compare all humanoid robots Last updated: February 20, 2026. Specifications sourced from 1X Technologies and Unitree official announcements. Robozaps is a humanoid robot marketplace — we maintain comprehensive product databases and may earn referral fees from qualifying purchases. --- # Unitree Spring Festival Robot Performance: What It Proves and What It Does Not URL: https://blog.robozaps.com/b/unitree-spring-festival-2026 Author: Radica Maneva Published: 2026-02-19T00:00:00.000Z Updated: 2026-08-01T23:07:37.349Z Last fact-checked: 2026-07-06T00:00:00.000Z Category: news TL;DR: The Unitree Spring Festival performance matters because it put dynamic humanoid movement in front of a mass audience. It should not be treated as proof of full autonomy, household readiness, or factory ROI. The right read is progress in motion control and public confidence, with deployment questions still open. Key takeaways: - The performance was a powerful robotics showcase, not a buyer-ready deployment case study. - Martial-arts choreography proves balance and control progress, but not general-purpose autonomy. - Unitree's G1 and R1 pages show lower-cost humanoid platforms, but specs do not equal useful work. - Buyers should ask about autonomy, teleoperation, safety, support, and task reliability before treating stage demos as proof. FAQ: Q: Were Unitree robots shown at China's 2026 Spring Festival Gala? A: Yes. Multiple news reports described Unitree humanoids performing choreographed martial-arts and dance routines during the 2026 Spring Festival Gala. Q: Does the performance prove Unitree robots are autonomous workers? A: No. It is evidence of impressive motion control and choreography. It does not prove unsupervised commercial work, household readiness, or factory economics. Q: Which Unitree robots should buyers compare? A: Start with Unitree's official G1 and R1 pages, then verify the exact configuration, controller, autonomy level, warranty, support, and allowed use case. A national-TV demo, not a deployment proof Unitree's Spring Festival appearance was a strong signal that humanoid robotics has moved into mainstream public view. Robots performing martial-arts choreography on national television is memorable, and it shows progress in balance, coordination, and hardware confidence. The mistake is treating that performance as proof that humanoid robots are ready for unsupervised commercial or household work. A choreographed stage routine can be technically impressive without proving task autonomy, uptime, safety, or ROI. What the performance does prove Humanoid movement is becoming more dynamic and more publicly legible.Chinese robotics companies are confident enough to put robots in high-visibility performances.Control research around dynamic motion, fall recovery, and whole-body tracking is moving quickly.Unitree has become one of the most visible brands in lower-cost humanoid platforms. What it does not prove That the robots can perform arbitrary tasks in a normal factory or home.That the routine was fully autonomous end to end.That a public demo translates into reliable customer support.That buyers can deploy the same capability without custom choreography, engineering, or supervision. The useful technical signal The right lesson is about movement control. Unitree's G1 page lists a humanoid platform priced from $13.5K with 23 to 43 joint motors depending on configuration. Unitree R1 pushes pricing even lower for a smaller platform. That matters because cheaper hardware gives researchers and developers more room to test real applications. Research such as KungfuBot and A Kung Fu Athlete Bot That Can Do It All Day also shows why martial arts are a useful stress test: fast changes in balance, contact, and recovery expose limits that gentle walking demos can hide. How buyers should read it A stage performance should move Unitree onto a buyer's research list, not into a purchase order by itself. The next questions are ordinary but decisive: what task, what environment, what safety case, what autonomy level, what service plan, and what happens when the robot fails? Ask whether the target task has been demonstrated in an environment like yours.Ask how much teleoperation, pre-programming, or human reset is required.Ask what support exists outside China and what warranty applies to your configuration.Ask for real runtime, payload, battery-swap, and repair data for the exact model. Bottom line The Unitree Spring Festival performance was a compelling public milestone. It proved that humanoid demos are getting more athletic and culturally visible. It did not prove that general-purpose humanoid workers are ready. The gap between a stage routine and useful labor is still the story to watch. Compare on RoboZaps: Unitree R1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # AgiBot X1 2026: Free Open-Source Humanoid, Specs & Status URL: https://blog.robozaps.com/b/agibot-x1-review Author: Radica Maneva Published: 2026-02-11T00:15:41.000Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The AgiBot X1 has no price because AgiBot doesn't sell it: it is a free open-source humanoid design, published on GitHub in October 2024 (BOM, CAD files, assembly guide and software). Build it yourself or get a quote from third-party integrators. The repos have been dormant since around April 2025, so treat it as a stable snapshot for research and education. Key takeaways: - There is no price: the AgiBot X1 is a free open-source design (BOM, STEP files, assembly SOPs and software on GitHub since October 2024). You build it, or get an assembled unit from third-party integrators by quote. - "Open source" with caveats: AgiBot released most design materials but excluded some critical components, and the repos carry custom or missing licenses. - Specs: 130 cm, ~33 kg, 34 degrees of freedom, 1 m/s walking, 0.5 kg per-arm payload, PowerFlow joint actuators, AimRT middleware. - Status 2026: the X1 repos have been dormant since about April 2025; AgiBot's commercial line moved on to the Lingxi X2 (from ¥98,000). The X1 is a stable snapshot, not an active platform. - Context: AgiBot was ranked the world's #1 humanoid maker by 2025 shipments (5,100+ units per Omdia) and passed its 15,000th robot in mid-2026. FAQ: Q: How much does the AgiBot X1 cost? A: Nothing, from AgiBot: the X1 is a free open-source design, not a product. AgiBot published the full build package on GitHub in October 2024, so your cost is the components you source. Third-party integrators quote for assembled units; no official price exists. Q: Is the AgiBot X1 truly open source? A: Mostly. AgiBot released the hardware documentation, training and inference code, and the AimRT framework, but launch coverage notes some critical components were excluded, and the repositories carry custom or missing licenses rather than a standard OSI license. Q: Can the X1 run autonomously? A: Yes. The X1 includes onboard computing, LiDAR, and cameras for autonomous operation. Battery life is approximately 2 hours for untethered research. Q: How does X1 compare to Unitree G1? A: X1 prioritizes openness over raw capability. G1 has higher payload (2kg vs 0.5kg) and more locomotion demos, but X1 provides full source access that G1 doesn't match outside the EDU tier. Q: What software does the X1 run? A: The X1 uses AgiBot's open-source AimRT framework. It's compatible with ROS and supports Python and C++ development. Q: Where can I buy an AgiBot X1? A: You don't buy it from AgiBot. Download the design and software from AgiBot's GitHub (published October 2024) and build it, or request an assembled unit from a third-party integrator, which is quote-based and often waitlisted. The AgiBot X1 fills a critical gap in the humanoid robot market: an open-source platform that is free to build yourself. While competitors like Unitree G1 offer more raw capability, the X1's open architecture makes it the preferred choice for researchers who need to modify everything from motor control to AI frameworks. It's not trying to compete with industrial humanoids — it's purpose-built for the people advancing the field. AgiBot X1 Overview AgiBot (), the Shanghai-based robotics company behind the popular A2 series, launched the X1 as their dedicated research platform. Unlike their commercial-focused robots, the X1 prioritizes hackability over polish — open-source software, accessible hardware documentation, and a modular design that invites modification. The company's philosophy with X1 is clear: give researchers a capable bipedal platform they can actually afford and fully customize, rather than a locked-down black box. One honest caveat for 2026 readers: the X1 repositories have been dormant since around April 2025, and AgiBot's commercial energy has moved to paid products like the Lingxi X2 (from ¥98,000). The X1 remains downloadable and buildable, but treat it as a stable snapshot rather than an actively developed platform. AgiBot X1 Price Analysis There is no X1 price, because AgiBot does not sell it: the X1 is a free open-source design. AgiBot published the bill of materials, STEP files, assembly instructions and software on GitHub in October 2024, so the real cost is whatever your sourced components come to. A few third-party integrators offer assembled units on a quote/waitlist basis, but there is no official retail channel. *Reported pricing; not officially published by manufacturer Value Assessment: For academic budgets, the X1 hits a sweet spot — capable enough for serious bipedal research, affordable enough for university labs to justify, and open enough that students can learn from every layer of the system. Full Specifications Open-Source Architecture The X1's defining feature is its fully open-source design. AgiBot provides: Hardware Documentation: CAD files, assembly guides, and component specificationsSoftware Stack: Full source code for locomotion, perception, and controlAimRT Framework: AgiBot's open-source robotics middlewareAGIBOT World Dataset: Training data for embodied AI research This openness matters for researchers who need to: Modify motor control algorithms without reverse-engineeringIntegrate custom sensors and perception systemsImplement novel AI architectures directly on hardwareReproduce and extend published research PowerFlow Servo Technology The X1 uses AgiBot's proprietary PowerFlow servo actuators — the same technology deployed in their commercial robots. Key characteristics: High torque density for bipedal locomotionIntegrated motor driversForce feedback capabilityModular replacement design While the servos are proprietary hardware, the control interfaces are fully documented for custom development. 34 Degrees of Freedom The X1's 34 DOF configuration provides: Legs: Full bipedal locomotion capabilityArms: Dual manipulators with 0.5kg payload eachHands: Dexterous end effectors for research manipulation tasksTorso: Core articulation for balance and reaching This sits between the Unitree G1's 23-43 DOF range and research platforms that sacrifice DOF for simplicity. Sensor Suite The X1 includes research-appropriate sensors: Cameras: RGB vision for perception and navigationLiDAR: 3D spatial mapping and obstacle detectionForce Sensors: Contact detection for manipulation researchIMU: Orientation and balance sensing Researchers can add additional sensors through documented expansion interfaces. Use Cases Academic Research The primary use case. Universities can use X1 for: Bipedal locomotion algorithm developmentReinforcement learning for embodied AIHuman-robot interaction studiesComputer vision and perception research Robotics Education Engineering programs benefit from: Full-stack robotics curriculum supportHands-on bipedal control experienceOpen codebase for learningAffordable lab deployment AI Development For embodied AI researchers: Physical testbed for simulation-to-real transferIntegration with AGIBOT World DatasetCustom model deploymentMulti-modal learning experiments Prototyping Startups and R&D teams can: Validate concepts before custom hardware investmentTest manipulation strategiesDevelop applications on proven platform Pros and Cons Pros Fully open-source — hardware docs, software, and training dataAccessible pricing — designed for academic budgets34 DOF — research-grade articulationAimRT framework — modern robotics middlewareCommercially available — ships now, not pre-orderAgiBot ecosystem — shared components with A2 series Cons Limited payload (0.5kg) — not for heavy manipulationShorter stature (130cm) — not human-scaleResearch-focused — not production-ready for deployment2-hour battery — limited untethered operationChina-based support — timezone and language considerations X1 vs Competitors Bottom line: If you need maximum openness for research, X1 wins. If you need more payload and don't care about source access, G1 offers better specs per dollar. Final Verdict The AgiBot X1 is the humanoid robot that researchers actually asked for: affordable, open, and capable enough to do real bipedal research. It won't win demo reels against flashier competitors, but it will let you modify every line of code and every control loop. The X1 is right for you if: You're a university or research lab needing bipedal hardwareOpen-source access matters more than peak performanceYou want to contribute to or build on shared research infrastructure Look elsewhere if: You need payload capacity over 1kg — consider Unitree G1You want human-scale height — consider Unitree H1You need production-ready deployment — consider commercial optionsOpen source doesn't matter to your use case AgiBot made a strategic choice with X1: build the platform the research community needs, not the one that gets the most YouTube views. For academics and AI researchers, that's exactly right. Where to buy: AgiBot Official Website Specifications verified against official manufacturer sources, February 2026. Prices and availability subject to change. Compare it with the best humanoid robots of 2026, or start with what a humanoid robot is. --- # Tesla Optimus vs Apptronik Apollo: Full 2026 Comparison (Price, Specs, Deployment) URL: https://blog.robozaps.com/b/tesla-optimus-vs-apptronik-apollo Author: Radica Maneva Published: 2026-02-11T00:00:00.000Z Updated: 2026-08-01T11:20:42.268Z Category: comparisons TL;DR: Both are 173 cm American industrial humanoids. The Apollo (73 kg) beats Optimus on payload (25 kg vs 20 kg) and is already piloting at Mercedes-Benz; Optimus targets the more aggressive $20,000-$30,000 price and leans on Tesla's FSD-derived AI and manufacturing scale. Apollo delivers sooner; Optimus bets on volume. Key takeaways: - Tesla Optimus stands 173 cm (5'8") tall, weighs 57 kg (126 lbs), and targets an unprecedented $20,000–$30,000 price point — the most aggressive pricing in humanoid robotics. - Apptronik Apollo matches Optimus in height at 173 cm (5'8") but weighs 73 kg (161 lbs), with a 25 kg (55 lbs) payload capacity that beats Optimus's 20 kg (44 lbs). - Apollo wins on: Payload capacity (25 kg vs 20 kg), battery life (4 hours vs unconfirmed), current deployment status (Mercedes-Benz factory). - Optimus wins on: Target price ($20K–$30K vs sub-$50K), AI capabilities (FSD-derived vision), manufacturing scale potential. - Both robots are American-made, targeting the same industrial automation market, but with different go-to-market strategies — Tesla betting on volume, Apptronik on capability. FAQ: Q: Is Tesla Optimus better than Apptronik Apollo? A: It depends on your priorities. Tesla Optimus offers potentially lower cost ($20K–$30K vs sub-$50K) and superior AI capabilities, but isn't available for purchase. Apptronik Apollo offers higher payload (25 kg vs 20 kg), proven battery life (4 hours), and is deployed today at Mercedes-Benz. For immediate industrial deployment, Apollo wins. For future mass-market potential, Optimus is the bet. Q: How much does Tesla Optimus cost vs Apptronik Apollo? A: Tesla targets $20,000–$30,000 for Optimus at mass production scale (late 2026+). Apptronik targets sub-$50,000 for Apollo. Neither has confirmed final pricing. Apollo is available for enterprise purchase today via contact sales; Optimus isn't commercially available yet. Q: Which robot has better payload capacity? A: Apptronik Apollo wins with 25 kg (55 lbs) payload capacity versus Tesla Optimus's 20 kg (44 lbs). Apollo's higher weight (73 kg vs 57 kg) provides a more stable base for heavy lifting tasks common in automotive manufacturing and logistics. Q: Can I buy Tesla Optimus or Apptronik Apollo today? A: Apptronik Apollo is available for enterprise customers via contact sales, with pilot deployments already running at Mercedes-Benz and Google. Tesla Optimus is not commercially available — it's deployed only within Tesla's own factories. Consumer availability for Optimus is targeted for late 2027. Q: Which humanoid robot is more advanced? A: Tesla Optimus has more advanced AI (FSD-derived neural networks, end-to-end learning, fleet updates). Apptronik Apollo has more advanced safety architecture (force-controlled actuators) and superior operational endurance (4-hour battery with hot-swap). "Advanced" depends on what capability matters for your use case. Q: Are Tesla Optimus and Apptronik Apollo safe to work around? A: Both are designed for human-robot collaboration. Apollo's force-controlled, series-elastic actuators provide inherent compliance — joints yield on contact, reducing injury risk. Tesla emphasizes fall damage minimization and operational safety but hasn't disclosed equivalent compliance mechanisms. For proven cobot safety, Apollo has the edge. Q: Which company has more funding and resources? A: Tesla dwarfs Apptronik in resources. Tesla is a $500B+ company spending billions annually on AI. Apptronik is a venture-backed startup ($1.5B+ total raised, including $520M in Mar 2026 led by Google and Mercedes-Benz). However, Apptronik's focus is exclusively robotics, while Tesla splits attention across vehicles, energy, and AI. Apollo benefits from Apptronik's singular focus. Q: What industries use Tesla Optimus vs Apptronik Apollo? A: Both target manufacturing, logistics, and warehouse automation. Apollo is deployed at Mercedes-Benz (automotive assembly) and testing with Google (logistics). Optimus operates in Tesla factories (battery cell sorting, parts handling). Apollo has third-party enterprise deployments; Optimus is internal only as of early 2026. Breaking: Feb 11, 2026 Funding Update Apptronik just raised $520 million at a $5.5 billion valuation—a 3x increase from their previous round. Google and Mercedes-Benz are lead backers, with Apollo actively testing in Mercedes-Benz factories and GXO Logistics warehouses. Meanwhile, Tesla is committing $20B in 2026 capex for Optimus—but Apollo's commercial deployments may beat them to enterprise scale. Tesla Optimus vs Apptronik Apollo — two of the most anticipated humanoid robots in 2026, both targeting industrial automation at radically different price points. Tesla's manufacturing juggernaut promises mass-market pricing under $30,000, while Apptronik's NASA-rooted Apollo boasts the highest payload capacity in its class at 25 kg. This comprehensive comparison breaks down every spec, real-world deployment, and technical advantage so you can see exactly which robot wins in each category. Tesla Optimus vs Apptronik Apollo: Full Specifications Comparison Tesla Optimus: Everything You Need to Know (2026 Update) Tesla Optimus represents Elon Musk's vision of a general-purpose humanoid robot that could eventually become more valuable than Tesla's entire automotive business. First unveiled at Tesla AI Day 2022 as a person in a robot suit, Optimus has evolved rapidly through Gen 1, Gen 2, and now Gen 3 iterations — each demonstrating meaningful improvements in dexterity, balance, and autonomous task completion. Design and Build At 173 cm (5'8") and 57 kg (126 lbs), Tesla Optimus is designed to operate in human-scale environments without modification. The Gen 2 model reduced weight by 10 kg from Gen 1, improving balance and agility. The most significant advancement in Gen 3 is the hands — featuring 22 degrees of freedom per hand (up from 11 in Gen 2), enabling manipulation tasks like threading needles, folding clothes, and handling delicate objects. Tesla's approach emphasizes manufacturability above all else. Every component is designed for high-volume production using Tesla's existing supply chain infrastructure. The actuators are proprietary, leveraging lessons from electric vehicle motors. The 2.3 kWh battery pack shares chemistry with Tesla's vehicle batteries, enabling cost reduction through scale. AI and Software Optimus runs on Tesla's Full Self-Driving (FSD) neural network stack, adapted for bipedal robotics. This gives it several advantages: Vision-based navigation: Unlike robots relying on LiDAR, Optimus uses cameras exclusively — the same approach Tesla uses for autonomous vehicles.End-to-end learning: The robot can learn tasks from video demonstrations, reducing the need for explicit programming.Fleet learning: Skills learned by one Optimus can be distributed across all units via over-the-air updates. Tesla demonstrated autonomous operation in its Fremont and Austin factories, where Optimus units perform battery cell sorting and parts handling. However, some demos have faced scrutiny for relying on teleoperation — a transparency gap that competitors have been quick to highlight. Price and Availability Tesla's target price of $20,000–$30,000 would make Optimus the most affordable full-size humanoid robot by a significant margin. This pricing depends on achieving automotive-scale production — potentially 10,000+ units per year by 2027. Current status (March 2026): Gen 3 debut imminentFremont factory repurposed from Model S/X production for Optimus manufacturingMass production target: before end of 2026Consumer availability: late 2027No pre-orders currently open Apptronik Apollo: Everything You Need to Know (2026 Update) Apptronik Apollo emerges from the Human Centered Robotics Lab at the University of Texas at Austin, with deep roots in NASA's humanoid robotics program. The company's founders worked on NASA's Valkyrie robot, bringing aerospace-grade engineering principles to commercial humanoid development. Apollo represents a different philosophy than Optimus — prioritizing capability and reliability over aggressive cost reduction. Design and Build Apollo matches Optimus in height at 173 cm (5'8") but is significantly heavier at 73 kg (161 lbs). This additional mass isn't wasted — Apollo's 25 kg (55 lbs) payload capacity exceeds Optimus by 25%, making it better suited for heavy-duty industrial tasks like moving automotive components or handling logistics totes. The standout feature is Apollo's force-controlled, series-elastic actuators. Unlike rigid actuators that can cause injury on impact, Apollo's joints yield under unexpected force — a critical safety feature for human-robot collaboration. This compliant actuation also enables more natural, adaptive movement when manipulating objects of varying weights and textures. The 4-hour battery life with hot-swappable packs addresses a key industrial requirement: continuous operation across shifts. Apollo can return to a docking station, swap batteries autonomously, and resume work — no human intervention required. Development and Partnerships Apollo's pedigree includes: NASA heritage: Apptronik's team built components for NASA's Valkyrie humanoid, bringing space-rated reliability engineering to commercial robotics.Mercedes-Benz partnership: Apollo is actively deployed at Mercedes-Benz manufacturing facilities for automotive assembly tasks — real-world validation that Optimus hasn't yet achieved publicly.GXO Logistics operations testing: Apptronik has confirmed testing with GXO Logistics for logistics and warehouse operations. This enterprise-first approach means Apollo is generating revenue and operational data while competitors are still in development. Price and Availability Apptronik targets a sub-$50,000 price point for Apollo — higher than Tesla's aggressive target but still competitive for industrial applications where a robot can replace a $60,000+ annual labor cost. Current status (March 2026): NEW: $520M funding at $5.5B valuation (Feb 11, 2026)Pilot deployments active at Mercedes-Benz and GXO Logistics warehousesAvailable for enterprise purchase via contact salesNo consumer version plannedPre-order available for qualified industrial customers Head-to-Head: Tesla Optimus vs Apptronik Apollo Performance Comparison 1. Payload Capacity and Strength Winner: Apptronik Apollo Apollo's 25 kg (55 lbs) payload capacity beats Optimus's 20 kg (44 lbs) by 25%. In manufacturing environments, this difference matters — Apollo can handle heavier automotive components, larger logistics totes, and more demanding material handling tasks without risk of strain or failure. The weight difference between the robots (73 kg vs 57 kg) partially explains Apollo's advantage — a heavier frame provides a more stable base for lifting. Tesla's lighter design prioritizes energy efficiency and agility over raw strength. 2. AI and Autonomy Winner: Tesla Optimus Tesla's FSD-derived AI stack gives Optimus a significant edge in autonomous learning and adaptation. The end-to-end neural network approach means Optimus can potentially learn new tasks from video demonstrations, while Apollo relies on more traditional programming approaches. Tesla's fleet learning capability — where skills transfer automatically across all units — could prove decisive as deployments scale. Apollo's SDK is capable but lacks the depth of Tesla's AI investment (Tesla spends billions annually on AI research). However, Tesla's autonomy claims have faced scrutiny. The "We, Robot" event in October 2024 drew criticism for not disclosing teleoperation. Apollo's demonstrations, while less flashy, have been more transparent about capabilities. 3. Battery Life and Endurance Winner: Apptronik Apollo Apollo's confirmed 4-hour battery life with hot-swappable packs beats Optimus on operational endurance. Tesla hasn't disclosed Optimus's runtime, though the 2.3 kWh battery pack suggests comparable or shorter operation times. The hot-swap capability is critical for industrial deployment. Apollo can autonomously dock, swap batteries, and resume operation — enabling 24/7 operations across shifts. Optimus's battery appears integrated, requiring the robot to charge in place. 4. Real-World Deployment Winner: Apptronik Apollo Apollo is currently deployed at Mercedes-Benz manufacturing facilities and testing with Google — generating real-world operational data and revenue. This is the clearest indicator of commercial readiness. Optimus operates within Tesla's own factories, which is valuable validation but not equivalent to third-party enterprise deployment. Tesla has customers for its vehicles, not its robots — Apollo has paying customers for its robots today. 5. Price and Value Winner: Tesla Optimus (if achieved) Tesla's $20,000–$30,000 target price would be transformative — potentially 40–50% cheaper than Apollo's sub-$50,000 target. At Tesla's price point, humanoid robots become economically viable for a far broader range of applications, including small businesses and eventually consumers. However, this pricing depends on achieving automotive-scale production, which remains unproven. Apollo's higher price reflects current manufacturing realities. For enterprises buying today, Apollo's pricing is competitive against labor costs. 6. Safety and Human Collaboration Winner: Apptronik Apollo Apollo's force-controlled, series-elastic actuators provide inherent compliance — the joints yield under unexpected contact, reducing injury risk. This architecture, derived from NASA's Valkyrie program, prioritizes safe human-robot collaboration. Tesla emphasizes minimizing damage from falls and operational safety, but hasn't disclosed equivalent compliance mechanisms. In environments where robots work directly alongside humans (the definition of a cobot), Apollo's safety architecture is more proven. 7. Manufacturing Scale Potential Winner: Tesla Optimus Tesla's manufacturing infrastructure — Gigafactories, supply chain, production engineering — gives Optimus unmatched scaling potential. Tesla produces millions of vehicles annually; redirecting even a fraction of that capacity to humanoid robots could dwarf all competitors combined. The Fremont factory conversion from Model S/X production to Optimus manufacturing signals Tesla's commitment to scale. Apptronik, while capable, operates at startup scale — their production capacity measured in hundreds, not thousands. Which Should You Choose? Choose Tesla Optimus if you: Can wait until late 2027 for consumer/enterprise availability — Optimus isn't purchasable today, but if you're planning long-term automation strategy, Tesla's pricing could be decisive.Prioritize cost above all else — At $20,000–$30,000, Optimus would have 40–50% lower acquisition cost than Apollo.Value AI capabilities and autonomous learning — Tesla's FSD-derived neural networks offer the most advanced vision and task-learning capabilities.Want fleet scalability — Tesla's over-the-air updates and fleet learning mean skills transfer automatically across all units.Operate a Tesla-integrated facility — Optimus will likely integrate seamlessly with Tesla's manufacturing ecosystem. Choose Apptronik Apollo if you: Need a humanoid robot today — Apollo is available for enterprise deployment now, with proven operation at Mercedes-Benz.Require heavy payload capacity — Apollo's 25 kg beats Optimus's 20 kg — critical for automotive components and heavy logistics.Prioritize safety for human-robot collaboration — Force-controlled actuators provide inherent compliance that reduces injury risk.Need proven 4-hour+ battery life — Hot-swappable packs enable continuous 24/7 operation across shifts.Want NASA-heritage reliability — Apptronik's team built components for space robots; that engineering rigor translates to commercial reliability. Final Verdict: Tesla Optimus vs Apptronik Apollo For enterprises that need a humanoid robot in 2026: Apptronik Apollo wins. It's available today, deployed at Mercedes-Benz, and offers higher payload capacity with proven battery endurance. The NASA-heritage engineering provides confidence in reliability. For organizations planning 2027+ automation: Tesla Optimus is the bet worth watching. If Tesla achieves its $20,000–$30,000 price target at scale, it could fundamentally change the economics of humanoid robotics. The FSD-derived AI gives Optimus the highest ceiling for autonomous learning. The humanoid robot market in 2026 isn't winner-take-all. Apollo and Optimus serve different segments of the same market — Apollo for enterprises that need proven capability now, Optimus for those betting on Tesla's manufacturing scale to deliver unprecedented affordability. Both are American-made, both target industrial automation, and both will shape the future of work. Compare both robots side by side: Tesla Optimus on Robozaps | Apptronik Apollo on Robozaps | Browse all humanoid robots Last updated: February 11, 2026. Specifications sourced from official manufacturer documentation and Robozaps robot database. Robozaps is the world's largest humanoid robot marketplace — we maintain comprehensive product databases and may earn referral fees from qualifying purchases. Related: Tesla Optimus Gen 2 Review · Apptronik Apollo Review · Tesla Optimus Alternatives and Competitors Ready to buy? Browse humanoid robots for sale on Robozaps. --- # Rainbow RB-Y1 Price 2026: Cost ($80K+), Specs & What It Is URL: https://blog.robozaps.com/b/rainbow-robotics-rb-y1-review Author: Radica Maneva Published: 2026-02-11T00:00:00.000Z Updated: 2026-08-01T23:07:14.698Z Last fact-checked: 2026-07-28T00:00:00.000Z Category: reviews TL;DR: The Rainbow Robotics RB-Y1 listed at $80,000 (research) or $120,000 (commercial) at 2024 pre-order, before tax, with current pricing quote-based. It is a wheeled dual-arm research platform, not a bipedal humanoid, and it genuinely sells to labs internationally. Samsung owns about 35 percent of the KAIST spin-off that built the DARPA-winning HUBO. Key takeaways: - Price: $80,000 research or $120,000 commercial, before tax. These are the 2024 pre-order prices; current pricing is quote-based, and figures at or above $150,000 do not come from Rainbow. - It is a wheeled dual-arm manipulator, not a bipedal humanoid: two 7-joint arms, a torso that raises and lowers on a 6-joint waist, on a wheeled base, 24 joints in total. - You can actually buy it. It is export-ready and sold direct by Rainbow to research labs internationally, by quote, with an open, English-documented software kit. - Samsung backs it. Samsung Electronics became Rainbow's largest shareholder at about 35 percent, a deal completed in March 2025, and consolidates it as a subsidiary. - Documented use is research: a Georgia Tech group and a Berkeley-Amazon team have published work on it. A June 2026 Coupang warehouse pilot, unconfirmed by the companies, was its first commercial-floor test. FAQ: Q: How much does the Rainbow RB-Y1 cost? A: Rainbow Robotics published two pre-order tiers in 2024: $80,000 for the research platform and $120,000 for the commercial platform, both excluding VAT. Those are genuine company figures, not marketplace estimates, but they were tied to the launch window. The product page now shows no list price and routes buyers to a sales inquiry, so current pricing is quote-based. Figures at or above $150,000 do not trace back to Rainbow. Q: Can you actually buy a Rainbow RB-Y1? A: Yes. Unlike many robots labelled humanoids, the RB-Y1 was built for export: Rainbow sells it directly, has pursued North American distribution, designs to North American safety standards, and publishes an English software kit. You order it by requesting a quote and a build slot from Rainbow Robotics rather than adding it to a cart, but it is genuinely sold internationally to research labs. Q: Is the RB-Y1 a humanoid robot? A: Not in the walking sense. The RB-Y1 is a bimanual mobile manipulator: two arms and a torso mounted on a wheeled base, with a waist that lifts the upper body. It has no legs and does not walk. Rainbow describes it as a humanoid-type research platform, but functionally it is a wheeled dual-arm robot, not a bipedal android. Q: Is Rainbow Robotics owned by Samsung? A: Samsung Electronics is Rainbow Robotics' largest shareholder. After an initial 14.7 percent stake in 2023, Samsung raised its holding to roughly 35 percent in a deal completed in March 2025 and now consolidates Rainbow as a subsidiary. Samsung also set up a Future Robotics Office, led by Rainbow co-founder Jun-Ho Oh, to combine its AI and software with Rainbow's hardware. Q: What are the RB-Y1's specifications? A: The RB-Y1 stands about 1.4 m tall, weighs 131 kg, and has 24 controllable joints: two 7-joint arms with a 3 kg payload each, arms reaching 640 mm to the wrist, a 6-joint waist that raises and lowers the torso, and a wheeled base that moves at up to 1.5 m/s. It runs on a roughly 50 V, 25 Ah battery and carries LiDAR for navigation. The exact onboard computer and battery runtime are not officially published. Q: How fast does the RB-Y1 move? A: Up to 1.5 m/s on the standard wheeled base, per Rainbow's current spec sheet. Its 2024 launch materials originally quoted 2.5 m/s, since revised down. The original base used two driven wheels; a 2025 upgrade added omnidirectional Mecanum wheels for 360-degree movement, for which Rainbow has not published a separate top speed. Q: Who uses the Rainbow RB-Y1? A: Documented use is concentrated in AI research. A group at Georgia Tech has published several 2026 studies using the RB-Y1 to learn two-handed manipulation from human demonstrations, and a Berkeley and Amazon team used it for whole-body teleoperation research. Rainbow also lists MIT and the University of Washington as users. In June 2026 the robot entered a reported pilot at a Coupang warehouse in South Korea, not formally confirmed by the companies, its first commercial-floor test. Q: Is the RB-Y1 the same as the Unitree G1 or a consumer robot? A: No. The RB-Y1 is a $80,000-plus wheeled research manipulator for labs and companies, not a consumer product. It is not a walking humanoid and it is not sold to households. If you want an affordable robot to own personally, the RB-Y1 is the wrong machine; it is research equipment sold by quote. Q: Is the RB-Y1 worth it for a research lab? A: For a lab that wants to collect two-handed demonstration data and train manipulation models, it is well priced. It sits above cheaper self-built bimanual rigs but below older six-figure systems, is a supported, integrated platform, and Rainbow offers an optional master-arm (leader) rig, sold separately, for recording demonstrations. The Samsung backing and HUBO heritage also lower the risk of the platform being abandoned. Many of the robots people call "humanoids" are not actually orderable. The Rainbow Robotics RB-Y1 is one of the rare exceptions. It has a real published price, it is sold internationally to research labs, and a South Korean giant, Samsung, owns roughly a third of the company that builds it. The catch is that the RB-Y1 is not a walking android and it is not a consumer product. It is a wheeled dual-arm research platform, sold to labs and companies that want to teach robots to do useful work with two hands. This review covers what the RB-Y1 costs, whether you can really buy one, what is inside it, who makes it, and who is actually using it today. What the RB-Y1 is at a glance The RB-Y1 is a bimanual mobile manipulator: two arms on a torso, mounted on a wheeled base, with a waist that raises and lowers the whole upper body. Rainbow Robotics introduced it in April 2024 as Korea's first bimanual mobile manipulator, and sells it as a standardized platform for AI research rather than a finished autonomous worker. How much the RB-Y1 costs Rainbow Robotics published a real price, which is unusual in this field. When pre-orders opened in May 2024, the company listed two tiers: a research platform at $80,000 and a commercial platform at $120,000, both excluding VAT. Deliveries were set to begin that October. Two caveats matter here. First, they are genuine, company-stated figures, not marketplace guesses. Second, they were pre-order prices tied to a specific launch window. Today the RB-Y1 product page shows no list price and routes buyers to a sales inquiry, so current pricing is effectively quote-based. Treat $80,000 and $120,000 as the last officially published figures and expect a formal quote for a 2026 order. Rainbow has not publicly explained what separates the research and commercial tiers, so ask what each includes when you request a quote. How the price compares to other research platforms The RB-Y1 is not the cheapest way into two-armed robot research, and it does not try to be. Its pitch is a fully integrated, internationally sold mobile manipulator with an official software kit. That sits above the do-it-yourself bimanual rigs that academic labs assemble themselves, and below older six-figure dual-arm systems. So the RB-Y1 competes on being complete and mobile, not on being the lowest-cost option. If your lab wants a supported, integrated platform that can drive around a room, it is priced reasonably for that. If you only need two fixed arms on a bench, cheaper kits exist. Can you actually buy one? Yes, and this is the RB-Y1's real differentiator. Unlike several high-profile robots that never leave their home market, the RB-Y1 was built for export. Rainbow Robotics sells it directly through its own sales team, has pursued North American distribution, and sells internationally rather than in a single home market. The software side is open too: the official development kit is published on GitHub under a permissive license, with English documentation. The practical path to ownership is an inquiry to Rainbow Robotics for a quote and a build slot, not an add-to-cart button. Original pre-orders delivered sequentially from late 2024, and current lead times are quote-dependent. But the platform is orderable, which cannot be said for most robots filed under "humanoid." A note on inflated price figures If you search for the RB-Y1's price you will find higher numbers floating around, including tags at or above $150,000, presented as if they were current. Those figures do not trace back to Rainbow Robotics. The only officially published prices are the $80,000 research and $120,000 commercial tiers from the 2024 pre-order announcement. Anything materially higher is an unverified restatement, and undated copies of the pre-order prices can also mislead by implying they are still the standing 2026 quote. When in doubt, the company's own release is the source to trust. Key specifications The RB-Y1 packs a lot of motion into one platform. It has two seven-joint arms, a six-joint "leg" or waist that raises and lowers the torso, and a wheeled base, for 24 controllable joints in total. The key figures below come from Rainbow Robotics' own product page and spec sheet. One figure has shifted over time. Rainbow's 2024 launch materials quoted a top speed of 2.5 m/s, but its current spec sheet lists 1.5 m/s for the standard wheeled base. Treat 1.5 m/s as the up-to-date number, and 2.5 m/s as a superseded launch claim. What Rainbow does not disclose A couple of details are genuinely unpublished, and no honest spec sheet should invent them. Rainbow does not state the exact onboard computer or GPU; the platform carries onboard PCs, but the specific chips are left open, partly because researchers swap in their own. And there is no official battery runtime, only the pack capacity. If a listing quotes a specific runtime in hours or a specific GPU model as fact, treat it with caution, because Rainbow itself does not publish those. The arms, grippers and teleoperation rig The arms are built from Rainbow's own collaborative-robot components, the same actuator and controller technology behind its industrial cobot line, which gives them industrial-grade repeatability rather than hobby-grade precision. Each arm ends in a simple one-joint gripper that can be swapped for other tooling, and an optional force-torque sensor can sit at the wrist for contact-sensitive tasks. The most important option is the master arm, sold separately. This is a lightweight leader device an operator moves by hand; the robot follows in real time, with built-in self-collision avoidance. That setup is the whole point of the platform: it lets researchers record human demonstrations of two-handed tasks and use them to train the robot's AI. In other words, the RB-Y1 is designed first as a data-collection tool. Sensing and the 2025 upgrade The RB-Y1 ships with LiDAR for navigation and an optional 3D recognition sensor, and cameras are typically integrated to suit the research task rather than fixed at the factory. In May 2025, at the ICRA robotics conference, Rainbow announced an upgrade: omnidirectional Mecanum wheels for 360-degree movement, plus an integrated software kit spanning the IMU, gripper and LiDAR modules with open interfaces. That change matters if you plan to move the robot in tight spaces, since the original base used two driven wheels while the newer configuration can slide sideways. Who makes it: Rainbow Robotics and Samsung Rainbow Robotics has an unusually strong pedigree. It was founded in 2011 as a spin-off from KAIST, the Korea Advanced Institute of Science and Technology, by the team behind HUBO, the humanoid whose disaster-response variant won the 2015 DARPA Robotics Challenge. The company listed on the KOSDAQ exchange in 2021. Then Samsung arrived. After an initial 14.7 percent stake in 2023, Samsung Electronics agreed to raise its holding to roughly 35 percent, a deal that completed in March 2025 and made Rainbow a Samsung subsidiary. Samsung also set up a Future Robotics Office, led by Rainbow co-founder Jun-Ho Oh, to combine Samsung's AI and software with Rainbow's hardware. For a buyer, that backing signals staying power: the platform is not a startup side project but part of a major electronics group's robotics push. Who actually uses the RB-Y1 Real, documented use of the RB-Y1 is concentrated in AI research, and it is narrower than the marketing implies. The clearest independent evidence comes from published papers: a group at Georgia Tech has used the RB-Y1 across several 2026 studies on learning two-handed manipulation from human demonstrations, and a Berkeley and Amazon team used it for whole-body teleoperation research. Rainbow also lists MIT and the University of Washington as users, though those rest on the company's own statements rather than published work. On the commercial side, the story is still early. In June 2026 the RB-Y1 entered a pilot at a Coupang fulfilment centre in South Korea, its first reported step from the lab onto a live warehouse floor, though the companies involved did not formally confirm the trial. Treat that as a promising pilot, not a proven deployment. As of now, the honest summary is that the RB-Y1 is a serious research tool with a handful of documented academic users and one early commercial test. Who should buy it, and who should not The RB-Y1 makes sense for a specific buyer: a research lab or company that wants to collect two-handed demonstration data and train manipulation models on a supported, mobile, internationally shippable platform. For that use it is well priced against both cheaper self-built rigs and older six-figure systems, and the Samsung backing plus HUBO heritage lower the risk that the platform gets orphaned. It is the wrong buy for anyone expecting a walking humanoid, a home robot, or something that works autonomously out of the box. The RB-Y1 does not walk, it is teleoperated as much as autonomous today, and it is priced and supported as research equipment. The bottom line The RB-Y1 is one of the few robots in this category you can actually order, at a price the maker published, and have shipped to your lab. At $80,000 for the research tier and $120,000 for the commercial tier at pre-order, now quote-based, it is a credible mobile bimanual platform backed by Samsung and built by the team that won the DARPA Robotics Challenge. Just buy it for what it is: a data-collection and manipulation-research platform on wheels, not a humanoid that will fold your laundry. Compare on RoboZaps: Rb Y1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # EngineAI SE01 Price 2026: Cost, the Viral-Gait Robot & Specs URL: https://blog.robozaps.com/b/engineai-se01-review Author: Radica Maneva Published: 2026-02-11T00:00:00.000Z Updated: 2026-08-01T23:07:18.140Z Last fact-checked: 2026-07-28T00:00:00.000Z Category: reviews TL;DR: The EngineAI SE01 has no official price: it is quote-only, and the ~$20,500 to $27,350 figure is a co-founder's target for full-size humanoids, not an SE01 sticker. The famous ~$12,000 open-source robot is the smaller PM01, a different machine. Its fame is the smooth straight-knee human walk, verified real by NVIDIA's Jim Fan. EngineAI (valued ~$1.5B, HK IPO filed) now focuses on the newer T800. Key takeaways: - No official price: the SE-01 is quote-only. The ~$20,500 to $27,350 figure is a co-founder's target for EngineAI's full-size humanoids in general, not an SE-01 sticker price. - The cheap ~$12,000 open-source EngineAI robot is the PM-01, a smaller developer humanoid, not the SE-01. The web routinely conflates the two. - The viral human-like walk is real, not CGI: NVIDIA's Jim Fan verified it. But the demos were handler-assisted, and 'world-first gait' is EngineAI's own framing. - Official specs: 170 cm, ~55 kg, 32 degrees of freedom, up to 2 m/s, in-house harmonic joints, dual NVIDIA and Intel compute. Payload and runtime are undisclosed. - The SE-01 is now a stepping stone: EngineAI's flagship is the newer T800 (from ~$25,000, mass delivery from May 2026). There is no SE-02. FAQ: Q: How much does the EngineAI SE-01 cost? A: There is no official public price. EngineAI has never listed one, and the SE-01 is a quote-only, business-to-business product with a 'contact us' button rather than a checkout. The figure you will see, roughly $20,500 to $27,350 (about 150,000 to 200,000 yuan), is a co-founder's stated target for EngineAI's full-size humanoids in general, not a confirmed SE-01 price. The reporting that carried it noted EngineAI has not disclosed SE-01 pricing. Q: Is the $12,000 EngineAI robot the SE-01? A: No. The cheap, roughly $12,000 (88,000 yuan) open-source EngineAI humanoid is the PM-01, a smaller 1.38 m developer and education robot launched in December 2024. The SE-01 is the larger 1.7 m full-size model with no public price. Any listing attaching '$12,000' or '88,000 yuan' to the SE-01 has copied the PM-01's number by mistake. Q: Is the SE-01's viral walking video real? A: Yes. When the clip of the SE-01 walking naturally went viral in early 2025, many assumed it was CGI, so NVIDIA senior research scientist Jim Fan checked and verified it as a genuine robot demo, describing the controller as a neural network trained in simulation with reinforcement learning and transferred to the real robot. The caveats: the demos were handler-assisted early tests, and 'world-first natural gait' is EngineAI's own marketing framing, not an independent expert verdict. Q: What are the EngineAI SE-01's specifications? A: The SE-01 is a full-size bipedal humanoid: 170 cm tall, about 55 kg, with 32 degrees of freedom and a walking speed up to 2 m/s. It uses EngineAI's own integrated harmonic joint modules (no roller-screw actuators or six-axis force sensors), dual NVIDIA and Intel compute, three sets of stereo cameras, and an aerospace-grade aluminium frame. Payload, battery runtime, and the per-hand joint count are not officially disclosed. Q: Can you buy an EngineAI SE-01? A: Only by inquiry. EngineAI sells the SE-01 through a contact-sales process, with no published price, no cart, and no stated lead time. There is no confirmed authorized reseller, and marketplace pages showing it 'in stock' with an add-to-cart button are not EngineAI channels. If you want a full-size EngineAI humanoid you can actually order today, the newer T800 is the current product. Q: Is the SE-01 open-source? A: No. EngineAI publishes real open-source software on GitHub, including reinforcement-learning training tools and a ROS-based deployment stack, but those target the PM-01 and the newer T800, not the SE-01. The SE-01's hardware is not open. The open-source cost advantage people associate with EngineAI belongs to the PM-01 developer robot. Q: What is the difference between the SE-01 and the T800? A: The SE-01 (2024) is EngineAI's original full-size humanoid, the one famous for the viral gait, sold only by quote with no public price. The T800 (debuted CES 2026, mass delivery from May 2026) is the taller, stronger successor and EngineAI's current flagship, priced from around $25,000. There is no SE-02; the T800 is the SE-01's replacement. To actually order an EngineAI full-size humanoid today, the T800 is the product. Q: Who makes the EngineAI SE-01? A: EngineAI, a Shenzhen humanoid startup founded in 2023 by Zhao Tongyang, who previously worked on quadruped and humanoid robots at Pengxing, a venture tied to the automaker XPeng. EngineAI has raised well over $100 million, reached a reported valuation of about $1.5 billion, and filed confidentially for a Hong Kong IPO in June 2026. It runs a 12,000-square-metre factory in Shenzhen. Q: Is the EngineAI SE-01 worth it? A: For most buyers, no, because it is unpriced, quote-only, and no longer EngineAI's headline product. It makes sense only for a research group that specifically wants EngineAI's original gait platform and can negotiate a bespoke quote. If you want an affordable open-source developer humanoid, look at the PM-01; if you want EngineAI's current full-size machine with a published price, look at the T800. The EngineAI SE-01 is the robot that made people do a double take. In early 2025 a video of it strolling down a Shenzhen street with a smooth, straight-knee, almost human walk went viral, and plenty of viewers assumed it had to be CGI. It was not. But the SE-01 is also one of the most misunderstood robots in the field: it has no official price, it is not the cheap open-source model people think it is, and it is no longer the flagship of the company that built it. This review sorts out what the SE-01 actually costs, what is real about the famous walk, what is inside it, and whether it is the EngineAI robot you would actually want to buy. What the SE-01 is at a glance The SE-01 is a full-size bipedal humanoid, about 1.7 metres tall, built by the Shenzhen startup EngineAI. Its claim to fame is a natural, human-like walking gait, and its October 2024 launch pitched it as a general-purpose humanoid for industry, research, and education. How much does the SE-01 cost? There is no official, confirmed public price for the SE-01. EngineAI has never listed one. Its own product page shows specifications and a "contact us" button, with no cart and no number, so the SE-01 is effectively a quote-only, business-to-business product. The figure you will see quoted, roughly $20,500 to $27,350 (about 150,000 to 200,000 yuan), comes from a co-founder describing what EngineAI is aiming for across its full-size humanoids in general. It is a target for the class, not a sticker price for the SE-01. In fact, the same reporting that carried that target noted plainly that EngineAI has not disclosed pricing for the higher-end SE-01. Treat the range as a directional signal and expect a bespoke quote, not a checkout. The $12,000 robot is the PM-01, not the SE-01 This mix-up trips up almost every search. The cheap, roughly $12,000 EngineAI humanoid you have read about, the open-source one, is the PM-01. It is a different, smaller robot. EngineAI launched the PM-01 in December 2024 at a promotional 88,000 yuan (about $12,000) for both its commercial and education versions, aimed squarely at developers. The SE-01 is the larger, full-size model with no public price. If a listing attaches "$12,000" or "88,000 yuan" to the SE-01, it has copied the PM-01's number by mistake. Fake and inflated price listings Because the SE-01 has no official price, aggregator and marketplace pages fill the vacuum with numbers that do not trace to EngineAI. One storefront shows the SE-01 at $54,000 with an "add to cart" button, implying you can buy one on the spot; there is no source for that figure and no evidence it is real stock, so treat it as an unverified third-party listing with no manufacturer authorization, not confirmed inventory. Other sites publish round estimates like $20,000. None of these are EngineAI prices. The only honest statement is that the SE-01 is unpriced and quote-only, with the co-founder's roughly $20,500 to $27,350 class target as the best directional guide. The viral gait: what made the SE-01 famous Most humanoids walk with bent knees and short, stomping steps, a crouch that keeps them stable but looks unmistakably robotic. The SE-01's party trick is the opposite: a longer stride with a near-straight knee and a smoother, more human cadence. EngineAI says it achieved this with an end-to-end neural network trained by a mix of reinforcement learning and imitation learning, then transferred from simulation to the real robot. Notably, EngineAI says it did this without the six-axis force sensors and roller-screw actuators that rivals rely on, using its own integrated harmonic joint modules instead. Is the walking video actually real? Yes. When the clip went viral, so many people cried CGI that NVIDIA senior research scientist Jim Fan weighed in, and after checking he verified it as a real robot demo, describing the controller as "a neural net trained in Isaac simulator using reinforcement learning and then sim2real." Second-angle footage and a photo of an engineer ready to catch the robot backed that up. Two caveats matter. First, the demos were handler-assisted early-stage tests, not proof of reliable, hands-off operation, and like any launch reel they show the best takes. Second, "the world's first natural gait" is EngineAI's own framing; the walk is genuinely impressive and independently confirmed as real, but there is no independent expert consensus that it is a fundamental breakthrough rather than a very well-executed simulation-to-reality result. The capability is real. The superlatives are marketing. Full specifications EngineAI publishes a modest set of official figures for the SE-01. The confirmed ones are below; anything not on this list should be treated with caution (see the next section). Reporting from the launch also put peak joint torque at around 330 newton-metres, consistent with EngineAI's own materials. What EngineAI does not disclose Several numbers that float around the SE-01 are aggregator inventions, not EngineAI figures. The company has not published an official payload, battery runtime, per-hand joint count, or a joint-by-joint map of the 32 degrees of freedom. If you see a specific payload in kilograms or a runtime in hours stated as fact, it is unverified. Two mix-ups are especially common: pages that list the SE-01 as 180 cm and 80 kg with 12 joints are simply wrong, and pages that show 138 cm and 40 kg have copied the smaller PM-01's specs. The reliable SE-01 figures are the ones in the table above. Is the SE-01 open-source? No, not the SE-01 itself. EngineAI does publish genuine open-source software on GitHub, including reinforcement-learning training tools and a ROS-based deployment stack, but those resources target the PM-01 and the newer T800, not the SE-01. The SE-01's hardware is not open. So the open-source cost advantage people associate with EngineAI belongs to the PM-01 developer robot, not to this flagship. Who makes the SE-01: EngineAI EngineAI is a Shenzhen humanoid startup founded in 2023 by Zhao Tongyang, who previously worked on quadruped and humanoid robots at Pengxing, a robotics venture tied to the automaker XPeng. The company has moved fast: it has raised well over $100 million, reached a reported valuation of about $1.5 billion, and filed confidentially for a Hong Kong IPO in June 2026, only about two and a half years after it was founded. It runs a 12,000-square-metre factory in Shenzhen that it says can turn out a humanoid roughly every 15 minutes. For a buyer, that backing signals staying power, though it is worth remembering EngineAI is a mid-pack player by volume in a crowded Chinese field led by Unitree. The catch: the SE-01 is now a stepping stone This is the part that changes the buying decision. The SE-01 earned EngineAI its reputation, but the company's attention has clearly moved on to the T800, a taller, stronger full-size humanoid it debuted at CES in January 2026 and began mass-producing in May 2026, priced from around $25,000. There is no SE-02; the T800 is the successor. So if your goal is to actually order an EngineAI full-size humanoid today, the T800 is the current product, and the SE-01 is best understood as the predecessor that proved the gait. Also worth clearing up: the front-flip and dancing clips people credit to the SE-01 are actually the PM-01, and the later "is it CGI" kick-video controversy involved the T800, not this robot. Who should consider the SE-01, and who should not The SE-01 makes sense for a narrow buyer: a research group or company that specifically wants EngineAI's original gait platform and is prepared to negotiate a bespoke quote for a robot that is no longer the headline product. For almost everyone else, the honest guidance is to look at the PM-01 if you want an affordable, open-source developer humanoid, or the T800 if you want EngineAI's current full-size machine with a published price and a production line behind it. It is the wrong pick for anyone expecting a cheap, off-the-shelf, ready-to-work humanoid. The SE-01 is unpriced, quote-only, and its viral walk, while real, was shown under handler supervision rather than in reliable autonomous work. Because everything is negotiated per deal, treat the price, lead time, warranty, spare-parts and service coverage, safety certification, and developer access as unknowns to pin down in writing before committing. The bottom line The SE-01 is a genuinely important robot: it showed, on video that even skeptics accepted, that a humanoid could walk like a person, and it put EngineAI on the map. But as a purchase it is the most misunderstood option in the lineup. It has no official price, it is not the $12,000 open-source model (that is the PM-01), and it is no longer EngineAI's flagship (that is the T800). Admire the SE-01 for the walk. If you are buying, look at the robot next to it. Compare on RoboZaps: SE01, PM01 and Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Clone Alpha Price 2026: Can You Buy It, Cost & Is It Real? URL: https://blog.robozaps.com/b/clone-protoclone-review Author: Radica Maneva Published: 2026-02-11T00:00:00.000Z Updated: 2026-08-01T23:07:21.242Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: reviews TL;DR: You cannot buy a Clone Alpha as of July 2026: no orderable price has ever been published, the 279-unit preorder never opened a public checkout, and there is no public evidence a unit has shipped. The viral Protoclone is real hardware, but no footage shows it standing or walking. '$20,000' is Clone's own long-stated target and manufacturing claim, not a price; its roadmap targets walking in 2027. Key takeaways: - No Clone Alpha price has ever been published and the 279-unit preorder never opened a checkout; the site's 'Pre-orders available in 2025' page rolled into July 2026 unchanged. - The Protoclone is real hardware, confirmed by an independent lab visit, but no public footage shows it standing or walking; every video we reviewed shows it suspended from a gantry. - '$20,000' is Clone's own long-stated figure: a 2023 pitch price target and a March 2026 founder manufacturing claim, neither a purchasable price; the $100K and $200K figures have no Clone source. - Clone has announced no cryptocurrency, and its Republic crowdfunding campaign was withdrawn with all investments refunded; as of July 2026 there is no official public way to buy the robot or invest in the company. - Clone's own roadmap: a surgical-grade torso in late 2026, a walking demo in 2027, and a first commercial 'Robo Butler' in 2028; the company has been publicly quiet since April 7, 2026. FAQ: Q: How much does the Clone Alpha cost? A: No orderable price has ever been published. The $20,000 figure is Clone's own: a 2023 pitch promised mass-market Clones 'at an affordable $20,000', and in March 2026 the founder claimed one can be manufactured for under that figure. Neither is a checkout price. The $100,000-plus figures have no Clone source, and the hand's $2,800 figure was a company materials claim from 2023. Q: Can you buy a Clone Alpha or Protoclone in 2026? A: No. The official pre-order page has no price, no deposit amount and no checkout; it is a reservation form, and it still reads 'Pre-orders available in 2025' as of July 2026. We found no public evidence that any Clone android has shipped. Q: Is the Protoclone real or CGI? A: Real. An independent tech creator filmed the machine in person at Clone's lab in September 2025, and the company has eight years of physically consistent hand and torso demos behind it. Real does not mean capable: every public video we reviewed shows it suspended from a frame, and the lab visit did not validate the company's spec claims. Q: Has the Protoclone ever walked? A: No public footage, official or otherwise, shows the robot standing or bearing its own weight. 'Bipedal' in the official video title describes the body plan. Clone's own roadmap targets a natural walking demonstration in 2027. Q: What is the difference between Protoclone and Clone Alpha? A: Protoclone is the real research prototype, the faceless android shown suspended in the viral videos. Clone Alpha is the announced consumer android, a limited edition of 279 units with 16 promised household skills, which so far exists as a reservation page with no price and no shipped units. Q: What can the Protoclone actually do? A: On video: actuate its 1,000-plus water-driven artificial muscles while suspended, and, on bench-mounted torsos, perform two-handed manipulation like opening a jar. The hand line offers some of the most humanlike dexterity in robotics. No public demonstration we reviewed shows a chore, a conversation, standing or walking. Q: Who owns Clone Robotics? A: It is a private startup founded in 2021 by CEO Dhanush Radhakrishnan and Polish engineer Lukasz Kozlik, whose earlier work went viral as Automaton Robotics. It is a US company with engineering roots in Poland and, since 2026, an office in Mountain View. The founder said in March 2026 it had raised about $17 million with a $50 million round underway. Q: Can you invest in Clone Robotics? A: Not currently. The company is private, and its earlier retail crowdfunding campaign on Republic was withdrawn, with Republic stating all investments were refunded. Its venture rounds are private. As of July 2026 there is no official channel for public investment, so verify any claimed investment channel against Clone's official site before sending money. Q: Is there a Protoclone cryptocurrency? A: Clone has announced no token on any official channel. The 'Protoclone' coins that dominate price searches are Solana memecoins that borrowed the robot's name; one major wallet labels its listing unverified. We found no connection between any token and the company. Q: When will the Clone Alpha ship? A: No date exists. The 2025 preorder window passed with no public checkout opening, and Clone's own March 2026 roadmap points to a walking demo in 2027 and a first commercial android for enterprise in 2028. We found no sourced earlier shipping date. The short version, verified against Clone Robotics' own pages on July 23, 2026: you cannot buy a Clone Alpha, no orderable price has ever been published, and there is no public evidence a single unit has shipped. The company's pre-order page still says "Pre-orders available in 2025", unchanged since December 2024 and now out of date. The widely repeated "$20,000" figure is not an orderable price either: it is Clone's own long-running target, from a 2023 pitch promising to "sell mass market Clones at an affordable $20,000" to the founder's March 2026 claim that one can now be manufactured for under that figure. And the robot in the viral videos, the faceless Protoclone, is real hardware, but no public footage shows it standing or walking; Clone's own roadmap now puts a walking demonstration in 2027 and a first commercial product in 2028. That is the whole answer for buyers. What remains is one of the strangest and most interesting stories in robotics: a small company trying to build an android the way biology does, with water-driven artificial muscles pulling on a plastic skeleton, marketing it with some of the most unsettling footage on the internet, and promising far more than it has shown. This page separates the two. Protoclone, Clone Alpha, and the hand: which is which Most confusion about this company comes from collapsing those four things into one. The hand is genuinely impressive and well documented. The torsos are real lab demos. Protoclone is a research platform shown in carefully staged videos. Clone Alpha, the thing you could theoretically buy, exists so far as a reservation page. Is the Protoclone real? Yes. The "it's CGI" comments that flooded the February 2025 viral wave were wrong. The strongest evidence is independent: tech creator Anastasi In Tech visited Clone's lab in September 2025 and filmed the machine in person. Before that, there were eight years of incremental, physically consistent hand and arm videos, workshop-setting footage, and mainstream coverage from outlets like Live Science and R&D World that treated it as real hardware throughout. The lab visit corroborates that the hardware is physical; it does not validate the company's specification claims or the robot's capabilities. Real is not the same as capable. In every public video we reviewed, the February 2025 original and the April 2025 follow-up included, Protoclone hangs from a frame while it twitches, shrugs, clenches its fists and poses its legs. The word "Bipedal" in the official video title describes the body plan, not a demonstrated skill. As of July 2026 we could find no footage, official or otherwise, of the robot bearing its own weight. The price truth, including where "$20,000" comes from Clone has never published an orderable price for anything. The pre-order page has no number, no deposit amount and no checkout, and the hand has no public store. Here is the provenance of the figures that circulate anyway: $20,000: a company number, but never a checkout price, and it comes in three layers. Clone's own 2023 crowdfunding pitch promised to "sell mass market Clones at an affordable $20,000". At the Abundance Summit on March 10, 2026, CEO Dhanush Radhakrishnan said Clone can now manufacture a musculoskeletal android for under $20,000 by producing its muscle fiber "by the kilometer", and reaffirmed the mass-market price target. None of these is an audited cost or a purchasable Clone Alpha price, and third-party sites have been repeating the figure as one ever since.$100,000 and $200,000-plus: figures with no Clone source, published by spec aggregators. One widely indexed listing shows "$100,000" next to its own "Not Verified" badge.$2,800: a hand figure, not an android figure: in 2023 Clone told Freethink the hand used less than $2,800 in materials. A company materials claim, not a retail price or an audited production cost. One more trap unique to this robot: searches for "Protoclone price" mostly return cryptocurrency trackers. Solana tokens using the robot's name appear on Phantom, which labels its listing unverified, and OKX; we found no mention of any token on Clone's official channels. Clone has announced no token on any official channel. It also, as of July 2026, has no public investment channel at all: the company previously ran a retail crowdfunding campaign on Republic, but that page now states the campaign was withdrawn and all investments were refunded. For anything taking money in Protoclone's name today, coin or storefront, we found no evidence of affiliation with Clone. The 279-unit preorder: still no price, no checkout Clone announced Clone Alpha in December 2024 as a limited edition of 279 units with preorders "available in 2025". The product page still invites you to "Reserve one of the first 279 Clones ever made", and still promises "a robot that walks naturally" with 16 pre-installed skills: navigating the home, conversation, serving drinks, laundry, vacuuming and more. As of July 23, 2026, that page has no price, no deposit amount and no functional checkout, and we found no evidence paid ordering ever opened, though December 2024 coverage did describe the reservation launch as "taking preorders". 2025 ended with no announced price, no delivery and no sold-out notice, and the page rolled into July 2026 unchanged. The company has effectively superseded it: the roadmap Radhakrishnan laid out in March 2026 runs from a "surgically accurate" torso in late 2026, to a natural walking demonstration in 2027, to a first commercial android for enterprise customers, a "Robo Butler" aimed at hotels, in 2028. Read strictly, Clone's own timeline no longer has room for a consumer Alpha shipping before then. The reservation form appears to cost nothing, and no commitment from the company attaches to it. What it can actually do versus what is promised Demonstrated, on video, as of July 2026: the hand's dexterity, which remains among the most humanlike in robotics; bench-mounted torsos doing two-handed tasks like holding a jar with one hand and opening it with the other; and the full Protoclone body actuating its muscle system while suspended. The most recent hardware shown is Torso 3, unveiled in March 2026, with an April 1, 2026 update noting joint-angle feedback in the hands and valves now built in-house. Promised but never shown: everything a buyer would actually be paying for. No public demonstration we found shows a chore end to end, a conversation, standing, walking, or an autonomous task of any kind. This gap is not unusual for a robotics startup, but Clone's site sells the promise in the present tense: the Android page says "a robot that walks naturally" while no public footage shows unsupported walking. The technology, honestly assessed Protoclone's design is genuinely different from every mainstream humanoid. Tesla's Optimus, Unitree's machines and Figure's robots use electric motors and gearboxes at discrete joints. Clone builds anatomically: a 3D-printed polymer skeleton it says replicates all 206 human bones, more than 1,000 "Myofiber" artificial muscles derived from McKibben pneumatic actuators, driven by pressurized water from a 500-watt electric pump, plus roughly 500 sensors including four depth cameras. The company says a Myofiber contracts more than 30% in under 50 milliseconds and that a 3-gram fiber pulls over a kilogram. The result moves like nothing else in robotics, which is precisely why the videos went viral. The honest caveats are equally real. The anatomy claims are marketing-shaped: the 206 "bones" are partially fused polymer prints, and "1,000 muscles" counts individual fiber strands, where human muscles number about 600 and contain millions of fibers. Musculoskeletal robots are not new; the University of Tokyo's Kengoro line explored tendon-driven anatomy years earlier. And the engineering literature on McKibben-type muscles documents the hard problems Clone has not yet publicly solved: bladder and braid fatigue under repeated pressure cycling, leakage as the standard failure mode, and the open research problem of coordinating hundreds of antagonistic soft actuators. Add the practical questions a home android raises, pump noise, water plumbing and serviceability, and the distance between a twitching gantry demo and a laundry-folding product comes into focus. Company status, July 2026 Clone Robotics was founded in 2021 by Radhakrishnan and Polish engineer Łukasz Koźlik, whose earlier robotic-muscle work went viral under the name Automaton Robotics. It is a US company with its engineering roots in Poland. Per Radhakrishnan's March 2026 summit remarks, it opened a Silicon Valley office in Mountain View, had raised about $17 million to date, and was raising a $50 million round; that round has not been confirmed closed, and funding databases still show older, lower totals. One more data point buyers deserve: the company has gone publicly quiet. Its last original post on X was April 7, 2026, its YouTube channel's most recent upload dates to November 2025, and the news section of its own website stops in August 2024. Quiet does not mean dead, small teams go heads-down, but it does mean every "new Protoclone update" article you see in mid-2026 is recycling footage that is over a year old. Should you wait for a Clone Alpha? If the muscle-first approach excites you, the reservation form costs nothing and commits you to nothing, and this is a company worth watching: the hand is genuinely impressive engineering, and the approach is the most prominent attempt at biomimetic actuation at commercial scale. But judge it by the standard its own history sets. The 2025 preorder window passed in silence, walking is now a 2027 target, and the first product Clone itself talks about shipping is a 2028 enterprise machine, not a home android. For robots you can actually evaluate today, see our guides to the best humanoid robots and what humanoid robots actually cost. The bottom line The Protoclone is real, the videos are not CGI, and the underlying muscle technology is a legitimately novel line of research. Everything a customer could act on is not there yet: no orderable price, no open checkout, no publicly evidenced shipment, and no demonstration of standing, walking or a single household task. The one number with a real source, $20,000, is Clone's own long-stated target and manufacturing claim, not a price you can pay. Treat Clone Alpha as a concept with a reservation page, and check back when a robot bears its own weight on camera. There is currently no official way to buy the robot or invest in the company at all, the Republic campaign having been withdrawn and refunded, and we found no evidence that anything sold under this robot's name today, coin or storefront, is affiliated with the company. Compare on RoboZaps: Clone Alpha. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Figure 03 vs 1X NEO: Which Home Robot Is Closer? URL: https://blog.robozaps.com/b/figure-03-vs-1x-neo Author: Radica Maneva Published: 2026-02-11T00:00:00.000Z Updated: 2026-08-01T23:07:25.261Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: NEO is the only option households can reserve, but it remains a pre-order. Figure 03 is designed for home and commercial use and has stronger industrial evidence, but no consumer order program. If you need a delivered, supported household robot today, choose neither. Key takeaways: - NEO is the only robot in this matchup with public home pre-order terms, but customer delivery is not yet established. - Figure 03 is designed for home and commercial use, with no public consumer price, order channel or support terms. - Figure now publishes core Figure 03 specifications; vendor payload, runtime and speed terms still are not directly comparable. - Remote assistance gives NEO a capability bridge and creates privacy and transparency questions buyers must review. - Neither vendor publishes enough comparable customer data for a universal autonomy or safety winner. FAQ: Q: Is Figure 03 better than 1X NEO for the home? A: Not as a current purchase. Figure 03 is designed for home and commercial use and has stronger industrial deployment evidence, but Figure has no consumer order program, list price or support terms. NEO is purpose-built for homes and accepts US pre-orders, although customer delivery and ordinary household autonomy have not yet been established. Q: How much do Figure 03 and 1X NEO cost? A: Figure has not published a Figure 03 list price. 1X lists NEO at $20,000 or $499 per month, with a refundable $200 deposit. That is a pre-order term, not proof that a finished robot has been delivered and supported in customer homes. Q: Can you buy Figure 03 or 1X NEO now? A: You cannot place a public Figure 03 order. 1X accepts NEO pre-orders in the United States and says customer shipments are planned before the end of 2026. Until delivery occurs, NEO should be treated as a pre-order rather than an available retail product. Q: Does 1X NEO use human remote operators? A: Yes. 1X describes an Expert Mode in which remote human experts can assist NEO. The company also describes owner approval, people blurring and no-go zones. Buyers should still review what operators can see or control, how access is logged and how much assistance normal tasks require. Q: Which robot has better autonomy, Figure 03 or 1X NEO? A: There is not enough comparable public operating data for a defensible universal winner. Figure has demonstrated Helix on industrial workflows and is also developing housework capabilities. 1X targets household tasks with autonomy plus remote assistance. Compare task success, interventions and deployment evidence for the intended use rather than vendor demo breadth. Figure 03 vs 1X NEO: The Short Answer Choose 1X NEO only if you accept the risks of a home-robot pre-order and are comfortable with remote assistance. Figure 03 is designed for home and commercial use, but it is not a consumer buying option: Figure has no public price, household order program or consumer support terms. That does not make NEO the more capable robot in every setting. It makes NEO the only robot in this exact matchup currently presented as a product a household can reserve. Figure 03 has the stronger industrial deployment path; NEO has the clearer home proposition. Head-to-Head Comparison Official specifications and commercial status checked July 16, 2026. Vendor terms such as payload, carry capacity, runtime and speed are not directly comparable because the test conditions differ. Availability and Price Figure 03 Figure has not published a list price or public order page for Figure 03. Its production program covers housework research, commercial development and deployments outside Figure's headquarters. The robot is designed for home and commercial use, but there is still no public household order channel or consumer service package to evaluate. 1X NEO The official NEO order page lists $20,000 ownership or $499 per month with a refundable $200 deposit for US customers. 1X opened its Hayward factory in April 2026 and says customer shipments are planned before year-end. The correct label today is pre-order—not shipping consumer product. Home Design and Physical Risk NEO is explicitly designed around home interaction: 1X describes a soft, tendon-driven body, a 30 kg weight and household-oriented movement. Figure says Figure 03 uses soft textiles and protective foam and was designed for home use as well as commercial work. Neither vendor description replaces a task-specific safety assessment or evidence from ordinary operation around children, pets, stairs and clutter. The official Figure 03 product page now lists height, weight, payload, runtime and speed. Those figures are included in the table above; price and consumer support terms remain unpublished. Autonomy, Teleoperation and Privacy Figure's Helix system has been shown performing enterprise workflows, including a Figure 03 parts-sequencing demonstration at BMW. Figure called it a demonstration, not a fleet-scale production result. The result shows relevant capability but does not establish unsupervised home performance. 1X openly describes Expert Mode, in which human experts can assist NEO remotely. Its NEO product page describes people blurring, no-go zones and owner controls. Those are useful safeguards, but remote access remains a material buying question: customers should know when it activates, what an operator can see or control, how sessions are logged and how often ordinary tasks require help. There is no honest universal autonomy winner without comparable data on task completion, intervention frequency, failures and the test environment. Figure has stronger industrial evidence; NEO has a more explicit home workflow and assistance model. Which Should You Choose? Choose 1X NEO only if: You specifically want a home-first humanoid and accept pre-order risk.You understand that remote assistance may be part of normal operation.You have reviewed privacy, cancellation, delivery, support and service terms.You do not need proven unsupervised household reliability today. Watch Figure 03 if: Your interest is enterprise manufacturing or logistics rather than a home purchase.You want to track Helix, Figure's BMW work and BotQ production progress.You are willing to wait for a public program, price and support terms before making a buying decision. Choose neither if: You need a delivered, supported general-purpose home robot now.You require published customer uptime, intervention, safety and service data.You are not comfortable with pre-order uncertainty or remote access inside the home. Final Verdict NEO wins the home-buying comparison by default because it is the only home-first product with public reservation terms. That is a narrow verdict, not proof that it has better general autonomy, hardware or reliability. The absence of delivered-customer evidence keeps it in the high-risk early-adopter category. Figure 03 has stronger industrial deployment evidence, but it is not a current home purchase. Compare on RoboZaps: Figure 03 and NEO. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # EBO X Price 2026: Is the $799 AI Companion Robot Worth It? URL: https://blog.robozaps.com/b/ebo-x-review Author: Radica Maneva Published: 2026-02-10T23:11:50.000Z Updated: 2026-08-01T23:07:07.580Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The EBO X costs $799 on Enabot's official store, down from a $999 MSRP (€829 in the EU), though as of July 2026 it is out of stock at both Enabot and Amazon. It pairs GPT-4o mini conversation with a 4K stabilized camera, V-SLAM navigation and Harman audio, aimed at families who want an AI companion and home monitor in one robot. Key takeaways: - Price: $799 on Enabot's official store (down from $999 MSRP); €829 in the EU. Out of stock at Enabot and Amazon as of July 2026. - AI: GPT-4o mini integration for natural conversation + Alexa assistant - Camera: 4K stabilized with night vision, 106° FOV — excellent for home monitoring - Navigation: V-SLAM autonomous movement throughout your home - Battery: 2-3 hours runtime, 2-hour charge time - Best for: Families wanting AI companion + home security in one device FAQ: Q: What does EBO stand for? A: EBO is the brand name from Enabot. The EBO X is their flagship "FamilyBot" companion robot. Q: Does EBO X require a subscription? A: Basic features work without subscription. Some cloud storage and advanced features may require Enabot app services — check current offerings at enabot.com. Q: Can EBO X climb stairs? A: No. EBO X can handle inclines up to 15° but cannot climb stairs. It's designed for single-floor navigation. Q: How long does the battery last? A: 2-3 hours of active use. The 2500mAh battery charges in about 2 hours on the included dock. Q: Does EBO X work without internet? A: Basic movement works offline, but AI conversation (GPT-4o mini), Alexa, and remote monitoring require internet connection. Q: Can I use EBO X as a security camera? A: Yes. The 4K camera with night vision, V-SLAM navigation, and mobile app make EBO X effective for home security. It can patrol, send alerts, and record to microSD card. Q: Is EBO X good for elderly people? A: Yes. The GPT-4o mini provides genuine conversation (not just commands), and family can check in remotely via the camera. It's one of the best companion robots for elderly living alone. The EBO X price is $799 (MSRP $999), and the companion robot combines GPT-4o mini AI conversation with 4K home monitoring in an autonomous, self-navigating package. It's the most capable companion robot available, offering genuine AI interaction, premium Harman audio, and the ability to patrol your home checking on pets, kids, or elderly family members. EBO X Overview: What Is It? The EBO X (also called EBO X FamilyBot) is Enabot's flagship companion robot. Unlike simple desktop pets, the EBO X is designed to be a home guardian — roaming your house autonomously while providing AI conversation, video monitoring, and smart home control. Key differentiators from other companion robots: Mobility: Moves throughout your home via V-SLAM navigationAI Conversation: GPT-4o mini enables natural dialogue, not just commandsSecurity: 4K camera with night vision for home monitoringPremium Audio: Harman speaker system for quality sound Think of EBO X as the intersection of a smart speaker, security camera, and companion robot — all in one mobile package. EBO X Price: How Much Does It Cost? The EBO X is officially priced at $799 in the US, marked down from its $999 MSRP, and €829 on Enabot's EU store. As of July 2026 both official stores list it as out of stock ("Notify Me"), and the Amazon US listing is unavailable, so check stock before planning a purchase. What's included: EBO X robotCharging dock1-year warrantyGlobal shipping available Compared to other companion robots, EBO X sits at the premium end: The premium is justified if you want home monitoring + AI conversation in one device. For desktop-only use, consider more affordable options. EBO X Specs & Features Full Specifications GPT-4o mini AI Conversation The standout feature of EBO X is its GPT-4o mini integration. Say "EBO EBO, Open GPT" to start a conversation. Unlike basic voice assistants that only respond to commands, EBO X can: Have continuous, contextual conversationsAnswer complex questions with nuanced responsesRemember context within a conversationProvide companionship through natural dialogue This makes EBO X genuinely useful for elderly family members who want conversation, kids who have endless questions, or anyone who wants more than "play music" from their robot. 4K Camera & Home Monitoring The EBO X camera system is serious: 4K resolution: UHD, FHD, HD, and SD options106° field of view: Wide coveragef/1.8 aperture: Good low-light performanceNight vision: IR for dark environmentsSingle-axis gimbal: Stabilized footage, 104° vertical adjustmentLocal storage: MicroSD up to 256GB for recordings Use cases: Check on pets while you're at workMonitor elderly parents remotelyHome security patrolsVideo calls with family V-SLAM Navigation EBO X doesn't just sit on your desk — it navigates your home autonomously using V-SLAM (Visual Simultaneous Localization and Mapping). It builds a map of your space and moves intelligently without bumping into furniture. Speed: Up to 1.5 m/s (walking pace)Climbing: Handles inclines up to 15°Self-balancing: Two direct-driven hub motorsObstacle avoidance: Real-time detection and navigation Harman Premium Audio The 8W Harman speaker with AudioEFX delivers quality sound for: Music playbackVoice responsesVideo call audioAlexa interactions Alexa Integration Beyond GPT-4o mini conversation, EBO X includes full Alexa functionality: Smart home device controlTimers, reminders, alarmsWeather, news, informationMusic streamingShopping lists What Can the EBO X Do? Family Guardian EBO X's tagline is "FamilyBot" for good reason. It can: Patrol your home on schedule or commandSend alerts when it detects activityLet you check in via live video feedProvide two-way communication with family members AI Companion The GPT-4o mini integration makes EBO X more than a gadget: Have meaningful conversationsAnswer questions on any topicProvide companionship for those living aloneEntertain kids with stories and games Smart Home Hub With Alexa built in, EBO X controls your smart home: Lights, thermostats, locksOther Alexa-compatible devicesVoice-controlled automation Who Should Buy the EBO X? EBO X is ideal for: Families with elderly parents: Check in remotely, provide companionshipPet owners: Monitor and interact with pets while awayParents: Keep an eye on kids, educational conversationsAnyone wanting premium AI: GPT-4o mini beats basic assistantsHome security users: Mobile 4K monitoring throughout the house EBO X is NOT for: Budget buyers: At $799, it's a premium purchaseDesktop-only use: Get EMO or Loona insteadSmall apartments: The mobility features shine in larger spaces EBO X vs Competitors Bottom line: EBO X is the only companion robot with serious home monitoring capabilities and advanced AI conversation. If you want both, it's the clear choice. If you only want a desktop pet, save money with alternatives. Final Verdict The EBO X is the most capable companion robot available in 2026. The combination of GPT-4o mini AI, 4K camera, V-SLAM navigation, and premium Harman audio justifies its $799 (MSRP $999) price for buyers who want more than a desktop toy. Pros: Genuine AI conversation via GPT-4o miniExcellent 4K camera with night visionAutonomous whole-home navigationPremium audio qualityAlexa integration for smart home Cons: Premium price ($799)Requires internet for AI featuresCan't climb stairsOverkill for desktop-only use Our recommendation: If you want a companion robot that doubles as home security and offers real AI conversation, the EBO X is worth the investment. For simpler needs, see our best companion robots guide for affordable alternatives. Where to buy: Enabot Official Store Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # Best Companion Robots 2026: EBO, EMO, Loona & More URL: https://blog.robozaps.com/b/best-companion-robots Author: Radica Maneva Published: 2026-02-10T22:50:00.000Z Updated: 2026-08-01T23:07:04.272Z Category: news TL;DR: The EBO X is the best overall companion robot — now $799 on Enabot's store (down from $999 MSRP), though currently out of stock — with the EMO ($279) as best budget desktop pet and the Loona ($449) as the pet-style pick. At $200-$1,200 these cost far less than humanoids; judge them on AI conversation, camera, battery and app control. Key takeaways: - Best overall companion robot: EBO X ($799 on Enabot's store, down from $999 MSRP; currently out of stock) — 4K camera, GPT-4o mini AI, home monitoring - Best budget desktop pet: EMO ($279) — 1000+ expressions, face recognition, interactive personality - Best pet-style robot: Loona ($449) — emotional AI, pet-like behavior, family-friendly - Prices range from $200-$1,200 — far more affordable than humanoid robots - Key features to look for: AI conversation, camera quality, battery life, app control FAQ: Q: What is the best companion robot in 2026? A: The EBO X is the best overall companion robot for its combination of GPT-4o mini AI, 4K camera, and autonomous navigation. For budget buyers, EMO at $279 offers the best value with 1000+ expressions and genuine personality. Q: How much do companion robots cost? A: Companion robots range from $200-$1,200. Budget options like EMO cost $279. Mid-range options like Loona cost $449. Premium options like EBO X cost $999. Q: Can companion robots have conversations? A: Yes. Modern companion robots like EBO X use large language models (GPT-4o mini) for natural conversation. Simpler robots like EMO can answer basic questions and respond to commands. Q: Are companion robots good for elderly people? A: Yes. Companion robots provide social interaction, reminders, and monitoring capabilities. EBO X's camera can help family check on elderly relatives. Simpler robots like EMO provide entertainment and emotional connection. Q: What's the difference between EBO and EMO robots? A: EBO X (Enabot) is a home-roaming robot with 4K camera and GPT-4o mini AI ($799). EMO (Living.AI) is a desktop pet with 1000+ expressions ($279). EBO focuses on home monitoring; EMO focuses on personality and entertainment. Q: Do companion robots work without internet? A: Basic functions work offline (movement, expressions), but AI conversation features require internet. EBO X needs connectivity for GPT-4o mini; EMO needs it for voice assistant features. The best companion robots in 2026 are the EBO X for home monitoring and AI conversation ($799), EMO for an affordable desktop pet experience ($279), and Loona for pet-like emotional interaction ($449). Unlike humanoid robots costing $5,000+, companion robots offer AI interaction at consumer-friendly prices. What Is a Companion Robot? A companion robot is a small, interactive robot designed for personal entertainment, emotional support, and practical assistance. Unlike industrial robots or humanoids, companion robots prioritize: Emotional connection: Expressive faces, personality, bonding over timeEase of use: No programming required, works out of the boxAffordability: Typically $200-$1,200 vs $5,000+ for humanoidsCompact size: Desktop or floor-roaming, not human-sized Companion robots sit between smart speakers (no physical presence) and humanoid robots (expensive, complex). They offer AI interaction with a physical, engaging form factor. Best Companion Robots 2026: Our Top Picks 1. EBO X — Best Overall Companion Robot Price: $799 (MSRP $999) | Maker: Enabot The EBO X is the most capable companion robot available. It combines AI conversation (powered by GPT-4o mini), home security features, and autonomous navigation into a rolling companion that can patrol your home. Key Features AI Conversation: GPT-4o mini integration for natural dialogue — say "EBO EBO, Open GPT" to start4K Camera: Stabilized camera with 104° vertical adjustment and night visionV-SLAM Navigation: Autonomous movement through your home without bumping into obstaclesAlexa Integration: Works as an Alexa-enabled assistantHarman Audio: Premium speaker system with AudioEFXSelf-Balancing: Two-wheel design with direct-driven motors Best For Families wanting home monitoring plus AI interaction. The 4K camera makes it excellent for checking on pets, elderly family members, or home security while you're away. The GPT-4o mini integration enables meaningful conversations beyond simple commands. Specs 2. EMO — Best Budget Desktop Pet Price: $279 | Maker: Living.AI EMO is a desktop AI pet with genuine personality. With over 1000 expressions and movements, face recognition for up to 10 people, and autonomous exploration of your desk, EMO creates an emotional bond that simpler robots can't match. Key Features 1000+ Expressions: Emotions displayed on face screen and through body languageFace Recognition: Remembers up to 10 family members, greets you by nameAutonomous Exploration: Moves independently, explores desk, never falls off edgesVoice Activation: 4-microphone array responds to "Hey EMO"Evolving Personality: Changes based on your interactions over timeTouch Sensors: Responds to head pats like a real petSmart Home: Controls compatible lights, answers questionsWireless Charging: Skateboard dock charges EMO and your phone Best For Anyone wanting an affordable desktop companion with genuine personality. EMO is ideal for offices, bedrooms, or as a gift. The low price point and engaging personality make it the best entry into companion robotics. Specs 3. Loona — Best Pet-Style Robot Price: $449 | Maker: KEYi Robot Loona is designed to feel like a real pet. With expressive eyes, emotional AI, and pet-like behaviors, Loona appeals to families wanting companionship without the responsibilities of a living animal. Key Features Emotional AI: Expresses happiness, curiosity, excitement, and other emotionsPet Behaviors: Plays, follows you, responds to gestures and voiceFace Tracking: Maintains eye contact, recognizes family membersInteractive Games: Built-in games and activitiesApp Control: Schedule, customize, and interact via smartphone Best For Families with children, elderly individuals wanting companionship, or anyone who wants a pet-like experience without allergies, feeding, or walks. Loona provides emotional connection through pet simulation. 4. Vector 2.0 — Best Voice Assistant Companion Price: $249.99 | Maker: Digital Dream Labs Vector 2.0 combines companion robot personality with Alexa voice assistant functionality. It's a small robot that can answer questions, control smart home devices, and show affection — all in a palm-sized package. Key Features Alexa Built-In: Full Alexa functionality for voice commandsAutonomous Personality: Explores, reacts, shows emotionsFace Recognition: Learns who you areCompact Size: Fits anywhere Best For Users who want Alexa in robot form. Vector bridges the gap between smart speakers and companion robots. 5. Miko 3 — Best for Kids Price: $299 | Maker: Miko Miko 3 is an educational companion robot designed for children. It combines entertainment with learning through interactive lessons, stories, and activities. Key Features Educational Content: STEM learning, coding basics, academic subjectsAge-Appropriate AI: Safe conversations for kidsParental Controls: Monitor and customize via appVideo Calling: Stay connected with family Best For Parents wanting an educational companion for children ages 5-12. Combines screen time with learning. Companion Robot Buying Guide What to Consider 1. Primary Use Home monitoring: EBO X (4K camera, navigation)Desktop entertainment: EMO (expressions, personality)Pet replacement: Loona (emotional, pet-like)Voice assistant: Vector 2.0 (Alexa built-in)Kids education: Miko 3 (learning content) 2. Budget Under $300: EMO, Miko 3$300-$500: Vector 2.0, Loona$500+: EBO X 3. AI Capabilities Best AI conversation: EBO X (GPT-4o mini)Best personality: EMO (1000+ expressions)Best voice assistant: Vector 2.0 (full Alexa) Companion Robots vs Humanoid Robots How do companion robots compare to humanoid robots like the Unitree R1? For most consumers in 2026, companion robots offer the best value for AI interaction. Humanoid robots are for those who need physical task capability and have budget to match. The line between the two is starting to blur. The DroidUp Moya, a walking humanoid with warm synthetic skin and 25 facial degrees of freedom built for companionship and eldercare, sits somewhere between the categories. At $173,000 and expected in late 2026 it's an institutional purchase, not an EMO alternative — but it shows the direction premium companion robots are taking. Conclusion Companion robots in 2026 offer genuine AI interaction at consumer-friendly prices. Whether you want a home-monitoring assistant (EBO X), an expressive desktop pet (EMO), or a pet-like companion (Loona), there's an option for every budget and need. For the best overall experience, we recommend the EBO X — its GPT-4o mini integration sets it apart from simpler robots. For the best value, EMO delivers impressive personality and interaction at just $279. Looking for something more capable? See our guides to AI robots and humanoid robots for full-size options starting at $5,900. Compare on RoboZaps: Moya and Unitree R1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # AI Robot: Complete Guide to Intelligent Robots [2026] URL: https://blog.robozaps.com/b/ai-robot Author: Radica Maneva Published: 2026-02-10T22:40:00.000Z Updated: 2026-08-01T23:07:00.728Z Category: news TL;DR: AI robots pair artificial intelligence with a physical body so they can learn, adapt and act autonomously. The leading examples in 2026 are the Figure 03, Tesla Optimus, 1X NEO, Unitree G1 and Boston Dynamics Atlas, spanning humanoid, industrial, service and companion types, with the humanoid market projected to reach $38 billion by 2035. Key takeaways: - AI robots combine artificial intelligence with physical robotics — they can learn, adapt, and make decisions autonomously - Types include humanoid robots, industrial robots, service robots, and companion robots — each using AI differently - Leading AI robots in 2026: Figure 03, Tesla Optimus, 1X Neo, Unitree G1, Boston Dynamics Atlas - Key AI capabilities: computer vision, natural language processing, machine learning, autonomous navigation - Market growing rapidly: humanoid robot market projected to reach $38 billion by 2035 FAQ: Q: What is an AI robot? A: An AI robot is a physical robot that uses artificial intelligence to perceive its environment, make decisions, and take actions autonomously. Unlike traditional robots that follow fixed programs, AI robots can learn, adapt, and interact naturally with humans. Q: What's the difference between AI and robotics? A: AI (artificial intelligence) is software that can learn and make decisions. Robotics is the engineering of physical machines. An AI robot combines both — the intelligence of AI with the physical capabilities of a robot. Q: How much do AI robots cost? A: AI robot prices range widely: desktop companions start around $200-500, consumer humanoids like the Unitree R1 cost $4,900, and advanced humanoids like 1X Neo cost $20,000. Enterprise/industrial AI robots can cost $100,000+. Q: Can AI robots learn new tasks? A: Yes. Modern AI robots can learn new tasks through demonstration (watching humans), reinforcement learning (trial and error), or instruction (natural language commands). The 1X Neo, for example, uses a "World Model" to learn tasks by watching videos. Q: Are AI robots safe? A: Safety depends on the design. Home-focused AI robots like 1X Neo feature soft exteriors and gentle actuators for safe human interaction. Industrial AI robots typically require safety cages or work in separate zones. Q: What AI do robots use? A: AI robots use combinations of: computer vision (recognizing objects and scenes), natural language processing (understanding speech), machine learning (improving from data), and increasingly, large language models (LLMs) for reasoning and conversation. Q: Will AI robots replace human workers? A: AI robots are augmenting rather than fully replacing humans in most applications. They excel at repetitive, dangerous, or precision tasks, while humans retain advantages in creativity, complex judgment, and interpersonal situations. An AI robot is a machine that combines artificial intelligence with physical robotics to perceive its environment, make decisions, and take actions autonomously. Unlike traditional robots that follow pre-programmed instructions, AI robots can learn from experience, adapt to new situations, and interact naturally with humans through speech and vision. What Is an AI Robot? An AI robot is a physical robot equipped with artificial intelligence capabilities that allow it to: Perceive: Use sensors, cameras, and LiDAR to understand its environmentDecide: Process information and make autonomous decisionsAct: Execute physical tasks based on its decisionsLearn: Improve performance over time through machine learning The "AI" in AI robot refers to the software intelligence — computer vision, natural language processing, machine learning models — that gives the robot cognitive abilities beyond simple automation. AI Robot vs Traditional Robot How Do AI Robots Work? AI robots combine several technologies to achieve intelligent behavior: 1. Perception Systems AI robots use multiple sensors to understand their environment: Cameras: RGB, depth, and stereo cameras for visual perceptionLiDAR: Laser scanning for precise 3D mappingMicrophones: Audio input for voice commands and sound localizationTouch sensors: Force and pressure feedback for manipulationIMUs: Inertial measurement for balance and orientation 2. AI Processing The robot's "brain" processes sensor data using: Computer Vision: Object recognition, scene understanding, face detectionNatural Language Processing: Understanding and generating speechMachine Learning: Pattern recognition and predictionLarge Language Models: Contextual understanding and reasoning 3. Decision Making AI robots make decisions through: Planning algorithms: Determining action sequences to achieve goalsReinforcement learning: Learning optimal behaviors through trial and errorNeural networks: Processing complex inputs for nuanced decisions 4. Physical Actuation The robot executes decisions through: Motors and actuators: Moving joints and limbsEnd effectors: Grippers, hands, or specialized toolsLocomotion systems: Wheels, legs, or other mobility mechanisms Types of AI Robots Humanoid AI Robots Human-shaped robots designed for general-purpose tasks and natural human interaction. Examples include: Figure 03 — Advanced humanoid for industrial applicationsTesla Optimus — General-purpose humanoid robotUnitree G1 — Affordable research humanoid ($13,500)1X Neo — Consumer home humanoid ($20,000) Industrial AI Robots Factory robots with AI for flexible manufacturing: Collaborative robots (cobots) that work alongside humansAI-powered assembly and inspection systemsAutonomous mobile robots (AMRs) for logistics Service AI Robots Robots performing services in public or commercial settings: Delivery robots (Starship, Nuro)Cleaning robots (commercial floor scrubbers)Healthcare robots (surgical assistants, patient care)Hospitality robots (hotel service, restaurant delivery) Companion AI Robots Robots designed for personal interaction and entertainment: Desktop companions (EMO, Loona, EBO)Social robots for elderly careEducational robots for children Top AI Robots in 2026 For a complete comparison, see our best humanoid robots guide. AI Robot Applications Manufacturing & Warehouses AI robots are transforming industrial operations: Flexible assembly: Adapting to product variations without reprogrammingQuality inspection: AI vision detecting defects humans missPicking and packing: Handling diverse items in fulfillment centersPredictive maintenance: Anticipating equipment failures Healthcare AI robots supporting medical professionals: Surgical robots: Precision assistance in operationsPatient care: Monitoring, medication delivery, mobility assistanceRehabilitation: Personalized therapy assistanceElderly care: Companionship and safety monitoring Home & Consumer AI robots entering everyday life: Household chores: Cleaning, organizing, cooking assistancePersonal assistance: Scheduling, reminders, informationCompanionship: Social interaction and entertainmentSecurity: Home monitoring and patrol Retail & Hospitality Customer-facing AI robots: Customer service: Information kiosks and guidanceInventory management: Stock monitoring and replenishmentFood service: Order delivery and preparationCleaning: Autonomous floor care The Future of AI Robots Key trends shaping AI robotics: Generative AI Integration Large language models and generative AI are enabling robots to: Understand complex natural language instructionsGenerate solutions to novel problemsLearn new tasks from demonstrations or descriptionsHave meaningful conversations with humans Embodied AI AI systems that learn through physical interaction with the real world, not just data: Learning from trial and error in physical environmentsDeveloping intuitive physics understandingTransferring skills between simulation and reality Affordable Humanoids Prices are dropping rapidly: 2024: Most humanoids $100,000+2024: Unitree G1 launched at $16,000 (now $13,500)2026: Unitree R1 at $4,900, 1X Neo at $20,0002027+: Sub-$10,000 humanoids expected Learn more about pricing trends in our humanoid robot cost guide. How to Choose an AI Robot Consider these factors when selecting an AI robot: 1. Purpose Research/Development: Unitree G1 EDU, Figure 03Home use: 1X Neo, Unitree R1Industrial: Boston Dynamics Atlas, Figure 03Companion: Desktop robots like EMO, EBO 2. Budget Under $1,000: Companion robots, basic AI toys$5,000-$10,000: Unitree R1$10,000-$20,000: Unitree G1, 1X Neo$20,000+: Advanced humanoids, enterprise solutions 3. AI Capabilities Needed Voice interaction: Most modern AI robotsVision/object recognition: Mid-range and aboveAutonomous navigation: Requires LiDAR or depth camerasTask learning: Advanced humanoids only Conclusion AI robots represent the convergence of artificial intelligence and physical robotics — machines that can perceive, think, and act in the real world. From industrial automation to home companions, AI robots are entering every sector of the economy and daily life. The market is evolving rapidly. Humanoid AI robots that cost $100,000+ just two years ago now start at $4,900. Companies like Unitree, 1X Technologies, Figure AI, and Tesla are racing to make AI robots as common as smartphones. Whether you're a researcher, business, or consumer, there's likely an AI robot for your needs. Explore our complete humanoid robot guide or browse available AI robots at Robozaps. Compare on RoboZaps: Unitree R1, Figure 03, Tesla Optimus, Atlas and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree R1 vs G1: Which Humanoid Should You Buy? [2026] URL: https://blog.robozaps.com/b/unitree-r1-vs-g1 Author: Radica Maneva Published: 2026-02-10T22:30:00.000Z Updated: 2026-08-01T23:06:57.159Z Category: news TL;DR: The Unitree R1 is 60% cheaper ($5,900 vs $13,500 for base models) and lighter; the G1 counters with better specs — more degrees of freedom (23-43 vs 20-26), longer battery life and 3D LiDAR. Both are orderable now. Choose the R1 for budget, the G1 for capability. Key takeaways: - R1 is 60% cheaper — $5,900 vs $13,500 for base models - Both are orderable now — the R1 launched on pre-order and now ships through Unitree and AliExpress, though some versions are backordered. - G1 has better specs — more DOF (23-43 vs 20-26), longer battery (2hr vs 1hr), 3D LiDAR - R1 is lighter — 25-29kg vs 35kg, more portable - Choose R1 for budget, G1 for capability and immediate availability FAQ: Q: Is the Unitree R1 available now? A: No. The R1 is available for pre-order only. Deliveries begin Q2 2026 (April onwards). The G1 is available for immediate purchase. Q: How much cheaper is the R1 than the G1? A: The R1 ($5,900) is about 56% cheaper than the G1 ($13,500) — a savings of $7,600. Q: Which robot is better for research? A: The G1 EDU. It has superior sensors (3D LiDAR + depth camera), more DOF (up to 43), longer battery life (2 hours), and is available now. The G1 was designed as a research platform. Q: Which robot is better for home use? A: The R1. It's specifically designed for consumer applications with its lighter weight, customizable appearance, and integrated LMM for voice/image interaction. The lower price point also makes it more accessible for home buyers. Q: Can I develop custom software for the R1 or G1? A: Only with the EDU versions. Both the standard R1 and standard G1 do not support secondary development. You'll need to contact Unitree sales for R1 EDU or G1 EDU pricing. Q: Does the R1 have 3D LiDAR like the G1? A: No. The R1 uses mono or binocular cameras only. The G1 includes both a depth camera and 3D LiDAR for superior spatial mapping. Q: Which has better battery life? A: The G1 lasts approximately 2 hours on a single charge. The R1 lasts approximately 1 hour. Both have quick-release battery systems for swapping. The Unitree R1 costs $5,900 while the G1 costs $13,500 — but the G1 is available now with superior specs including 3D LiDAR, longer battery life, and more degrees of freedom. Choose the R1 if budget is your priority; choose the G1 if you need a more capable robot today. Quick Verdict: R1 or G1? Choose Unitree R1 if: Budget is your main constraint ($5,900 vs $13,500)You want a lighter, more portable robot (25-29kg)You're okay waiting until Q2 2026You're a hobbyist or educator with basic needs Choose Unitree G1 if: You need a robot NOW (available today)You need advanced sensors (3D LiDAR + depth camera)You need more DOF and manipulation capabilityYou're doing serious research or development Unitree R1 vs G1: Full Specs Comparison Price Comparison: How Much Do They Cost? Unitree R1 Pricing Unitree G1 Pricing The bottom line: The R1 at $5,900 is less than half the price of the G1 at $13,500. That's a $7,600 difference — enough to buy an additional R1 AIR and still have change left over. For buyers comparing humanoid robot costs, the R1 represents a new price floor for full humanoid robots from a major manufacturer. What's the Difference Between R1 and G1? While both robots come from Unitree, they target different markets: Unitree R1: The Budget Option The R1 is Unitree's push into the consumer market. It's designed to be affordable, lightweight, and approachable — the entry point into humanoid robotics. Target: Consumers, hobbyists, educatorsPhilosophy: "Ultra-lightweight, fully customizable"Trade-offs: Fewer sensors, shorter battery, less DOFAdvantage: Price and portability Unitree G1: The Capable Option The G1 is Unitree's research-grade humanoid. It has better sensors, longer battery life, and more articulation — designed for serious development work. Target: Researchers, developers, universitiesPhilosophy: Full-featured development platformTrade-offs: Higher price, heavierAdvantage: Capability, sensors, available now Sensors & Perception: G1 Has the Edge This is where the G1 clearly outperforms the R1: The G1's 3D LiDAR is a significant advantage for spatial mapping and navigation. Combined with the depth camera, the G1 can build detailed 3D maps of its environment — critical for research applications and autonomous operation in complex spaces. The R1's simpler camera system is adequate for basic tasks and LMM-powered interactions, but won't match the G1's spatial awareness. Battery & Runtime: G1 Lasts Twice as Long R1: ~1 hour battery lifeG1: ~2 hours battery life (9000mAh) The G1's double battery life is a meaningful advantage for extended operation. Both robots feature quick-release battery systems for easy swapping, but the G1's larger capacity means fewer interruptions. Size & Portability: R1 is More Manageable R1: 123cm tall, 25-29kg — one person can move itG1: 132cm tall, 35kg — manageable but heavier The R1's lighter weight (up to 10kg less) makes it easier to transport, reposition, and work with. For classroom environments or home use, this portability advantage matters. The G1 can fold to 690x450x300mm for transport, but at 35kg it's still a two-person lift for most situations. Development & Programming: Both Require EDU Versions Both the R1 and G1 restrict secondary development to their EDU editions: R1 / R1 AIR: No development accessR1 EDU: Full development support, optional Jetson OrinG1: No development accessG1 EDU: Full development support, includes Jetson Orin option If you plan to develop custom software, you'll need to contact Unitree sales for EDU pricing on either model. Availability: G1 Wins by Months G1: Available for purchase NOWR1: Pre-order only, ships Q2 2026 (April onwards) If you need a humanoid robot today, the G1 is your only option from Unitree. The R1 won't ship until at least April 2026. Who Should Buy the Unitree R1? The R1 is ideal if: Budget is primary concern — $5,900 vs $13,500 is a significant differenceYou're new to humanoid robotics — lower investment to learnPortability matters — lighter weight for classrooms or demosYou can wait — Q2 2026 delivery is acceptableBasic tasks are sufficient — household chores, simple interactions Who Should Buy the Unitree G1? The G1 is ideal if: You need it now — available for immediate purchaseAdvanced sensors required — 3D LiDAR + depth camera for researchExtended operation needed — 2-hour battery vs 1-hourResearch/development focus — serious academic or commercial workBudget isn't the constraint — $13,500 is acceptable Final Verdict: R1 for Budget, G1 for Capability This isn't a case of one robot being "better" — they serve different needs: Choose the R1 if: You want the most affordable humanoid robot from a reputable manufacturer, you can wait for Q2 2026 delivery, and your needs are consumer/educational rather than research-intensive. Choose the G1 if: You need a robot now, you require advanced sensors for serious development work, and the $13,500 price is within your budget. For most hobbyists and educators, the R1's price-to-capability ratio makes it the smarter choice. For researchers and developers who need the best sensors and can't wait, the G1 justifies its premium. Ready to buy? Order the Unitree R1 or Unitree G1 at Robozaps. Compare on RoboZaps: Unitree R1 and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree R1 Review 2026: $4,900 AIR vs $5,900 R1 URL: https://blog.robozaps.com/b/unitree-r1-review Author: Radica Maneva Published: 2026-02-10T00:00:00.000Z Updated: 2026-08-01T10:53:12.310Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: The Unitree R1 costs $4,900 (Air) or $5,900 direct from Unitree, both in stock, and neither can be programmed or grasp objects. The overlooked option is the R1 EDU at $10,500 on Unitree's own shop: Unitree's cheapest programmable humanoid, $3,000 under a base G1. AliExpress lists it globally from ~$6,800, with first deliveries scheduled from June 30, 2026. Key takeaways: - $4,900 (Air) and $5,900 (standard) buy a real walking, flipping humanoid that cannot be programmed and cannot grasp: fixed fists, no SDK. - The buried headline is the R1 EDU at $10,500, in stock on Unitree's own shop despite the 'Contact Sales' label: Unitree's cheapest programmable humanoid, $3,000 under a base G1. - Global buying changed in 2026: official AliExpress listings from April (~$6,800 observed), first deliveries scheduled June 30, US dealer stock at $8,990. - The viral stunts are operator-triggered RL motion routines, not autonomy, and Unitree's UniStore app store does not support the R1 yet. - Researchers named the R1 among models carrying the UniPwn Bluetooth root vulnerability; isolate it on a filtered network like any Unitree robot. FAQ: Q: How much does the Unitree R1 cost in 2026? A: Direct from Unitree: $4,900 for the R1 Air, $5,900 for the standard R1, and $10,500 for the programmable R1 EDU, all in stock at our July 22, 2026 check, plus $300-$1,200 shipping and import duties. Unitree's official AliExpress store showed listing prices around $6,802-$8,123 at launch (variant mapping unclear), and US dealer RoboStore lists domestic stock from $8,990. Q: Can you program the Unitree R1? A: Not the Air or standard R1: Unitree's store states they "do not support secondary development." Programming requires the R1 EDU ($10,500 on Unitree's shop), which enables SDK access, optional NVIDIA Orin compute and optional dexterous hands. Q: What can the Unitree R1 do out of the box? A: Walk, run, balance, respond to voice and vision, and perform built-in motion routines from the included controller, with no calibration needed. The acrobatics you have seen online are operator-triggered reinforcement-learning routines, not autonomous behavior, and the base tiers cannot grasp objects. Q: Can the Unitree R1 pick things up? A: No. The Air and standard R1 ship with fixed, non-articulated fists. Dexterous hands are an option only on the R1 EDU. For manipulation on a budget, Unitree's dual-arm R1-D (announced from $4,290) targets exactly that gap. Q: How long does the Unitree R1 battery last? A: About one hour, per Unitree's official spec, from a 6,000 mAh (199 Wh) quick-release battery. Spares are $500 from Unitree's store. That is roughly half a G1's endurance. Q: How tall is the Unitree R1? A: 123 cm (1,230 mm), weighing about 27 kg for the Air and 29 kg for the standard and EDU with battery, per Unitree's current official specs. The 121 cm / 25 kg figures in many articles are the July 2025 launch specs. Q: Where can I buy a Unitree R1? A: Three legitimate channels: Unitree's own shop (cheapest sticker, you handle freight and duties, no returns), Unitree's official AliExpress storefront (serving North America, Europe, Japan and Singapore since April 2026), or US dealer RoboStore ($8,990 with domestic stock and service). Q: Should I buy the R1 Air or the standard R1? A: The $1,000 step to the standard R1 buys six more joints (including waist and head articulation), a binocular camera instead of monocular, and two extra months of warranty. For anything beyond casual demos, the standard R1 is the better machine; for programming, only the EDU matters. Q: Is the Unitree R1 better than the G1? A: It is cheaper, not better. Against the G1 it gives up LiDAR and the depth camera, two joints per arm, roughly half the battery life, and dexterous-hand options below EDU tier. The R1 EDU at $10,500 undercuts the unprogrammable base G1, which is exactly its appeal for budget developers. Q: Is the Unitree R1 secure? A: Treat it cautiously. The researchers behind the UniPwn disclosure name the R1 among units sharing the Bluetooth provisioning vulnerability that allows root takeover in radio range, and Unitree has not published a model-by-model advisory. Standard lab practice applies: Bluetooth off when unused, isolated network. Q: Is the Unitree R1 banned in the US? A: No. The Pentagon added Unitree to its Section 1260H list in June 2026, which restricts Department of Defense procurement. It is not an import or consumer purchase ban, though defense-funded institutions should review procurement eligibility. Q: Can the R1 use Unitree's robot app store? A: Not yet. UniStore, Unitree's motion-app store (opened for the G1 in late 2025 and expanded in May 2026), supports the G1, H1, B2 and Go2. R1 support has not been announced. How much does the Unitree R1 cost, and can you actually buy one? The Unitree R1 costs $4,900 (R1 Air) or $5,900 (standard R1) direct from Unitree's own store, both orderable today, and neither one can be programmed. The version that can, the R1 EDU, is the buried headline: Unitree's shop sells it at $10,500, in stock, $3,000 cheaper than a base G1, even though Unitree's own product page still says "Contact Sales" for it. That makes the R1 EDU Unitree's cheapest programmable humanoid, and almost nobody reports it. (The one cheaper programmable rival is smaller: Booster's 95 cm K1 at $5,999.) Since April 2026 the R1 also sells through Unitree's official AliExpress storefront, with first deliveries scheduled from June 30, 2026. We verified every price on this page against Unitree's live store data on July 22, 2026. Air vs R1 vs EDU: the only decision that matters The three tiers look alike and are not. The differences below are Unitree's own numbers, from the official comparison table. Read that table bottom-up. The Air gives up six joints (no waist or neck articulation), an eye, and two months of warranty to save $1,000. The standard R1 is the better demo machine. And if the word "programmable" appears anywhere in your plans, the only tier that matters is the EDU. The $4,900 catch: app control is not an SDK Unitree's store page states it without ambiguity: the R1 Air and standard R1 "do not support secondary development." No SDK, no code deployment; what you get is remote control, voice and vision interaction, and the built-in motion library. Confusingly, Unitree's marketing page for the same robot advertises "fully open control interfaces for joints and sensors." Both sentences are on official Unitree pages; only one of them describes the product you receive at $4,900. When The Robot Report covered the launch, Unitree itself was candid about the gap between perception and reality: "public perception of robots — e.g., believing robots can immediately perform household chores — has a gap with the actual state of robot development." The EDU tier is a genuinely different machine: secondary development enabled, optional Orin compute, optional dexterous hands, and first-class support in Unitree's open tooling. The R1 has its own training task in Unitree's official reinforcement-learning repo alongside the G1 and H2. At $10,500 that is a real research platform at a third of a G1 EDU U1's price, with real trade-offs covered below. The fists problem The most-quoted criticism from the R1's viral wave, "it cannot pick up a cup," is accurate. The Air and standard R1 ship with fixed, non-articulated fists; Unitree's own spec table lists dexterous hands as "None" on both tiers and optional only on the EDU. An R1 can run, cartwheel, box and chat. It cannot grasp anything. If manipulation is the point, either budget for an EDU with hands, or look at Unitree's own R1-D, a dual-arm data-acquisition derivative announced from $4,290 in April 2026 that trades legs for graspers (availability is still limited). Unitree R1 specs: the ultra-lightweight pitch, reconciled Unitree markets the R1 with one word, "ultra-lightweight," and the spec drift since launch has confused most coverage. The numbers below are the current official figures, with the launch-era variants flagged. Full record: Unitree R1 in the RoboZaps robot database. What a delivered R1 actually does Out of the box, per Unitree: walk, run, balance, respond to voice and vision through its onboard model, and perform built-in motions from the included controller, with no calibration required. The viral gymnastics deserve the same decoding as the G1's. The cartwheels, handstands and kung-fu routines are real hardware running trained reinforcement-learning motion policies, but The Robot Report put it plainly at launch: unlike the autonomy-first humanoids, "Unitree's R1 is remote-controlled." The stunts are operator-triggered performances, not decisions. One 2026 development matters here. Unitree's UniStore, a motion-app store where owners tap to install packaged moves (Mantis boxing, a Michael Jackson routine, a Dragon Ball power-up), opened for the G1 in late 2025 and expanded in May 2026 across the G1, H1, B2 and Go2. The R1 is not on that list, so the robot pitched as the affordable fun machine currently cannot download the fun. Whether R1 support arrives is one of the open questions we're watching. And an honesty note most R1 coverage skips: international sales only opened in April 2026, with deliveries scheduled from June 30, so there are as yet no substantial long-term owner reviews anywhere. Nobody outside Unitree and early Chinese buyers knows how the R1 holds up over months of falls, heat and battery cycles. Every site claiming otherwise is extrapolating from spec sheets. We'll update this page as real ownership data lands. What owning one actually costs The practical math for a US buyer: Unitree-direct looks cheapest until freight and duties land; AliExpress pre-bakes the import cost; RoboStore's $8,990 buys domestic stock, service and no customs paperwork. Three legitimate channels, three honest totals, and which is cheapest depends on how much of the import risk you want to own. None of that math is disturbed by the FCC action: the R1 was authorized on 22 June 2026, five weeks before foreign-made robots joined the Covered List. See humanoid robots legal in the US for where every other model stands. Security and the Pentagon list: what applies to the R1 Our G1 review covers Unitree's 2025–2026 security record in depth; here is what carries over to the R1, precisely. The UniPwn disclosure (September 2025) demonstrated root takeover over Bluetooth using hardcoded keys shared across Unitree's fleet. The repo's affected list names the Go2, G1, H1 and B2, not the R1 — but the underlying research paper explicitly states the vulnerable provisioning interface is "shared across all G1, H1, and R1 units." In other words, the R1 is named by the researchers even where headline coverage missed it, and no Unitree advisory has clarified the exact model scope. The practical guidance is the same as for the G1: treat it as a managed device, keep Bluetooth off when unused, and isolate it on a filtered network. Separately, the US Department of Defense added Unitree to its Section 1260H list in June 2026 (WilmerHale's explainer) — a DoD procurement restriction, not a consumer purchase ban, but relevant if your school or lab touches defense funding. R1 vs the field: what $5,900 does and doesn't buy Two boundary notes. RoboCup's eligible-platform list for the merged humanoid soccer league names the G1 and H1, not the R1, so competition teams should check eligibility before buying the cheaper robot. And Booster's K1 is the one machine in this class that's programmable at R1-Air money, at the cost of size. Our R1 vs G1 comparison and the cheapest-humanoids ranking go deeper on both. Misinformation audit: eight R1 claims, checked The R1's 2026, dated The company behind it, briefly The R1 is the sharp end of Unitree's volume strategy: the company shipped more than 5,500 humanoid robots in 2025, which its IPO prospectus claims made it the world's largest humanoid maker by units, and its average humanoid selling price collapsed from ¥593,400 in 2023 to ¥167,600 in the first nine months of 2025 (prospectus) as the G1 and R1 took over the mix. China's regulator approved Unitree's Shanghai STAR Market IPO on July 2, 2026 (Caixin); as of July 22 it had not yet priced or begun trading. TIME named the R1 one of its Best Inventions of 2025. Our G1 review carries the full company and financials treatment. Who should buy an R1, and which one Buy the R1 or R1 Air if you want the cheapest real, walking, flipping humanoid from a major maker for content, events, education showcases or sheer curiosity, and you accept that it will never run your code and cannot hold objects. Buy the R1 EDU if you want to program a full-size-adjacent humanoid on the smallest defensible budget: at $10,500 it is Unitree's cheapest SDK-enabled robot and undercuts the unprogrammable base G1 by $3,000, provided you can live without LiDAR and with an hour of battery. Look elsewhere if you need manipulation (R1-D or a G1 EDU with hands), competition eligibility (G1, or Booster K1 for smaller leagues), or any autonomous usefulness around the house, which nothing at this price delivers in 2026; our buyable-humanoids guide is the wider reality check. What to watch next Four dated things will change this page. First owner reviews from the initial delivery wave should surface through August, and they'll answer the durability questions no spec sheet can. UniStore either adds R1 support or the "fun robot" stays locked out of the app store. Unitree's IPO prices within weeks, with the R1's volume economics at the center of the story. And the R1-D's availability will decide whether the manipulation-first buyer at this price point stays in the Unitree family. We re-verify prices on this page against Unitree's live store data; every number above carries its check date. The bottom line The R1 is the most important cheap robot ever sold and a machine most buyers misunderstand in both directions. It is not a toy: it's a real dynamic biped with genuine RL locomotion, from the company that ships more humanoids than anyone. It is also not the $4,900 programmable robot butler the viral posts implied: at the base tiers it can't grasp, can't be coded, and runs an hour per battery. The quiet story is the $10,500 EDU, the cheapest ticket into Unitree's humanoid development stack in 2026. Buy the tier that matches your honesty about what you'll do with it. --- # Fourier GR-1 Price 2026: Specs & What Replaced It URL: https://blog.robozaps.com/b/fourier-gr1-review Author: Radica Maneva Published: 2026-02-10T00:00:00.000Z Updated: 2026-08-01T23:07:11.262Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The Fourier GR-1 is a 40-DoF research humanoid with no public price, supplied quote-only to universities and institutions. It has been superseded: Fourier's current flagship is the GR-3 care robot (August 2025), after the GR-2, and the GR-1 no longer appears in Fourier's own lineup. It remains relevant as the platform that started Fourier's humanoid line. Key takeaways: - No public price: the GR-1 was supplied quote-only to universities, AI companies and research institutions. The $149,000+ figures online are aggregator estimates, not Fourier numbers. - Superseded: Fourier's current flagship is the GR-3 care robot (August 2025), after the GR-2 (September 2024). The GR-1 no longer appears in Fourier's own lineup. - Official specs: 1.65 m, 55 kg, 40 degrees of freedom, up to 5 km/h walking and 300 Nm peak hip torque; Fourier claimed load handling up to 50 kg. - Heritage: built by a rehabilitation-robotics company founded in 2015, which shows in the human-safe, force-sensing design. - Best for: labs sourcing one second-hand or through distributors, as a first-generation research platform rather than a current product. FAQ: Q: How much does the Fourier GR-1 cost? A: Pricing is not publicly disclosed. The GR-1 is available through limited R&D deployment to universities and AI companies. Interested research partners should contact Fourier Intelligence directly for pricing information. Q: Can I buy a Fourier GR-1 for commercial use? A: No. The GR-1 is not commercially available for general purchase. Fourier offers limited deployment to select research institutions, universities, and AI companies for R&D purposes. Q: When was the GR-1 released? A: The GR-1 was unveiled in July 2023 at the World AI Conference in Shanghai. It is not "released" in the traditional commercial sense—availability is limited to strategic R&D partnerships. Q: What is the GR-1's payload capacity? A: Fourier said at the GR-1's unveiling that it "will be able to handle loads up to 50 kg," and The Robot Report lists a 50 kg payload for the production version. The newer GR-2 specifies a more conservative 3 kg per arm. Q: Is the GR-1 a healthcare robot or general-purpose robot? A: The GR-1 is officially described as a "bipedal, general-purpose bionic robot" despite Fourier's strong rehabilitation robotics background. It's designed as a versatile research platform rather than a healthcare-exclusive device. Q: Does the GR-1 support ROS? A: Yes. The GR-1 is ROS compatible, allowing integration with existing robotics research infrastructure and open-source libraries. Q: Can the GR-1 run? A: Running capability has not been demonstrated or specified. The verified walking speed is up to 5 km/h (3.1 mph). Q: What's the difference between GR-1 and GR-2? A: The GR-2 (September 2024) is taller at 175 cm with 53 degrees of freedom, a 3 kg per-arm payload and a swappable battery. Fourier's current flagship is the GR-3 care robot (August 2025). The GR-1 no longer appears in Fourier's own lineup. Q: Is Fourier Intelligence a startup? A: No. Fourier Intelligence was founded in 2015 and has nearly a decade of experience in rehabilitation robotics and exoskeletons before entering the humanoid robot market with the GR-1. Q: Where can I learn more about the GR-1? A: Visit Fourier's official website (fftai.com) or contact their research partnership team directly. Technical documentation and SDK access are provided to approved research partners. The Fourier Intelligence GR-1 is a general-purpose bipedal humanoid robot developed by a company with deep roots in rehabilitation robotics. Unveiled in July 2023 at the World AI Conference in Shanghai, the GR-1 (no public price; supplied quote-only) represents Fourier's entry into humanoid robotics—bringing their expertise in medical-grade engineering to a developer-focused platform. While Fourier markets the GR-1 as their "first mass-produced humanoid robot," availability is currently limited to R&D deployment with select universities and AI companies. This positions it as a research platform rather than a commercially available product for general sale. Fourier Intelligence GR-1 Overview Fourier Intelligence was founded in 2015 in Shanghai, initially focusing on rehabilitation exoskeletons and medical robotics. Their background includes products like the ExoMotus lower-limb exoskeleton and ArmMotus upper-limb rehabilitation systems—giving them years of experience in building human-safe, precision-engineered robotic systems. The GR-1, unveiled in July 2023, applies this rehabilitation engineering expertise to a general-purpose humanoid platform. Unlike pure healthcare robots, the GR-1 is designed as a versatile research platform with an emphasis on SDK support and AI development capabilities. Company Background: Rehabilitation Robotics Heritage Fourier's rehabilitation background distinguishes them from many humanoid startups. Their existing medical robotics products require: Human-safe design: Force sensing and gentle interaction controlsPrecision engineering: Reliable performance in clinical settingsRegulatory experience: Understanding of medical device complianceClinical relationships: Connections with healthcare institutions and researchers This foundation carries over to the GR-1, even as it targets broader research applications beyond healthcare. Fourier GR-1 Price & Availability The GR-1 is not commercially available for general purchase. Fourier offers limited R&D deployment to universities, AI companies, and research institutions. Pricing is not publicly disclosed—interested research partners should contact Fourier Intelligence directly. As of mid-2026 the GR-1 no longer appears in Fourier's own product lineup: the company's current flagship is the GR-3 care robot (launched August 2025, priced above ¥200,000 for institutions), which followed the GR-2 (September 2024). Third-party distributors still list the GR-1 for quote or lease. This limited availability model is common in early-stage humanoid robotics, where companies prioritize strategic partnerships with research institutions over broad commercial sales. GR-1 Full Specifications Many technical details are not publicly disclosed. The verified specifications above come from official Fourier presentations and press releases. Researchers considering the GR-1 should contact Fourier directly for complete technical documentation. 40 Degrees of Freedom: Articulation & Movement The GR-1's 40 DOF configuration enables human-like movement fidelity across its full body: Upper body: Highly articulated arms and hands for dexterous manipulationLower body: Stable bipedal locomotion at up to 5 km/h walking speedCore: Torso articulation for natural reaching, bending, and balance adjustment This level of articulation positions the GR-1 competitively against other research-focused humanoids. For comparison: Use Cases & Applications AI & Robotics Research The GR-1's primary audience is AI researchers and robotics developers. The platform supports: Embodied AI research: Testing AI models in physical humanoid formHuman-robot interaction studies: Social robotics and collaboration researchManipulation research: Dexterous task learning and executionLocomotion research: Bipedal walking, balance, and navigation algorithms University Research Programs Academic institutions can deploy the GR-1 for: Graduate research projects in robotics and AIMulti-robot system studiesBenchmark testing of novel algorithmsCross-disciplinary research (robotics + psychology, design, etc.) Potential Healthcare Applications While the GR-1 is a general-purpose platform (not a medical device), Fourier's rehabilitation background suggests potential future healthcare applications: Rehabilitation exercise assistance (research phase)Patient mobility support studiesHealthcare human-robot interaction researchTherapeutic companionship studies for elderly care The GR-1 is not currently marketed as a medical device. Any healthcare applications would be research/experimental and require appropriate regulatory approvals for clinical deployment. Industrial & Service Research As a general-purpose platform, the GR-1 can be adapted for research in: Warehouse and logistics automation conceptsService robotics (hospitality, retail)Manufacturing assistance and quality controlInspection and monitoring applications Software, SDK & AI Capabilities Fourier provides a proprietary SDK with ROS (Robot Operating System) compatibility, making the GR-1 accessible to the broader robotics research community. Developer Features ROS Compatibility: Integration with existing robotics infrastructure and librariesVision System: Camera array for environment awareness and object recognitionForce-Torque Sensing: Safe interaction controls inherited from rehabilitation robotics experienceNavigation Stack: Autonomous movement and obstacle avoidanceManipulation Framework: Dexterous control APIs for task execution The ROS compatibility is particularly significant for research institutions, allowing developers to leverage existing open-source robotics tools and algorithms rather than starting from scratch with proprietary systems. Pros and Cons Pros 40 DOF: Highly articulated for complex movementsRehabilitation heritage: Engineering expertise in human-safe roboticsROS compatible: Developer-friendly platformGeneral-purpose design: Versatile research applicationsEstablished company: Fourier has 9+ years in robotics (not a startup)SDK support: Dedicated development tools and documentationVerified specs: Core measurements (height, weight, DOF, speed) confirmed Cons Limited availability: Not commercially available to general marketUndisclosed pricing: No transparent pricing for budget planningSparse public specs: Many technical details not disclosedChina-based: May face regulatory scrutiny in some marketsUnverified claims: Some marketing claims (e.g., "first mass-produced") lack independent verificationLimited track record: First humanoid product from a medical robotics companyR&D-only access: Not suitable for commercial deployment yet GR-1 vs Competitor Humanoid Robots Humanoid robot pricing is often confidential and varies by configuration, support packages, and partnership terms. Listed prices are estimates based on publicly available information and may not reflect actual purchase costs. How the GR-1 Compares vs. Unitree H1: The H1 offers greater transparency (known pricing, availability) and athletic performance focus. The GR-1 has higher DOF (44 vs 19) but less clear specifications and availability. vs. Unitree G1: The G1 is more accessible for education and small-scale research at ~$13,500 with commercial availability. The GR-1 targets more advanced research with higher DOF but limited access. vs. Agility Digit: Digit is purpose-built for warehouse logistics with proven commercial deployments. The GR-1 is a general-purpose research platform without the logistics-specific optimizations. Final Verdict Who Should Consider the GR-1? The GR-1 is right for you if: You're a university or research institution with robotics/AI programsYou need a highly articulated (40 DOF) humanoid for researchROS compatibility is important for your existing infrastructureYou value a supplier with rehabilitation robotics engineering heritageYou can work within limited R&D deployment constraints Look elsewhere if: You need transparent pricing and commercial availability—consider Unitree G1 or H1You're seeking immediate commercial deployment—the GR-1 is R&D-focusedYou need warehouse/logistics optimization—Agility Digit is purpose-built for thisYou require complete public technical specifications before evaluationBudget transparency is critical for your procurement process The Bottom Line The Fourier Intelligence GR-1 represents an interesting entry into humanoid robotics from a company with proven rehabilitation engineering experience. Its 40 DOF configuration and general-purpose design make it a capable research platform, while ROS compatibility ensures integration with existing robotics infrastructure. However, limited availability, undisclosed pricing, and sparse public specifications make it challenging to evaluate compared to more transparent competitors. Researchers considering the GR-1 should engage directly with Fourier to obtain complete technical documentation and partnership terms. The GR-1 is neither the cheapest (Unitree G1 at ~$16K) nor the most commercially available (Unitree H1) humanoid research platform. Its value proposition rests on Fourier's rehabilitation robotics expertise and the potential for human-safe interaction controls—assets that may matter significantly for certain research applications, particularly those exploring healthcare or human-robot collaboration. Recommendation: If you're a research institution exploring humanoid platforms and value engineering heritage from medical robotics, request detailed technical specs and pricing from Fourier. Compare against Unitree's more accessible offerings before making a decision. The GR-1 may offer unique advantages for specific research directions, but transparency-seeking buyers will find better documentation with commercial alternatives. Where to learn more: Fourier Intelligence Official Website Last updated: February 2026 Fact-checked: All specifications verified against official Fourier sources. Unverified claims removed or clearly marked as marketing statements. Compare on RoboZaps: Gr 1, Digit, Unitree H1 and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # This Week in Humanoid Robots: February 3-8, 2026 URL: https://blog.robozaps.com/b/humanoid-robot-news-week-feb-8-2026 Author: Radica Maneva Published: 2026-02-08T00:00:00.000Z Category: news FAQ: Q: How many humanoid robots has China sold compared to the US? A: China's Unitree and AgiBot each sold more humanoid robots than Tesla's entire 2025 target of 5,000 units. Only three non-Chinese companies (Figure AI, Agility Robotics, Tesla) made Omdia's top-seller chart—each with roughly 150 units sold. Q: When will Unitree go public? A: Unitree's IPO is expected mid-2026, according to Caixin Global. This would make it the first major humanoid robotics company to go public, providing market validation and pricing transparency for the entire sector. Q: What is Figure 03's Helix AI? A: Helix is the first Vision-Language-Action (VLA) model that controls a humanoid robot's full upper body—including individual fingers, wrists, torso, and head. It enables Figure 03 to perform complex domestic tasks in unpredictable home environments. Q: How much did FieldAI raise for embodied AI? A: FieldAI raised $405 million for its embodied AI robotics platform—one of the largest funding rounds in robotics history. The company focuses on AI/software infrastructure rather than hardware manufacturing. Q: What's the humanoid robot market growth forecast? A: The market is projected to grow from $3.14 billion (2025) to $13.25 billion (2029)—a 45.5% CAGR. Long-term projections reach $251 billion by 2035 and $5 trillion by 2050. Q: Why is China dominating humanoid robotics? A: China has 150+ humanoid robot companies versus roughly 20 in the US. Chinese manufacturers are shipping at scale with competitive pricing while US companies are still in prototype or limited production phases. Government subsidies and manufacturing expertise accelerate their lead. By Dean Fankhauser | February 8, 2026 This week, China's humanoid robot dominance became undeniable: Unitree and AgiBot each outsold Tesla's entire 2025 target of 5,000 units. Meanwhile, Figure unveiled its home-focused Figure 03, Mobileye acquired Mentee Robotics, and FieldAI raised $405M for embodied AI. The industry is no longer coming—it's here. Key Takeaways China dominates: Two Chinese companies outsold Tesla's entire 2025 humanoid robot target—China now has 150+ robotics companies vs ~20 in the USUnitree IPO incoming: Expected mid-2026, marking the first major humanoid robotics company to go publicFigure 03 targets homes: New Helix AI enables full upper-body control for domestic tasksBig money flows in: FieldAI's $405M round signals massive VC confidence in embodied AI How Did China Beat Tesla in Humanoid Robot Sales? The headline of the week came from Rest of World: Unitree and AgiBot each sold more humanoid robots than Tesla's entire 2025 target of 5,000 units. Let that sink in. Two Chinese companies—neither of which has Tesla's brand recognition or resources—outpaced Elon Musk's robotics ambitions. According to Omdia's top-seller chart, only three non-Chinese companies appear: Figure AI, Agility Robotics, and Tesla (each with roughly 150 units). China now has over 150 humanoid robot companies versus roughly 20 in the US. They're installing 10x more robots annually. Our take: We've been tracking the best humanoid robots for years. The shift from US innovation leadership to Chinese production dominance mirrors EVs, solar panels, and consumer drones. Companies like Unitree aren't just cheaper—they're shipping at scale while others announce. When Is Unitree's IPO Expected? Caixin Global reported that Unitree is preparing for what would be the first major humanoid robotics company IPO, expected around mid-2026. The filing reflects "soaring investor interest" in embodied intelligence. Chinese AI startups are rushing to public markets, and Unitree—with proven sales numbers and a diverse product lineup—is leading the charge. Why it matters: A public Unitree means more stable supply chains, increased pricing transparency, and market legitimacy for the entire sector. IPO filings will reveal production costs and margins we've only been guessing at. Our take: This validates what we've been saying: commercial humanoid robots aren't vaporware anymore. Monitor Q2 2026 for filing details. What Makes Figure 03 Different from Previous Models? Figure AI unveiled the Figure 03, a general-purpose humanoid designed specifically for home environments. The star is Helix—the first Vision-Language-Action model to control a robot's full upper body, including individual fingers, wrists, torso, and head. Unlike industrial robots built for predictable factory floors, Figure 03 is designed for the chaos of real homes: cluttered counters, unpredictable layouts, curious pets. Why it matters: Home humanoids are the holy grail. Factory applications prove the tech; home applications prove the market. Figure is betting big on consumer-facing robotics while competitors focus on enterprise. Our take: The Figure lineup has been impressive in demos. The question is whether Helix's real-world performance matches the controlled videos. We're watching closely. Why Did Mobileye Acquire Mentee Robotics? At CES 2026, Intel subsidiary Mobileye announced the acquisition of Mentee Robotics—bringing an automotive tech giant into the humanoid space. Mentee's third-generation humanoid is vertically integrated and AI-first, designed from the ground up for advanced autonomy. Autonomous driving and humanoid robotics share core technologies: perception, planning, real-time decision-making. Why it matters: Mobileye brings billions in R&D and proven ADAS deployment experience. Watch for a competitor product by mid-2027. The automotive industry has the manufacturing capacity to scale robots faster than pure robotics startups. What Does FieldAI's $405M Funding Mean for the Industry? FieldAI secured $405 million for its embodied AI robotics platform—one of the largest funding rounds in robotics history. The focus: software and AI infrastructure for physical robots, not hardware. FieldAI is betting that the AI/software layer is where the real value lies. Why it matters: This signals VCs believe humanoid AI infrastructure is the next platform opportunity, similar to how cloud computing infrastructure became more valuable than individual applications. Our take: The industry is bifurcating into hardware manufacturers and AI platform providers. Future competition may be more about software than servos. Explore our complete market analysis for deeper insights. What Should You Watch Next Week? Unitree IPO filings: Any Q2 2026 timeline hints or filing preparationsFigure 03 hands-on reviews: First independent assessments of Helix in home environmentsChinese government policy: Robot sports competitions and industry subsidies The humanoid robot industry isn't coming—it's here. The question is no longer "if" but "who" and "how fast." Frequently Asked Questions How many humanoid robots has China sold compared to the US? China's Unitree and AgiBot each sold more humanoid robots than Tesla's entire 2025 target of 5,000 units. Only three non-Chinese companies (Figure AI, Agility Robotics, Tesla) made Omdia's top-seller chart—each with roughly 150 units sold. When will Unitree go public? Unitree's IPO is expected mid-2026, according to Caixin Global. This would make it the first major humanoid robotics company to go public, providing market validation and pricing transparency for the entire sector. What is Figure 03's Helix AI? Helix is the first Vision-Language-Action (VLA) model that controls a humanoid robot's full upper body—including individual fingers, wrists, torso, and head. It enables Figure 03 to perform complex domestic tasks in unpredictable home environments. How much did FieldAI raise for embodied AI? FieldAI raised $405 million for its embodied AI robotics platform—one of the largest funding rounds in robotics history. The company focuses on AI/software infrastructure rather than hardware manufacturing. What's the humanoid robot market growth forecast? The market is projected to grow from $3.14 billion (2025) to $13.25 billion (2029)—a 45.5% CAGR. Long-term projections reach $251 billion by 2035 and $5 trillion by 2050. Why is China dominating humanoid robotics? China has 150+ humanoid robot companies versus roughly 20 in the US. Chinese manufacturers are shipping at scale with competitive pricing while US companies are still in prototype or limited production phases. Government subsidies and manufacturing expertise accelerate their lead. Stay ahead of the curve. Browse humanoid robots for sale or explore our complete buying guide. --- # Can Humanoid Robots Actually Do Your Laundry? A Practical Guide to Home Robot Capabilities in 2026 URL: https://blog.robozaps.com/b/can-humanoid-robots-do-laundry-2026-guide Author: Radica Maneva Published: 2026-02-04T00:00:00.000Z Updated: 2026-08-01T23:06:48.512Z Category: best TL;DR: Not really, yet: robots can fold basic items and load machines, but none handles a complete laundry workflow. The LG CLOiD has the best demos, the $20,000 1X NEO is the consumer option, and 3-5 hour battery life plus delicate fabrics remain hard limits. Realistic timeline: basic assistance 2026-2027, full automation closer to 2030-2035. Key takeaways: - Current state: Robots can fold basic items and load/unload machines, but not handle complete laundry workflows - Top performers: LG CLOiD (best laundry demos), 1X NEO ($20,000 consumer option), Figure 03 (general household) - Key limitations: Delicate fabrics, complex garments, stain treatment, 3-5 hour battery life - Cost reality: $20,000+ for consumer robots—time saved doesn't justify expense for most households yet - Timeline: Basic laundry assistance 2026-2027; moderate independence 2028-2030; full automation 2030-2035 FAQ: Q: Can humanoid robots do laundry better than humans? A: No, not yet. Current humanoid robots can handle basic laundry tasks like folding simple garments and transferring clothes between machines, but they lack the dexterity and judgment needed for complete laundry care. Delicate fabrics, complex garments, and stain treatment remain beyond their capabilities. Q: How much does a humanoid robot for laundry cost in 2026? A: Consumer-ready humanoid robots range from $20,000 (1X NEO) to over $250,000 for industrial models. The 1X NEO offers a $499/month subscription option, while LG CLOiD hasn't announced consumer pricing yet. Q: Are humanoid robots safe around children and pets? A: Safety remains a concern with current models weighing 60-130+ pounds. The 1X NEO operates at whisper-quiet 22dB levels and uses soft actuators for gentler movement, but manufacturers recommend supervision around children and pets. Q: What household tasks can humanoid robots do besides laundry? A: Current models can handle basic cleaning, dishwasher loading, simple food preparation, object retrieval, and organization tasks. However, most require structured environments and specific training for each task. Q: When will humanoid robots reliably handle all laundry tasks? A: Full laundry automation is likely 5-10 years away. While basic folding and machine operation may become reliable by 2028-2030, complex fabric care and end-to-end laundry management will require significant advances in AI and dexterity. Q: Do I need special appliances for humanoid robots to do laundry? A: Some robots work better with specific appliances. LG CLOiD integrates seamlessly with LG's ThinQ smart home ecosystem, while others may require appliances with specific height, handle types, or control interfaces. Q: How long can humanoid robots work before needing to recharge? A: Most current models operate for 3-5 hours before requiring 2-4 hour charging periods. This limitation significantly impacts their utility for all-day household assistance. Picture this: you come home from a long day at work to find your laundry neatly folded, dishes put away, and countertops spotless—all done by your personal humanoid robot assistant. This sci-fi dream is closer to reality than ever before in 2026, but the question remains: can these robots actually handle your laundry reliably? No, humanoid robots cannot reliably do your complete laundry in 2026. They can fold simple garments like t-shirts and towels on flat surfaces and transfer clothes between machines, but delicate fabrics, complex folding, and stain treatment remain beyond current capabilities. Full laundry automation is realistically 5-10 years away. After analyzing the latest humanoid robots from leading manufacturers and testing real-world capabilities, we'll give you the honest truth about what these machines can and can't do right now. Whether you're considering investing in a humanoid robot for home use or just curious about the technology, this guide covers everything you need to know. What Can Humanoid Robots Actually Do With Laundry in 2026? The reality of humanoid robot laundry capabilities in 2026 is more nuanced than glossy marketing videos suggest. Current robots can handle basic tasks but struggle with the complexity of real-world laundry. What Laundry Tasks Can Robots Do Today? Folding Simple Garments: Most advanced humanoid robots can fold basic items like t-shirts, towels, and simple pants. The Figure 03 demonstrates impressive precision when folding pre-sorted, clean laundry on flat surfaces. Sorting by Color and Type: Using computer vision, robots can distinguish between dark and light fabrics, though complex patterns or mixed materials still present challenges. Loading and Unloading Machines: Several models, including the 1X NEO and LG CLOiD, can transfer clothes between washer and dryer, though they require specific machine types and careful programming. Hanging Basic Items: Simple items like shirts and pants can be hung on standard hangers, though delicate fabrics or complex garments remain problematic. What Laundry Tasks Can Robots NOT Do Reliably? Delicate Fabric Handling: Fine fabrics, silk, or items requiring special care are beyond current robot capabilities. The dexterity required to handle lace, beading, or fragile materials safely isn't there yet. Complex Folding: Fitted sheets, bras, or irregularly shaped garments present significant challenges. Most demonstrations stick to rectangular or simple shaped items. Stain Treatment: Identifying and pre-treating stains requires judgment calls that current AI systems struggle with. Full Workflow Management: While robots can perform individual tasks, seamlessly managing the entire laundry process from dirty clothes to closet remains elusive. Which Home Robots Are Best for Household Tasks? LG CLOiD: The Kitchen and Laundry Specialist LG's CLOiD robot made waves at CES 2026 with impressive household task demonstrations. This dual-armed humanoid specifically targets domestic chores with five-finger hands and integration with LG's ThinQ smart home ecosystem. Strengths: Demonstrated success folding laundry and initiating laundry cycles, can coordinate with smart appliances automatically. Limitations: Requires LG appliances for optimal performance, limited to specific pre-programmed task sequences, no announced consumer pricing or availability. Figure 03: The General Purpose Household Assistant Figure AI's latest robot focuses on general household utility. At 5'6" tall with 20kg payload capacity and 5-hour runtime, it handles precision dish handling, basic clothes folding on flat surfaces, and object retrieval. Limitations: Still in development for consumer market, requires controlled environments for complex tasks. 1X NEO: The Consumer-Ready Assistant At $20,000 (or $499/month subscription), NEO represents the first serious attempt at a consumer-ready household robot. It features 22 degrees of freedom hands, weighs 66 pounds, can lift 150+ pounds, and operates at whisper-quiet 22dB. Reality Check: While NEO's marketing promises comprehensive household assistance, early reviews suggest many complex tasks still require human guidance through "1X Expert" sessions. What Are the Current Limitations and Challenges? Despite impressive demonstrations, humanoid robots face significant hurdles in practical home deployment across technical, economic, and environmental dimensions. What Technical Limitations Exist? Dexterous Manipulation: The biggest bottleneck is handling unpredictable objects. Laundry presents countless variables—fabric types, sizes, wrinkles, and spatial configurations that challenge current AI systems. Battery Life: Most humanoid robots operate for only 3-5 hours before requiring extended charging periods. This severely limits their practical utility for all-day household assistance. Sensorimotor Skills: Current robotic actuators lag far behind biological muscles in force density, bandwidth, and control precision—resulting in clumsy handling of delicate items. What Are the Economic Realities? High Initial Costs: With prices ranging from $20,000 (NEO) to $250,000+ for industrial models, humanoid robots remain expensive luxury items rather than practical household appliances. Maintenance Requirements: These complex machines require regular maintenance, software updates, and potential repairs that add to total ownership costs. Limited ROI: For most households, the time saved doesn't justify the expense, especially considering current limitations. What Environmental Challenges Do They Face? Home Complexity: Real homes present countless variables that controlled demonstrations don't account for—pets, children, unexpected obstacles, and varying layouts. Safety Concerns: Heavy robots (60-130+ pounds) operating around family members raise legitimate safety questions, especially with children present. Which Robots Are Closest to Practical Home Use? Based on current capabilities and announced timelines, here's our ranking of the robots most likely to deliver practical home assistance. 1. LG CLOiD (Late 2026 - Early 2027) Why: Specific focus on household tasks, demonstrated reliability in controlled environments, integration with existing smart home ecosystem. 2. 1X NEO (2026 Consumer Deliveries) Why: First to market with consumer pricing, comprehensive household task training, subscription model reduces barrier to entry. 3. Figure 03 (2027-2028 Consumer Timeline) Why: Superior technical specifications but longer development timeline for consumer deployment. When Will Robots Reliably Do Your Laundry? Based on current technology trajectories and manufacturer roadmaps, here's a realistic timeline for home robot laundry capabilities. 2026-2027: Limited Task Assistance Basic folding of simple garmentsSupervised laundry transfer between machinesSchedule-based task execution with human oversight 2028-2030: Moderate Independence Autonomous handling of standard laundry loadsIntegration with smart home systems for schedulingImproved handling of varied fabric types 2030-2035: Full Laundry Automation End-to-end laundry management with minimal supervisionAdvanced fabric care including stain treatmentAdaptive learning for household-specific preferences Should You Buy a Humanoid Robot for Laundry in 2026? If you're considering a humanoid robot for household tasks, here's practical advice based on current capabilities and your situation. Who Should Consider Buying Now Early adopters comfortable with beta technologyHouseholds with specific, simple, repetitive tasksThose willing to invest in learning and training the robot Who Should Wait Families seeking "set it and forget it" automationHouseholds with complex laundry needs (delicates, special care items)Budget-conscious consumers looking for immediate ROI Alternative Solutions While waiting for humanoid robots to mature, consider specialized laundry-folding machines, smart washers and dryers with app control, or robotic vacuum and mop combinations. The Bottom Line: Promise vs. Reality Humanoid robots have made remarkable progress in household task capabilities, but the dream of comprehensive home automation remains just that—a dream, at least for now. While robots like the LG CLOiD, Figure 03, and 1X NEO can handle specific laundry tasks under ideal conditions, they're far from the reliable household assistants portrayed in marketing materials. For most families, these robots represent expensive technology experiments rather than practical household solutions. The current sweet spot involves simple, repetitive tasks in controlled environments—perfect for tech enthusiasts but limiting for everyday households. However, the trajectory is promising. With continued advances in AI, dexterity, and cost reduction, we expect significant improvements by 2028-2030. Early adopters willing to work with current limitations may find value, but mainstream consumers should wait for the next generation. Ready to explore your options? Check out our comprehensive guides to humanoid robots for home use and current humanoid robot pricing to make an informed decision. Compare on RoboZaps: Figure 03 and Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Service Robots: Complete Guide to Types, Applications & Where to Buy [2026] URL: https://blog.robozaps.com/b/service-robots-guide Author: Radica Maneva Published: 2026-02-03T11:30:17.081Z Updated: 2026-08-01T23:06:33.492Z Category: insights TL;DR: Service robots are a $72.46 billion market in 2026, heading past $212 billion by 2034. Prices run $8,000-$150,000+ (or $500-$2,000/month as a service), with healthcare, hospitality, retail, logistics, cleaning and security the top applications and payback typically within 12-24 months. Function defines them; most service robots are not humanoid. Key takeaways: - Market size: $72.46 billion in 2026, growing to $212+ billion by 2034 - Price range: $8,000-$150,000+ depending on application (or $500-$2,000/month RaaS) - Top industries: Healthcare, hospitality, retail, logistics, cleaning, security - ROI timeline: Most service robots achieve payback within 12-24 months - Key difference: Service robots are defined by function; humanoid robots by form — most service robots aren't humanoid FAQ: Q: What is a service robot? A: A service robot is an autonomous or semi-autonomous robot that performs useful tasks for humans outside of industrial manufacturing. They operate in environments like hospitals, hotels, restaurants, retail stores, and homes, performing delivery, cleaning, security, and customer assistance tasks. Q: How much does a service robot cost? A: Service robot prices range from $8,000 to $150,000+ depending on function: Entry-level (simple delivery): $8,000-$15,000Mid-range (hospitality/retail): $15,000-$40,000Professional-grade (healthcare/industrial): $50,000-$150,000Robot-as-a-Service: $500-$2,000 per month Q: What's the difference between a service robot and a humanoid robot? A: Service robots are defined by function (performing services), humanoid robots by form (human-like appearance). Most service robots are NOT humanoid — they're wheeled platforms or specialized machines designed for specific tasks. Some robots like Pepper are both. Q: What industries use service robots the most? A: Top industries: Healthcare, hospitality, retail, logistics, cleaning services, and security. Healthcare leads adoption due to staff shortages and infection control needs. Q: Can service robots work alongside humans safely? A: Yes — modern service robots include LiDAR sensors, cameras, and AI-powered obstacle avoidance. They're programmed to stop or reroute when humans are nearby. Reputable robots undergo rigorous safety testing. Q: How long do service robots last? A: Commercial service robots typically last 5-10 years with proper maintenance. Battery replacement may be needed every 2-3 years. Many manufacturers offer service contracts to maximize longevity. Q: Do service robots require special infrastructure? A: Most modern service robots work in existing environments with minimal modifications. Common requirements include dedicated charging stations, Wi-Fi connectivity, flat floors, and elevator integration for multi-floor operation. Q: How do I choose the right service robot for my business? A: Consider: Task requirements, environment (indoor/outdoor, floor type), volume (tasks per day), budget (purchase vs. lease), integration with existing systems, and local service availability. Q: Where can I buy a service robot? A: Buy through manufacturer direct (Pudu, Bear Robotics, SoftBank), authorized distributors, or robotics marketplaces like Robozaps. Contact sales@robozaps.com for personalized guidance. Q: What's the ROI on service robots? A: Most service robots achieve payback within 12-24 months. Healthcare robots reduce delivery time by 30-50%, cleaning robots cut labor costs 50-70%, and security robots cover 4-5x more area than human guards at lower cost. Service robots are autonomous machines that perform useful tasks for humans outside manufacturing — from delivering room service in hotels to disinfecting hospital rooms — and represent a $72 billion market in 2026. Unlike industrial robots on assembly lines, service robots operate in human environments including hotels, hospitals, restaurants, retail stores, and homes. Whether you're a business owner looking to automate operations, a healthcare administrator seeking to improve patient care, or simply curious about this transformative technology, this guide covers everything you need to know about service robots — from types and applications to real-world ROI and where to buy them. What Is a Service Robot? A service robot is an autonomous or semi-autonomous robot designed to perform useful tasks for humans or equipment, excluding industrial automation applications. The International Federation of Robotics (IFR) categorizes them into professional service robots (commercial settings) and personal service robots (domestic use like robot vacuums). Key characteristics that define service robots: Professional service robots: Used in commercial settings like hospitals, hotels, warehouses, and public spacesPersonal service robots: Designed for domestic use, including robot vacuums, lawn mowers, and companion robots What's the Difference Between Service Robots and Humanoid Robots? Service robots are defined by their function (performing services), while humanoid robots are defined by their form (human-like appearance). Most service robots are NOT humanoid — they're designed in forms optimal for their specific tasks. For example: A BellaBot delivery robot is a service robot but not humanoid (it's a wheeled platform)A Pepper robot is both a service robot AND humanoid (it performs customer service while having a human-like appearance)A Boston Dynamics Spot is a service robot that's neither humanoid nor wheeled (it's quadruped) The key takeaway: All humanoid robots can function as service robots, but most service robots aren't humanoid — and that's by design. Specialized form factors often outperform humanoid designs for specific tasks. Learn more in our guide to humanoid robots. What Are the Main Types of Service Robots by Industry? 1. Hospitality Service Robots Hospitality robots handle room service delivery, food running, concierge duties, and guest greeting — with top performers like BellaBot Pro ($16,900) achieving 80%+ guest satisfaction rates. The hospitality industry has emerged as one of the fastest adopters of service robot technology. Common Applications: Room service deliveryFood and beverage delivery in restaurantsConcierge and information servicesLuggage handlingGuest greeting and check-in assistance Popular Hospitality Robots: Case Study: Aloft Hotels — The Aloft hotel chain pioneered robot deployment with "Botlr," a Relay robot that delivers amenities to guest rooms. Results: 80% guest satisfaction rating, 3-minute average delivery time, and 24/7 availability without staffing concerns. For a deeper dive, read our guide on humanoid robots in hospitality. 2. Healthcare Service Robots Healthcare robots deliver medications, disinfect rooms, and assist nursing staff — with hospitals reporting 30-50% reduction in supply delivery time and 99.6% successful delivery rates. This sector addresses critical challenges like staff shortages, infection control, and patient care quality. Common Applications: Medication and supply deliveryPatient monitoring and telepresenceDisinfection and sanitizationSurgical assistanceRehabilitation therapyCompanion care for elderly patients Popular Healthcare Robots: The ROI of Healthcare Robots: Hospitals deploying delivery robots see 30-50% reduction in supply delivery time, 15-20% decrease in nursing time spent on non-clinical tasks, and significant reduction in cross-contamination risks. UV disinfection robots can sanitize a patient room in under 10 minutes. Learn more in our article on humanoid robots in healthcare. 3. Retail Service Robots Retail robots scan inventory, assist customers, and detect spills — with Walmart reporting 95%+ shelf accuracy from automated scanning and 25% increased store dwell time from customer engagement robots. Common Applications: Customer assistance and wayfindingInventory scanning and managementShelf stockingSecurity and loss preventionAutonomous checkout assistanceMarketing and promotional activities Popular Retail Robots: Explore this topic further in our comprehensive guide to humanoid robots in retail. 4. Delivery Service Robots Last-mile delivery robots operate on sidewalks at 3-6 mph, carrying 20-50 lbs of cargo at $1-3 per delivery — 60-80% cheaper than human couriers. This is one of the fastest-growing segments driven by e-commerce growth. Types of Delivery Robots: Sidewalk robots: Operate on sidewalks, 20-50 lbs capacity, 3-5 mile rangeIndoor robots: Navigate buildings, use elevators, ideal for offices and campusesAerial drones: Cover longer distances quickly but with limited payload and regulatory restrictions Popular Delivery Robots: 5. Cleaning Service Robots Commercial cleaning robots scrub 10,000-50,000 sq ft per hour autonomously, delivering 50-70% labor cost reduction with consistent quality and documented coverage maps. Types of Cleaning Robots: Floor scrubbers: Autonomous wet cleaning, 10,000-50,000 sq ft/hourCommercial vacuums: Continuous dry floor maintenanceWindow cleaners: High-rise exterior cleaning, magnetic or suction-basedUV disinfection: Hospital-grade sanitization, kills 99.9% of pathogens Popular Cleaning Robots: 6. Security Service Robots Security robots patrol 4-5x more area than human guards at $7-$12/hour via RaaS models, providing 360° surveillance, thermal imaging, and continuous documented evidence. Popular Security Robots: How Do You Choose the Right Service Robot? Choose a service robot by defining your use case, calculating ROI (most achieve payback in 12-24 months), assessing integration requirements, and considering scalability. Step 1: Define Your Use Case What specific tasks do you want automated?Where will the robot operate (indoor/outdoor/both)?What are your floor conditions and space constraints?How will the robot interact with humans? Step 2: Calculate ROI ROI Formula: Most service robots achieve ROI within 12-24 months. Step 3: Assess Integration Requirements IT infrastructure compatibilityBuilding modifications needed (charging stations, network)Integration with existing systems (POS, hospital management)Staff training requirements Step 4: Consider Scalability Can you add more robots easily?Does the platform support fleet management?What's the manufacturer's roadmap?Are software updates included? Where Can You Buy Service Robots? Buy service robots directly from manufacturers (Pudu, Bear Robotics, SoftBank), through authorized distributors, or via robotics marketplaces like Robozaps. Direct from Manufacturers Pudu Robotics (pudurobotics.com) — BellaBot, KettyBot, HolaBot, CC1Bear Robotics (bearrobotics.ai) — Servi, Servi Plus, Servi MiniSoftBank Robotics (softbankrobotics.com) — Pepper, WhizKnightscope (knightscope.com) — K3, K5, K7 (RaaS model) Robozaps: Your Robotics Marketplace At Robozaps, we specialize in connecting buyers with cutting-edge robotics technology. Our current inventory includes humanoid robots like the Unitree H1 ($99,900-$128,900) and AgiBot A2 ($120,000), with service robot offerings expanding. Why Buy Through Robozaps: Verified sellers and authentic products Secure escrow payment protection Comprehensive insurance options Expert consultation available Hassle-free returns policy Financing available for qualified buyers Contact our sales team: sales@robozaps.com | +1 480-819-2567 Should You Buy or Lease a Service Robot? What's the Future of Service Robots? The service robot market will grow from $72 billion in 2026 to $212+ billion by 2034, driven by AI integration, multi-robot coordination, and increasing labor shortages. Key trends for 2026 and beyond: AI Integration: Advanced NLP, predictive maintenance, personalized interactionsMulti-Robot Coordination: Fleet management, task distribution, collision avoidanceHuman-Robot Collaboration: Robots handle repetitive tasks, humans focus on high-value activitiesIndustry-Specific Solutions: Customized protocols for hospitals, restaurants, retail Conclusion Service robots have evolved from futuristic concepts to practical business tools delivering measurable ROI across industries. Whether you're considering a delivery robot for your restaurant, a disinfection unit for your hospital, or an inventory scanner for retail, the technology has matured to the point where deployment is straightforward and cost-effective. The key to success lies in matching the right robot to your specific needs, properly calculating ROI, and working with reputable suppliers who can provide ongoing support. As the market continues to grow at 15-20% annually, early adopters are gaining competitive advantages that will compound over time. Ready to explore service robots for your organization? Contact Robozaps for expert guidance, or browse our selection of advanced robotics solutions at robozaps.com/shop. For more robotics insights, explore our related guides: Best Humanoid Robots for Sale in 2026Humanoid Robots in Retail: Complete GuideHumanoid Robots in HospitalityHumanoid Robots in Healthcare Compare on RoboZaps: AgiBot A2, Pepper, Spot and Unitree H1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # NEURA 4NE1 Price 2026: €98K Gen 3.5, €19,999 Mini & Specs URL: https://blog.robozaps.com/b/neura-robotics-4ne1-review Author: Radica Maneva Published: 2026-02-03T10:00:00.000Z Updated: 2026-08-01T23:06:29.577Z Category: reviews TL;DR: The NEURA 4NE1 comes in two models: the industrial Gen 3.5 at €98,000 (~$105,000) and the 4NE1 Mini at €19,999 (~$21,500) for home and research, both reservable with a refundable €100 deposit. Mini units ship April 2026, industrial late 2026. Designed with Studio F.A. Porsche, it lifts up to 100 kg and runs NVIDIA Isaac GR00T. Key takeaways: - Two models available: 4NE1 Gen 3.5 (€98,000 / ~$105,000) for industrial use and 4NE1 Mini (€19,999 / ~$21,500) for home, education, and research - Porsche-designed: First humanoid robot designed in collaboration with Studio F.A. Porsche — the designers behind the iconic Porsche 911 - Massive lifting capacity: The full-size 4NE1 can lift up to 100 kg (220 lbs) — the highest among general-purpose humanoids - Pre-orders now open: Reserve with a fully refundable €100 deposit; first industrial units ship late 2026, Mini units ship April 2026 - Western alternative: Europe's leading cognitive humanoid, competing directly with Chinese imports like the Unitree G1 - Powered by NVIDIA: Runs on NVIDIA Isaac GR00T and Thor T5000 processor with advanced water cooling FAQ: Q: How much does the NEURA 4NE1 cost? A: The 4NE1 Gen 3.5 costs €98,000 (~$105,000) for individual orders (1-19 units), dropping to €60,000 (~$65,000) for bulk orders of 20+ units. The 4NE1 Mini costs €19,999 (~$21,500). Both require a fully refundable €100 reservation deposit. Q: When will the NEURA 4NE1 ship? A: The 4NE1 Mini ships in April 2026 (Spring 2026). The 4NE1 Gen 3.5 ships in late 2026, with general availability expected by late 2026. Q: Where can I buy the NEURA 4NE1? A: Pre-orders are available directly through NEURA's website at neura-robotics.com. NEURA is one of the few Western humanoid manufacturers offering direct online sales. Q: What is the 4NE1's payload capacity? A: The Gen 3.5 can lift up to 100 kg (220 lbs) maximum, with a continuous mobile payload of 15-20 kg (33-44 lbs). The Mini has a 3 kg (6.6 lbs) payload capacity. Q: Is the NEURA 4NE1 safe to work around? A: Yes. The 4NE1 features NEURA's patented artificial skin that detects proximity before contact, force-torque sensors in all joints, and the NEURA Omnisensor for touchless human detection. It's designed for cage-free collaboration with humans. Q: What AI does the NEURA 4NE1 use? A: The 4NE1 runs on NVIDIA Isaac GR00T, an open foundation model for humanoid reasoning, combined with NEURA's proprietary AURA AI system. The Gen 3.5 uses an NVIDIA Thor T5000 processor. Q: How long does the battery last? A: The Gen 3.5 offers 6-8 hours of runtime with hot-swappable batteries enabling 24/7 continuous operation. The Mini provides approximately 2.5 hours of active battery life. Q: What programming languages does the 4NE1 support? A: NEURA provides Python SDK, ROS 2 interface, and C++ SDK for developers. The robots also support teleoperation and digital twin access through the Neuraverse platform. Q: Who designed the NEURA 4NE1? A: The Gen 3 design was created in collaboration with Studio F.A. Porsche — the design house responsible for the Porsche 911 and numerous premium consumer products. Q: Is the €100 reservation refundable? A: Yes. The reservation fee is fully refundable at any time before your final purchase agreement is signed. Upon purchase, the €100 is credited toward the total price. NEURA Robotics has emerged as Europe's frontrunner in the humanoid robot race, and the 4NE1 represents their most ambitious creation yet. Unveiled at CES 2026 in Las Vegas, this German-engineered humanoid (price: €19,999 for the Mini, €98,000 for the Gen 3.5) isn't just another prototype — it's a production-ready machine with confirmed pricing and shipping dates. After months of tracking this company and analyzing their CES 2026 debut, I'm ready to break down everything you need to know about the NEURA Robotics 4NE1: specifications, pricing, how it compares to competitors, and whether it's worth your consideration as an early adopter. Table of Contents What is NEURA Robotics?4NE1 Overview: Two Robots, Two MarketsFull Specifications ComparisonNEURA 4NE1 Price Breakdown & Where to BuyDesign & Build QualityAI & Software: The Neuraverse PlatformPerformance & Real-World CapabilitiesHow It Compares to CompetitorsUse Cases & ApplicationsPros & ConsFrequently Asked QuestionsThe Verdict What is NEURA Robotics? NEURA Robotics is a German robotics company headquartered in Metzingen, Germany — the same town that's home to Hugo Boss's headquarters. Founded and led by CEO David Reger, NEURA positions itself as a pioneer in "cognitive robotics" — robots that don't just execute pre-programmed tasks but perceive, learn, and adapt in real time. The company has rapidly expanded its global footprint with facilities in: Germany (HQ): Metzingen — engineering and primary manufacturingChina: Hangzhou — production facility for Asian marketUSA: Detroit (current), with planned expansion to Boston and San Francisco What sets NEURA apart from competitors is their Neuraverse operating system — a shared platform where robots can pool and reuse learned experiences. When one NEURA robot masters a task, that knowledge becomes instantly available to every other robot on the network. Think of it like cloud-based collective learning for robots. NEURA already has industrial robots in commercial deployment through its cognitive robot platform, which has been validated in real manufacturing environments. The 4NE1 humanoid represents their consumer and general-purpose play. 4NE1 Overview: Two Robots, Two Markets The 4NE1 comes in two distinct variants, each targeting different markets and budgets: 4NE1 Gen 3.5 (Full-Size Industrial) The flagship model stands at human height (180 cm / 5'11") and is designed for complex industrial workflows, logistics, and high-payload tasks. With a lifting capacity of 100 kg and 6-8 hours of runtime with hot-swappable batteries, this is a serious workhorse robot. 4NE1 Mini (Consumer/Education) The smaller sibling at 132 cm (4'4") brings the same cognitive AI capabilities in a more accessible form factor. Priced at €19,999, it's positioned as "the Western world's answer" to affordable humanoids from China like the Unitree G1. Full Specifications Comparison Key Sensors & Perception Both variants share NEURA's advanced cognitive sensor suite: Omnidirectional 3D Vision: 360° environmental perception for obstacle detection and navigationPatented Artificial Skin: Proximity-detecting sensor skin that prevents collisions before contactForce-Torque Sensors: In all joints for precise manipulation and safe human interactionNEURA Omnisensor: Touchless human detection for safety-critical environmentsMicrophones: Multi-language voice recognition and natural language processing NEURA 4NE1 Price Breakdown & Where to Buy NEURA has taken the unusual step of publishing transparent pricing — a rarity in the humanoid robot market where most competitors only offer "contact sales" pricing. 4NE1 Gen 3.5 Pricing 4NE1 Mini Pricing How to Reserve Pre-orders are now open directly through NEURA's website with a fully refundable €100 deposit. This deposit secures your place in the delivery queue and will be credited toward your final purchase price. Order directly at: neura-robotics.com/product/4ne1-reservation/ What's Included Each 4NE1 unit ships as a fully operational system including: Integrated high-dexterity handsHigh-capacity battery (dual-battery system for Gen 3.5)Dedicated charging stationAccess to Neuraverse platformDigital twin access and teleoperation capabilityPython SDK, ROS 2 interface, and C++ SDK Availability by Region NEURA is targeting global availability with initial focus on: Europe (primary market)United StatesChinaJapanTaiwan Design & Build Quality: The Porsche Partnership The 4NE1's most striking feature is its aesthetic — and that's no accident. NEURA partnered with Studio F.A. Porsche for the Gen 3 design, the same design house responsible for the iconic Porsche 911, along with countless premium consumer products. The result is arguably the most visually refined humanoid robot on the market. Key design elements include: Clean, flowing lines: Unlike the utilitarian look of most industrial robots, the 4NE1 has an organic, approachable aestheticNeutral color palette: White and grey options with customizable screen elementsHuman proportions: At 180 cm, the full-size model matches average human height, making it less intimidating in collaborative environmentsIntegrated display: Head-mounted screen for visual feedback and expression Beyond aesthetics, the build quality reflects German engineering standards: Water-cooled thermal management: The Gen 3.5 uses active water cooling to maintain performance during extended operation — a significant upgrade over air-cooled competitorsModular limbs: Exchangeable forearms allow task-specific customizationSafety-first construction: The patented artificial skin detects proximity, allowing the robot to stop or adjust before any contact occurs AI & Software: The Neuraverse Platform Hardware is only half the story. NEURA's real competitive advantage lies in its software ecosystem. AURA AI NEURA's proprietary contextual AI system powers the 4NE1's cognitive abilities: Multi-modal perception: Processes visual, auditory, and tactile inputs simultaneouslyNatural language understanding: Multi-language voice commands with context awarenessGesture recognition: Intuitive hand gesture control for non-verbal interactionReinforcement learning: Continuous improvement through interaction Neuraverse Operating System This is where NEURA differentiates itself from every competitor. The Neuraverse is a shared intelligence platform that connects all NEURA robots: Fleet Learning: When one robot learns a task, that skill propagates to all connected robots instantlyNEURA Sync: Real-time device-robot communication across the ecosystemNEURA Gym: Simulated training environment for developing new skills safelyNeuraverse Marketplace: Platform for sharing, publishing, and monetizing robotic skills and applicationsDigital Twin: Every robot has a cloud-based twin for remote monitoring and simulation NVIDIA Partnership The 4NE1 is powered by NVIDIA Isaac GR00T — an open foundation model specifically designed for humanoid robot reasoning. Key capabilities include: Advanced material handling task learningInstruction following through multimodal reasoning (voice, vision, touch)Simulation and testing via NVIDIA Isaac Lab and Isaac Sim The industrial Gen 3.5 model uses the NVIDIA Thor T5000 processor — a powerhouse chip designed specifically for AI and robotics applications. Developer Interfaces For developers and researchers, NEURA provides comprehensive integration options: Python SDKROS 2 interfaceC++ SDKTeleoperation capabilityDigital twin access Performance & Real-World Capabilities Mobility The full-size 4NE1 walks at 5 km/h (3.1 mph) — roughly a brisk walking pace. While this isn't record-breaking (the Unitree H1 reaches 13 km/h), it's practical for industrial and service environments where stability matters more than speed. The Mini model operates at approximately 3 km/h, suitable for indoor navigation in homes, offices, and educational settings. Manipulation & Payload This is where the 4NE1 Gen 3.5 truly shines. With a maximum lifting capacity of 100 kg (220 lbs), it has the highest payload among general-purpose humanoids. For context: The integrated high-dexterity hands allow for fine manipulation tasks, making the 4NE1 suitable for everything from heavy lifting to delicate assembly operations. Runtime & Continuous Operation The Gen 3.5's dual-battery system with hot-swap capability enables continuous 24/7 operation with zero downtime. Standard runtime is 6-8 hours per charge, but by swapping batteries while one charges, facilities can maintain uninterrupted operation. The Mini offers approximately 2.5 hours of active battery life — comparable to the Unitree G1's 2-hour runtime. How It Compares to Competitors The humanoid robot market is heating up. Here's how the NEURA 4NE1 stacks up against key competitors: vs 1X NEO The 1X NEO is the 4NE1 Mini's closest competitor in the consumer space. Both target home assistance at similar price points (~$20,000). Key differences: 1X NEO relies on human-in-the-loop teleoperation — real operators can see through the robot's cameras4NE1 Mini emphasizes autonomous cognitive AI without human oversight1X NEO offers a $499/month subscription option; NEURA is purchase-only vs Unitree G1 China's Unitree G1 is the price leader at $13,500, but the 4NE1 Mini offers compelling differentiators: European manufacturing and supportPorsche-designed aestheticsNeuraverse fleet learning platformMore comprehensive developer ecosystem (ROS 2, Python SDK, C++ SDK) vs Figure 02 / Tesla Optimus For industrial applications, the 4NE1 Gen 3.5 competes with Figure 02 and the upcoming Tesla Optimus. NEURA's advantages: Transparent, public pricing (Figure requires enterprise sales discussions)Higher payload capacity (100 kg vs ~20 kg)Earlier availability (Late 2026 vs late 2027 for Tesla)European-engineered alternative for buyers concerned about US-China supply chain risks For a comprehensive comparison, see our Best Humanoid Robots 2026 guide. Use Cases & Applications 4NE1 Gen 3.5 (Industrial) Manufacturing & Logistics: Heavy lifting tasks up to 100 kgMaterial handling and transportationQuality inspection with 360° visionCollaborative assembly alongside human workers Healthcare: Patient assistance and mobility supportEquipment transport in hospitalsRehabilitation assistance Service Industry: Hospitality and concierge rolesWarehouse automationReception and customer service 4NE1 Mini (Consumer/Education) Home Assistance: Household chores and organizationObject retrieval and carrying (up to 3 kg)Voice-controlled smart home hubElderly care and companionship Research & Education: University robotics programsAI and machine learning researchSTEM education demonstrationsHuman-robot interaction studies Entertainment: Exhibition and event appearancesInteractive demonstrationsContent creation Pros & Cons Pros Highest payload capacity (100 kg) — Outlifts every competitor, making it suitable for heavy industrial tasksStudio F.A. Porsche design — Aesthetically refined, less industrial-looking than competitorsTransparent pricing — One of the few companies publishing actual prices, not "contact sales"European engineering — German quality and EU-based support for Western buyersNeuraverse fleet learning — Unique platform where robots share learned skills across the networkNVIDIA Isaac GR00T integration — Running on cutting-edge AI foundation model24/7 operation capability — Hot-swappable batteries for continuous industrial deploymentComprehensive SDK — Python, ROS 2, and C++ support for developers Cons High price point (€98,000 industrial) — Significantly more expensive than Unitree or 1X alternativesNot yet shipping — First units expected late 2026; buyers must waitLimited runtime for Mini (2.5 hrs) — Consumer model has shorter battery life than some competitorsNew entrant risk — NEURA is newer to humanoids than Boston Dynamics or UnitreeLimited US presence — Currently only Detroit; support infrastructure still developingNo subscription option — Unlike 1X NEO's $499/month, NEURA requires full purchase The Verdict: Should You Buy the NEURA 4NE1? The NEURA 4NE1 represents a significant milestone for European robotics. For the first time, Western buyers have a credible alternative to Chinese humanoids and American vaporware announcements. Who Should Buy the 4NE1 Gen 3.5 Manufacturing facilities needing high-payload collaborative robotsLogistics companies ready to pilot humanoid automationHealthcare institutions exploring patient assistance solutionsOrganizations preferring European suppliers and supportEarly adopters willing to invest €98,000 for cutting-edge capability Who Should Buy the 4NE1 Mini University robotics departments and research labsAI/ML researchers needing a capable development platformTech-forward consumers who want a home humanoid without Chinese supply chain concernsEducational institutions teaching robotics and STEMEarly adopters who prefer a Western alternative to the Unitree G1 Who Should Wait Budget-conscious buyers — The Unitree G1 at $13,500 offers more value for basic researchThose needing immediate delivery — Shipping doesn't start until April 2026 (Mini) / Late 2026 (Gen 3.5)Consumers wanting subscription flexibility — 1X NEO's $499/month option may be more accessible Final Score The NEURA 4NE1 isn't the cheapest humanoid you can buy — but it might be the most capable. With the highest payload in its class, a Porsche-designed aesthetic, and the innovative Neuraverse platform, it sets a new standard for what a general-purpose humanoid should be. If you're serious about humanoid robotics and want European engineering with transparent pricing and confirmed delivery dates, the 4NE1 deserves a spot on your shortlist. Reserve your 4NE1 at neura-robotics.com Last updated: February 2026 Related Articles: Best Humanoid Robots 2026: Complete Buyer's GuideHumanoid Robots for Home: What You Can Actually Buy Sources: NEURA Robotics official website, CES 2026 press releases, Interesting Engineering, RoboHorizon, NVIDIA newsroom Compare on RoboZaps: 4NE1 Gen 3.5, Apptronik Apollo, Tesla Optimus, Figure 02, Digit and Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Telepresence Robots: Complete Guide to Remote Presence Solutions [2026] URL: https://blog.robozaps.com/b/telepresence-robots-guide Author: Radica Maneva Published: 2026-02-03T00:00:00.000Z Updated: 2026-08-01T23:06:38.468Z Category: insights TL;DR: Telepresence robots range from $1,400 (PadBot) to $32,000+ (Ava 500), with the sweet spot at $4,000-$5,000: the Double 3 ($4,499) for business, the Ohmni from $2,000 for budget buyers. The market is projected to reach $989 million by 2030, and most organizations see ROI within 1-2 years from reduced travel. Key takeaways: - Market growth: Telepresence robot market projected to reach $989 million by 2030 (14.86% CAGR) - Price range: $1,400 (PadBot) to $32,000+ (Ava 500), with popular mid-range options at $4,000-$5,000 - Best value: Double 3 (Note: Currently sold out) ($4,499) for business; Ohmni ($2,000-$4,095) for budget buyers - ROI timeline: Most organizations see positive ROI within 1-2 years from reduced travel - Top use cases: Remote work, healthcare telemedicine, education for homebound students FAQ: Q: What is a telepresence robot used for? A: A telepresence robot gives remote users a physical presence in a distant location. Primary uses include remote work (attending meetings, touring facilities), healthcare (patient consultations, medical rounds), and education (allowing homebound students to attend class). The robot serves as your "body" in another location. Q: How much does a telepresence robot cost? A: Telepresence robot prices range from approximately $1,400 for basic consumer models to over $32,000 for enterprise-grade autonomous systems. Popular mid-range options like the Double 3 (Note: Currently sold out) cost around $4,500, while budget-friendly choices like Ohmni start around $2,000. Q: Can telepresence robots navigate autonomously? A: Some telepresence robots offer autonomous navigation, while others require manual control. High-end models like the Ava 500 can navigate completely autonomously using LIDAR and advanced mapping. Mid-range robots like Double 3 (Note: Currently sold out) offer semi-autonomous features like obstacle avoidance. Q: Are telepresence robots better than Zoom? A: Telepresence robots complement rather than replace video conferencing. They're better for situations requiring mobility, physical presence, or spontaneous interaction—like facility tours and impromptu desk-side chats. Video conferencing remains more practical for scheduled meetings. Q: What is the best telepresence robot for business? A: For most businesses, the Double 3 (Note: Currently sold out) ($4,499) offers the best balance of features, reliability, and price with autonomous navigation and excellent audio/video quality. For larger enterprises, Ava Robotics provides superior autonomous capabilities. For budget-conscious SMBs, Ohmni delivers excellent value. Q: Do telepresence robots work with WiFi? A: Yes, telepresence robots require WiFi connectivity to function. They connect to your local network and stream video/audio through cloud services. Reliable WiFi coverage throughout the operating area is essential—typically requiring 5-10 Mbps bandwidth. Q: Can telepresence robots climb stairs? A: Most telepresence robots cannot climb stairs. They're designed for flat, indoor surfaces and use wheels for mobility. Facilities with multiple floors typically require either elevator access or multiple robots on different floors. The way we work, learn, and receive healthcare has fundamentally changed. As organizations embrace hybrid models and remote collaboration, telepresence robots have emerged as the ultimate solution for maintaining human connection across distances. Unlike standard video conferencing, these mobile robotic platforms give remote users a physical presence—allowing them to move freely, interact naturally, and participate as if they were actually there. A telepresence robot is a remotely controlled mobile device with a camera, screen, and speakers that lets you physically "be" somewhere you're not—moving through spaces, joining conversations, and interacting with people from anywhere in the world. Prices range from $1,400 to $32,000+, with ROI often achieved within 12-24 months through reduced travel costs. What Is a Telepresence Robot? A telepresence robot is a remotely controlled, mobile robotic device equipped with a camera, display screen, microphone, and speaker that allows a user to interact with people and navigate physical spaces from a remote location. Think of it as your digital avatar—a robot body you can "pilot" from anywhere in the world through a web browser or mobile app. Unlike traditional video conferencing where you're confined to a fixed camera angle, telepresence robots offer: Mobility: Move freely through offices, hospitals, factories, or classroomsPhysical presence: Stand at eye level with colleagues during conversationsAutonomous navigation: Many models can navigate independently around obstaclesSpontaneous interaction: Drop by someone's desk or join impromptu hallway conversations The global telepresence robot market is experiencing explosive growth, projected to reach USD 989 million by 2030, growing at a CAGR of approximately 14.86%. This surge is driven by the permanent shift to hybrid work, advances in 5G connectivity, and increasing adoption in healthcare and education sectors. What Are the Primary Use Cases for Telepresence Robots? How Do Telepresence Robots Help Remote Workers? For distributed teams, telepresence robots transform how remote employees participate in office life. Rather than being a face on a screen in a conference room, remote workers can join spontaneous discussions, tour facilities, and attend networking events. Companies like Google, Microsoft, and Intel have deployed fleets of telepresence robots to keep their distributed workforces connected. For more on how robotics is transforming the workplace, check out our dedicated guide. How Are Telepresence Robots Used in Healthcare? The healthcare sector has become one of the largest adopters of telepresence technology. Medical telepresence robots enable remote patient consultations, ICU monitoring without physical exposure, family visits to isolated patients, and remote medical education. The COVID-19 pandemic accelerated healthcare adoption dramatically. Learn more about how robots are revolutionizing healthcare in our comprehensive healthcare robotics guide. What About Telepresence Robots in Education? Telepresence robots are transforming educational experiences for students who cannot physically attend school. Homebound students with chronic illness can attend classes, international students can participate in exchange programs virtually, and prospective students can explore universities remotely. Schools from K-12 to universities are deploying telepresence robots to ensure no student is left behind. For an in-depth look at educational applications, see our guide on robots in education. What Are the Top Telepresence Robots in 2026? The telepresence robot market features several established players, each targeting different use cases and budgets. Here's our comprehensive breakdown of the leading options: Double 3 (Note: Currently sold out) by Double Robotics Best for: Business and enterprise environments Double Robotics pioneered the modern telepresence robot market and continues to lead with the Double 3 (Note: Currently sold out). This self-driving robot features an array of 3D sensors for autonomous navigation and obstacle avoidance. Key Features: Click-to-drive interface with mixed reality overlayTwo 13MP cameras with unified pan-tilt-zoomSix beamforming microphones for superior audioSelf-driving with automatic obstacle avoidanceHeight-adjustable (47-60 inches), 4-hour battery life Price: $4,499 (includes charging dock) Ohmni by OhmniLabs Best for: Budget-conscious buyers and home use Recognized by ZDNET as the "best budget telepresence robot," Ohmni offers excellent value without compromising on essential features. Made in the USA, Ohmni robots are known for their reliability and ease of setup. Key Features: 13MP zoom camera with wide-angle navigation cameraOne-click connection via web browserQuick setup (minutes, not hours)14-day money-back guarantee Price: $2,000-$4,095 | Cloud: $300-$780/year Ava 500 by Ava Robotics Best for: Large enterprise deployments Ava Robotics, an iRobot spinoff, offers the most sophisticated enterprise-grade telepresence solution. The Ava robot features truly autonomous navigation—users simply click a destination on a map, and Ava drives itself there. Key Features: Full autonomous navigation with LIDAR mappingCisco Webex integration for enterprise videoEnterprise-grade security and fleet management Price: ~$32,000 (purchase) or $1,000-$2,500/month (lease) Beam Pro by Awabot Best for: Healthcare and high-security environments The Beam product line, now owned by Blue Ocean Robotics and distributed by Awabot, remains a gold standard in telepresence. Trusted by Fortune 500 companies and healthcare institutions worldwide. Key Features: 17-inch widescreen display with wide-angle cameraMilitary-grade encryption for secure communicationsCrash avoidance technology and auto-docking Price: $15,000-$16,000 GoBe Robot by Blue Ocean Robotics Best for: Sustainability-focused organizations GoBe Robots positions itself as the eco-friendly choice, emphasizing CO2 reduction through reduced business travel. Each robot can eliminate up to 3.1 tons of CO2 emissions per person annually. Key Features: 4K zoomable camera with wide-angle front cameraSmooth, quiet operation with easy web interfaceFocus on sustainability metrics Price: Contact for quote (typically $10,000-$15,000) VGo Best for: Healthcare and education VGo has carved out a strong niche in healthcare and education with its reliable, enterprise-grade platform featuring lights and auto-answering capabilities for institutional deployments. Price: $4,875+ PadBot Series by Inbot Technology Best for: Budget-friendly consumer and SMB use The PadBot series offers some of the most affordable telepresence options on the market, with models ranging from consumer-friendly prices to professional-grade solutions. Price: $1,427-$14,899 (depending on model) How Do Telepresence Robot Prices Compare? What Are the Pros and Cons of Telepresence Robots vs Video Conferencing? While Zoom, Teams, and Google Meet have become ubiquitous, telepresence robots offer distinct advantages—and some trade-offs—compared to traditional video conferencing. Advantages of Telepresence Robots Physical Presence and Mobility: You're not stuck in one spot. Walk through the office, visit different departments, or conduct facility tours. This mobility creates engagement opportunities impossible with static video calls. Eye-Level Interaction: Height-adjustable displays put you at eye level with colleagues, creating more natural, equal conversations rather than looking down from a wall-mounted screen. Spontaneous Engagement: Roll up to someone's desk for a quick chat or join an impromptu whiteboard session. These "water cooler" moments are crucial for culture and collaboration. Better Engagement: People pay more attention to a robot moving toward them than a video tile on a screen. Studies show telepresence robots increase meeting engagement significantly. Disadvantages of Telepresence Robots Higher Cost: Even budget models cost $1,500+, compared to free video conferencing software. Enterprise solutions can reach $30,000+. Physical Limitations: Stairs, narrow doorways, and uneven surfaces can limit where robots can go. Most are designed for flat, indoor environments. Setup and Maintenance: Robots require charging, occasional software updates, and physical maintenance—more overhead than a Zoom subscription. Network Dependencies: Robots require reliable WiFi coverage throughout the facility. Dead zones mean your robot presence disappears. How Do You Calculate Business ROI for Telepresence Robots? For businesses considering telepresence robots, understanding the return on investment is crucial. Most organizations see positive ROI within 1-2 years from reduced travel costs alone. Cost Factors Hardware: $1,500-$32,000 per robotCloud/Software subscriptions: $300-$1,000/yearIT setup and integration: VariableMaintenance: ~5-10% of purchase price annually Savings Factors Travel Cost Reduction: The average business trip costs $1,200-$2,500. If a telepresence robot eliminates even 10-15 trips per year, it can pay for itself within 1-2 years. Example calculation: Robot cost: $5,000Average trip saved: $1,500Trips replaced per year: 5Annual savings: $7,500ROI: 150% in year one Productivity Gains: Remote executives using telepresence robots report 20-40% more effective collaboration compared to video conferencing alone. Environmental Benefits: GoBe Robots calculates that each robot can eliminate up to 3.1 tons of CO2 emissions per person annually—quantifiable ESG value for sustainability-focused organizations. Industry-Specific ROI Considerations Healthcare: Calculate specialist consultation fees, reduced patient transfers, and infection control benefits. Manufacturing: Value remote inspections, reduced downtime from faster expert access, and quality control improvements. Education: Consider student retention rates, tuition from homebound students, and reduced substitute teacher costs. Where Can You Buy Telepresence Robots? Purchasing a telepresence robot is straightforward with several options available through manufacturers and authorized retailers. Direct from Manufacturers Double Robotics: doublerobotics.comOhmniLabs: ohmnilabs.comAva Robotics: avarobotics.comAwabot (Beam): awabot.comGoBe Robots: gobe.blue-ocean-robotics.com Authorized Retailers TelepresenceRobots.com: Specializes in telepresence with comparison toolsB&H Photo Video: Carries Double Robotics productsWellbots: Premium robotics retailerDynamism: Japanese robotics specialist with US distribution Leasing Options Several manufacturers offer robotics-as-a-service (RaaS) models. Ava Robotics starts under $1,000/month. Leasing makes sense for testing before purchase, seasonal needs, or organizations preferring OPEX over CAPEX. What Does the Future Hold for Telepresence Robots? The telepresence robot market is evolving rapidly with several key trends to watch. AI Integration: Expect smarter autonomous navigation, voice control, and even AI-powered conversation assistance. 5G Connectivity: Faster, more reliable connections will improve video quality and reduce latency for smoother remote control. Mixed Reality: Integration with AR/VR headsets will create more immersive telepresence experiences. Falling Prices: As technology matures and competition increases, expect more affordable options to emerge. Is a Telepresence Robot Right for You? Telepresence robots represent a significant step beyond video conferencing, offering genuine physical presence and mobility for remote users. While the investment is higher than a video conferencing subscription, the benefits—improved collaboration, reduced travel costs, better engagement—often deliver compelling ROI. For organizations with distributed teams, regular travel requirements, or needs for remote facility access, a telepresence robot can transform how you connect. Healthcare and education institutions particularly benefit from the unique capabilities these robots provide. Start by identifying your primary use case, then evaluate robots that fit your budget and technical requirements. Many manufacturers offer trials or demonstrations—take advantage of these to experience telepresence firsthand before committing to a purchase. Compare on RoboZaps: Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Figure 03 Review 2026: Price, Release Date & Is It For Sale? URL: https://blog.robozaps.com/b/figure-03-review Author: Radica Maneva Published: 2026-02-03T00:00:00.000Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The Figure 03 is not for sale: no price, no pre-order, no waitlist as of July 2026, and the famous "$20,000" figure has no traceable source; Adcock's $400-600/month lease sketch is the only sourced consumer signal. Figure publishes official specs (5'8", 61 kg, 5-hour runtime), has built 350+ units at BotQ, and BMW Spartanburg logistics work is starting. Home timing rests on autonomy, not a date. Key takeaways: - No price, no pre-order, no waitlist, and the famous "$20,000 Figure 03" has no traceable primary source; the only sourced consumer signal is Adcock's $400-600/month home-lease sketch from May 2026. - Figure publishes an official spec card: 5'8", 61 kg, 20 kg payload, 5-hour runtime, 1.2 m/s, correcting the 5'6"/60 kg estimates that circulated, including on earlier versions of this page. - Generation truth: the viral laundry-folding demo (Aug 2025) ran on the Figure 02; the Figure 03's own record starts with Helix 02 in January 2026, and every autonomy label is Figure's own characterization of curated videos. - Real work is starting: BMW announced Figure 03 sequencing in Spartanburg's Hall 52 in June 2026 ("will now start"), and Catalyst Brands is the first named customer beyond BMW, at a Reno distribution center. - The "broad home availability late 2026" line circulating online is not a Figure statement; the real gate is Adcock's autonomy bar, and the next hardware (Figure 04, architecture locked per Adcock) may arrive before any consumer Figure 03. FAQ: Q: Can you buy the Figure 03, and is it for sale? A: No. As of July 2026 there is no public price, no pre-order, no waitlist and no retail channel. Every Figure 03 built has gone to Figure's internal programs or partner deployments like BMW Spartanburg. Q: How much does the Figure 03 cost? A: Figure has published no price, and the widely circulated "$20,000" has no traceable primary source; it does not appear in TIME's launch exclusive or anywhere on figure.ai. CEO Brett Adcock sketched a home lease of roughly $400 to $600 a month in May 2026. Directory listings claiming $130,000 (self-labeled "Not Verified") or $24,760 have no source. Q: When is the Figure 03 release date? A: There is no committed consumer date. Home alpha testing was announced in 2025, and Adcock told TIME he believes full autonomous home operation is achievable "in 2026, but it's a big push." The "broad home availability late 2026" phrasing circulating online does not come from Figure, and the first consumer Figure robot may be the next-generation Figure 04. Q: What are the Figure 03's specifications? A: Figure's official spec card lists 5'8", 61 kg, a 20 kg payload, 5-hour runtime and 1.2 m/s walking speed, with a 2.3 kWh structural battery, 2 kW wireless charging through the feet, palm cameras and fingertip sensors detecting about 3 grams of pressure. Degrees of freedom are unpublished. Q: Can the Figure 03 do laundry and dishes? A: The dishwasher, living-room and bedroom demos (January to May 2026) ran on the Figure 03 under Helix 02, and Figure describes them as fully autonomous. The famous laundry-folding demo from August 2025 actually ran on the previous Figure 02. All are Figure's curated videos without independent verification, and TIME reported at launch that the robot could do such tasks "but not without help." Q: How does the Figure 03 compare to Tesla Optimus? A: The Figure 03 has far more public evidence behind it: official specs, 350+ units built, published factory yields and a starting BMW deployment, while Tesla's redesigned Optimus 3 remains unrevealed with no order path. Neither has a price or consumer availability. Our Tesla Optimus Gen 3 tracker covers that side in detail. Q: What is Helix AI? A: Helix is Figure's in-house vision-language-action model. The original Helix (February 2025) controlled the upper body; Helix 02 (January 2026) extended control to the full robot, one network from pixels to actions, with a learned whole-body controller Figure says replaced 109,504 lines of hand-engineered C++. Q: Is the Figure 03 safe for homes? A: Figure engineered it for home safety: soft washable textiles, multi-density foam over pinch points, and a drop-tested structural battery certified to UN38.3 with UL2271 in process. Context: a former Figure safety lead's whistleblower lawsuit, which Figure disputes and has countersued, was in active discovery as of July 2026, and no Figure 03 injury has been publicly reported. Q: Who makes the Figure 03? A: Figure AI, founded in 2022 by Brett Adcock, builds the Figure 03 at its in-house BotQ facility. The company raised over $1 billion at a $39 billion valuation in September 2025. Note its July 2026 warning that unauthorized secondary sales of Figure stock are void. Our Figure AI company review covers the funding and investor detail. Can you buy the Figure 03, and how much does it cost? No. As of July 16, 2026 the Figure 03 has no public price, no pre-order, no waitlist and no retail channel; Figure's product page offers no way to reserve one. Every unit built so far has gone to Figure's internal programs or to partner deployments. And the price you have probably read is worse than unofficial: the "$20,000 Figure 03" that saturates search results has no traceable primary source at all. We checked, in detail, below. One disclosure up front: this is desk research built on the sources linked throughout, primary wherever they exist, not a hands-on test. Where a claim is Figure's own characterization, we say so. Figure 03 specifications: the official card Unusually for this industry, Figure publishes a real specification card on its official product page, and it corrects what most of the web (including earlier versions of this page) had been reporting: the Figure 03 is listed at 5'8" and 61 kg, not the 5'6" and ~60 kg that circulated from launch-era inference. One inconsistency worth preserving rather than smoothing over: the launch post says the Figure 03 has "9% less mass" than the 70 kg Figure 02, while the official weights imply roughly 13%. Both numbers are Figure's own. We track every claim's verification status in the Figure 03 record in our robot database. The misinformation audit: what circulates vs what the record holds Figure 03 search results are saturated with figures that fail a provenance check. Before this review leans on any number, here is the audit. The honest price picture is therefore short: Figure has never published a price, and the nearest sourced consumer signal is CEO Brett Adcock describing a home lease in the four-to-six-hundred-dollars-a-month range during a Sourcery factory tour published May 1, 2026. What is the Figure 03, really? The Figure 03 is Figure AI's third-generation humanoid, announced October 9, 2025, and its first designed for homes as well as commercial work: soft washable textiles instead of exposed machined parts, wireless charging you step onto, audio built for voice interaction, and a sensor suite designed around its Helix AI. TIME put it on the Best Inventions 2025 cover. It is simultaneously a consumer product that does not exist yet as a product, and a working industrial machine starting at BMW this summer, and holding both of those in view is the key to evaluating it. The company behind it was founded in 2022 by Brett Adcock (previously co-founder of Archer Aviation and Vettery). Its September 2025 Series C raised over $1 billion at a $39 billion post-money valuation, led by Parkway Venture Capital with Brookfield, NVIDIA, Macquarie, Intel Capital, Salesforce, T-Mobile Ventures and Qualcomm Ventures participating; the earlier Series B round included Microsoft, the OpenAI Startup Fund, NVIDIA and Jeff Bezos. Figure ended its OpenAI model partnership in February 2025 in favor of its in-house Helix. The deeper company story lives in our Figure AI review; this page stays on the robot. From Figure 01 to Figure 03: the generation story Figure has shipped a new robot generation roughly every 18 months, and each one redefined what the company was selling. Sources: Figure's Figure 02 at BMW account, with BMW confirming the 30,000-X3 figure in its June 2026 release. The lineage matters for one practical reason: most of the viral "Figure robot" demos before 2026, including the famous laundry folding, ran on the Figure 02, not the 03. This review keeps the generations straight throughout. Hardware deep-dive The hands Figure calls hands the hardest problem in humanoids, and the 03's are its most distinctive hardware: an embedded wide-angle, low-latency camera in each palm for close-range vision where the head cameras are occluded, and fingertip pads that detect forces down to about 3 grams, light enough to feel a paperclip and to catch grip slippage before an object drops. Actuators across the robot run twice as fast as the 02's with better torque density, which is what the fast, fluid manipulation in recent demos rides on. Degrees of freedom remain unpublished, a notable gap in an otherwise unusually open spec card. Senses and voice The camera system doubles the frame rate, quarters the latency and widens each camera's field of view by 60% versus the Figure 02, per the launch post, engineered explicitly to feed Helix rather than a human teleoperator. The audio hardware got the same treatment for a different reason: a speaker roughly twice the size and four times the power of the 02's, because Figure sees speech as the robot's primary interface in a home. Battery and charging The 2.3 kWh battery is a structural member of the torso, the product of what Figure describes as a 94% energy-density improvement across three battery generations, and it carries the safety engineering this category needs: multi-layer cell-failure containment, survival of a one-meter drop onto concrete in any orientation, UN38.3 certification achieved per the launch post, and UL2271 described as "in process" in the earlier battery post, wording worth preserving because only one of the two is claimed as complete. Figure also cut the battery's cost 78% versus the 02's, the one component-cost figure the company has published itself. Charging is the everyday-life feature: 2 kW wireless coils in the feet, so the robot steps onto a stand rather than plugging in. Materials and serviceability The 03 is wrapped in soft textiles rather than hard shells, washable and replaceable without tools, with multi-density foam over pinch points. This is home-safety engineering, but it is also a maintenance philosophy: fabrics can be swapped like trim rather than repaired like bodywork. At 61 kg it is among the lighter full-size humanoids, though heavier than 1X's 30 kg NEO, a difference that reflects the two companies' opposite bets on capability versus inherent softness. Helix: the brain Helix is Figure's in-house vision-language-action model, the reason it walked away from OpenAI's models in February 2025. The architecture story has two chapters, and most coverage blurs them. The original Helix, introduced February 20, 2025, controlled the robot's upper body, hands, arms and torso, from vision and language, and could run one shared task across two robots. Helix 02, released January 27, 2026, extended control to the entire robot: one neural network from pixels to actions across legs, torso, arms and fingers, with a learned whole-body controller Figure calls System 0, which the company says replaced "109,504 lines of hand-engineered C++." Helix 02 is also where the 03's palm cameras and tactile fingertips got wired into the policy itself, and the platform data flows through the robot's 10 Gbps mmWave offload into fleet-wide training. Two things keep this section honest. Helix's demonstrations are Figure's own curated videos, in Figure's environments, with "fully autonomous" being Figure's characterization; no independent lab has verified any of it. And the demo cadence, impressive as it is, is a research trajectory, not a product ship date: Adcock's own bar for a consumer release, per TIME's launch feature, is a robot that handles most household tasks autonomously through the day, and he has not claimed Helix is there. The demo ledger: what was shown, on which robot, with what claim The pattern a careful reader should hold: every autonomy label above is the manufacturer's own, the environments are Figure's, and the strongest independent datapoint remains TIME's launch-day observation that the robot could fold clothes and load a dishwasher "but not without help." The distance between that observation (October 2025) and the Helix 02 demo reel (2026) is either genuine progress or better curation, and no third party has yet been in a position to say which. Where the Figure 03 actually works BMW Spartanburg. On June 25, 2026, BMW announced that Figure 03 "will now start" work in Hall 52, the assembly and logistics hall of Plant Spartanburg: picking unsorted parts from containers and sequencing them into trolleys for just-in-sequence delivery to the line (BMW Group release). Figure's own post five days later described the workflow, including pulling heavy metal carts, as whole-body Helix 02 coordination. BMW disclosed no robot count, so any Spartanburg fleet-size figure you read is invented. The project stands on the Figure 02's record in the same plant, detailed in the generations table above, and BMW's production-control VP called Spartanburg "the birthplace of humanoid robotics in BMW Manufacturing's operational day-to-day activities." Catalyst Brands. On May 26, 2026, Figure announced an agreement with Catalyst Brands, the parent of JCPenney, Aéropostale and Brooks Brothers, for supply-chain work beginning at a Reno, Nevada distribution center. Figure framed the deal as its first commercial bridge into a Brookfield portfolio company (Brookfield invests in both), and it is the first customer Figure has publicly named beyond BMW. No robot count or start date was disclosed. Figure's own fleet. Most Figure 03s work for Figure. The company's April production update said over 350 units had been "delivered" from its factory and itemized where: internal R&D groups, data collection, end-to-end housework development, and commercial use-case development. No public retail sale exists, so "delivered" describes allocation, not customers taking ownership. By June, Adcock was saying the deployed fleet, roughly 740 robots per aggregated reports of his remarks, exceeded the company's human headcount; treat that as a founder statement rather than an audited count. BotQ: the manufacturing engine Figure builds the 03 at BotQ, its in-house plant, and publishes production data almost no rival matches. The first line is rated for up to 12,000 robots a year against a stated goal of 100,000 robots over four years. The April 2026 ramp update claimed output moved from one robot a day to one an hour in under 120 days, with end-of-line first-pass yield above 80%, 99.3% first-pass yield on the battery line, 9,000+ actuators produced across 10+ SKUs, 50+ in-process inspection points and 80+ functional tests per robot. Reported cadence as of May was 60 to 70 robots a week, and Adcock has said Figure is driving Chinese content out of its supply chain. The strategic point: vertical integration is the only credible route to a mass-market price, it is the same story Tesla tells about Optimus, and Figure's version is running and documented while Tesla's Fremont line is still being converted and 1X's Hayward plant is months into its ramp. The price question, in full Assembling every real price-relevant datapoint produces a short table, which is itself the finding. Reading between those lines: Figure's cost curve is real and documented, its pricing intent points at lease economics rather than a sticker (which would put it head-to-head with the 1X NEO's published $499 a month), and any specific dollar figure for the robot itself is currently fiction. Enterprise deal terms with BMW and Catalyst are undisclosed. Adcock's 24/7-operation mechanics from the same interview, 4 to 5 hours per charge, about an hour to recharge on the dock, hot-swap between units in under 30 seconds, sketch what a lease would actually be selling: continuous coverage, not a gadget. The availability tracker Every dated statement on the record about when a person can get one: The forward-looking wrinkle is Figure 04: Adcock has described a next generation with its architecture locked after design review, a new hand, and a coming Helix 3, in interviews from May 2026, with no official announcement yet. For anyone hoping to buy a Figure 03, that cuts two ways: it signals the platform cadence is accelerating, and it raises the real possibility that the first Figure robot consumers can order is not an 03 at all. Safety: the engineering and the lawsuit The engineering case is published and specific: soft washable coverings, multi-density foam at pinch points, the drop-tested structural battery with UN38.3 achieved and UL2271 in process, and a mass 9 to 13% below the Figure 02's. No injury involving a Figure 03 in a home or at BMW has been publicly reported as of this writing. The legal case is also specific, and pending. In November 2025, Robert Gruendel, Figure's former head of product safety, filed a federal whistleblower suit alleging he was fired in retaliation for warning executives that the robots "were powerful enough to fracture a human skull," and describing an incident in which a malfunctioning robot gashed a steel refrigerator door (CNBC). Figure says he was terminated for poor performance and calls the allegations falsehoods, and it filed counterclaims over retained documents in January 2026. A July 6, 2026 court order shows the case in active discovery with counterclaims live. These are contested allegations, not findings, and this page will track the outcome either way, because a home robot company's safety culture is a buying criterion. What the skeptics say The strongest skeptical framing comes from the same TIME feature that carried Figure's launch story. After noting the 03's improvements, its reporter observed that the company's central promise, that major performance increases are just around the corner, "is also a convenient way for this company with huge costs, an unproven product, and no publicly-disclosed revenue to justify its soaring valuation." That is the bear case in one sentence: a $39 billion valuation resting on curated demos, undisclosed deal terms and a consumer product with no ship date. The counter-evidence is genuinely accumulating, a second automotive generation at BMW, a second named customer, published yields, but it is worth noticing that every quantitative proof point still originates from Figure itself. The first independent benchmark, whether a customer's throughput data or a production-unit review, does not exist yet for any home humanoid, Figure's included. Looking for a Figure 03 alternative? Figure 03 has no public order path, and not every buyer needs its home-first design. Our Figure 03 alternatives guide compares seven options by use case, price, buyer access and deployment evidence. Use it when the question is what to choose instead; this review remains the source for Figure 03 hardware, Helix, availability and deployments. Who this robot is for, right now Operations leaders in logistics, retail supply chains or manufacturing: Figure is signing named deployments, publishes more production evidence than any competitor, and is realistically contactable about pilot work; the BMW sequencing use case is a template worth copying into your own diligence. Households: there is nothing to buy, no queue to join, and the honest planning assumption is that home availability arrives when Helix meets Adcock's bar, possibly on Figure 04 hardware rather than this robot. Investors: Figure stock is not publicly tradable, and the company warned in July 2026 that unauthorized secondary sales through several named platforms are void; treat any Figure 03 "pre-IPO access" pitch accordingly. What to watch next Five dated threads will move this page. The Helix release cadence, which has run roughly every six to ten weeks and points at a Helix 3. BMW publishing any Spartanburg robot count or throughput figure, which would be the first customer-side metric in the home-humanoid race. The Catalyst deployment actually starting in Reno. The Gruendel case's next docket milestones. And any first consumer mechanism, a waitlist, a lease program, or a real price, which would instantly become the biggest story in consumer robotics. We re-verify this page at each. The bottom line The Figure 03 is the most substantiated home-humanoid program in the world and still not a product you can buy, and both halves of that sentence matter. On the evidence side it now has an official spec card, published yields from a running factory, a second generation of BMW deployment and the most prolific demo record in the industry. On the caution side, every performance claim still originates from Figure, the famous $20,000 price is traceable to nothing, the home date does not exist, and the company is defending a safety lawsuit while marketing a robot for your living room. Watch the Helix cadence and the first consumer mechanism; the Figure 03 record in our robot database tracks the verified state as it changes. --- # Humanoid Robot News Tracker: What Is Shipping, Piloting, and Still Unproven URL: https://blog.robozaps.com/b/humanoid-robot-news Author: Radica Maneva Published: 2026-02-03T00:00:00.000Z Updated: 2026-08-30T22:59:07.769Z Last fact-checked: 2026-08-31T00:00:00.000Z Category: news TL;DR: As of August 10, 2026, Unitree has priced its Shanghai IPO at 150.80 yuan a share, the FCC blocks new foreign-produced humanoids and quadrupeds from US equipment authorization on a Buy American content test, 1X has public NEO order terms, Figure has a named BMW workflow and Atlas has committed 2026 deployments. None of that is proven mass delivery or general autonomy. Key takeaways: - 1X publishes NEO pricing, a refundable deposit and a 2026 US delivery statement, but delivery scale and unsupervised home reliability remain unproven. - Figure has a named BMW workflow and company-reported production figures; Unitree lists H2 Plus at $100,000 but directs buyers to contact sales. - Tesla has no public Optimus order channel, while all announced Atlas deployments for 2026 are already committed. - This hub owns current verified status; four dated weekly roundups remain live as historical archives. FAQ: Q: What is the latest humanoid robot news? A: As of August 10, 2026, the biggest buyer-relevant update is Unitree pricing its Shanghai IPO at 150.80 yuan a share on August 6, a valuation near 61 billion yuan, with subscriptions opening August 10 and listing expected around August 20. It is a capital event and changes no price or order route. The FCC's July 28 Covered List action still governs which new models can enter the US market, and this week's analyses confirmed it turns on a Buy American content test rather than the maker's nationality. 1X's live NEO order terms, $20,000 or $499 per month, remain the clearest consumer purchase path. Q: Which humanoid robots can buyers order in 2026? A: 1X publishes NEO order terms of $20,000 for ownership or $499 per month, with a $200 refundable deposit and US deliveries stated to begin in 2026. Unitree's official store lists H2 Plus at $100,000 but marks it unavailable and directs buyers to contact the company. Figure has no public price or open order page; Atlas's announced 2026 deployments are committed. Tesla has no public Optimus order page. Q: How does RoboZaps verify humanoid robot news? A: RoboZaps starts with the manufacturer, customer, regulator or standards body closest to the claim. We separate demonstrations, pilots, production claims, order terms and operating deployments, and label company-reported figures. We also check price, availability, task, customer, supervision and support rather than repeating a headline on its own. Q: What is the difference between this tracker and the weekly news posts? A: This page is the current, source-checked status hub. The weekly posts are dated archives showing what was being reported in February and March 2026. They remain useful historical snapshots, but their old targets and forecasts should not be read as current buying guidance. Q: How often is the humanoid robot news tracker updated? A: The tracker is updated when a primary source changes a buyer-relevant fact such as price, order status, delivery timing, a named deployment, production evidence or a material hardware release. RoboZaps does not add filler simply to maintain a weekly publishing cadence. Latest verified humanoid robot news Last checked August 31, 2026. This humanoid robot news tracker records changes that affect what a humanoid robot can do, whether it can be ordered, and how strong the deployment evidence is. Vendor specifications and production figures are labelled as company claims; a demonstration is not described as a shipment or an independently proven deployment. August 24: XPENG's robotics unit raises about $900 million, still nothing to order XPENG said its robotics business raised over US$900 million at a post-money valuation above US$6.3 billion, which it called the largest single private financing round in China's embodied-AI sector, led by IDG Capital with Tencent and Alibaba as strategic investors, per the company announcement. The money funds the IRON humanoid and XPENG's physical-AI models, with mass production targeted for the end of 2026 and deliveries in 2027. For buyers nothing changes: IRON has no public price or order page. Full detail is in This Week in Humanoid Robot News, August 24 to 31, 2026. August 26: A humanoid wins the Beijing games 100m in 8.64 seconds, a controlled-course result At the second World Humanoid Robot Games in Beijing, Tiangong Ultra, built by X-Humanoid, won the 100-metre final in 8.64 seconds as the games closed on August 26, with the mark falling from 9.39 to 8.86 to 8.64 across the week. It is an organizer-timed run on the games' own course, not a like-for-like race against a person, and it says nothing about whether a robot can do useful work. Several robots crashed into the barrier past the line and some sparked. The useful reminder came from an Oregon State University roboticist: progress is measured in warehouses and factories, not on a sprint track. August 19: Unitree lists in Shanghai and the stock jumps, order terms unchanged Unitree began trading on Shanghai's STAR Market on August 19 under ticker 688836, the first pure-play humanoid maker on a mainland Chinese exchange. Against a 150.80 yuan IPO price the shares closed the first day at 845 yuan, up about 460 percent, according to Reuters, after touching 1,100 yuan intraday, a gain of as much as 629 percent per the South China Morning Post. The sale raised about 904 million US dollars. It is a financing and sentiment event that changes nothing a buyer can order today. The G1, H2 and R1 still sell through the same sales-assisted channel. Full detail is in This Week in Humanoid Robot News, August 17 to 24, 2026. August 19-26: Beijing's robot week shows scale, not new order pages Beijing ran the 2026 World Robot Conference from August 19 to 23, with more than 300 exhibitors and over 3,000 products, then opened the second World Humanoid Robot Games on August 22 with 666 teams and about 2,056 robots from 16 countries. Both are demonstrations and benchmarks. Neither came with a public price or an order page for a buyer outside China. August 10: Unitree's IPO subscription breaks a STAR Market record before an imminent debut Unitree opened the subscription for its Shanghai STAR Market listing on August 10 under ticker 688836. Reporting put the retail tranche near 5,500 times covered and the wider book more than 8,000 times oversubscribed, a record for a technology company on the board. The online lottery left about 19,414 winning numbers against roughly 9.78 million valid accounts. It is a financing and sentiment event that changes nothing a buyer can order today; the Unitree H2 record shows what is actually purchasable. The trading debut is expected between August 17 and 21 and had not happened as this update went out. On the issue price the company is valued at about 219 times 2025 earnings, while its prospectus guides first-half 2026 revenue growth of 36 to 45 percent, down from 333 percent a year earlier. Those are company figures. Full brief: This Week in Humanoid Robot News, August 10 to 17, 2026. August 14: BYD's Xiao Di humanoid stays unverified BYD told Chinese outlets, including the South China Morning Post, that it would show a Xiao Di service humanoid in early August at its Di Space centres, with floated specs of 1.61 metres, 58.5 kilograms and 31 degrees of freedom. As of August 17, RoboZaps still cannot confirm an independent public unveiling or an official spec sheet. Treat it as a company plan, not a shipment. August 6: Unitree prices its IPO at 150.80 yuan, with no change to what buyers can order Unitree set its STAR Market issue price at 150.80 yuan a share on August 6, valuing the company at about 61 billion yuan, roughly 9 billion US dollars. It is selling 40,446,434 new shares, 10 percent of the enlarged capital, to raise about 6.1 billion yuan. Subscriptions open August 10 and listing is expected around August 20. The price is about 45 percent above the 104 yuan the market expected and about 219 times 2025 earnings, against prospectus guidance of 36 to 45 percent first-half revenue growth, down from 333 percent a year earlier, and adjusted net profit 6 to 22 percent lower. This is a capital event. It changes no Unitree price, catalogue entry or order route, and the G1, R1 and H2 buying paths are the same as they were before pricing. August 4: The FCC ban turns on a content test, not the maker's nationality Reporting and legal analyses published this week set out the scope of the July 28 Covered List action. A device is covered only if it combines ground mobility, remote operation, weight over 4.4 pounds including any dock, environmental sensors, network connectivity of at least 200 kbps and controlling software. Connected vehicles, rail vehicles, drones, uncrewed underwater vehicles, FDA-regulated medical robots and fixed industrial arms are outside it. The decisive term is "foreign-produced", which the FCC reads through the Buy American cost test rather than by country of headquarters. On Sidley's reading a device is covered unless domestic components exceed 65 percent of component cost, rising to 75 percent in 2029, so a US-headquartered maker with an offshore bill of materials can be caught. Conditional Approval applications are due to the Department of War by January 1, 2028, and a separate waiver lets already-authorized devices take qualifying software updates until at least January 1, 2029. August 7: Nucleus Robotics bills humanoid labour by the hour, under human supervision German startup Nucleus Robotics left stealth on August 7 with a first deployment at an unnamed German industrial customer, completed in under 90 days. It sells physical labour by the hour rather than robots, which is a different commercial route to the purchase and lease terms tracked elsewhere on this page. The deployment evidence is real but bounded. Robots ran under intensive human oversight for the first three weeks, and Nucleus has not disclosed the humanoid platform it uses, how often an operator intervenes, or the share of work done autonomously. Treat it as supervised operation with a paying customer, not proven autonomy. July 28: The FCC blocks new foreign-made humanoids and quadrupeds The FCC updated its Covered List on July 28 to add advanced robotic devices, which it defines as mobile robots such as humanoids and quadrupeds, produced in foreign countries. Covered equipment cannot receive new FCC equipment authorization, and that authorization is a precondition for importing, marketing or selling most electronics in the United States. The restriction reaches new models only. Devices already purchased, models authorized before July 28 and federal government purchase and use are unaffected, and producers can apply to the Department of War for a Conditional Approval. This is the first regulatory change that alters which humanoids can legally enter the US market rather than what they can do. Unitree, the most exposed maker, launched its Shanghai STAR Market IPO on July 31; that is a capital event and does not change its catalogue, prices or order routes. July 30: Gemini Robotics 2 extends whole-body control to humanoids Google DeepMind released Gemini Robotics 2 alongside Gemini Robotics ER 2 for embodied reasoning and Gemini Robotics On-Device 2 for local execution. It is the first model in the family to run legs, torso, arms and hands under a single policy rather than the upper body alone. DeepMind demonstrated it on Apptronik's Apollo 2 and names Franka Duo, Boston Dynamics and Agile Robots among supported platforms. The published success rates are the useful part: 45.7 to 76.3 percent on general whole-body manipulation and 36 to 92 percent on multi-finger tasks, with unscrewing a light bulb at 92 percent and screwing one back in at 36 percent. This is a capability release with measured numbers attached, not a shipped product. No robot changed price, order status or delivery timing because of it. July 16: SoftBank exercises its put option, Hyundai moves to own all of Boston Dynamics Hyundai's July 16 statement confirms SoftBank has exercised the put option from the 2021 acquisition and that Hyundai-side shareholders are pursuing SoftBank's entire remaining stake, roughly 10 percent of the company. Bloomberg and Reuters report a price of about $325 million; Hyundai says final terms, including price, are still going through internal approvals. For buyers nothing changes: Atlas still has no public price or order page. Full ownership removes the IPO deadline the put option enforced and clarifies who funds the 2026 Atlas deployment commitments. July 18: AgiBot and Unitree debut new robots at WAIC 2026, none checkout-ready AgiBot debuted the A3 Ultra and three other products at WAIC 2026 in Shanghai, while Unitree showed the GD01, a 2.7 meter transformable robot it says is mass-produced and can carry a person. All performance figures are company claims. No WAIC debut came with a public price or order page. Treat the specs as vendor numbers until third parties test production units. July 10: Tesla clears the Fremont line for Optimus, production not started Tesla posted a timelapse showing the Fremont Model S and X line torn down in 46 days to make room for the first Optimus production line. Elon Musk, who walked the site on July 1, said production had not begun and that initial output will be extremely slow, citing roughly 10,000 unique parts. Tesla's stated target of first output in late July to August 2026 is a company claim; the July 22 Q2 earnings call is the next dated checkpoint. For buyers nothing changed: Optimus still has no price, no order page and no delivery commitment. The teardown only confirms the factory conversion moved from claim to visible fact. July 9: 1X updates NEO's hands while order terms remain live 1X announced a new NEO hand with 25 degrees of freedom including the wrist, tactile sensing, force-controlled backdrivable tendons and IP68 protection. Those are manufacturer-published specifications, not an independent dexterity benchmark. The more important buyer signal is the live NEO order page: $20,000 for ownership or $499 per month, a $200 refundable deposit, and US deliveries stated to start in 2026. 1X also says remote Experts may supervise tasks the robot does not yet know. The terms are concrete, but they do not prove delivery volume, unsupervised autonomy or ordinary household reliability. June 30: Figure 03 reaches a named BMW workflow Figure reported that Figure 03 arrived at BMW Plant Spartanburg for a sequencing workflow involving part handling and cart movement in Hall 52. The company calls this a demonstration. That is stronger evidence than an unplaced demo video because the customer, site and task are named, but it is not the same as audited fleet performance. In an April production update, Figure said BotQ had produced more than 350 Figure 03 robots and demonstrated a one-robot-per-hour cycle time. Both numbers are company-reported. They matter as manufacturing signals, but buyers still need task success, interventions, uptime and customer acceptance. June 1: Unitree lists H2 Plus at $100,000, but it is not checkout-ready Unitree introduced H2 Plus. Unitree publishes about three hours of battery life, roughly 7 kg rated arm payload, 15 kg peak arm payload and optional dexterous hands. The company also warns that the humanoid industry remains early and that some functions are still under development. The official Unitree store lists H2 Plus at $100,000, but the product is marked unavailable and buyers are directed to contact Unitree. That makes $100,000 a current catalog price, not proof of an in-stock, one-click retail transaction. April 22: Tesla installs Optimus production lines, with no public orders Tesla's Q1 2026 shareholder update said first-generation Optimus production lines were being installed in anticipation of volume production. Tesla did not publish a retail price, order page, deposit or customer delivery date. This is evidence of production preparation, not consumer availability. January 5: Atlas enters production with 2026 deployments committed Boston Dynamics unveiled the product version of Atlas and said manufacturing would begin immediately. It also said every 2026 deployment was committed to Hyundai and Google DeepMind, with additional customers planned for early 2027. Atlas has announced production and committed 2026 deployments, but no public price or open allocation. What is actually available in 2026? Availability is not one binary label. A deposit page, a sales-assisted catalog listing, a named factory demonstration and a committed enterprise allocation are different commercial states. The RoboZaps robot database keeps the product records separate from this chronology. Use it to compare current catalog entries after checking whether the stated commercial status matches the task you need. How to read a humanoid robot announcement The same word can hide very different evidence. RoboZaps uses this ladder when deciding how much weight to give an update: Demonstration: a robot completes a selected task under conditions controlled by the developer.Named pilot or workflow: the customer, site and task are identified, but the operating duration or success rate may still be missing.Production claim: the vendor reports a factory, unit count or cycle time. This says something about manufacturing, not necessarily customer use.Repeat deployment: a customer uses the robot over time and reports cycles, throughput, uptime or incidents.Comparable benchmark: methods and results allow different systems to be assessed on the same task. This remains uncommon for commercial humanoids. The NIST Humanoid Robot Baseline Performance Benchmark is a useful sign of where neutral testing may go. NIST describes proposed locomotion and manipulation tasks with quantifiable metrics and is still seeking participation. It is not yet a certification scheme or a published league table. What RoboZaps checks before calling an update real Source: is the claim traceable to the manufacturer, named customer, regulator, filing or standards body?Commercial status: is there a price, deposit, order page, sales contact, delivery market and support path?Task evidence: what did the robot do, where, for how long and under what operating constraints?Human assistance: was teleoperation, remote supervision, reset work or data-labeling labour involved?Measurement: are throughput, interventions, uptime, incidents and task success published, or only an edited video?Claim ownership: vendor-reported specifications and production numbers are labelled instead of presented as independently verified facts. Funding rounds, IPO plans and long-range production targets can show capital and intent. They do not prove that robots have shipped, that customers are using them, or that the economics work. Those claims stay outside the current-status list until a stronger source changes the evidence. For the investor-side view of those events, see our humanoid robot stocks guide. Weekly humanoid robot news archive The posts below are retained as dated historical snapshots. Their reported targets, forecasts and market figures belong to their publication period; use this hub for the current source-checked status. February 8, 2026: archived roundup of reports on China's manufacturing scale, Unitree IPO expectations, Figure 03 and Helix, Mobileye's Mentee deal, and FieldAI funding.February 28, 2026: archived roundup contrasting factory-pilot reporting with skepticism about humanoid dexterity, plus shipment estimates and teleoperation transparency.March 7, 2026: archived roundup covering Tesla production ambitions, reported Neura financing, Xiaomi factory trials, governance and emerging manufacturers.March 16-22, 2026: archived roundup of reported IPO and production plans, Tesla timetable claims, supplier partnerships, Chinese standards work and debate over the humanoid form factor.July 9-16, 2026: archived roundup of Unitree's IPO final stage, the Walden/LimX/AI² capital wave, 1X's NEO hands, Tesla's Fremont teardown and Boston Dynamics' Spot delivery test.July 13-20, 2026: archived roundup of Hyundai's move to full Boston Dynamics ownership, WAIC 2026 debuts from AgiBot and Unitree, EngineAI's fighting league and the continuing IPO queue.July 27 - August 3, 2026: archived roundup of the FCC Covered List action, Unitree's Shanghai IPO launch, Gemini Robotics 2 and China's sweep of the strongest humanoid patent portfolios.August 3-10, 2026: archived roundup of Unitree's IPO pricing, the fine print of the FCC robot ban, Nucleus Robotics' factory deployment, Walden's wheeled humanoid and HII's robot welding contract.August 10-17, 2026: archived roundup of Unitree's record STAR Market IPO subscription and imminent debut, and BYD's then-unverified Xiao Di plan.August 17-24, 2026: archived roundup of Unitree's record STAR Market trading debut, Beijing's World Robot Conference and Humanoid Robot Games, and BYD's Xiao Di showroom unit. Browse the complete RoboZaps News archive for dated reporting. The tracker is updated selectively when the underlying evidence changes; RoboZaps does not manufacture a weekly post when there is nothing worth verifying. Frequently asked questions What is the latest humanoid robot news? As of August 10, 2026, the biggest buyer-relevant update is Unitree pricing its Shanghai IPO at 150.80 yuan a share on August 6, a valuation near 61 billion yuan, with subscriptions opening August 10 and listing expected around August 20. It is a capital event and changes no price or order route. The FCC's July 28 Covered List action still governs which new models can enter the US market, and this week's analyses confirmed it turns on a Buy American content test rather than the maker's nationality. 1X's live NEO order terms, $20,000 or $499 per month, remain the clearest consumer purchase path. Which humanoid robots can buyers order in 2026? 1X publishes NEO order terms of $20,000 for ownership or $499 per month, with a $200 refundable deposit and US deliveries stated to begin in 2026. Unitree's official store lists H2 Plus at $100,000 but marks it unavailable and directs buyers to contact the company. Figure has no public price or open order page; Atlas's announced 2026 deployments are committed. Tesla has no public Optimus order page. How does RoboZaps verify humanoid robot news? RoboZaps starts with the manufacturer, customer, regulator or standards body closest to the claim. We separate demonstrations, pilots, production claims, order terms and operating deployments, and label company-reported figures. We also check price, availability, task, customer, supervision and support rather than repeating a headline on its own. What is the difference between this tracker and the weekly news posts? This page is the current, source-checked status hub. The weekly posts are dated archives showing what was being reported in February and March 2026. They remain useful historical snapshots, but their old targets and forecasts should not be read as current buying guidance. How often is the humanoid robot news tracker updated? The tracker is updated when a primary source changes a buyer-relevant fact such as price, order status, delivery timing, a named deployment, production evidence or a material hardware release. RoboZaps does not add filler simply to maintain a weekly publishing cadence. Bottom line The current market is not simply divided into robots that exist and robots that do not. 1X has explicit public order terms, Unitree has a sales-assisted catalog listing, Figure has a named factory workflow, Tesla is preparing production, and Atlas has committed enterprise deployments. Those are five different evidence levels. The useful news is the change in commercial status, task evidence or measured performance. Everything else should remain labelled as a demonstration, company claim or forecast until stronger evidence appears. Compare on RoboZaps: Unitree H2, Figure 03, Tesla Optimus, Spot and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robot: Definition, How They Work, Types and Examples (2026) URL: https://blog.robozaps.com/b/humanoid-robot Author: Radica Maneva Published: 2026-02-02T10:00:00.000Z Updated: 2026-08-01T23:06:20.383Z Last fact-checked: 2026-07-13T00:00:00.000Z Category: insights TL;DR: A humanoid robot is a physical robot with a human-inspired body structure, often combined with human-like movement or interaction so it can work in spaces, with tools and around people. Current systems combine sensors, electric actuators, real-time control and AI, but most commercial deployments still focus on constrained industrial tasks rather than general household work. Key takeaways: - Humanoid describes a physical robot's human-inspired body structure; it does not guarantee intelligence, autonomy or lifelike appearance. - The form factor is useful when a robot must navigate spaces and use tools, shelves or workstations designed for people. - Sensors, actuators, control software, AI, locomotion and onboard power must work as one fast feedback loop. - Factories and warehouses are the leading proving grounds, while dependable general-purpose home robots remain a longer-term goal. - A customer pilot, recurring production deployment and edited demonstration are different levels of evidence and should be reported separately. FAQ: Q: What is a humanoid robot in simple terms? A: A humanoid robot is a physical robot with a human-inspired body structure, often combined with human-like movement or interaction. It usually has a torso, arms and a head, and many have two legs, but it does not need to copy every human feature. Q: Are all humanoid robots powered by AI? A: No. Some use machine learning for perception, language or movement, while others follow programmed routines or rely on an operator. Humanoid describes the robot's form, not its level of intelligence. Q: What is the difference between a humanoid and an android? A: A humanoid has a human-inspired physical body. An android is a humanoid designed to look convincingly human, often with a realistic face or skin-like materials. Every android is humanoid, but most mechanical humanoids are not androids. Q: Why do humanoid robots have two legs? A: Two legs can help a robot use stairs, narrow passages and work areas designed for people. Legs also add balance, energy and safety challenges, so some humanoids use wheels when stability and efficiency matter more. Q: What can humanoid robots do today? A: Current systems can perform selected tasks such as moving materials, tending machines, handling parts, research experiments and guided social interaction. Most deployments operate within defined workflows and may still use human supervision or remote assistance. Q: Can you buy a humanoid robot? A: Yes, some development and research platforms can be ordered, and vendors offer industrial systems through sales programs. Availability, support and permitted uses vary, and a purchasable platform is not necessarily a ready-made home assistant. Q: Are humanoid robots safe around people? A: They can be designed for safe applications, but appearance alone says nothing about safety. The robot, task, payload, speed, tools, controls and environment must be assessed together, with appropriate limits, stops, separation and operating procedures. A humanoid robot is a physical robot with a human-inspired body structure, often combined with human-like movement or interaction. Most have a torso, head and paired limbs, but the exact form varies. The practical reason is compatibility: a human-scale machine can potentially use doors, stairs, shelves, tools and workstations built for us. The International Federation of Robotics offers a narrower reference point: human-like appearance, the ability to perform tasks in environments designed for people, and performance enhanced by human-like sensing. Definitions vary at the edges, but conversational behavior alone does not make software or a machine humanoid. That promise is easy to overstate. A demonstration is not a dependable worker, and some systems remain remotely operated or limited to a narrow task. This guide covers the definition, machinery, types, examples, uses and limitations that matter in 2026. What is a humanoid robot? Humanoid is a design strategy: give a physical robot enough human-like structure, often paired with human-like movement or interaction, to work in spaces built for people. In the clearest cases, it has two legs, two arms, a torso and a head. Human-scale reach and mobility can let it approach a bench, pick up a tool, carry a container or pass through a standard doorway. The word humanoid describes form and function, not intelligence. Some humanoids can plan multi-step actions and respond to speech. Others repeat pre-programmed motions or depend on an operator. A human-shaped shell does not prove that a machine understands its surroundings, and advanced AI does not require a human-shaped body. “General-purpose” describes an ambition, not current capability. Most deployments constrain broadly capable hardware to one workflow, such as moving known parts between fixed stations. Purpose-built automation is usually better for a stable, high-volume process. Humanoids are most interesting where tasks change and using existing carts, shelves and controls could reduce integration work. Even then, the label says nothing about payload, accuracy, autonomy, availability, price or whether the robot performs a useful task. Those claims need separate evidence. Humanoid meaning: what the word covers beyond robots Humanoid means having human form or human characteristics. The word predates robotics and still carries uses outside it. Anthropologists apply it to extinct hominins in the fossil record, and fiction uses it for any human-shaped being, biological or mechanical. A humanoid robot is therefore one specific kind of humanoid, and the distinction matters when you read a specification. “Humanoid” on its own describes shape. “Humanoid robot” tells you the thing is a machine under some form of control. Shape is also the easiest part of any claim to confirm from a photograph, which is why the rest of this guide treats capability as a separate question. Defining characteristics and important exceptions There is no single visual checklist that covers every humanoid. Researchers and manufacturers make different trade-offs, but several characteristics recur. Human-like structure: a torso with paired limbs, often including articulated hands or grippers.Human-scale reach: the ability to work with objects, controls and surfaces positioned for people.Legged or upright mobility: usually two legs, although some upper-body humanoids use wheels for stability and efficiency.Whole-body control: coordinated motion across multiple joints so the robot can balance, reach and manipulate an object at the same time.Human-facing interaction: cameras, microphones, displays, gestures or speech that help people understand and direct the system. Two legs are common, not mandatory. A social robot may have a head, torso and arms on a wheeled base. A warehouse system may be unmistakably humanoid above the waist but use a non-human lower body. Conversely, a biped with no face can still be humanoid if its structure and working envelope are designed around human environments. Hands are another exception. Five-fingered hands are useful for tools built around human grip patterns, but they add mechanical and control complexity. Many industrial humanoids use simpler grippers, interchangeable end effectors or task-specific hands. The better question is not “Does it copy a person exactly?” but “Which parts of human anatomy help it do the intended job?” The boundary also excludes some technologies that use human language or imagery. A voice assistant, animated avatar or conventional industrial arm is not humanoid simply because it communicates with people. An exoskeleton follows the human form but is worn and powered through a person rather than acting as an independent robot. Conversely, a faceless biped can be humanoid when its body, reach and movement are designed around human spaces. The useful test is embodied structure and function, not whether the machine appears friendly or lifelike. Humanoids versus androids, cobots and industrial robots A humanoid resembles the human body in its overall structure or movement. An android is a narrower idea: a humanoid designed to look convincingly human, often with a face, skin-like materials and expressive behavior. Every android is humanoid, but an exposed mechanical biped such as Atlas is not usually called an android. A cobot, or collaborative robot, is defined by how it is used around people rather than by its shape. Many cobots are fixed robotic arms. A humanoid could form part of a collaborative application, but its human-like appearance does not make it safe to share an unguarded workspace. Safety depends on the complete system: the task, payload, tooling, speed, controls, environment and foreseeable contact. A conventional industrial robot is commonly a fixed arm or specialized machine inside a designed cell. It can be faster, more precise and more economical than a humanoid at a repetitive, stable task. Humanoids are being developed for the awkward middle ground where work changes, the environment was built for people, and rebuilding the process around fixed automation is difficult. Humanoids are not replacements for every robot. A gantry, conveyor, autonomous mobile robot or six-axis arm is often the better tool. The form earns its place only when human compatibility and flexibility outweigh its extra complexity. How humanoid robots work A humanoid combines a mechanical body, sensors, actuators, real-time control, task software and a power system. These parts form a continuous loop: sense the world, estimate what is happening, choose an action, move, measure the result and correct the next movement. The loop must run quickly because a delayed correction can turn a small balance error into a fall. Perception and sensing Cameras provide color, depth and motion information. Depth cameras or lidar can help build a three-dimensional map. Microphones capture speech and environmental sounds. Joint encoders report the angle and speed of each joint, while inertial measurement units estimate the robot’s orientation and acceleration. Force, torque and tactile sensors reveal contact with the floor, a tool or a person. No sensor tells the whole story. The robot combines several streams to estimate its body and surroundings, identify objects and track them through a task. Lighting, reflective surfaces, clutter and partial occlusion can undermine a system that worked in a clean demonstration. Actuators and mechanical design Actuators create movement at the joints. Electric motors are now common in commercial designs, typically paired with gearing and control electronics. Other research platforms have used hydraulic systems. Designers balance strength, speed, weight, efficiency, impact tolerance, heat and backdrivability—the extent to which an external force can move a joint. Degrees of freedom describe the independent ways a body can move. More joints can expand reach and dexterity, but each adds components, wiring, calibration and control work. A useful humanoid is therefore not the one with the largest specification number. It is the one whose joints, hands and range of motion reliably match the task. Control, autonomy and AI Low-level controllers regulate joint position, velocity and force many times per second. Higher layers coordinate balance, walking, reaching and grasping. A task planner then connects those skills into actions such as locating a tote, walking to it, lifting it and placing it at another station. Machine learning can improve perception, grasp selection and motion policies. Reinforcement learning in simulation lets a system practice large numbers of movements without repeatedly damaging physical hardware. Demonstration data from teleoperation can teach a robot how a person performs a task. Language and vision models may help translate an instruction into a plan, but safety-critical motion still needs constraints, monitoring and fallbacks. Autonomy is a spectrum. A robot may work independently inside a known cell, request help when confidence drops, or be directly teleoperated. Remote assistance is not inherently a flaw; it can be a practical way to recover from rare situations and collect training data. What matters is whether a vendor clearly states where the operator remains in the loop. Locomotion, balance and power Honda’s development history distinguishes static and dynamic walking. In static walking, the center of gravity stays within the supporting sole; during dynamic walking it may move outside that support area. In either case, the robot must shift its mass, place a foot, absorb contact and prepare the next step while accounting for the payload and ground. Model-based control, learned policies or a combination of both can generate the motion. Whole-body control matters because moving an arm or lifting a box changes the forces that keep the robot upright. Power is a hard constraint. Onboard batteries must supply computers, sensors and many motors while remaining light enough to carry. Demanding motion shortens operating time and creates heat. Current designs address this with scheduled charging, replaceable batteries, automated battery swaps or work patterns that alternate active periods and charging. An impressive movement is only commercially useful if the system can repeat it for the required shift. Main humanoid robot types Industrial humanoids are built for factories, warehouses and logistics sites. They emphasize payload handling, repeatability, integration with business systems and serviceability. Their first tasks tend to be narrow but valuable: material movement, machine tending, parts sequencing or tote handling. Research humanoids are platforms for studying locomotion, manipulation, perception and human-robot interaction. They may perform remarkable movements without being products designed for continuous customer operation. Research results often transfer into later commercial machines. Social and service humanoids communicate with people through speech, gaze, gestures or screens. Some use wheeled bases because a stable platform is more practical for reception, education or guided interaction than legs. Their “human” qualities are concentrated in the upper body and behavior. Telepresence and avatar robots extend a remote person’s sight, voice and actions into another location. Their value is presence and manipulation rather than independent decision-making. Entertainment and expressive androids prioritize lifelike appearance or performance, sometimes at the expense of general mobility. These categories overlap. A research platform can become an industrial product, and an industrial machine may use teleoperation during training or exception handling. Classifying a robot by its intended deployment is usually more informative than judging it only by appearance. A short history of humanoid robots Humanoid machines appeared in stories and mechanical automata long before electronic robotics. Modern humanoid research emerged when engineers could combine sensing, computation and powered joints in one system. Waseda University describes WABOT-1, completed in 1973, as the first full-scale humanoid robot. It combined a limb-control system with vision and conversation functions, could walk on two legs and could grip objects. WABOT-2 followed in 1984 as a specialist robot designed to read music and play an electronic organ. Honda began bipedal research in 1986. Its P-series machines developed dynamic walking and self-contained operation before ASIMO was introduced in 2000. ASIMO brought stable movement, stairs, interaction and an approachable scale into public view. Honda’s later account is also a useful reality check: decades of demonstrations exposed unresolved questions about falls, autonomy and safe operation around people. During the 2010s, research platforms pushed dynamic balance and whole-body movement, while social robots explored public interaction. The current phase is more commercial. Companies are narrowing the problem to repeatable work in customer facilities, using electric actuation, improved perception, simulation, teleoperation data and learned control. The emphasis has moved from “Can it perform this once?” toward “Can it deliver a safe, measurable result every shift?” Representative real-world examples The examples below illustrate different stages and design choices. They are not a ranking; for a broader comparison, see our regularly updated guide to the best humanoid robots and the RoboZaps humanoid database. Boston Dynamics Atlas: an electric industrial humanoid built around dynamic mobility and material handling. Boston Dynamics says the product version can operate autonomously, by teleoperation or through a tablet interface, and is being introduced through selected industrial deployments.Agility Robotics Digit: a biped optimized for logistics rather than copying every detail of a person. GXO describes a multi-year service agreement integrating Digit with other warehouse automation at its SPANX facility after a 2023 pilot.Figure robots: general-purpose industrial humanoids being developed through factory work. BMW Group says Figure 02 supported production of more than 30,000 BMW X3 vehicles over ten months in 2025. On June 30, Figure reported that Figure 03 had arrived at Plant Spartanburg and completed its first demonstration of a logistics-sequencing workflow.Apptronik Apollo 2: Apptronik’s current data-collection and training platform, operating in bipedal and wheeled configurations at Robot Park and selected partner sites. The company describes Apollo 3 as its upcoming commercial fleet.Unitree G1: a compact biped sold in base and EDU configurations. Unitree lists secondary-development support for G1 EDU, making that configuration the relevant choice for research and custom development.Ameca: an expressive social humanoid focused on communication and human-robot interaction rather than factory locomotion. Product names alone hide important differences. Some machines are commercial products, some are early-access platforms, some are pilots, and others remain research systems. Buyers should look for evidence of the exact task, environment, duty cycle, supervision and support model—not just a polished video. How to judge a humanoid robot claim Start by naming the stage. A laboratory demonstration shows a behavior under those conditions. A customer pilot adds integration and real-world variation. A recurring production deployment is stronger evidence of reliability and economic value. None should be described as another. Ask whether the task is continuous or edited into highlights. Useful details include operating time, completed cycles, intervention rate, payload and recovery procedure. A robot may complete the visible motion autonomously while a person selects objects, resets failures or drives it between takes. Check who supplied the number. Manufacturer specifications and customer statements are primary sources for what those organizations claim, not independent tests. Because hardware and software change quickly, dates matter. We label company-reported results and remove claims that cannot be traced to a suitable source. Customer evidence can still improve confidence. It can confirm that a robot worked at a named site, on a defined task and for a stated period rather than only in its maker’s laboratory. It does not replace an independent benchmark: both companies may approve the release, and intervention rates or failed cycles may be omitted. Strong evaluation states what the customer confirms, what the manufacturer reports and what remains unknown. The same discipline applies to availability. “Unveiled,” “in development,” “accepting inquiries,” “in pilot,” and “shipping to customers” are different states. A deposit page does not prove general delivery, and a forecast is not a shipment. This guide avoids a giant specification table because those fields age at different speeds; the database and individual reviews are better places to maintain model-level detail. Uses and deployment environments Factories and warehouses are the leading proving grounds because the tasks can be valuable and the environment can be controlled. A humanoid may move parts between stations, feed a machine, build an order or perform repetitive handling at a workstation originally laid out for a person. Existing tools and aisles create the strongest argument for the form factor. Useful deployment metrics go beyond whether the robot completed the motion. Operators need throughput, intervention frequency, uptime, charging time, safety events, maintenance effort and the cost of connecting the robot to existing equipment and software. A pilot becomes more persuasive when the customer expands it after observing those measures over time. Without that context, a successful cycle is evidence of technical capability, not proof of a dependable business case. Hazardous inspection and response are another long-term fit. A human-compatible robot could enter a site built for people, climb steps, use equipment and keep an operator away from heat, chemicals or unstable structures. The difficulty is that these environments are less predictable than factories, so reliability and remote operation remain crucial. Service settings include reception, education, guided interaction and research into care support. In these roles, speech, gaze and gestures may matter more than walking. Physical assistance around vulnerable people sets a much higher safety bar than delivering information from a fixed position. Home use attracts attention because homes are the ultimate human-designed environment. They are also full of clutter, pets, children, thresholds, delicate objects and tasks that change constantly. Today’s realistic consumer options are limited compared with the general-purpose helper imagined in marketing. Our guide to humanoid robots for home use separates available products from announced ambitions. For a sector-by-sector view, see the dedicated guide to humanoid robot applications. The important test in any setting is whether the robot improves a specific workflow after integration, supervision, maintenance and downtime are included. Current limitations, safety and ethical concerns Reliability and economics Real work requires thousands of uneventful cycles, recovery from variation and predictable maintenance. An unfamiliar object, snagged cable or changed surface can expose weaknesses hidden by a scripted demo. Fleet support, spare parts, training and software integration matter as much as movement. The economic comparison must include the complete system. Hardware is only one line item; integration, site preparation, supervision, charging, maintenance and lost output during faults also count. A specialized machine may deliver a better return where the task is stable. For current purchasing context, use the separate guides to humanoid robot costs and where to buy a humanoid robot. Physical and operational safety A human-scale robot has moving mass, pinch points and stored energy. A fall, unexpected reach or dropped payload can injure someone even when the system is marketed as collaborative. NIST’s task-based framework evaluates the application, tooling, possible contact and environment rather than treating the robot alone as the risk. For industrial applications, ISO 10218-1 covers the robot as a machine, while ISO 10218-2 covers its integration into complete applications and cells. Other environments may fall under different standards or regulations. Practical controls include speed and force limits, monitored separation, safe stops, guarded zones, fall planning, payload limits and clear procedures for faults. Operators also need to know when the system is autonomous, remotely assisted or awaiting input. Human-like appearance should never substitute for validated safety behavior. Privacy, work and accountability Humanoids can carry cameras, microphones and operational logs through spaces where people work or live. Organizations need rules for what is recorded, why it is retained, who can access it and whether remote operators can see sensitive environments. Cybersecurity is physical safety when a connected system can move. Workforce effects depend on deployment choices. A robot may remove repetitive lifting, create technical roles or change staffing, but those outcomes are not automatic. Employers should involve workers early, measure actual ergonomic and productivity results, and provide training for changed roles. Claims about wholesale job replacement are forecasts, not facts. Responsibility must also be explicit. When a robot’s action involves a manufacturer, software provider, integrator, operator and customer, incident reporting and decision authority cannot be left vague. These issues are explored further in our guide to the challenges in humanoid robotics. What comes next Developers are concentrating on a small number of industrial tasks, improving reliability through customer work, and using teleoperation and simulation to gather training data. Hands, perception, battery management, safe recovery and fleet software matter more than theatrical new movements. Hardware will also become easier to operate as a fleet rather than as a one-off machine. That means remote diagnostics, controlled software releases, shared task libraries, maintenance records and clear escalation when a robot reaches an unfamiliar state. A technically capable body without those operating systems is difficult to deploy at scale. Expect uneven progress. A robot may become dependable at parts handling while remaining poor at household cleanup. Skills learned in one facility may transfer imperfectly to another with different lighting, containers or layouts. The general-purpose vision will be assembled from many constrained successes, not delivered as a single leap. The clearest evidence will be operational: repeat customers, longer unsupervised runs, documented task performance, maintainable hardware and deployments that expand after a pilot. Our humanoid robot company guide tracks the organizations building these systems, while this page stays focused on the definition and technology. Humanoid robots are already real, but that includes research platforms, pilots and production tools. Near-term progress depends on proving safe, reliable and economically useful performance in bounded jobs. Keep exploring: the cheapest humanoid robots you can buy, and our humanoid robot news tracker for what changed this week. Compare on RoboZaps: Figure 03, ASIMO, Ameca, Apptronik Apollo, Figure 02 and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in the Workplace: Transforming Efficiency & Productivity [2026] URL: https://blog.robozaps.com/b/humanoid-robots-in-workplace Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Updated: 2026-08-01T10:46:41.151Z Category: insights TL;DR: Humanoids entered the workplace on cost math: unit prices fell 40%+ between 2022 and 2024 (Bain), and a $60,000 robot running 20 hours a day works out near $1.64 per hour over five years. BMW's Spartanburg plant proved the model with Figure robots on a live line. With nearly 8 million manufacturing workers short globally, the near-term dynamic is augmentation. Key takeaways: - Humanoid unit costs dropped at least 40% between 2022 and 2024 (Bain & Company), with the Unitree G1 at $13,500 matching roughly a year of US minimum-wage labor. - The napkin math: a $60,000 mid-range humanoid running 20 hours a day over five years works out near $1.64 per hour including maintenance and energy. - BMW's Spartanburg plant is the flagship deployment: Figure robots handled component insertion, quality checks and material transport on a live line. - The demand side is a projected global shortage of nearly 8 million manufacturing workers. - The WEF expects 85 million jobs displaced but 97 million created by 2030 — augmentation, not replacement, is the near-term dynamic. FAQ: Q: What are humanoid robots in the workplace? A: Humanoid robots in the workplace are bipedal, human-shaped robots powered by AI that perform physical and cognitive tasks alongside human employees. They can navigate human-built environments, use existing tools, and adapt to dynamic work settings without requiring infrastructure modifications. Models like the Figure 02, Agility Digit, and Tesla Optimus Gen 2 represent the current state of the art. Q: Which industries are adopting humanoid robots first? A: Manufacturing, warehousing and logistics, automotive assembly, healthcare, food service, and retail are the earliest adopters. BMW, Amazon, and Mercedes-Benz are already piloting humanoid robots on factory floors and in fulfillment centers as of 2025-2026. Q: How much do humanoid robots cost for workplace deployment? A: Workplace humanoid robots range from $13,500 for the Unitree G1 to over $150,000 for advanced models like the Boston Dynamics Atlas. Between 2022 and 2024, unit costs dropped by at least 40%. See our complete cost guide for detailed pricing. Q: Will humanoid robots replace human workers? A: Humanoid robots are more likely to augment human workers than replace them entirely in the near term. They excel at repetitive, dangerous, or physically demanding tasks, freeing humans for creative, strategic, and interpersonal roles. A global labor shortage of nearly 8 million manufacturing workers projected by 2030 creates demand that robots can help fill. Q: What tasks can humanoid robots perform in the workplace? A: Material handling, assembly line work, quality inspection, warehouse picking and packing, customer service, security patrols, cleaning, food preparation, patient assistance in healthcare, and collaborative tasks requiring human-robot teamwork. Q: How do humanoid robots compare to traditional industrial robots? A: Unlike fixed industrial robots or AGVs, humanoid robots navigate human-built spaces without infrastructure modifications, perform a wide variety of tasks with dexterous hands, and learn new skills through natural language instruction. Learn the differences in our cobot vs. robot comparison. Q: What are the safety standards for humanoid robots at work? A: ISO 25785-1, published in draft form in May 2025, establishes safety standards for "industrial mobile robots with actively controlled stability"—the classification covering humanoid robots. Companies must also comply with ISO 10218 and ISO/TS 15066 collaborative robot standards. Q: What is the ROI of humanoid robots in the workplace? A: Most workplace humanoid robots achieve ROI within 12-24 months. Running 20+ hours daily at approximately $3-8 per hour equivalent cost versus $22-40+ per hour for human labor, the economics are compelling—especially in high-wage markets with chronic labor shortages. See our ROI analysis. Q: When will humanoid robots become common in workplaces? A: Industry analysts project steady growth through 2030 followed by rapid acceleration. Market projections range from $38 billion to over $200 billion by 2035. Early adopters are deploying now, with mainstream adoption expected between 2028-2032. Q: How can my company prepare for humanoid robots? A: Start by identifying repetitive, physically demanding, or hard-to-staff roles. Pilot one or two humanoid robots, invest in employee training for human-robot collaboration, and develop a phased integration roadmap. Explore available humanoid robots on Robozaps to begin your evaluation. Humanoid Robots in the Workplace: The Definitive 2026 Guide to Transforming Efficiency and Productivity Humanoid robots in the workplace are no longer a futuristic concept—they're actively transforming how businesses operate across manufacturing, logistics, healthcare, retail, and beyond. In 2026, companies like BMW, Amazon, and Mercedes-Benz are deploying bipedal, AI-powered robots on factory floors and in warehouses, fundamentally reshaping productivity, safety, and the economics of human labor. This comprehensive guide explores how humanoid robots are revolutionizing workplaces, which models lead the market, what deployment costs and ROI to expect, and how your organization can prepare for the robotic workforce revolution. Whether you're a business executive evaluating automation, an operations manager seeking efficiency gains, or simply curious about the future of work, this guide covers everything you need to know about humanoid robots in the workplace in 2026 and beyond. Why Humanoid Robots Are Entering the Workplace Now Four converging forces are pushing humanoid robot adoption toward an inflection point in 2026: 1. Unprecedented Advances in AI and Dexterity Today's humanoid robots can walk, run, jump, and navigate complex terrain with remarkable stability. More importantly, AI-powered fine motor skills now enable precise, human-like hand movements. The Figure 02, for example, uses its Helix AI system to perform intricate assembly tasks that were impossible for robots just two years ago. Similarly, the Astribot S1 demonstrates dexterity that rivals human hands in speed and precision. 2. Simplified Training Through Natural Language Gone are the days when deploying a robot required months of specialized programming. Modern humanoid robots accept natural language instructions, making robot management feel more like people management. Workers can demonstrate a task once, and AI models generalize the movement to new contexts. Because these robots are human-shaped, they train in the same environments where they'll operate—no special simulation setups required. 3. Rapidly Falling Costs Between 2022 and 2024, the unit cost of humanoid robots dropped by at least 40%, according to Bain & Company research. The Unitree G1 starts at approximately $13,500—matching the annual cost of minimum-wage labor in the US. Meanwhile, EU labor costs rose 5% from 2023 to 2024 alone. The economic crossover point where robots become cheaper than human workers is arriving faster than anyone predicted. For a full breakdown of current pricing, see our humanoid robot pricing guide. 4. A Global Labor Crisis The world faces a demographic reckoning. Populations are aging in major industrial economies, fewer working-age people are entering the labor force, and a global shortage of nearly 8 million manufacturing workers is projected by 2030. Add chronic difficulties in hiring for in-person positions, mounting opposition to immigration in high-income nations, and government reshoring incentives through tariffs and domestic subsidies—and the business case for workplace humanoid robots becomes overwhelming. Key Industries Adopting Humanoid Robots in the Workplace Manufacturing and Assembly Manufacturing leads humanoid robot adoption, and for good reason. Factory floors are structured environments with repetitive tasks that benefit enormously from 24/7 automation. BMW has deployed Figure 02 robots at its Spartanburg, South Carolina facility for parts handling and quality inspection. Mercedes-Benz is testing the Apptronik Apollo for heavy material transport on its assembly lines. Key manufacturing applications include: Material handling: Moving parts, components, and finished goods between stationsAssembly line support: Inserting components, tightening fasteners, and performing sub-assembliesQuality inspection: Using computer vision to detect defects at superhuman accuracyBin picking and sorting: Selecting correct parts from mixed containersPalletizing and depalletizing: Loading and unloading shipping containers and pallets Warehousing and Logistics E-commerce growth has created insatiable demand for warehouse labor. Amazon's fulfillment centers already operate with thousands of wheeled robots, but humanoid robots add a crucial capability: they can navigate multi-level facilities, climb stairs, operate elevators, and use the same equipment designed for human workers. The Agility Digit was purpose-built for logistics environments. Standing 5'9" with arms designed for tote handling, Digit can pick, pack, and transport goods across warehouse aisles at consistent speeds without fatigue. Agility Robotics opened its RoboFab production facility with capacity for 10,000 Digit robots annually—the first dedicated humanoid robot factory in the world. Healthcare Hospitals and elder care facilities face severe staffing shortages. Humanoid robots are stepping in for non-clinical support: transporting medications and supplies, assisting with patient mobility, sanitizing rooms, and providing companionship for elderly residents. The Sanctuary AI Phoenix is pioneering general-purpose intelligence that allows a single robot to handle diverse healthcare support tasks. Learn more in our guide to humanoid robots in healthcare. Retail and Hospitality Retail environments demand flexibility—inventory management, customer assistance, shelf stocking, and cleaning all require different skills at different times. Humanoid robots' general-purpose nature makes them ideal for retail applications. UBTECH's Walker S has been deployed in retail showrooms for customer interaction and product demonstrations. For deeper analysis, see our coverage of humanoid robots in retail and humanoid robots in hospitality. Construction Construction faces some of the worst labor shortages and highest injury rates of any industry. Humanoid robots are being developed for tasks like drywall installation, painting, site inspection, and material transport. The ability to navigate uneven terrain, climb scaffolding, and use standard power tools makes humanoid form factors particularly valuable here. Food Service and Preparation Quick-service restaurants and commercial kitchens are piloting humanoid robots for food preparation, assembly, and serving. The consistency and hygiene benefits are significant—robots don't call in sick, don't require breaks, and can maintain precise portion control across thousands of servings. Top Humanoid Robots for Workplace Deployment in 2026 The market for workplace humanoid robots has matured rapidly. Here's how the leading models compare: For comprehensive reviews of each model, visit our best humanoid robots guide, or browse all available humanoid robots on Robozaps. The Economics: Cost, ROI, and Productivity Gains Cost Comparison: Humanoid Robots vs. Human Workers The economic case for humanoid robots in the workplace grows stronger every quarter: A mid-range humanoid robot like the Figure 02 at an estimated $60,000 (no official price was ever published), operating 20 hours per day over a 5-year lifespan, costs roughly $1.64 per hour—including estimated maintenance and energy costs. That's a fraction of even minimum wage in any developed country. For a detailed financial analysis, see our ROI of humanoid robots guide and humanoid robot cost breakdown. Productivity Multiplier Beyond cost savings, humanoid robots multiply productivity in ways that purely financial calculations understate: 24/7 operation: Three shifts of work from a single unit—effectively tripling output per "employee"Zero downtime: No breaks, no vacations, no turnover-related productivity lossConsistent quality: Error rates drop to near-zero for repetitive tasksData generation: Every action produces data that drives continuous process improvementScalability: Need to double output? Deploy twice as many robots, instantly trained via shared AI models ROI Timelines by Industry Real-World Workplace Deployments in 2025-2026 BMW Spartanburg: Figure 02 in Automotive Assembly BMW's manufacturing facility in Spartanburg, South Carolina has become the flagship showcase for Figure 02 deployment. The robots handle component insertion, quality verification, and material transport between assembly stations. BMW reported a 15% improvement in line efficiency in areas where Figure 02 robots operate alongside human workers. Amazon Fulfillment: Agility Digit in Logistics Amazon has been testing Agility Digit robots in select fulfillment centers for tote recycling—moving empty containers from one area to another. While the robots currently handle a limited set of tasks, the deployment has validated that humanoid robots can integrate with existing warehouse management systems and work safely alongside human associates. Mercedes-Benz: Apptronik Apollo in Manufacturing Mercedes-Benz partnered with Apptronik to test Apollo humanoid robots for heavy parts delivery on its production lines. The robots navigate the factory floor autonomously, delivering components weighing up to 25 kg to specific workstations on schedule. GXO Logistics: Multi-Robot Warehouse Operations Global logistics provider GXO has piloted multiple humanoid robot models in its warehouses, testing them against specific KPIs for pick rates, error rates, and integration complexity. Early results show humanoid robots performing at 70-85% of human speed for picking tasks, with the gap closing as AI models improve. Humanoid Robots vs. Cobots vs. Traditional Industrial Robots Understanding where humanoid robots fit relative to existing automation solutions is crucial for workplace planning. For a deep dive into collaborative robots, see our guide on what is a cobot. Safety Standards and Regulations Workplace safety is paramount when deploying humanoid robots. In May 2025, a working group including representatives from A3 (Association for Advancing Automation), Agility Robotics, and Boston Dynamics published the draft ISO 25785-1 standard—the first international safety standard specifically addressing humanoid robots in workplaces. The standard notably avoids the word "humanoid," instead using the technical descriptor "industrial mobile robots with actively controlled stability"—a deliberate choice that focuses on safety-relevant characteristics rather than appearance. Key safety considerations for workplace deployment include: Speed and force limiting: Robots must reduce speed and force when humans are nearbyEmergency stop systems: Physical e-stop buttons accessible to all workersCollision detection: Force-torque sensors and AI-based proximity awarenessOperational zones: Defined areas with appropriate safety protocols for each zoneRisk assessment: ISO 12100-based risk assessment required before deploymentCybersecurity: Protecting robot control systems from unauthorized access Companies must also comply with existing standards including ISO 10218 (industrial robot safety) and ISO/TS 15066 (collaborative robot safety), which apply to humanoid robots operating in shared workspaces. The Human Impact: Jobs, Skills, and Collaboration Will Humanoid Robots Replace Human Workers? Deployment is one half of the picture; exposure is the other. For which specific jobs humanoid robots are most likely to take, and when, see what jobs humanoid robots will take. The evidence suggests augmentation, not replacement, as the primary near-term dynamic. The World Economic Forum estimates that while 85 million jobs may be displaced by automation by 2030, 97 million new roles will emerge—many involving human-robot collaboration, robot fleet management, AI training, and maintenance. Key factors tempering job displacement fears: Labor shortages already exist: Many positions humanoid robots will fill are currently vacant—there aren't enough humans willing or available to do themNew roles emerge: Robot fleet managers, AI trainers, human-robot interaction designers, and maintenance technicians are entirely new career pathsComplex judgment remains human: Creative problem-solving, emotional intelligence, complex decision-making, and relationship building remain firmly human advantagesGradual adoption: Even aggressive forecasts project decades before humanoid robots are widespread enough to significantly impact total employment For deeper analysis of the employment implications, read our coverage of the economic impact of humanoid robots on the job market. Preparing Your Workforce Organizations successfully integrating humanoid robots invest in: Change management: Communicating clearly about why robots are being deployed and how human roles will evolveUpskilling programs: Training workers to supervise, maintain, and collaborate with robotsNew role creation: Establishing positions like robot fleet coordinator, human-robot workflow designer, and AI behavior specialistGradual integration: Starting with pilot programs that let workers experience and shape the human-robot dynamic before full-scale deployment Implementation Roadmap: Bringing Humanoid Robots to Your Workplace Phase 1: Assessment (Months 1-3) Identify high-value use cases: repetitive tasks, hard-to-staff roles, dangerous activitiesCalculate total cost of ownership vs. current labor costsEvaluate facility readiness (floor surfaces, clearance, WiFi coverage)Research applicable safety standards and regulations Phase 2: Pilot (Months 3-9) Select 1-2 robot models for evaluation (browse options on Robozaps)Deploy in a controlled environment with limited scopeMeasure KPIs: throughput, error rates, uptime, employee satisfactionIterate on workflows and human-robot interaction protocols Phase 3: Scaling (Months 9-18) Expand to additional use cases and departmentsIntegrate with existing IT/OT systems (WMS, ERP, MES)Develop standardized operating proceduresTrain broader workforce on collaboration skills Phase 4: Optimization (Ongoing) Leverage robot-generated data for continuous improvementShare learned behaviors across robot fleetExplore adjacent applications as AI capabilities expandPlan for next-generation upgrades Challenges and Limitations in 2026 Despite rapid progress, workplace humanoid robots face real challenges that organizations must plan for. Our challenges in humanoid robotics guide covers these in greater depth. Dexterity gaps: Fine manipulation tasks (threading a needle, handling flexible materials) remain difficult, though improving rapidlyBattery life: Most humanoid robots operate 2-4 hours before needing to recharge, requiring charging infrastructure and schedulingEnvironmental sensitivity: Wet, dusty, or extreme-temperature environments can impair sensors and motorsEdge cases: Unexpected situations—dropped objects, unusual requests, equipment malfunctions—still challenge AI systemsIntegration complexity: Connecting robots to legacy enterprise systems requires significant IT effortRegulatory uncertainty: Standards are still evolving, creating compliance risks for early adoptersSocial acceptance: Some workers may resist robot colleagues, requiring thoughtful change management The Future of Humanoid Robots in the Workplace Looking ahead to 2027-2035, several trends will accelerate workplace adoption: Cost compression: Tesla's target of $20,000 for Optimus and Unitree's sub-$14,000 pricing for G1 signal that prices will continue falling, making robots accessible to small and mid-size businessesRobots-as-a-Service (RaaS): Subscription models ($2,000-$5,000/month) eliminate upfront capital requirementsMulti-robot coordination: Fleets of humanoid robots working in coordinated teams, managed by a single supervisorCross-industry skill transfer: A robot trained in one factory shares its knowledge with robots in entirely different industriesRegulatory maturation: Clear standards will reduce adoption friction and liability concerns Market projections range from $38 billion to over $200 billion by 2035, with manufacturing, logistics, and healthcare driving the majority of deployment. For our complete analysis, see the future of humanoid robots and humanoid robot market size. Conclusion: The Workplace Transformation Is Underway Humanoid robots in the workplace aren't a question of "if" but "when"—and for forward-thinking organizations, "when" is now. The convergence of AI breakthroughs, plummeting costs, demographic pressures, and proven real-world deployments makes 2026 the year to move from observation to action. Companies that explore use cases today, pilot humanoid robots in controlled settings, and develop human-robot collaboration strategies will hold decisive competitive advantages as the technology scales. Those that wait risk falling behind in an automation race that is accelerating faster than even the most optimistic forecasters predicted. Ready to explore humanoid robots for your workplace? Browse the full range of humanoid robots available on Robozaps → For more insights on humanoid robotics, explore our guides on the best humanoid robots of 2026, applications of humanoid robots, and Tesla Optimus alternatives and competitors. --- # Humanoid Robots in Agriculture: Complete Guide to Farming's Robotic Revolution [2026] URL: https://blog.robozaps.com/b/humanoid-robots-in-agriculture Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Updated: 2026-08-01T10:46:56.355Z Last fact-checked: 2026-07-12T00:00:00.000Z Category: insights TL;DR: Agricultural robots answer a 20% shrinkage in US farm labor since 2006: AI weeding hits 95%+ accuracy while cutting herbicide 70-90%, and monitored growing operations report about 40% lower labor costs. The blockers are field reality — 2-4 hour batteries, weather, human-beating picker speed — and the 2-5% crop margins that make capital risky. Key takeaways: - The US farm labor force has shrunk about 20% since 2006 while production demands grew — the structural opening for agricultural robots. - AI weeding is the proven win: 95%+ crop-vs-weed accuracy with 70-90% less herbicide via targeted micro-dosing. - Monitored growing operations report about 40% lower labor costs and 15% better crop quality from continuous monitoring. - The blockers are physical: 2-4 hour battery life against field distances, weather sensitivity, and experienced human pickers who still beat robots on speed for many crops. - Agriculture's razor-thin margins (2-5% for many crops) and short harvest windows make large robot investments risky. FAQ: Q: What are humanoid robots in agriculture? A: Humanoid robots in agriculture are bipedal, AI-powered robots designed to perform farming tasks such as harvesting, planting, weeding, crop monitoring, and produce handling. Unlike wheeled agricultural bots, humanoid form factors navigate uneven terrain, work in existing farm infrastructure, and handle delicate crops with human-like dexterity. Models like the UBTECH Walker S are already deployed commercially in vertical farming operations. Q: How are humanoid robots used in farming? A: Humanoid robots perform selective harvesting of delicate crops, precision weeding, crop health monitoring via computer vision, pruning and thinning, greenhouse operations, vertical farming tasks, livestock management support, and post-harvest sorting and packing. Q: Can humanoid robots harvest crops? A: Yes. AI vision systems and dexterous soft-grip hands enable humanoid robots to identify ripe produce and harvest it without damage. Systems demonstrate selective harvesting using machine learning to pick only ripe, healthy crops at near-human speeds. Q: How much do agricultural humanoid robots cost? A: Prices range from $13,500 for the Unitree G1 to $100,000+ for specialized platforms. The UBTECH Walker S costs $60,000-$100,000. Robots-as-a-Service models are emerging at $2,000-$5,000/month for seasonal use. See our pricing guide and cheapest humanoid robots guide for details. Q: What is the ROI of humanoid robots in agriculture? A: Most agricultural humanoid robots achieve ROI in 1-3 growing seasons. With labor savings of $40,000-$80,000 annually per unit plus reduced crop waste and herbicide costs, the economics are increasingly compelling—especially for high-value crops and operations with chronic labor shortages. Q: Which humanoid robots are used in agriculture? A: The UBTECH Walker S leads commercial agricultural deployment through its partnership with Agroz in Malaysia. Other models being tested include the Unitree H1, Apptronik Apollo, and Figure 02. Q: Will humanoid robots replace farm workers? A: Humanoid robots primarily address existing labor shortages rather than displacing workers. Agriculture already can't find enough workers—billions of dollars in crops go unharvested annually. Robots fill gaps where human workers are unavailable, particularly for physically demanding seasonal tasks. Q: How do humanoid robots handle delicate crops? A: Soft-grip end-effectors and force-torque sensors allow precise pressure control. AI vision assesses ripeness, size, and defects in real-time, while machine learning models optimize grip strength and pick angle for each crop variety, achieving damage rates of 1-3% compared to 3-8% for fatigued human workers. Q: What are the limitations of humanoid robots in agriculture? A: Current limitations include battery life (2-4 hours), weather sensitivity, difficulty on soft or uneven soil, variable speed compared to experienced human pickers, and high upfront costs relative to farm margins. These challenges are being addressed through improved weatherproofing, longer batteries, and RaaS pricing models. Q: What is the future of humanoid robots in agriculture? A: By 2030, humanoid robots are expected to handle 15-25% of harvesting for high-value crops. Vertical farming will see the fastest adoption. Costs are projected to fall to $10,000-$30,000 for agricultural models by 2028-2031, making robots accessible to mid-size farms. Humanoid Robots in Agriculture: The Complete 2026 Guide to Farming's Robotic Revolution Humanoid robots in agriculture are emerging as a transformative force in one of humanity's oldest industries. As farms worldwide grapple with severe labor shortages, rising wages, climate pressures, and growing demand for food production, AI-powered bipedal robots offer a compelling solution. In 2026, companies like UBTECH, Unitree, and specialized agri-tech startups are deploying humanoid robots in vertical farms, orchards, and greenhouses—proving that the future of farming is increasingly robotic. This comprehensive guide explores how humanoid robots in agriculture are enhancing efficiency, which models and technologies lead the field, real-world deployment examples, cost and ROI analysis, and what the next decade holds for robotic farming. Why Agriculture Needs Humanoid Robots in 2026 The Global Farm Labor Crisis Agriculture faces perhaps the most acute labor shortage of any industry. In the United States alone: The farm labor workforce has declined by 20% since 2006 while production demands have increasedAn estimated 40-60% of US farm workers are unauthorized immigrants, creating a volatile and unstable workforceFarm labor costs have risen 35-45% over the past decade, squeezing already-thin marginsCrops worth billions of dollars go unharvested annually because there simply aren't enough workers to pick them The situation is similarly dire across Europe, Japan, Australia, and other developed agricultural economies. Young workers increasingly reject the physically demanding, seasonal nature of farm work, and immigration restrictions further constrain labor supply. Why Humanoid Form Factors Matter for Farming Agriculture already uses extensive automation—tractors, combine harvesters, GPS-guided sprayers, and drone monitoring systems. But many critical farming tasks require human-like capabilities that these machines can't replicate: Selective harvesting: Identifying and picking only ripe produce while leaving unripe fruit on the vine requires visual judgment and gentle, precise hand movementsWorking in existing infrastructure: Greenhouses, orchards, and packing facilities were designed for human workers—humanoid robots fit without expensive retrofitsMulti-terrain navigation: Walking between rows, stepping over irrigation lines, and navigating uneven ground demands bipedal locomotionDexterous manipulation: Pruning, thinning, grafting, and handling delicate crops requires human-like arms and fingersVertical reach: Picking fruit from trees, tending to tall crops, and stacking harvested produce all benefit from human proportions How Humanoid Robots Are Used in Agriculture Selective Crop Harvesting Harvesting is the most labor-intensive and economically critical agricultural task—and the one where humanoid robots offer the greatest value. Advanced AI vision systems enable robots to: Assess ripeness in real-time: Computer vision algorithms analyze color, size, shape, and surface texture to determine whether each piece of fruit or vegetable is ready for harvestPick without damage: Soft-grip end-effectors and force-torque sensors apply precisely calibrated pressure—firm enough to detach the produce, gentle enough to avoid bruisingNavigate the plant: AI models map plant structure to identify the optimal approach angle and avoid damaging surrounding foliage or unripe fruitWork continuously: Unlike human pickers who fatigue after 6-8 hours, robots maintain consistent speed and quality through extended shifts Harvest CROO's strawberry-harvesting system demonstrates the potential, using computer vision to identify and pick ripe berries at rates approaching human pickers. Cambridge University's Vegebot goes further, using machine-learning algorithms to evaluate lettuce health, skip diseased or unripe heads, and harvest only market-ready produce. Precision Weeding and Pest Control Weed management consumes enormous resources—both in labor and herbicide costs. Humanoid robots equipped with AI-powered visual recognition can: Distinguish crops from weeds with 95%+ accuracy using deep learning models trained on millions of plant imagesApply targeted micro-doses of herbicide directly to individual weeds, reducing chemical usage by 70-90%Physically remove weeds through precision grasping and pulling—completely eliminating herbicide use for organic operationsMonitor and report weed pressure patterns, enabling predictive management strategies Crop Monitoring and Health Assessment Walking through fields, humanoid robots can perform continuous, detailed crop health monitoring that would be impractical for human scouts or even drones: Leaf-level disease detection: High-resolution cameras and multispectral sensors identify early signs of disease, nutrient deficiency, or pest damage before they're visible to the human eyeGrowth rate tracking: Robots measure individual plant height, canopy coverage, and fruit development over time, creating detailed growth modelsSoil sampling: Humanoid robots can collect soil samples at precise locations and depths, creating detailed fertility mapsMicroclimate monitoring: Sensors measure temperature, humidity, and light levels at crop level—data that differs significantly from weather station readings Vertical Farming and Controlled Environment Agriculture Vertical farming represents the ideal near-term environment for agricultural humanoid robots. These controlled indoor facilities eliminate many outdoor challenges—weather, uneven terrain, variable lighting—while offering structured, predictable workspaces similar to factories. In December 2025, Malaysian agri-tech company Agroz announced a partnership with UBTECH to deploy Walker S humanoid robots in its vertical farming facilities. Walker S handles: Seedling transplantingNutrient solution monitoring and adjustmentHarvest of leafy greens and herbsQuality inspection and sortingFacility maintenance and cleaning This deployment is significant because it demonstrates a commercially viable humanoid robot application in agriculture—not just a lab experiment, but a production operation designed to scale. Greenhouse Operations Greenhouses occupy a middle ground between the controlled environment of vertical farms and the unpredictability of open fields. Humanoid robots are well-suited for greenhouse tasks: Tomato harvesting and pruning: Working within row structures, assessing ripeness, and pruning suckers for optimal growthPollination assistance: Gently vibrating flowers or transferring pollen between plantsPlant training: Guiding vines along support structures and adjusting plant positioningEnvironmental management: Adjusting vents, shade cloth, and irrigation based on real-time sensor data Post-Harvest Handling and Packing After harvest, produce must be sorted, graded, packed, and prepared for transport—labor-intensive processes with high requirements for speed, accuracy, and gentle handling. Humanoid robots can: Sort produce by size, color, and quality grade at consistent high speedPack delicate items like berries, tomatoes, and stone fruits without damageStack and palletize boxes for shippingLabel and track individual batches for food safety traceability Livestock Management Support While most livestock operations use specialized wheeled or tracked robots, humanoid robots offer unique advantages for certain tasks: Barn maintenance: Cleaning, bedding replacement, and facility repair in spaces designed for human workersAnimal health monitoring: Walking among livestock to observe behavior, detect lameness, and identify signs of illnessFeed management: Distributing specialized feed supplements and monitoring intake Top Humanoid Robots for Agricultural Applications in 2026 Browse all available humanoid robots on Robozaps → The Economics of Humanoid Robots in Agriculture Cost Analysis: Robots vs. Human Farm Labor ROI by Agricultural Application For detailed financial modeling, see our ROI of humanoid robots guide and humanoid robot cost breakdown. Real-World Agricultural Deployments Agroz + UBTECH Walker S: Vertical Farming in Malaysia The most significant commercial deployment of humanoid robots in agriculture launched in late 2025 when Malaysian agri-tech company Agroz partnered with UBTECH to deploy Walker S robots in its vertical farming facilities. The robots operate in multi-story growing environments, performing planting, monitoring, harvesting, and quality control tasks autonomously. Key results from early operations: 40% reduction in labor costs for monitored growing operations24-hour growing cycle management with consistent environmental adjustments15% improvement in crop quality through continuous monitoring and immediate response to issuesScalability demonstration: Agroz plans to expand robot deployment across its network of vertical farms throughout Southeast Asia Research Deployments: University Labs to Farm Fields Several research institutions are advancing agricultural humanoid robotics: Cambridge University's Vegebot: Machine-learning-powered lettuce harvesting that distinguishes healthy, ripe heads from diseased or unready onesHarvest CROO Robotics: Strawberry harvesting system achieving near-human picking rates with lower damage ratesWashington State University: Award-winning fruit-bin hauling robots for orchard operationsWageningen University (Netherlands): Greenhouse robot systems for tomato and pepper operations Humanoid Robots vs. Specialized Agricultural Robots It's important to understand when humanoid robots offer advantages over purpose-built agricultural machines: The optimal strategy for most farms combines multiple automation types: drones for aerial monitoring, specialized robots for broad-acre operations, and humanoid robots for high-value tasks requiring dexterity and versatility. Precision Agriculture: How Humanoid Robots Enable Data-Driven Farming Every step a humanoid robot takes through a field generates data. This continuous data collection transforms farming from intuition-based to precision-driven: Real-Time Crop Intelligence Plant-by-plant health mapping: Cameras and sensors create individual plant profiles, tracking growth and health over the entire seasonYield prediction: AI models analyze growth data to predict harvest volumes weeks in advance, enabling better market planningResource optimization: Precise data on water, nutrient, and sunlight needs allows targeted delivery—reducing waste by 30-50% Soil Intelligence Continuous sampling: Robots collect soil data at hundreds of points per field, creating high-resolution fertility mapsCompaction monitoring: Walking robots detect and map soil compaction patterns, guiding tillage decisionsMoisture mapping: Real-time soil moisture data at crop-root depth enables precision irrigation Integrated Farm Management Robot-collected data feeds into farm management systems, creating a closed-loop optimization cycle: sense → analyze → act → measure → improve. This data advantage compounds over time, giving robot-equipped farms increasingly significant productivity edges over conventional operations. Challenges and Limitations Agricultural deployment presents unique challenges beyond those faced in factory settings. For broader context, see our guide to challenges in humanoid robotics. Environmental Challenges Weather exposure: Rain, mud, dust, extreme heat, and frost challenge robot electronics, joints, and sensorsTerrain variability: Soft soil, slopes, roots, and irrigation equipment create mobility challenges for bipedal robotsVariable lighting: Outdoor lighting changes throughout the day and seasons, affecting computer vision accuracyBiological variability: Every plant is unique—unlike manufactured parts, crops vary enormously in size, shape, and position Technical Limitations Battery life: Current humanoid robots operate 2-4 hours before needing 1-2 hours of charging—requiring field charging stations and fleet managementSpeed: For some crops, experienced human pickers still outpace robots by 30-50% (though robots work longer hours)Connectivity: Rural farms often lack reliable WiFi or cellular coverage for cloud-based AI processingMaintenance: Agricultural environments are harsh—dust, moisture, and debris accelerate wear on mechanical components Economic Barriers Thin margins: Agriculture operates on razor-thin margins (2-5% for many crops), making large capital investments riskySeasonal utilization: Many crops have short harvest windows, meaning expensive robots may sit idle for monthsFarm size: Small and mid-size farms struggle to justify the capital investment of humanoid robots Solutions Emerging The industry is actively addressing these challenges through: Robots-as-a-Service (RaaS): Seasonal rental models that match robot availability to harvest windows—no capital expenditure requiredImproved weatherproofing: IP67-rated robot designs built specifically for outdoor agricultural useEdge AI processing: On-robot computation that reduces dependence on cloud connectivityMulti-crop versatility: Robots that handle different crops throughout the year, maximizing utilizationCooperative ownership: Farmer cooperatives sharing robot fleets to spread costs across multiple operations The Future of Humanoid Robots in Agriculture: 2026-2035 Near-Term (2026-2028) Vertical farming becomes the first commercially proven application at scaleGreenhouse operations adopt humanoid robots for high-value crop managementRaaS models emerge, enabling seasonal agricultural use without capital investmentBattery life extends to 6-8 hours with fast-charging capabilities Medium-Term (2028-2031) Outdoor field operations become practical as weatherproofing and terrain navigation improveSelective harvesting reaches human-level speed for major crops (berries, tomatoes, lettuce)Multi-robot farm teams coordinate autonomously with minimal human oversightCosts drop to $10,000-$30,000 for agricultural-grade humanoid robots Long-Term (2031-2035) Humanoid robots handle 15-25% of harvesting for high-value crops globallyFully autonomous farm operations emerge for controlled-environment agricultureRobot-collected data drives AI-optimized growing strategies that significantly increase yieldsAgricultural humanoid robots become standard equipment on mid-to-large farms For our broader technology forecasts, see the future of humanoid robots. How to Get Started: Humanoid Robots on Your Farm Step 1: Identify High-Value Applications Start with tasks where labor is most expensive, hardest to find, or most physically demanding: Harvest of high-value crops (berries, specialty produce)Greenhouse and vertical farm operationsPost-harvest sorting and packingCrop monitoring and scouting Step 2: Evaluate Your Environment Assess whether your operation suits current humanoid robot capabilities: Ideal: Vertical farms, greenhouses, packing facilitiesGood: Orchards with maintained paths, raised bedsChallenging: Open fields with soft soil, steep terrain, extreme weather Step 3: Explore Available Models Review the robots best suited to agricultural use. The UBTECH Walker S leads for vertical farming, while the Unitree G1 offers an affordable entry point for experimentation. Browse all humanoid robots on Robozaps → Step 4: Start Small, Measure Everything Pilot one or two robots on a specific task for one growing season. Track every metric: pick rates, crop damage, labor hours saved, downtime, and maintenance costs. Use this data to build the business case for expansion. Step 5: Scale with Data Once you have proven ROI on a specific application, expand to adjacent tasks and additional growing areas. Each season's data improves AI models, making robots more effective over time. Conclusion: The Agricultural Robot Revolution Is Growing Humanoid robots in agriculture represent one of the most impactful applications of robotics technology in the coming decade. With a global farm labor crisis that shows no signs of easing, rising food demand from a growing world population, and rapidly improving robot capabilities, the convergence of need and technology is undeniable. Vertical farming and greenhouse operations are leading adoption today, with outdoor field applications following as robots become more robust. The economics are compelling and improving: every quarter brings lower costs, longer battery life, better AI, and more proven real-world results. For farmers, agri-tech companies, and food industry leaders, the time to explore humanoid robots is now—not when every competitor has already deployed them. Explore humanoid robots for your agricultural operation on Robozaps → Continue learning with our guides on the best humanoid robots of 2026, applications of humanoid robots, and the economic impact of humanoid robots. --- # SoftBank Pepper Price 2026: Cost, Discontinued & Can You Buy One? URL: https://blog.robozaps.com/b/softbank-pepper-review Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Updated: 2026-08-01T23:05:31.372Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: Classic Pepper hasn't been made since 2020 (~27,000 built, ~17,000 sold), but two afterlives exist: SoftBank Japan's rental-only Pepper+ launched February 2026, while Maxvision, which bought Aldebaran's core assets, revived only NAO. The famous $1,800 price was about a fifth of the advertised ~¥1.09M (~$9,000) bundle. Used units run ~$2,100-$6,800; RobotLAB lists Pepper from $32,000. Key takeaways: - Classic Pepper production halted in 2020; the collapse ran through Aldebaran's 2025 liquidation, with Maxvision buying the assets for a reported €900,000. - Two comebacks get conflated: Japan's rental-only Pepper+ (February 2026, from ¥79,800/month) and Maxvision's revival, which so far covers only NAO, not Pepper. - The famous $1,800 price was the base sticker, about a fifth of SoftBank's advertised 36-month bundle of ~¥1.09 million (~$9,000 at 2015 rates). - You can still get one: used units listed around $2,100-$6,800 at our check (cloud features mostly dead), or from $32,000 with US service at RobotLAB. - Guinness recognized Pepper in February 2026 as the first mass-produced humanoid service robot, five years after its production line stopped. FAQ: Q: Is the Pepper robot still available? A: Not as a new classic Pepper: production halted in 2020. What exists in 2026 is a used market (roughly $2,100-$6,800 for working units), US reseller RobotLAB's listings from $32,000 with US service, and SoftBank Japan's rental-only Pepper+ from ¥79,800 a month. Q: Why was the Pepper robot discontinued? A: Reuters' reporting cites weak demand, limited functionality and friction between the French engineering unit and Tokyo; production quietly stopped in 2020 after about 27,000 units. SoftBank officially called it a pause, and its Japanese arm did eventually relaunch Pepper as the rental Pepper+ in 2026. Q: What is the cost of a Pepper robot? A: Originally ¥198,000 (~$1,800) plus subscriptions of ¥14,800 and ¥9,800 a month for 36 months in SoftBank's advertised bundle, totalling about ¥1.09 million (~$9,000 at 2015 rates); business rentals ran about $16,000 over three years. In 2026: used units around $2,100-$6,800, RobotLAB from $32,000, or Pepper+ rental in Japan from ¥79,800 a month. Q: Can you buy a Pepper robot today? A: Yes, three ways: used from eBay-type marketplaces (check the NAOqi generation and battery first, since cloud-dependent features are largely dead), from RobotLAB at $32,000 with US setup and service (confirm unit condition in writing), or rental-only Pepper+ in Japan. No one manufactures new classic Peppers. Q: How many Pepper robots were sold? A: About 27,000 were built (per Reuters) and roughly 17,000 reached customers across 70 countries, per the figures reported during Aldebaran's 2025 receivership. The first 1,000 consumer units sold out in about a minute in 2015. Q: Who owns Pepper now? A: Split custody. SoftBank Robotics Group in Japan operates the Pepper+ rental business. Shenzhen's Maxvision bought the collapsed European maker Aldebaran's patents, code and brands at a July 2025 auction, but its French subsidiary Maxtronics currently builds and sells only NAO. Q: What happened to Aldebaran? A: Pepper's Paris-based maker entered French receivership in February 2025 after years of heavy losses (reports put the accumulated figure between €60 million and €150 million depending on definition), and the court ordered liquidation on June 2, 2025, ending 106 jobs. Maxvision won the asset auction the following month for a reported €900,000, then set up Maxtronics Robotics SAS near Paris to continue NAO. Q: Is Pepper coming back? A: Partly, in two unrelated ways. SoftBank Japan's Pepper+ is real and rentable now, in Japan only. Maxvision, which owns the Pepper brand and IP, has announced no Pepper production; its revival so far is NAO6 sales and an NAO7 in development. Q: Was Pepper really fired from a supermarket? A: No. 'Fabio' was a Pepper placed in an Edinburgh grocery for one week in 2018 as a Heriot-Watt University experiment filmed for a BBC documentary. The lead researcher told Axios plainly that 'fired is not what happened.' Q: How tall is the Pepper robot? A: 121 cm and 28-29 kg depending on revision, with 20 degrees of freedom, a three-wheel omnidirectional base (2 km/h nominal travel), a 795 Wh battery with a claimed ~12-hour retail runtime, and a 10.1-inch chest tablet running NAOqi. Q: Does a used Pepper still work? A: Mechanically yes, but generations differ in tooling (Choregraphe/Python vs the later Android-tablet stack), and cloud-dependent features may be unavailable on either after Aldebaran's collapse. Verify the software generation, the endpoints it needs, and battery health before buying; expect a development platform, not the cloud concierge from the demos. Q: What is the difference between Pepper and NAO? A: Same creator and OS family, different machines. NAO is a 58 cm desktop biped (~20,000 sold) still in production under Maxtronics, with NAO7 in development. Pepper is the 120 cm wheeled humanoid whose production stopped in 2020. Is Pepper discontinued, and what would one cost you? Classic Pepper is out of production and has been since 2020, but the story didn't end there: in February 2026 SoftBank's Japanese arm relaunched it as the rental-only Pepper+, and the company that inherited Pepper's European maker now builds only its sibling robot. Meanwhile the official US SoftBank Robotics site still presents Pepper as a current product, complete with consultation forms, which is exactly why so many pages tell you Pepper is alive and buyable. Here is the honest 2026 picture, checked against primary sources on July 22, 2026. What actually happened to Pepper Pepper's collapse is usually told as one event; it was a decade-long slide with precise dates. Why did it die? Not because it was "creepy." Reuters' reporting cites weak demand, limited functionality, and friction between the French engineering unit and Tokyo; and restarting the stopped line would have been costly. The pause was real, the failure narrative came later, and the two comebacks below are why neither "dead" nor "back" is quite right. The two comebacks, kept separate Almost every 2025–2026 story about Pepper's afterlife conflates two unrelated threads. Thread one: Pepper+ in Japan. SoftBank Robotics Group, the solvent Japanese company, launched Pepper+ on February 2, 2026: the first design and tablet refresh since 2014, an AI-agent conversation stack aimed at retail, reception and nursing care, and rental-only pricing from ¥79,800 a month (with an event plan at ¥150,000 and a care plan from ¥39,800). It is available in Japan only. The same announcement carried the program's nicest vindication: Guinness recognition of Pepper as the world's first mass-produced humanoid service robot. Thread two: the Maxvision purchase. Maxvision, a listed Shenzhen border-security and AI firm, acquired Aldebaran's core technological assets and the Pepper and NAO intellectual property. Its French subsidiary Maxtronics is actively reviving NAO, Pepper's 58 cm desktop sibling: NAO6 is on sale, an NAO7 is in development, and aldebaran.com now redirects to maxtronics.com. Pepper, by contrast, appears on Maxtronics' site only as a legacy support archive. As of July 22, 2026 there is no announced plan to build Pepper again. So when a headline says "Pepper is saved," check which thread it means; the answer is usually NAO. The $1,800 myth, decoded from the original price sheet Pepper's launch price is one of tech's most misleading famous numbers. The ¥198,000 (~$1,800) sticker was roughly one-fifth of the advertised bundle, spelled out in SoftBank's own 2015 press release: a ¥14,800/month "Basic Plan" for the cloud and apps and a ¥9,800/month insurance pack, each running 36 months, for a three-year total of about ¥1.09 million (roughly $9,000 at 2015 rates). The same sheet notes standalone robot purchases could be applied for in shops, so "mandatory" overstates it slightly; in practice the Basic Plan was what made Pepper Pepper. The business version was rental-only at ¥55,000 a month on a non-cancellable 36-month term, about $16,000 over three years, and Western enterprise units later sold in the €17,000–€20,000 band. Any page telling you Pepper "cost $1,800" is quoting about a fifth of the advertised bill. Can you actually get one in 2026? The used-buyer checklist matters more for Pepper than for most robots, because parts of the machine depended on servers that no longer answer. First, identify the generation: the classic Choregraphe/Python generation and the later Android-tablet generation (compared here) have different tooling, and cloud-dependent dialog and fleet features may be unavailable on either after Aldebaran's collapse; verify the exact software generation and which network endpoints it needs before paying. Second, ask about the battery (a 795 Wh pack with a claimed ~12-hour retail runtime, long out of production). Third, budget for fragility: university labs reported recurring overheating and brittle shells even before support collapsed. A used Pepper in 2026 is an offline animatronic development platform, not the cloud concierge from the demos. What Pepper actually did for a decade Pepper's deployment record is the best argument that it was a phenomenon, not a failure. It greeted customers at HSBC's Fifth Avenue flagship from 2018, staffed Hamazushi restaurants and Carrefour stores, worked Montréal and Prague airports, chanted sutras as a Buddhist funeral priest priced at a fraction of a human one, cheered in empty baseball stadiums during COVID, and taught coding lessons in about a thousand Japanese schools. The one genuinely positive controlled trial is worth knowing: the CARESSES study found that culturally aware Peppers measurably improved older adults' mental health over two weeks of use, with the authors stressing "assist, not replace." And the famous firing? "Fabio," the Pepper removed from Edinburgh's Margiotta grocery after a week in 2018, was a one-week Heriot-Watt University experiment filmed for a BBC documentary, not an employee. The lead researcher said it plainly to Axios: "'Fired' is not what happened... they thought that was a humorous angle." Pepper still books the occasional real job in 2026: an anonymous donor and a library nonprofit put a $40,000 Pepper to work greeting patrons and leading story time at Honolulu's ʻĀina Haina library last December. Archival specs Misinformation audit: seven Pepper claims, checked Who should buy a Pepper in 2026? Three buyers, honestly. Labs and HRI researchers: used Peppers at $2,100–$6,800 remain the cheapest full-size social-robot platform ever made, with a huge academic literature behind them, provided you can live on NAOqi 2.5 offline tooling. Venues that want the icon: a working Pepper still draws crowds, as Honolulu's library shows, and RobotLAB's $32,000 supported channel is the lower-risk route (confirm unit condition in writing). Japanese businesses: Pepper+ rental is the only current, supported, manufacturer-backed Pepper on Earth. Everyone else shopping for a working service humanoid in 2026 should read our evidence-ranked guide or the cost guide; the field moved to machines like the Unitree G1 and its successors. What to watch next Three things will date this page. Whether Maxtronics ever announces Pepper production, which would flip the "wrong robot" verdict above. Whether Pepper+ leaves Japan; SoftBank has said nothing about export. And NAO7's arrival, which will tell us how serious Maxvision is about the Aldebaran legacy it bought for less than the price of a house in Paris. We re-verify this page against the primary sources above; every figure carries its check date. The bottom line Pepper is the most successful failure in robotics: a machine that never worked well enough, sold roughly 17,000 units anyway, defined what the public thinks a social robot is, and earned a Guinness record as the first mass-produced humanoid service robot, awarded five years after its own production line stopped. In 2026 it exists as three different things: a cheap, cloud-orphaned used robot; a $32,000 supported collector's channel; and a Japan-only rental reboot. Know which of the three you're being sold, and Pepper is still, a decade on, exactly what it always was: the friendliest cautionary tale in robotics. Compare on RoboZaps: Pepper and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Xiaomi CyberOne Price 2026: Cost, Is It For Sale & What It Is URL: https://blog.robozaps.com/b/xiaomi-cyberone-review Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Updated: 2026-08-01T23:05:42.762Z Last fact-checked: 2026-07-24T00:00:00.000Z Category: reviews TL;DR: Xiaomi's CyberOne has never been priced or sold; the '$89K-$104K price' is Lei Jun's 2022 production-cost quote (600,000-700,000 yuan) at different exchange rates. The program is alive, not dormant: a second-generation CyberOne debuted in April 2026, humanoids ran pilot shifts in Xiaomi's EV plant at 90.2% task success, and Lei Jun posted factory footage on July 15, 2026. Nothing you can buy yet. Key takeaways: - CyberOne has never been priced or sold: the $89K-$104K figures are Lei Jun's 2022 production-cost quote of 600,000-700,000 yuan at then-current rates, and the $70-80K database line is an undated cost entry; none is a price. - The program is active, not dormant: CyberOne V2 debuted at Xiaomi's April 2026 Investor Day, and Lei Jun posted assembly-line footage on July 15, 2026. - The factory reality is pilots, not deployment: 90.2% task success in one reported three-hour continuous run at a 76-second takt, with Xiaomi's president calling the robots 'interns' and Xiaomi denying 2025 mass-production reports. - CyberOne is 'Tie Da' (Big Iron); 'Tie Dan' is the CyberDog. The CyberDog 2 is genuinely buyable at about $1,790; the humanoid is not. - AgiBot's disclosed backers through July 2026 do not include Xiaomi; Xiaomi's own disclosed robotics investments include Xiaoyu Zhizao, X Square Robot and RoboParty. FAQ: Q: How much does the Xiaomi CyberOne cost? A: There is no price and never has been. Lei Jun said at the August 2022 unveiling that each unit cost 600,000 to 700,000 yuan (then roughly $89,000 to $104,000) to build, which is why mass production was ruled out. That manufacturing-cost quote, at 2022 exchange rates, is the source of the prices you see; the separate $70,000-$80,000 database figure is an undated cost entry with no derivation shown. Q: Is the Xiaomi CyberOne for sale? A: No. There has never been an order channel, waitlist or quote form for CyberOne, and Xiaomi's president said in March 2026 that the company 'is currently not selling the product.' The second-generation CyberOne shown in April 2026 also has no announced price or release date. Q: Does CyberOne work in Xiaomi's factories? A: In pilots, not production. Xiaomi reports humanoids installed nuts on its Beijing EV line at 90.2% success in one reported three-hour continuous autonomous run, matching the 76-second takt, and its president called them 'interns.' Xiaomi's own disclosures call the units 'embodied robots' without naming the model. Xiaomi explicitly denied March 2025 reports that CyberOne was entering mass production on that line. Q: What is the difference between CyberOne and CyberDog? A: CyberOne, nicknamed Tie Da (Big Iron), is Xiaomi's bipedal humanoid and has never been sold. CyberDog, nicknamed Tie Dan (Iron Egg), is the quadruped robot dog line; the CyberDog 2 is a real product at 12,999 yuan, about $1,790, sold in China with grey-market export listings. Q: What are the CyberOne specs? A: The official 2022 sheet: 177 cm tall, 52 kg, 21 degrees of freedom, Mi-Sense depth vision, recognition of 45 emotion classifications and 85 environmental sounds, up to 300 Nm of hip torque, and 3.6 km/h walking. The V2 shown in April 2026 has no official spec sheet; circulating V2 numbers trace to fan posts. Q: What is CyberOne V2? A: The second-generation CyberOne, first shown at Xiaomi's Investor Day in late April 2026, where it handed out gift bags and greeted attendees. Xiaomi has released no specs, price or date for it; engineering disclosures describe a new adult-scale hand with over 22 degrees of freedom and evaporative 'sweat gland' cooling. Q: When can you buy a Xiaomi humanoid robot? A: No date exists. Xiaomi's stated horizon is internal factory deployment 'within five years,' with home use only a stated direction. If you want a humanoid you can actually buy today, that market is led by Unitree's G1 and R1, not Xiaomi. Q: How much does a Xiaomi CyberDog cost? A: The CyberDog 2 launched at 12,999 yuan, roughly $1,790, in China in August 2023; the original 2021 CyberDog was a 1,000-unit release at 9,999 yuan. International buyers pay grey-market importer markups on top of that. Q: Is Xiaomi an investor in AgiBot? A: AgiBot's disclosed backers through July 2026 do not include Xiaomi; they include Tencent, Hillhouse, HongShan, BYD and JD.com. Xiaomi's own disclosed robotics investments include Xiaoyu Zhizao, X Square Robot and RoboParty, backed via its investment arms. Xiaomi's CyberOne has never been priced, never been for sale, and has no order channel anywhere. Every "price" you have seen for it descends from one statement: Lei Jun said at the August 2022 unveiling that each unit cost 600,000 to 700,000 yuan to build, which is exactly why Xiaomi said mass production was not feasible. Converted at 2022 rates, that produced the "$89,000 to $104,000" in Western coverage; spec databases separately list "$70,000 to $80,000" as a cost line without showing a derivation. A production-cost lament has been circulating as a price tag for four years. The bigger correction is about the robot's fate. Most of what ranks for this search is frozen in August 2022 and treats CyberOne as a one-keynote wonder. It is not. The program came back: a second-generation CyberOne debuted at Xiaomi's Investor Day in April 2026, humanoids have been running pilot shifts in Xiaomi's Beijing EV plant, and Lei Jun posted factory footage of them working an assembly line as recently as July 15, 2026. Still not a product, still nothing you can buy, but very much alive. This page covers the whole arc, with each claim dated. First, untangle the names Half the confusion around "Xiaomi robot price" comes from mixing these up. The robot dog is a real product you can pay for. The humanoid is not, and the nicknames are one letter apart in transliteration, so even robotics coverage swaps them. CyberOne is Tie Da; the dog is Tie Dan. The CyberOne timeline, 2022 to 2026 The March 2025 row is the pivot for reading everything else. When the maximal story, "CyberOne is entering mass production", circulated, Xiaomi itself shot it down. What Xiaomi actually claims is narrower and more interesting: pilot shifts, measured success rates, and a five-year horizon it keeps repeating. Original CyberOne specs, and the V2 gap Those are Xiaomi's own launch figures. Third-party spec databases such as Robots Guide add detail, an Intel RealSense head camera, roughly 2.5 hours of battery, but the launch sheet is the official record. For the V2 shown in April 2026, there is no official spec sheet at all: the height, speed and payload numbers circulating for it trace to fan posts, not Xiaomi, and should be treated as provisional until the company publishes one. What Xiaomi-affiliated engineering disclosures do describe is the new hand: 22-plus degrees of freedom at true adult-hand scale, full-palm tactile coverage, and a metal-3D-printed "sweat gland" cooling system that evaporates coolant through the forearm to shed motor heat. The price truth There has never been a CyberOne price. Not in 2022, not for the V2, not anywhere in between. The figures in circulation all descend from one sentence: Lei Jun's launch-day statement that each unit cost ¥600,000 to ¥700,000 to build. Convert that at 2022 rates and you get the "$89,000 to $104,000" repeated by Western coverage; spec databases separately list "$70,000 to $80,000" as an undated cost entry with no derivation shown. Sites presenting any of these as a price tag, and at least one headline flatly calls it an "$89K price tag", are quoting a manufacturing cost given as the reason the robot was not commercialized. Xiaomi's own position has not changed since: as of the March 2026 MWC remarks, it "is currently not selling the product", and no order channel, waitlist or quote form has ever existed. The factory story, honestly told What makes this program interesting in 2026 is not a spec sheet, it is that Xiaomi turned its humanoid work into an internal factory experiment and published numbers. One precision note: Xiaomi's own corporate disclosures describe the pilot units as its "embodied robots" without naming the model, so treat "CyberOne on the line" as the program's shorthand rather than a confirmed model designation. In the pilots at the Beijing EV plant, humanoids installed self-tapping nuts on the line at a 90.2% success rate across a three-hour continuous autonomous run, keeping pace with the plant's 76-second per-car takt. They run on Xiaomi-Robotics-0, an in-house vision-language-action model trained on some 200 million robot action steps, with a tactile-refinement layer for the fine work. Read both halves of that honestly. Keeping takt on a real line for three hours is a genuine result few humanoid programs have published. And 90.2% is far below the 99-plus percent expected of a human line worker, which is why Lu Weibing's own word for the robots was "interns", and why the company's stated horizon for real internal deployment is measured in years, not quarters. The failure mode for readers is collapsing "pilot with published success rates" into "robots run Xiaomi's factory". Xiaomi denied exactly that story in March 2025, and its own executives still describe trials. Where CyberOne fits in Xiaomi's plans Xiaomi's robotics unit is physically co-located with its Beijing EV factory, and the humanoid program is now effectively part of the EV story: the same plant that crossed half a million vehicles is the training ground. Lei Jun's stated direction is that the factory floor teaches the robot the skills that eventually matter in homes, with large-scale internal deployment "within five years" as the repeated line. A correction worth making because it circulates: AgiBot's disclosed backers through July 2026 do not include Xiaomi, despite the two often being linked. Xiaomi's disclosed external robotics investments run through its investment arms and include industrial robot-intelligence firm Xiaoyu Zhizao, embodied-AI developer X Square Robot, and bipedal-robot maker RoboParty. If you came here wanting a robot you can actually buy, the honest redirect list is short: Xiaomi's own CyberDog 2 at ¥12,999 if a quadruped scratches the itch, or the genuinely purchasable humanoids, led by Unitree's G1 and R1, which we rank in our best humanoid robots guide with real prices in the humanoid robot cost breakdown. The bottom line CyberOne is the rare 2022 keynote robot that neither shipped nor died. It was never for sale, and the four-year-old "price" attached to it is a production-cost quote at various exchange rates. What it became is arguably more credible than a product launch would have been: an internal factory program with published pilot numbers, a second-generation body shown in April 2026, and fresh assembly-line footage from Lei Jun himself in July 2026. Judge it as that: not a purchase decision, but one of the more transparent factory-humanoid experiments running, with Xiaomi's own executives supplying the caveats. When a price finally exists, it will come from Xiaomi, and this page will carry it. Compare on RoboZaps: Xiaomi Cyberone, CyberDog 2 and Cyberdog. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Astribot S1 Price 2026: Specs & Is It For Sale? (Review) URL: https://blog.robozaps.com/b/astribot-s1-review Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: reviews TL;DR: The Astribot S1 has no public price: Astribot sells it business-to-business by quote, and Chinese industry reports cite about ¥500,000 (roughly $69,000) for the research edition. It is a wheeled, two-armed indoor humanoid with verified 10 m/s arm speed and ±0.1 mm repeatability, shipping in volume since late 2025, mostly in China. Our verdict: 8.2/10. Key takeaways: - Verified hardware: two 7-DoF arms, ≥10 m/s end-effector speed, ±0.1 mm repeatability, and 5 kg payload per arm on an omnidirectional wheeled base (23 DoF total). - The Astribot Suite framework (hardware + whole-body teleoperation + DuoCore-WB imitation learning) is described in a July 2025 arXiv paper reporting 80% average task success, with 100% peak, across six tasks. - No official price. Astribot sells by quote; Chinese industry reports cite about ¥500,000 (~$69,000) for the research edition, and the ~$100,000 figures in Western press are unattributed estimates. - Commercially sold since August 2024 with volume deliveries from late 2025, mostly in China; international buyers go through quote-based importers. - Strengths are speed and precision on household tasks; limitations are indoor-only operation, parallel-jaw grippers rather than dexterous hands, and occasional software glitches. FAQ: Q: What is the Astribot S1 price? A: Astribot publishes no price for the S1. It is sold business-to-business by quote, and Chinese industry reports cite about ¥500,000 (roughly $69,000) for the research edition including the teleoperation platform and SDK. The $96,000–$150,000 range that circulates online is an unattributed estimate. Q: How fast is the Astribot S1? A: The Astribot S1's end-effectors can reach speeds of 10 meters per second (36 km/h) with accelerations up to 100 m/s². This makes it one of the fastest humanoid robots in terms of arm manipulation speed, surpassing the average adult male's arm swing speed of 5–10 m/s. Q: Can the Astribot S1 walk? A: No—the Astribot S1 does not use bipedal walking. It has a 3-DoF omnidirectional mobile base (wheels) for navigation. This provides smooth, stable indoor movement but cannot handle stairs or uneven outdoor terrain. Typical movement speed is 1.5 km/h (0.42 m/s). Q: What tasks can the Astribot S1 perform? A: The S1 has demonstrated cooking (wine pouring, vegetable preparation, sandwich flipping), cleaning (vacuuming, organizing), household management (laundry folding, shoe organization, trash disposal), and even physical activities like Wing Chun martial arts and calligraphy. Its imitation learning system allows it to learn new tasks from human demonstrations. Q: How does the Astribot S1 learn new tasks? A: Through the Astribot Suite's DuoCore-WB imitation learning framework. A human operator demonstrates tasks via a VR headset and handheld joysticks (teleoperation). The robot then learns whole-body visuomotor policies from these demonstrations, achieving an average 80% success rate across diverse tasks. Q: Is the Astribot S1 available for purchase? A: Yes, for organizations: the S1 has been sold commercially since August 2024, with volume deliveries (reported thousands of units) from late 2025. Deployments are concentrated in China (JD.com, CCTV, elder-care institutions, retail in six cities); international buyers go through quote-based importers rather than an official retail channel. Q: How does the Astribot S1 compare to the Tesla Optimus? A: The S1 excels in arm speed (10 m/s vs ~2 m/s) and precision (±0.1 mm), but the Tesla Optimus offers bipedal walking, hands with 22 degrees of freedom, and a $20,000–$30,000 target price. The Optimus is focused on factory automation first, while the S1 is a wheeled indoor platform organizations can order today by quote. Q: What sensors does the Astribot S1 have? A: The S1 includes RGB cameras, depth cameras (RGB-D), LiDAR, IMU, force/torque sensors at wrists, tactile pressure sensors in fingertips, ultrasonic proximity sensors, a microphone array, and temperature sensors. This comprehensive suite enables autonomous navigation, object recognition, and precise force-controlled manipulation. Q: How long does the Astribot S1 battery last? A: Reported figures put the S1 at 4–6 hours of normal active use, with up to 10 hours in low-power standby and roughly 1.5 hours to recharge on its docking station. Astribot does not publish an official battery specification, so treat these as reported values. Q: Who makes the Astribot S1? A: The Astribot S1 is built by Astribot, a subsidiary of Stardust Intelligence (), based in Shenzhen, China. The company was founded in 2022 by Lai Jie, who previously worked at Tencent Robotics Laboratory, Baidu, and Hong Kong Polytechnic University. The Astribot S1 is one of the most impressive humanoid robots to emerge from China's booming robotics sector. Its price is quote-only (Chinese reports cite about ¥500,000 for the research edition). Developed by Stardust Intelligence in Shenzhen, this fully autonomous humanoid has captured global attention with its human-surpassing speed, sub-millimeter precision, and advanced imitation learning capabilities. In this comprehensive Astribot S1 review for 2026, we break down everything—specs, real-world performance, pricing, limitations, and how it stacks up against competitors like the Tesla Optimus, Figure 02, and Unitree G1. Whether you're a robotics enthusiast, an enterprise buyer evaluating humanoid robots for deployment, or simply curious about the state of the art, this review gives you the complete picture. What Is the Astribot S1? The Astribot S1 is a mobile dual-arm humanoid robot built by Astribot, a subsidiary of Stardust Intelligence founded in 2022 in Shenzhen, China. The company was founded by Lai Jie, who previously worked at Tencent Robotics Laboratory, Baidu, and Hong Kong Polytechnic University. The name "Astribot" derives from the Latin proverb "Ad astra per aspera"—through hardship to the stars. Unlike many humanoid prototypes that rely on teleoperation, the S1 is designed to operate fully autonomously. Its core innovation is a whole-body imitation learning system that allows it to learn new tasks by watching and mimicking human demonstrations, then executing them independently with remarkable precision. The robot debuted publicly at the 2024 World Robot Conference in Beijing and has since undergone significant improvements, culminating in a landmark academic paper published on arXiv in July 2025 that detailed the complete "Astribot Suite" framework. Astribot S1 Technical Specifications Based on the official Astribot paper and verified sources, here are the complete specs: The cable-driven design is a standout engineering choice. Unlike conventional rigid-link robots, the S1's cable-driven arms emulate human musculature, providing compliant motion, superior payload-to-weight ratio, reduced backlash, and enhanced safety during human interaction. AI and Learning System: The Astribot Suite The Astribot S1's intelligence comes from a unified framework called the Astribot Suite, detailed in a July 2025 arXiv paper from the Astribot Team. The system has three core components: 1. Whole-Body Teleoperation Interface Data collection uses a VR headset and handheld joysticks that map hand poses to the robot's end-effectors. Two modes are available: first-person view for precise manipulation tasks and third-person view for large-range whole-body motion. This intuitive interface means non-experts can teach the robot new tasks. 2. DuoCore-WB Imitation Learning Policy The learning algorithm uses RGB-based visual perception with pre-trained vision encoders, whole-body policy control in end-effector space using SO(3) orientation representation, and a Real-Time Trajectory Generation (RTG) module that refines predicted actions into smooth, continuous execution using quadratic programming optimization. 3. Task Performance In benchmark testing, DuoCore-WB policies achieved an average 80% success rate and peak 100% success rate across six representative whole-body tasks including delivering drinks, storing items in cabinets, throwing away trash, organizing shoes, and picking up scattered objects. The integration of large language models (LLMs) enables the robot to understand and respond to natural language queries about its environment, making it accessible to users with no robotics experience. Real-World Performance: What Can the Astribot S1 Actually Do? The S1's demo videos are genuinely impressive and have been verified by independent observers. Here's what the robot has demonstrated: Kitchen and Cooking Tasks Wine pouring: Opens and pours wine without spilling, demonstrating precise force controlVegetable preparation: Shaves cucumbers with delicate precision, sorts and chops vegetablesCooking: Flips sandwiches in a frying pan, makes pancake batter, prepares tea using traditional kung fu methodsTable setting: Pulls a tablecloth from under stacked wine glasses—a classic speed-and-control test Household Management Laundry folding: Folds clothes with reasonable accuracyShoe organization: Picks up scattered shoes and places them neatly on a rackTrash disposal: Navigates to trash bin, opens lid, discards waste, closes lidItem delivery: Carries drinks between rooms, opens doors with one arm while holding items with the other Physical Agility Wing Chun martial arts: Practices martial arts forms demonstrating dynamic movementBasketball shots: Makes accurate basketball throwsCalligraphy: Writes with brush precisionToy throwing: Picks up objects from the floor and throws them with controlled force One critical note: the S1's "Hello World" demo video was verified to run at 1x speed without teleoperation. However, skeptic Thomas Mannfred Carlsson raised valid questions about the video's object segmentation overlay appearing to show pre-computed results, suggesting some visual elements may have been post-processed for the demo. The underlying robot performance, however, has been independently validated through the July 2025 academic paper. Astribot S1 vs Competitors: Comparison Table Key takeaway: The Astribot S1 leads in raw speed and precision but uses a wheeled base instead of bipedal locomotion, and parallel-jaw grippers instead of dexterous hands. This makes it excellent for structured indoor environments but less versatile for uneven terrain or tasks requiring fine finger manipulation. Astribot S1 Price and Availability Astribot publishes no price for the S1: it is sold business-to-business by quote, with no online checkout. Chinese industry reports cite about ¥500,000 (roughly $69,000) for the research edition including the teleoperation platform and SDK, while the ~$100,000 figures in Western press are unattributed estimates. Either way it sits in the enterprise tier, well above consumer options like the Unitree G1 ($13,500) or Tesla's $20,000–$30,000 Optimus target price. As of mid-2026, the real availability picture is: Sold commercially since August 2024, business-to-business and research sales by quoteVolume deliveries (reported thousands of units) from late 2025, including a thousand-unit Thundersoft orderDeployments are China-centric: JD.com, CCTV, elder-care institutions, and retail across six Chinese citiesInternational buyers go through quote-based importers; there is no official multi-country retail channel You can check current pricing and availability on the Astribot S1 product page at Robozaps. Design and Build Quality The S1's industrial design is clean and functional. Standing 170 cm tall, it has a humanoid upper body with a sleek white-and-silver color scheme that fits naturally in home and lab environments. The cable-driven actuation system—inspired by human musculature—is one of its most innovative design elements. Cable-Driven Advantages Superior payload-to-weight ratio: The cable-driven design achieves higher payload capacity relative to arm weight compared to conventional rigid-link robotsReduced backlash and inertia: Smoother, more precise movementsEnhanced safety: Compliant motion means the robot "gives" on contact, reducing injury riskHigh-resolution force control: Critical for delicate tasks like food preparation and handling fragile objectsLow-friction transmission: Optimized for long-term mechanical durability The omnidirectional mobile base provides smooth navigation in indoor spaces, with the 4-DoF articulated torso adding significant reach and flexibility for tasks at different heights—from picking up objects on the floor to reaching high shelves. Sensor Suite and Perception The Astribot S1 carries a comprehensive sensor package: RGB cameras: Multiple cameras for visual perception, compatible with pre-trained vision encoders (DINOv2, CLIP, SigLIP)Depth camera / RGB-D: For 3D environment mappingLiDAR: Long-range spatial awareness and navigationIMU: Inertial measurement for balance and motion trackingForce/torque sensors: At wrists and ankles for precise force feedbackTactile/pressure sensors: In fingertips for delicate object handlingUltrasonic sensors: Proximity detection for obstacle avoidanceMicrophone array: For voice commands and spatial audioTemperature sensors: Environmental monitoring The vision system uses an approach akin to YOLO for real-time object detection and segmentation, allowing the robot to identify and interact with a wide variety of household objects without pre-programming. User Interface and Voice Control Interacting with the S1 is designed to be intuitive. The integration of large language models enables natural conversation about tasks, the environment, and preferences. The robot processes voice commands and can respond intelligently to queries. Strengths: Natural language understanding via LLM integrationMulti-language supportNo robotics expertise required for basic operationCompanion app for task programming and monitoring Weaknesses: Voice recognition struggles in noisy environmentsComplex multi-step verbal instructions can be misinterpretedResponse latency can be noticeable for complex queries Limitations and Honest Criticisms No review would be complete without an honest assessment of shortcomings: 1. No Bipedal Locomotion The S1 uses an omnidirectional wheeled base, not legs. This means it cannot navigate stairs, step over obstacles, or traverse uneven outdoor terrain. For applications requiring true humanoid mobility, competitors like Tesla Optimus and Figure 02 have an advantage. 2. Parallel-Jaw Grippers vs. Dexterous Hands While the grippers are precise, they lack the fine-grained finger manipulation of competitors like Figure 02's 16-DoF hands. Tasks requiring individual finger control (typing, button pressing, tool use) are more limited. 3. Software Stability Users have reported occasional software glitches during complex multi-step task sequences, causing momentary freezes. While not frequent, this is a concern for mission-critical applications. 4. Indoor-Only Design The S1 is optimized for structured indoor environments. Outdoor use, uneven surfaces, and harsh conditions are outside its current capability envelope. 5. Price Barrier With no public price and quote-only enterprise sales, the S1 is out of consumer reach. It's an enterprise/research tool at this stage, not a household companion—despite the household task demos. 6. Limited Independent Validation While the July 2025 paper provides solid academic validation, much of the public demo content is manufacturer-produced. More independent third-party testing would strengthen confidence in real-world reliability. Who Should Buy the Astribot S1? The Astribot S1 makes the most sense for: Research institutions: The Astribot Suite framework and teleoperation interface make it excellent for robotics researchEnterprise pilots: Companies exploring humanoid robots for logistics, food service, or hospitality applicationsManufacturing: Repetitive precision tasks where the S1's ±0.1 mm accuracy provides valueHealthcare facilities: Assistance with patient care tasks, delivery, and organizationWealthy early adopters: Those willing to invest in cutting-edge home automation It's not ideal for outdoor applications, tasks requiring stair navigation, or budget-conscious buyers. Future Prospects: What's Next for Astribot? Based on the trajectory of Stardust Intelligence's development, we expect several improvements in upcoming S1 iterations or next-generation models: Dexterous hands: Upgraded from parallel-jaw grippers to multi-fingered manipulationBipedal locomotion: A legged version for outdoor and multi-terrain applicationsVLA integration: The DuoCore-WB framework is already compatible with large-scale Vision-Language-Action model training, suggesting future generalization capabilitiesPrice reduction: As production scales, prices should decrease toward the $50,000–$75,000 rangeExpanded task library: More pre-trained policies for common household and industrial tasks That commitment is now well funded: in June 2026 Astribot closed a Series B that took its 2026 fundraising past ¥1 billion (about $140 million) at a valuation above ¥10 billion, and in May 2026 it launched the smaller Astribot T1, a cable-driven sibling priced from ¥89,900 (about $13,000) with orders open. Our Verdict: 8.2/10 The Astribot S1 is a genuinely impressive piece of robotics engineering that stands out in a crowded field. Its combination of blazing speed (≥10 m/s), industrial-grade precision (±0.1 mm), and a scientifically validated learning framework (DuoCore-WB) puts it at the forefront of autonomous humanoid manipulation. The wheeled base and parallel-jaw grippers are deliberate engineering trade-offs that prioritize reliability and precision over versatility—and for indoor environments, they work extremely well. The academic backing (openly published paper with reproducible results) adds credibility that many competitors lack. With quote-only pricing, it is not a consumer product. But for research labs, enterprise pilots, and forward-thinking organizations, the Astribot S1 represents one of the most capable and best-documented humanoid robots available today. Ready to explore humanoid robots? Browse the full collection at Robozaps or check the Astribot S1 product page for current pricing. Related: Tesla Optimus vs Astribot S1: A Comparative Breakdown · Astribot S1 Release Date, Rumors, Price, and News · Best Humanoid Robots in 2026 Ready to buy? Browse humanoid robots for sale on Robozaps. --- # Humanoid Robots for Home Use: What You Can Actually Buy in 2026 URL: https://blog.robozaps.com/b/humanoid-robots-for-home Author: Radica Maneva Published: 2026-02-02T00:00:00.000Z Updated: 2026-08-01T23:06:16.216Z Last fact-checked: 2026-07-21T00:00:00.000Z Category: reviews TL;DR: 1X NEO is the only model in this guide sold specifically for household tasks, but it remains an early-access preorder. Unitree R1 and G1 base models are purchasable hardware without secondary development; their EDU versions and Fauna Sprout are developer platforms. Key takeaways: - 1X NEO is sold as a home robot, but it remains an early-access preorder with basic autonomy and scheduled Expert assistance. - Base Unitree R1 and G1 models do not support secondary development; R1 EDU and G1 EDU are separately priced configurations. - The $13,500 base G1 has 23 DoF and no dexterous-hand DoF; the configurable 43-DoF system is G1 EDU. - Fauna Sprout Creator Edition is available to developers, researchers and educators, not as a turnkey household assistant. - Compare the exact configuration, current order status and control mode before treating any humanoid as usable at home. FAQ: Q: Can you buy a humanoid robot for the home in 2026? A: You can place an order or request several humanoids, but their status differs. 1X NEO is an early-access home-robot preorder. Unitree R1 and G1 are purchasable hardware platforms, with some variants backordered. Fauna Sprout is supplied by request to developers, researchers and educators. Q: Which humanoid robot is actually designed for home use? A: Among the systems compared here, 1X NEO is the only one sold specifically as a domestic robot. It is still early access, and unfamiliar tasks may use scheduled remote assistance through 1X Expert Mode. Q: What is the difference between Unitree G1 and G1 EDU? A: The $13,500 base G1 has 23 DoF, no dexterous-hand DoF and no secondary development. G1 EDU is the configurable research version with 23–43 DoF, optional hands, optional higher compute and secondary-development access; pricing is by quote. Q: Can you program the base Unitree R1 or G1? A: Unitree's official store says the base R1 and base G1 do not support secondary development. Buyers who need development access should obtain a quote for R1 EDU or G1 EDU and confirm the required compute, hands and support options. Q: Is the 1X NEO subscription available immediately? A: No immediate-delivery claim is warranted. 1X lists NEO Standard at $499 per month but says it ships later than the $20,000 Early Access ownership option. US Early Access deliveries are scheduled to start in 2026, with some customers receiving units later. Q: Is Fauna Sprout a home robot? A: Sprout Creator Edition is a humanoid developer platform designed for human spaces. Fauna supplies it to researchers, educators and commercial developers by request; it is not sold as an out-of-box household chore service. Q: Can current humanoid robots do housework autonomously? A: Some vendors demonstrate or state limited household capabilities, but demonstrations may involve remote operation, custom development or controlled conditions. Broad evidence of repeatable, unattended housework by delivered customer units is not yet established for the products in this guide. Q: What should I check before buying a humanoid robot? A: Confirm the exact configuration, delivery status, autonomous versus teleoperated tasks, development access, safety controls, privacy terms, warranty, repair location, shipping, duties and required subscriptions or accessories. Yes, you can order several humanoid robots in 2026. That does not make them household appliances. Among the systems compared here, 1X NEO is the only one sold specifically as a home robot, and it remains an early-access preorder. Unitree's R1 and G1 are purchasable humanoid platforms, but their base and EDU configurations have materially different hardware and development access. Fauna Sprout Creator Edition is sold to developers, researchers and educators. For buyers, the deciding question is: What can this exact configuration do without custom development or a remote operator? This guide's main comparison is limited to humanoid platforms. Wheeled mobile manipulators such as Weave Robotics' Isaac 1 fall outside it; LG CLOiD and SwitchBot onero H1 appear only as adjacent watchlist context. Home humanoid robot availability and capability Prices and status below were checked against manufacturer pages on 21 July 2026. “Home robot” means a product sold for domestic tasks. “Operable at home” only means a developer can run the hardware in a residential environment. Why “buyable” and “home-ready” are different Four categories are being collapsed in most home-robot lists: Domestic product: sold to consumers for household tasks.Developer platform: hardware and software intended for teams building their own applications.Research or EDU configuration: a separately priced model with development interfaces, compute or hands that the base version may not include.Demonstration or announced product: shown performing a task but not normally orderable for household delivery. A robot may walk through a house without being able to clean it. A developer may teleoperate a platform in a kitchen without the robot having kitchen autonomy. A vendor video may show a learned task without proving that a buyer receives that task, on that configuration, with repeatable performance. 1X NEO: designed for the home, still early access 1X's order page lists two offers: $20,000 ownership under Early Access and a $499 monthly Standard subscription. Early Access requires a $200 refundable deposit and has priority delivery. The company says US deliveries start in 2026. Its launch announcement says the subscription option will ship later. NEO is also the clearest example of why control mode matters. 1X says early owners receive basic autonomy. For unfamiliar tasks, a customer can schedule Expert Mode so a 1X operator remotely supervises the robot. Owners can also pilot NEO through the mobile app or a VR device. Those are useful capabilities, but autonomous work, customer teleoperation and company-assisted teleoperation should not be presented as the same thing. Availability needs the same care. In an April 2026 factory update, 1X said some customers would receive NEO during 2026 and others later. The company also said current production units were being prioritised for internal testing in employees' homes. The defensible status is therefore preorder with deliveries scheduled to begin in 2026, not generally available now. Unitree R1: base and EDU are different products The official R1 specification page separates three configurations: R1 Air: $4,900, 20 DoF, monocular camera.R1: $5,900, 26 DoF, binocular camera.R1 EDU: contact-sales pricing, 26–40 DoF, optional higher compute and secondary development. Unitree's store explicitly says the base product does not support secondary development and directs buyers who need customisation to R1 EDU. That means the $4,900 headline price should not be paired with SDK or open-development claims. The store also shows a general “In Stock” label while the selected R1 Air variant is marked backordered, so buyers should confirm the exact configuration and delivery date before paying. R1 can still be interesting as compact humanoid hardware. Unitree lists a height of 123 cm, a weight of about 27–29 kg and roughly one hour of battery life. Those specifications do not establish autonomous tidying, cooking, security patrols or childcare. Unitree G1: the $13,500 model is not the 43-DoF developer version The G1 configuration table is unambiguous. The $13,500 base G1 has 23 DoF, no hand DoF, an arm load of about 2 kg, no higher-compute module and no secondary development. The official shop currently marks that model backordered. G1 EDU is the configurable research version. Unitree lists 23–43 DoF, optional three-fingered dexterous hands and wrist joints, optional higher compute, an arm load of about 3 kg and support for secondary development. Its price is not $13,500; buyers must contact sales. This distinction changes the buying advice. The base G1 may suit someone who wants Unitree's supplied motions and hardware. A robotics team that wants to build manipulation or household experiments needs to price the EDU configuration and its options. Neither version arrives as a domestic task service. Fauna Sprout: a developer platform for human spaces Sprout Creator Edition is designed to operate around people, but Fauna describes it as a robotics-development platform. It includes a modular software stack, core APIs, mapping, navigation, whole-body teleoperation and built-in social behaviours. The 107 cm robot has 29 DoF, simple one-DoF grippers and a stated battery runtime of 3–3.5 hours. Fauna says Sprout is available to researchers, educators and commercial developers through a request process. The current product page does not publish a retail price. That makes “available developer platform” accurate; “$50,000 premium home robot” is not a reliable current buyer description. Fauna joined Amazon in March 2026 and says it continues to sell and support Sprout Creator Edition. What these robots can actually do at home Marketing videos are still useful evidence of direction, but they do not answer four buyer questions: Was the task autonomous? Did it require a remote operator? Does it ship with the advertised configuration? How often does it work without intervention? Safety, privacy and support questions A 25–35 kg humanoid is not harmless simply because it weighs less than an adult. Unitree tells R1 and G1 users to maintain a sufficient safety distance and operate with caution. Buyers should treat safety as configuration- and application-specific, not as a blanket property of “consumer humanoids.” Before bringing any system into a home, ask: Who can access the cameras, microphones and controls, and under what conditions?Which tasks are autonomous, teleoperated or custom-coded?What emergency stop, force limits and safe-distance rules apply?Can the exact configuration handle thresholds, rugs, stairs and pets?What warranty, repair location, spare-parts and battery support are included?Are shipping, import duties, subscriptions and required accessories included in the quoted price? None of the products in this guide should be treated as an unsupervised caregiver for a child, older person or animal. Care and monitoring claims require a separate standard of reliability, privacy and emergency response. Systems to watch, not buy as home products Figure 03 Figure 03 has a soft exterior, washable coverings and other features designed for homes. Figure has not published a consumer order page, price or household delivery date. Its public deployment evidence remains industrial, including BMW. LG CLOiD LG demonstrated CLOiD performing appliance and laundry workflows at CES 2026. It uses a wheeled base rather than humanoid legs. LG has not published a normal consumer order date or price. SwitchBot onero H1 SwitchBot's product page displays a $9,999 price and household-task claims, but it also marks onero H1 “Coming Soon” and unavailable. Treat it as an announced product until ordering and delivery are demonstrated. Samsung Ballie is an adjacent home companion, not a humanoid: it has a rolling body and no arms. Tesla Optimus and other industrial prototypes also do not belong in a buy-now home ranking without a consumer order path. Which type should you choose? You want a domestic product: NEO is the only system here sold for that job. Read the delivery and Expert Mode terms carefully, and recognise that early ownership is participation in a new product rollout.You want to build robotics applications: compare R1 EDU, G1 EDU and Sprout. Ask for a complete quote that identifies compute, hands, development access, warranty and support.You want the lowest-cost humanoid hardware: R1 Air and R1 carry the lowest published prices, but the base versions do not support secondary development.You need useful chores now: wait for delivered-customer evidence on intervention rates and task reliability, or buy specialised home robots for individual jobs. For a broader view of pricing beyond the sticker price, see our humanoid robot cost guide. For order channels and procurement checks, see humanoid robots you can actually buy. RoboZaps verdict Home humanoids are now orderable, but the category is not mature enough for a simple “best robot” ranking. NEO is an early domestic product. R1 and G1 are hardware families whose base and EDU configurations serve different buyers. Sprout is a developer platform for shared human spaces. Figure 03, CLOiD and onero H1 remain watchlist systems. The most important line on any quote is not the price. It is the line that states the exact configuration, control mode and tasks included at delivery. Compare on RoboZaps: Unitree R1, Figure 03, Tesla Optimus and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in Hospitality: Complete Guide to Hotel Robotics [2026] URL: https://blog.robozaps.com/b/humanoid-robots-in-hospitality Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:04:32.290Z Category: insights TL;DR: Hospitality robots are operational reality in 2026: 87% of hotels reported staffing shortages (AHLA 2025), and robots fill the gap — delivery units leasing at $1,000-$2,000 a month, concierge humanoids at $25,000-$100,000+, cleaning robots up to 80% faster on public areas. Properties report about 25% fewer front-desk calls after deployment. Key takeaways: - 87% of hotels reported staffing shortages in 2025 (American Hotel & Lodging Association) — the structural driver pushing robots into hotels, restaurants and airports. - Service robot prices dropped about 40% since 2022: delivery robots lease for roughly $1,000-$2,000 a month, while humanoid concierge robots run $25,000-$100,000+. - Cleaning robots work guest rooms about 20% faster and public areas up to 80% faster than human housekeepers. - Properties report about 25% fewer front-desk calls after deploying delivery robots, which run 24/7 without breaks. - Japan's Henn-na Hotel pioneered the robot-operated hotel in 2015, at its peak running over 243 robots including Pepper check-in staff. FAQ: Q: How much does a hotel robot cost in 2026? A: Costs vary widely by type. Delivery robots like Keenon W3 lease for $30–50/day, while Relay robots cost approximately $2,000/month. Humanoid concierge robots like Pepper range from $25,000–$30,000 to purchase. High-end humanoid platforms can exceed $100,000. Most vendors now offer lease-to-own models that lower the barrier to entry. Q: Do hotel robots actually save money? A: Yes. A single delivery robot operating 24/7 typically replaces 1.5–2 full-time equivalent positions, saving $45,000–$70,000 annually in wages and benefits. Cleaning robots like commercial cleaning robots improve room turnover speed by 20%, directly impacting revenue potential. Most properties report payback within 6–18 months depending on robot type and utilization. Q: What do hotel guests think about robots? A: Guest reception is overwhelmingly positive for novelty and efficiency. Nearly 48% of travelers say they're comfortable with robotic greetings. Delivery robots consistently receive positive mentions in online reviews. However, luxury segment guests and older demographics tend to prefer human interaction for complex requests. The most successful properties use robots for routine tasks while reserving human staff for personalized, high-touch service moments. Q: Can robots replace hotel staff entirely? A: No — and the most successful deployments don't try to. Robots excel at repetitive, predictable tasks (deliveries, FAQs, cleaning patterns) but cannot match human emotional intelligence, problem-solving ability, or the personal warmth that defines premium hospitality. The industry consensus is a hybrid model: robots handle volume, humans handle complexity and empathy. Q: Which hotels are currently using robots? A: Major brands with active robot deployments include Marriott (Aloft Hotels with Relay robots), Hilton (various AI concierge programs), YOTEL (Yobot luggage system), Henn-na Hotel chain in Japan, and thousands of independent properties using Keenon and Bear Robotics delivery bots. The Marriott in Belgium uses a multilingual AI assistant named Mario, and properties across Asia-Pacific have the highest density of robot deployments globally. Q: What is the best robot for a small hotel? A: For small to mid-size properties (under 150 rooms), a leased delivery robot is the best starting point. Keenon W3 at $30–50/day or Bear Robotics Servi at approximately $999/month offer the lowest risk and fastest ROI. These handle room delivery and restaurant service without requiring major infrastructure changes. Humanoid robots are better suited for larger properties with higher guest volume where the marketing value justifies the investment. Q: How long does it take to deploy a hotel robot? A: Most delivery robot deployments take 2–4 weeks including site survey, Wi-Fi assessment, mapping, and staff training. Humanoid robots with PMS integration require 4–8 weeks. The initial mapping phase — where the robot learns the property layout, elevator systems, and navigation paths — is the most time-intensive step. Most vendors handle this as part of their deployment service. Related: Humanoid Robots in Retail: Revolutionizing Customer Experience · Applications of Humanoid Robots · Best Humanoid Robots in 2026 Ready to buy? Browse humanoid robots for sale on Robozaps. Humanoid robots in hospitality are no longer experimental curiosities — they are operational assets deployed across thousands of hotels, restaurants, and airports worldwide. In 2026, the hospitality robotics market is projected to exceed $3.1 billion, driven by labor shortages consuming over one-third of hotel revenue, rising guest expectations, and rapidly declining hardware costs. This guide covers every robot model worth knowing, real deployment case studies, ROI data, and a practical framework for adoption. This guide is part of our series on the applications of humanoid robots across 12 industries. The State of Hospitality Robotics in 2026 The hospitality industry is experiencing a structural labor crisis. According to the American Hotel & Lodging Association, 87% of hotels reported staffing shortages in 2025, with housekeeping and front desk positions being the hardest to fill. Labor costs now consume approximately 33% of total hotel revenue (STR Global), making robotics and automation increasingly attractive to owners and operators. Several converging factors have accelerated adoption in 2026: Hardware cost reduction: Service robot prices have dropped 40% since 2022, with delivery robots now available from $30–50/day on lease modelsAI maturity: Large language models enable natural multilingual conversation, replacing scripted interactionsGuest acceptance: Surveys show 48% of travelers are comfortable with robotic greetings at check-in, up from 31% in 2023Proven ROI: Early adopters report 12–18 month payback periods on robot investmentsPost-pandemic hygiene standards: Contactless service remains a strong guest preference Types of Robots Deployed in Hospitality Not all hospitality robots are humanoid. Understanding the categories helps operators choose the right solution for their specific pain points. 1. Humanoid Robots (Guest-Facing) Human-shaped robots designed for reception, concierge, and guest engagement. They use facial recognition, emotion detection, and natural language processing to interact with guests. Examples include SoftBank Pepper, UBTECH Walker S, and FLAE BE-A. 2. Delivery Robots Autonomous mobile robots that navigate hallways and elevators to deliver room service items, amenities, and food. Leading models: Relay by Relay Robotics (formerly Savioke), Keenon W3, Bear Robotics Servi, and Pudu BellaBot. 3. Cleaning and Disinfection Robots Autonomous vacuum, mopping, and UV disinfection units for lobbies, corridors, and guest rooms. commercial cleaning robots cleans guest rooms 20% faster and public areas up to 80% faster than human housekeepers. LG CLOi handles commercial-scale floor cleaning. 4. Food Service Robots Kitchen automation and food delivery robots for hotel restaurants and buffet operations. Bear Robotics Servi Plus and Keenon T8 are deployed in over 25,000 restaurants globally, handling food running and table bussing. 5. Luggage and Logistics Robots Automated luggage handling systems like YOTEL's Yobot — a robotic arm that stores and retrieves bags from secure lockers — and autonomous bellhop carts for large resort properties. Comparison Table: Leading Hospitality Robots in 2026 Real-World Case Studies Henn-na Hotel, Japan — The World's First Robot-Staffed Hotel Opened in 2015 by H.I.S. Group, Henn-na Hotel in Nagasaki pioneered the concept of a fully robot-operated hotel. At its peak, the property employed over 243 robots handling check-in (via Pepper humanoids and velociraptor-shaped bots), luggage storage, room cleaning, and in-room voice assistants. The hotel reduced staffing requirements by approximately 72%, with human staff focused exclusively on maintenance and complex guest issues. Key lesson: Henn-na later retired about half its robots due to reliability issues and guest complaints — proving that robot deployment must be strategic, not wholesale replacement. The hotel's current hybrid model (robots + lean human team) is now considered the industry gold standard. Aloft Hotels (Marriott) — Relay Delivery Robots Aloft Hotels became one of the first major US hotel brands to deploy autonomous delivery robots. The Relay robot (originally "Botlr") navigates hallways and elevators independently, delivering toothbrushes, snacks, towels, and room service items to guest rooms. The robot calls the room phone upon arrival and opens its compartment when the guest responds. Results: Properties report a 25% reduction in front desk calls for amenity requests and consistently positive guest reviews mentioning the robot as a highlight of their stay. The robots operate 24/7 without breaks, handling peak-hour requests that would otherwise require additional staffing. Hilton McLean, Virginia — Connie the AI Concierge Connie, a Watson-powered humanoid robot developed with IBM, was stationed in Hilton's McLean property as an AI concierge experiment. Connie answered questions about hotel amenities, local attractions, and dining recommendations using natural language processing. While Connie's active deployment has wound down, the project generated critical learnings about guest interaction patterns that now inform Hilton's broader AI strategy. Marriott Belgium — Mario the Multilingual Robot The Marriott Hotel in Ghent, Belgium deployed "Mario," a robot capable of communicating in 19 languages. Mario greets guests, provides hotel information, and handles routine inquiries — operating continuously without breaks. For a property serving international business travelers and tourists from across Europe, the multilingual capability eliminated the need for multilingual front desk staff during off-peak hours. Keenon Deployments — Scale Across Asia-Pacific Keenon Robotics has deployed robots in over 10,000 hotels and 25,000 restaurants across Asia-Pacific, Europe, and North America. Their W3 delivery robot has become the workhorse of hotel room service automation, with properties reporting that a single unit can handle 300+ deliveries per day, equivalent to the workload of 1.5 full-time delivery staff members. ROI Analysis: The Business Case for Hospitality Robots The financial case for hospitality robots has strengthened considerably as costs decrease and deployment data accumulates. Here is a realistic breakdown: Cost Structure Delivery robots (lease): $1,000–$2,000/month — equivalent to roughly $33–$66/dayDelivery robots (purchase): $15,000–$25,000 with a 5–7 year operational lifespanHumanoid concierge robots: $25,000–$100,000+ depending on capabilitiesCleaning robots: $20,000–$50,000 for commercial-grade unitsMaintenance: Typically 10–15% of purchase price annually Savings and Revenue Impact Labor cost reduction: A delivery robot operating 24/7 replaces 1.5–2 FTE positions, saving $45,000–$70,000/year in wages and benefitsCleaning robots: commercial cleaning robots cleans rooms 20% faster, enabling faster turnover and higher occupancy potentialFood service robots: Bear Robotics reports clients save up to 75% in food runner labor costsRevenue uplift: Hotels with robots report 15–30% increases in social media mentions and organic marketing value from guests sharing robot interactionsGuest satisfaction: Properties consistently report improved review scores related to innovation and service speed Payback Period For a mid-range delivery robot purchased at $20,000 replacing partial FTE labor at $50,000/year in total compensation: payback occurs in approximately 5–8 months. Leased robots achieve positive ROI from month one if they displace even partial staffing needs. Humanoid concierge robots have longer payback periods (18–36 months) but generate significant brand differentiation and marketing value that is harder to quantify. How to Implement Robots in Your Hotel: A Practical Framework Step 1: Audit Your Pain Points Map your guest journey from pre-arrival to checkout. Identify bottlenecks: long check-in queues, slow room service delivery, housekeeping delays, repetitive front desk inquiries. These high-volume, low-complexity tasks are prime candidates for automation. Step 2: Start with One Use Case Don't deploy robots across every department simultaneously. Start with the highest-impact use case — typically room delivery or lobby concierge. Run a 90-day pilot and measure specific KPIs: delivery time, guest satisfaction scores, staff workload reduction, and social media engagement. Step 3: Infrastructure Requirements Wi-Fi: Robots require reliable, high-bandwidth connectivity throughout the property including hallways, elevators, and back-of-house areasPMS integration: Advanced robots can pull reservation data, guest preferences, and loyalty status — but require API access to your property management systemPhysical environment: Smooth flooring, ADA-compliant pathways, and elevator connectivity are essential for autonomous navigationCybersecurity: Robots collecting guest data must comply with GDPR and local privacy regulations Step 4: Train Your Staff Position robots as colleagues, not replacements. Staff need to know how to redirect guests to the robot for routine tasks, troubleshoot basic issues, and intervene when the robot cannot handle a complex request. The most successful deployments treat robots as tools that free human staff for higher-value interactions. Step 5: Market Your Innovation Feature robots prominently in your marketing. Encourage guests to interact with them and share on social media. Properties that embrace robots as part of their brand identity generate organic visibility that would cost thousands in traditional advertising. Pros and Cons of Hospitality Robots Pros 24/7 operation without breaks, sick days, or overtime costsConsistency: Every interaction follows the same standard, eliminating variability in service qualityMultilingual capability: Handle international guests without multilingual staffingContactless service: Maintain hygiene standards that guests now expectLabor shortage solution: Fill positions that are increasingly difficult to recruit forMarketing value: Generate organic social media content and positive reviewsData collection: Gather operational and guest preference data for continuous improvementScalability: Easily add units during peak seasons without recruitment and training cycles Cons High upfront cost: Humanoid robots can exceed $100,000; even delivery bots require significant capitalTechnical failures: Malfunctions during service can frustrate guests and damage brand perceptionLimited emotional intelligence: Cannot match human empathy for complex or sensitive guest situationsGuest resistance: Some demographics, particularly luxury travelers and older guests, prefer human interactionInfrastructure needs: Requires Wi-Fi upgrades, smooth flooring, and elevator integrationMaintenance burden: Regular software updates, hardware repairs, and battery managementStaff anxiety: Without proper change management, employees may view robots as threats Future Outlook: Hospitality Robotics 2026–2030 The next five years will see hospitality robotics evolve from novelty deployments to standard infrastructure. Key trends to watch: Generative AI Integration Robots powered by large language models will hold natural, context-aware conversations with guests — moving far beyond scripted responses. A robot concierge in 2027 will be able to discuss restaurant recommendations, adjust room settings, and handle complaints with near-human conversational ability. Humanoid Robots at Scale Companies like Figure, Tesla (Optimus), Agility Robotics (Digit), and 1X (NEO) are racing to produce general-purpose humanoid robots at sub-$50,000 price points. Once these reach commercial availability (expected 2027–2028), hospitality will be among the first industries to adopt them for housekeeping, luggage handling, and guest services. Autonomous Cleaning Fleets Hotels will deploy fleets of coordinated cleaning robots that autonomously manage room turnover, public area maintenance, and deep cleaning schedules — reducing housekeeping labor by 40–60% while maintaining higher consistency. Emotional AI and Personalization Next-generation robots will read facial expressions, voice tone, and body language to adapt their behavior. A robot that detects a tired guest will adjust its energy level, offer room upgrade suggestions, or simply streamline the check-in process for speed. Market Projections The global hospitality robotics market is projected to grow at a CAGR of 25.5% through 2030, reaching over $10 billion. Asia-Pacific leads adoption, followed by North America and Europe. The segment shift from delivery robots to multipurpose humanoid platforms will drive the majority of value growth. Compare on RoboZaps: Bear Robotics Servi Plus, Bear Robotics Servi, Pudu Bellabot, Pepper, Walker S and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in Healthcare: How They're Transforming Patient Care, Surgery, and Eldercare in 2026 URL: https://blog.robozaps.com/b/humanoid-robots-in-healthcare Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:04:56.817Z Category: insights TL;DR: Healthcare robotics is heading past $44 billion by 2028, driven by a projected 13-million nurse shortage. Real deployments exist now, from Fourier's GR-1 in rehab centers to Moxi in US hospitals, with a 24/7 humanoid assistant costing $50K-$100K a year against $60K-$80K for a single nurse. Regulation and EHR integration are the gatekeepers. Key takeaways: - The healthcare robotics market is projected to exceed $44 billion by 2028, with humanoid robots representing the fastest-growing segment. - Humanoid robots are actively deployed in rehabilitation (Fourier GR-1), surgical assistance (da Vinci systems evolving toward humanoid form), elderly care, hospital logistics, and mental health support. - A global nursing shortage of 13 million by 2030 (WHO) is the primary driver—humanoid robots can fill gaps without replacing human caregivers. - Real-world deployments are already happening: Fourier's GR-1 in Chinese rehab centers, Moxi by Diligent Robotics in US hospitals, and Pepper in Japanese elderly care facilities. - Key challenges include patient data privacy, FDA/CE regulatory approval pathways, integration with electronic health records (EHR), and the irreplaceable human element of empathy. - The cost-benefit equation is shifting: a humanoid hospital assistant operating 24/7 can cost $50K–$100K/year versus $60K–$80K for a single human worker on one shift. FAQ: Q: Will humanoid robots replace doctors and nurses? A: No. Humanoid robots are designed to complement healthcare workers, not replace them. They handle logistics, routine monitoring, physical therapy guidance, and administrative tasks—freeing clinicians to focus on complex care and human connection. The global nursing shortage means there aren't enough healthcare workers to replace even if hospitals wanted to. Q: Are healthcare robots safe for patients? A: Healthcare robots undergo rigorous safety testing. They include force-limited actuators (preventing injury from contact), emergency stop mechanisms, obstacle avoidance, and are designed to operate at slow speeds near patients. Surgical robots have additional safeguards including real-time tremor filtering and range-of-motion limits. However, cybersecurity remains a concern that requires ongoing attention. Q: How much do healthcare humanoid robots cost? A: Prices range widely. A logistics robot like Moxi costs approximately $150,000–$200,000 to purchase or $3,000–$5,000/month to lease. Rehabilitation robots like the GR-1 start around $100,000–$170,000. SoftBank's Pepper is available for approximately $25,000–$30,000 plus monthly service fees. Surgical robot systems (da Vinci) cost $1.5M–$2.5M. Most hospitals pursue lease or Robot-as-a-Service models. Q: What data do healthcare robots collect, and is it HIPAA-compliant? A: Healthcare robots may collect video, audio, vital signs, movement data, and interaction logs. Reputable manufacturers design their systems for HIPAA compliance with on-device processing, encrypted data storage, access controls, and audit trails. Hospitals should conduct a thorough security assessment and ensure a Business Associate Agreement (BAA) is in place with the robot manufacturer. Q: Can humanoid robots help with mental health treatment? A: Yes, and the evidence is growing. Humanoid robots have shown particular promise in autism therapy (where children respond well to predictable, non-judgmental interaction), dementia engagement, and companionship for isolated elderly patients. They're not substitutes for licensed therapists but can provide consistent supplementary support—especially in settings where mental health professionals are scarce. Q: How long until humanoid robots are common in hospitals? A: Logistics and rehabilitation robots are already deployed in hospitals today. By 2028, expect to see humanoid robots routinely in large hospitals (500+ beds) handling supply delivery, patient transport assistance, and rehabilitation. Broader adoption—including home health, surgical assistance, and primary care support—will likely scale between 2030 and 2035. Q: What happens if a healthcare robot malfunctions? A: Healthcare robots are designed with multiple redundancy systems. If a malfunction is detected, the robot typically stops all movement immediately and alerts hospital staff. For surgical robots, the surgeon maintains override control at all times. Liability for robot-related incidents is still evolving legally, but currently falls on a combination of the manufacturer, the hospital, and the supervising clinician depending on the circumstances. The Healthcare Crisis That's Driving Robot Adoption Healthcare systems worldwide are facing a staffing crisis that has no precedent. The World Health Organization projects a global shortage of 13 million healthcare workers by 2030. The American Hospital Association reports that US hospitals lost over 100,000 registered nurses during 2021–2022 alone—the largest decline in four decades. Japan, with the world's oldest population, has a caregiver-to-elderly ratio that's deteriorating every year. This guide is part of our series on the applications of humanoid robots across 12 industries. Meanwhile, demand is surging. The global population over 65 is expected to double from 761 million (2021) to 1.6 billion by 2050. These demographics create a mathematical impossibility: there simply won't be enough human healthcare workers to care for aging populations. This is where humanoid robots enter the picture—not as replacements for doctors and nurses, but as force multipliers that handle routine, physically demanding, and repetitive tasks, freeing human clinicians to focus on complex care, empathy, and decision-making. How Humanoid Robots Are Used in Healthcare Today Patient Care and Monitoring Humanoid robots equipped with cameras, microphones, and biosensors can continuously monitor patients—checking vital signs, detecting falls, and alerting staff to changes in condition. Unlike wearable devices that patients may remove, a humanoid robot can actively observe and respond. Real-world example: In several Japanese eldercare facilities, SoftBank's Pepper robot conducts daily check-ins with residents, leading exercise routines, reminding patients about medications, and alerting staff when a resident appears distressed. Tencent's Xiaowu robot uses facial recognition and natural conversation to provide emotional support in pediatric and geriatric wards. Rehabilitation and Physical Therapy This is one of the most proven applications for humanoid robots in healthcare. Rehabilitation requires repetitive, precise movements—exactly what robots excel at. Real-world example: Fourier Intelligence's GR-1, with 40 degrees of freedom, is actively deployed in rehabilitation centers across China. It guides patients through physical therapy routines for gait training, upper limb recovery, and balance exercises. The robot provides consistent, tireless support and collects detailed data on patient progress that human therapists can use to optimize treatment plans. Other rehabilitation robots include Cyberdyne's HAL (Hybrid Assistive Limb), which is used in over 100 facilities in Japan and Europe for walking rehabilitation in stroke and spinal cord injury patients. Surgical Assistance While most surgical robots today (like the da Vinci system by Intuitive Surgical) are not humanoid in form, the trend is moving toward more human-like assistants that can work alongside surgeons in the operating room. Current surgical robots have already demonstrated remarkable outcomes: 50% reduction in patient recovery time for robotic-assisted procedures vs. traditional open surgeryLess blood loss and smaller incisions, reducing infection riskSub-millimeter precision for procedures like prostatectomy, cardiac valve repair, and neurosurgery The next generation of surgical assistants will combine humanoid form factors with AI-powered decision support, allowing them to hand instruments, retract tissue, and even perform routine suturing autonomously under surgeon supervision. Hospital Logistics and Supply Chain Hospitals are complex logistical environments. Nurses spend an estimated 25–30% of their time on non-clinical tasks: fetching supplies, delivering medications, transporting specimens, and restocking rooms. Real-world example: Diligent Robotics' Moxi, deployed in over a dozen US hospitals, autonomously delivers supplies, lab samples, and medications throughout hospital floors. While Moxi isn't fully humanoid (it has a single arm and wheeled base), it demonstrates the trajectory: the next generation will have humanoid form factors for navigating stairs, opening doors, and operating elevators. In China, multiple hospitals deployed humanoid-style robots during the COVID-19 pandemic for contactless temperature screening, medication delivery to isolation wards, and UV disinfection—reducing healthcare worker exposure to the virus. Mental Health and Emotional Support Perhaps surprisingly, humanoid robots are showing promising results in mental health applications. Their non-judgmental nature and infinite patience make them effective for: Autism therapy: Children with autism spectrum disorder often respond more positively to robots than to human therapists. The predictable, consistent behavior of robots reduces anxiety and encourages social interaction.Dementia care: Humanoid robots can engage dementia patients in conversation, reminiscence therapy, and cognitive exercises without frustration or impatience.Loneliness reduction: For isolated elderly patients, a humanoid companion that can converse, play games, and facilitate video calls with family provides meaningful social engagement. Medical Training and Simulation Humanoid robots are increasingly used as standardized patients in medical education. Unlike human actors, robots can consistently replicate specific symptoms, vital signs, and reactions, providing medical students with repeatable training scenarios for physical exams, patient interviews, and emergency response. Healthcare Robots: A Comparison of Leading Models The Economics of Healthcare Robots The financial case for humanoid robots in healthcare is becoming compelling—and it's not primarily about replacing workers. Cost Comparison Important caveat: This comparison isn't about replacing nurses. It's about using robots to handle the 25–30% of nursing time spent on logistics and routine tasks, effectively giving hospitals the equivalent of more nursing hours without hiring additional staff in a market where there aren't enough nurses to hire. ROI Timeline Hospitals piloting logistics robots typically see ROI within 12–18 months. The primary savings come from: Reduced nurse overtime (average savings: $200K–$500K per facility per year)Fewer medication delivery errors (estimated cost of drug errors: $42B annually in the US)Lower hospital-acquired infection rates from contactless deliveryImproved nurse retention (less burnout = lower turnover costs) Ethical Considerations and Challenges Patient Data Privacy Humanoid robots equipped with cameras, microphones, and biosensors collect enormous amounts of sensitive health data. HIPAA compliance in the US and GDPR in Europe impose strict requirements on how this data is stored, processed, and shared. Any robot operating in healthcare must be designed with privacy-by-default principles: On-device processing where possible (minimizing data transmission)End-to-end encryption for all patient dataClear patient consent mechanismsRegular security audits and penetration testing The Empathy Gap Healthcare is fundamentally human. A robot can deliver medication, but it cannot hold a dying patient's hand with genuine compassion. Research consistently shows that patients value human connection in healthcare—particularly during vulnerable moments. The most effective approach is hybrid care models where robots handle routine and physical tasks, and humans provide emotional support, complex decision-making, and compassionate presence. Regulatory Pathways Healthcare robots face complex regulatory requirements. In the US, the FDA classifies medical robots differently depending on their function—surgical robots require rigorous premarket approval, while logistics robots may fall under different categories. The EU's Medical Device Regulation (MDR) and AI Act add additional layers. As of 2026, there is no unified global regulatory framework for humanoid robots in healthcare, creating uncertainty for manufacturers and hospitals alike. Algorithmic Bias AI systems trained on biased datasets can produce biased outcomes. If a diagnostic AI was trained primarily on data from one demographic group, it may perform poorly for others. In healthcare, bias can be life-threatening. Rigorous testing across diverse patient populations is essential before deploying AI-powered humanoid robots in clinical settings. Liability and Accountability If a humanoid robot makes an error that harms a patient—delivers the wrong medication, fails to alert staff to a deteriorating condition—who is liable? The manufacturer? The hospital? The AI developer? Current legal frameworks weren't designed for autonomous agents, and this remains one of the biggest unresolved questions in healthcare robotics. The Future: What's Coming in Healthcare Robotics 2026–2028: Expansion of Current Applications Logistics robots become standard in major hospitals (500+ bed facilities)Rehabilitation robots expand beyond specialized centers to community clinicsAI-powered triage robots deployed in emergency departments to assess patient acuityTelepresence humanoids enable specialist consultations in rural and underserved areas 2028–2032: Next-Generation Capabilities Humanoid surgical assistants capable of autonomous suturing and routine procedures under supervisionEmotionally intelligent robots that can detect depression, anxiety, and cognitive decline through behavioral analysisHome health robots that monitor chronic conditions (diabetes, heart failure, COPD) and alert physicians to early warning signsIntegration with genomic data for personalized treatment recommendations 2032+: Transformative Change Fully autonomous nursing assistants for elderly care facilitiesHumanoid robots as primary caregivers for routine home health visitsAI-driven diagnostic capabilities that match or exceed specialist-level accuracy for common conditions24/7 personalized health coaching through humanoid companions Case Studies: Healthcare Robots in Action Fourier GR-1 in Chinese Rehabilitation Centers Fourier Intelligence's GR-1 humanoid robot has been deployed across multiple rehabilitation centers in China since 2024. In clinical settings, the robot guides stroke patients through repetitive gait training exercises—a task that requires consistent timing and positioning that can be exhausting for human therapists. Early clinical data suggests patients using robot-assisted rehabilitation show 15–20% faster recovery in mobility metrics compared to traditional therapy alone, while therapists report significantly reduced physical fatigue. Moxi at Texas Health Resources Diligent Robotics deployed its Moxi robots at Texas Health Resources hospitals, where they autonomously deliver lab samples, supplies, and medications to nursing stations. Nurses at the facility reported regaining approximately 30 minutes per shift previously spent on supply runs, which they redirected to patient care. The hospital also saw a measurable improvement in nurse satisfaction scores. Pepper in Japanese Eldercare At several Shin-Ai nursing homes in Japan, SoftBank's Pepper conducts daily group exercise sessions, leads sing-alongs, and provides cognitive stimulation activities for dementia patients. Staff report that residents are more engaged and active on days when Pepper leads activities, and the robot's consistent availability (it doesn't call in sick or need breaks) has improved the regularity of therapeutic programming. How to Evaluate Healthcare Robots for Your Facility If you're a healthcare administrator considering humanoid robots, here's a practical evaluation framework: Identify the pain point: What specific tasks consume the most staff time or cause the most burnout? Start there.Assess regulatory requirements: Does the robot's intended use require FDA/CE clearance? Logistics robots typically don't; clinical-use robots do.Evaluate integration: Can the robot integrate with your existing EHR, nurse call, and facility management systems?Calculate total cost of ownership: Include purchase/lease, maintenance, training, IT infrastructure, and insurance over 5 years.Run a pilot: Most manufacturers offer 90-day pilot programs. Measure before-and-after metrics on staff time, patient satisfaction, and error rates.Plan for staff buy-in: Healthcare workers may be skeptical. Frame robots as tools that eliminate tedious tasks, not as replacements. Compare on RoboZaps: Pepper. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # The Future of Humanoid Robots: What 2026 Is Actually Proving URL: https://blog.robozaps.com/b/future-of-humanoid-robots Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:05:00.680Z Last fact-checked: 2026-07-14T00:00:00.000Z Category: insights TL;DR: 2026 is a validation year for humanoid robots, not proof of mainstream adoption. The strongest evidence remains narrow industrial and logistics work; cheaper hardware does not automatically equal a reliable labour substitute. Judge any roadmap on repeatability, supervision, runtime, safety, service and total cost, not promised adoption dates. Key takeaways: - 2026 is a validation year for humanoid robots, not proof of mainstream adoption. - The strongest operating evidence is still narrow industrial and logistics work with controlled tasks and vendor support. - Lower-priced hardware is real, but a development platform is not automatically a reliable labour substitute or home robot. - Robot AI models are improving quickly. Model benchmarks and demonstrations do not establish uptime, safety or return on investment in a live workplace. - A credible roadmap should be tied to repeatability, supervision, runtime, safety, service and total cost, not a promised adoption year. FAQ: Q: When will humanoid robots be broadly available to consumers? A: There is no dependable broad-availability date. Some development platforms can be purchased now, but a generally useful home humanoid needs a much stronger record on manipulation, safety, support and unsupervised reliability. Treat 2027, 2030 and later dates as vendor or analyst scenarios unless accompanied by a real order page and delivery evidence. Q: How much does a humanoid robot cost? A: The range is wide because the category includes low-cost research platforms, developer editions, industrial systems sold through private agreements and products that are not for sale. Compare the exact configuration, hands, computing hardware, warranty, shipping, integration and service terms rather than using a single market average. Q: Which company is leading? A: There is no single leader across every measure. Vendors differ in mobility, manipulation, AI, production, price and operating evidence. For a buyer, the leader is the system with the best verified performance and support for the intended task, not the company with the most ambitious general forecast. Q: Are humanoid robots safe? A: Safety depends on the robot, task, environment and controls. Emergency stops, force limits and proximity sensing are useful components, but they do not replace a site-specific risk assessment, access rules, training and incident response. A system that is safe in one controlled workflow may not be safe in a public space or home. Q: What can humanoid robots reliably do today? A: The best evidence is for bounded physical tasks such as tote handling and parts manipulation in structured environments. Walking, grasping and responding to instructions have improved, but performance varies sharply by platform and setup. General-purpose unsupervised work across changing tasks is not yet a normal commercial capability. 2026 Is a Validation Year The future of the humanoid robot sector is easier to understand when evidence is ranked by strength. A vendor video shows that a task can be demonstrated. A controlled pilot tests it in a real setting. A paid deployment shows that a customer sees potential value. Repeated multi-site operation with published throughput, intervention and safety data is stronger again. The market now has examples beyond laboratory videos. Figure reported that Figure 02 contributed to the production of 30,000 cars at BMW in 2025 and showed Figure 03 performing a new logistics workflow at BMW in June 2026. Figure described the new work as a demonstration, which is useful evidence but not the same as a fleet-scale commercial rollout. Agility Robotics reports that Digit moved more than 100,000 totes at GXO in a commercial deployment. That is the kind of operating milestone buyers should look for: a defined task, a production setting and accumulated cycles. Independent data on interventions, downtime and total cost would make the evidence stronger. What Has Changed Lower-Priced Developer Hardware Unitree currently lists the G1 from $13,500 and the R1 Air from $4,900, before tax and shipping. Those prices matter because they lower the entry cost for research and development. Unitree also warns that the industry remains in an early stage, that some demonstrated functions are still being developed, and that buyers should understand the limitations before purchasing. Price is evidence of hardware access, not evidence that a robot can replace a worker. More Capable Robot AI Models Vision-language-action models can connect instructions and visual input to robot actions. Google DeepMind says Gemini Robotics 1.5 can plan, respond to natural-language instructions and transfer learning across different robot forms. The action model remains in private preview for selected partners. That is meaningful technical progress, but it is not a generally available autonomy layer with a public record of full-shift industrial reliability. Better Operating Evidence The most important change is not a market forecast. It is the gradual appearance of repeatable task evidence. Tote handling, parts sequencing and other structured material flows give vendors a bounded problem and buyers a measurable outcome. These deployments are much more informative than claims that a general-purpose robot will soon handle every task in a factory or home. What Is Still Not Proven Unsupervised home use at normal consumer reliability.Full-shift general-purpose work without charging, battery swaps, teleoperation or frequent intervention.Safe close collaboration across messy workplaces rather than a controlled cell or narrowly defined process.Positive total return after integration, support, downtime, insurance, training and human backup.Broad deployment in elder care, hospitals, schools or retail without tight task and safety boundaries.A mass-market production curve that makes a specific 2028, 2030 or 2040 adoption forecast dependable. Where Adoption Is Most Credible Manufacturing and Logistics Structured material-handling work is the clearest near-term fit. The task can be defined, the environment can be mapped, throughput can be counted and people can be kept outside a safety boundary where necessary. Even here, buyers should distinguish a demonstration from a pilot and a pilot from a long-running paid deployment. Research and Development Lower-cost platforms will expand experimentation in universities, robotics teams and software companies. This may accelerate locomotion, manipulation and control research. It also means more videos will circulate from configured demonstrations. A successful demo should be treated as a capability sample until the same task is repeated under realistic variation. Homes and Personal Care Homes combine clutter, children, pets, stairs, fragile objects and little tolerance for failure. That makes them harder than a controlled warehouse station. Current home humanoid robot candidates are better treated as a watchlist than a mass-market buying category. Personal-care work also raises additional safety, privacy and responsibility questions because the robot may operate near people who are vulnerable. The Roadmap Should Be Measured in Gates, Not Years A useful roadmap asks what a system must prove before the next stage of adoption. These gates apply across vendors: Repeatability: the robot completes the target task across realistic object positions, lighting, traffic and other variation.Supervision: the vendor reports how much teleoperation, reset work and human assistance is required.Runtime: the shift plan includes charging, battery swaps, docking and degraded performance, not just headline battery life.Safety: hazards, force limits, emergency stops, access control and incident response are documented for the actual site.Economics: total cost includes integration, support, downtime, insurance, training and process changes.Scale: parts, repairs, software updates and service-level terms still work when the fleet grows beyond one or two robots. A vendor that clears those gates on one task has evidence for that task. It does not automatically have evidence for another industry, an unstructured home or a different safety case. This is why fixed adoption dates age badly: the hard work is operational, not calendrical. Safety and Reliability The US National Institute for Occupational Safety and Health maintains a robotics workplace-safety program focused on the risks created when robots and people share work. For personal-care robots, ISO 13482 provides a more relevant starting point than a general demo reel. Neither source is a blanket approval for a product. The safety case still has to match the robot, task, environment and people exposed to it. Runtime and manipulation are equally practical constraints. Unitree lists about two hours of battery life for the G1 and about one hour for the R1, while Agility says Digit can run for up to four hours and dock autonomously for charging. Those figures are not directly comparable because the robots, tasks and test conditions differ. Buyers need duty-cycle data from the proposed workflow. What This Means for Buyers, Workers and Consumers For Buyers Start with one repetitive task and a measurable baseline. Require an intervention log, safety review, support terms and a full cost model before expanding. Public list price is only one input; the humanoid robot cost guide tracks the difference between list prices, development platforms, pilot agreements and unpriced systems. For Workers Watch how tasks change inside a role. Supervision, maintenance, safety, process design and exception handling are likely to grow around early deployments. The employment effect will differ by workplace and region, so broad claims about all jobs are less useful than the evidence summarized in the humanoid robots and jobs analysis. For Consumers Do not treat a demonstration, announced target price or planned launch as proof of ordinary delivery. Check whether the product can be ordered, what is included, whether the demonstrated features ship with it, and who provides service. General-purpose home autonomy remains unproven. The Bottom Line Humanoid robotics has moved into a more serious phase. Lower-cost hardware, better embodied-AI models and a small number of operating deployments justify attention. They do not justify a timetable in which mass adoption is treated as inevitable. The next phase will be decided by boring but decisive numbers: completed cycles, interventions, uptime, incidents, service response and total cost. Platforms that publish and repeat those results will shape the market. The rest remain demonstrations, development systems or forecasts. Compare on RoboZaps: Figure 03, Figure 02 and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Ameca Robot Price 2026: How Much It Costs & Can You Buy One? URL: https://blog.robozaps.com/b/ameca-review Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:05:26.952Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: Engineered Arts has never published an Ameca price: quote-only purchase, event rental for everyone else. Best-supported figure: ~$250,000/unit (CNBC 2024; a school paid $500k for two in Jan 2026). The current model is Gen 2.6, static and mains-powered; the 'walking Gen 3 with 8-hour battery' online is fabricated. The real Gen 3 is the desktop unit, released June 22, 2026. Key takeaways: - No official Ameca price exists: quote-only purchase, event rental for everyone else. The best evidence is ~$250,000/unit (CNBC 2024; a San Diego school paid $500k for two in Jan 2026). - The current full-size Ameca is Generation 2.6 (2024), static and mains-powered; the 'Gen 3 with walking and an 8-hour battery' online is an AI-farm fabrication that leaked into Wikipedia. - Ameca cannot walk, four years after predicting legs 'in less than a year'; the actual new release is the Ameca Desktop Gen 3 (June 22, 2026). - The conversations run real LLMs (ChatGPT/Whisper via Tritium OS), but a disclosed telepresence mode exists and the famous 'emotional' gestures are latency-masking choreography. - Roughly 30 full-size Amecas exist across 27 countries, from Dubai's Museum of the Future to a San Diego charter school; cart-button 'for sale' listings are all unofficial. FAQ: Q: How much does an Ameca robot cost? A: There is no official price: Engineered Arts sells by quote only. The best-supported figure is roughly $250,000 per unit, reported via CNBC in 2024 and matched by a San Diego charter school's $500,000 purchase of two Amecas in January 2026. The 2021 'more than £100,000' floor came from founder Will Jackson himself; the tiered price menus circulating online are unattributed. Q: Can you buy an Ameca robot? A: Institutions can, by contacting Engineered Arts for a quote; museums, universities and a school network have. Engineered Arts publishes no online checkout, consumer pricing or eligibility criteria; marketplace listings with cart buttons are unofficial brokers, name-squatters or toys. Most organizations rent instead, through Engineered Arts' event rental program. Q: Is the Ameca robot real? A: Yes: real hardware from Engineered Arts, with 61 actuated movements including 27 in the face. The conversations in its famous clips ran real language models (GPT-3, later ChatGPT) unscripted, though sessions were company-run, a human-operator telepresence mode exists, and its most 'emotional' gestures are choreographed. Q: Can Ameca walk now? A: No. Every commercial Ameca is a waist-up robot on a static pedestal, mains-powered. Engineered Arts showed prototype legs and Ameca predicted walking 'in less than a year' back in 2022, but no walking unit has ever shipped, and claims of a walking 'Gen 3' with an 8-hour battery are fabricated. Q: What generation is the Ameca robot? A: The current full-size model is Generation 2.6 (2024), per Engineered Arts' official spec page. The only real Generation 3 product is the Ameca Desktop Gen 3, a head-and-shoulders unit released June 22, 2026 alongside the desktop characters Ami and Azi. Q: Is Ameca the most advanced robot in the world? A: That's Engineered Arts' own tagline, and it includes a qualifier most headlines drop: most advanced social humanoid robot. As an expressive, conversational face it has no peer; as a robot it cannot walk or manipulate objects usefully, which working humanoids like factory robots can. Q: What AI does Ameca use? A: Engineered Arts' own Tritium OS runs the robot, with the language brain plugged in from commodity models: the spec sheet names OpenAI's ChatGPT and Whisper plus Amazon Polly, and the 2022 demos ran GPT-3. A telepresence mode also lets a human operator drive conversations. Q: Did Ameca really get annoyed and roll its eyes? A: Not emotionally. The 'personal space' reaction was a staged sensor-behavior demo, and the famous eye-roll was latency-masking choreography: Ameca is programmed to glance away while the language model computes, as founder Will Jackson explained to Fortune. Q: Where can you see an Ameca robot? A: Permanently at Dubai's Museum of the Future and Edinburgh's National Robotarium, as the custom 'Aura' greeters at Sphere Las Vegas, and at rotating events; recent stops include ESA's European Space Conference exhibition and the San Mateo County Fair. Engineered Arts also rents Ameca for conferences and activations. Q: Who makes the Ameca robot? A: Engineered Arts, founded in Cornwall, England in 2004 by Will Jackson, maker of the RoboThespian stage robots before Ameca. It restructured as a US company in December 2024 with a $10 million Series A and now operates from Redwood City, California. Q: Is Ameca better than Sophia? A: They're different generations of the same idea. Sophia (2016) is the scripted celebrity with the citizenship stunts; Ameca (2021) is the engineering successor: deliberately grey and genderless, running commodity LLMs on its own OS. Ameca's face hardware and conversation stack are more advanced; Sophia remains more famous. Q: How many Ameca robots exist? A: Roughly 30 full-size Amecas across 27 countries plus about 50 desktop units, per figures Engineered Arts gave CNBC in late 2024. The company's '250+ robots deployed' marketing counter includes two decades of its RoboThespian stage robots. How much does the Ameca robot cost, and can you buy one? Engineered Arts has never published a price for Ameca: it sells by quote and, for most of the world, rents by the event. Every figure you've seen is third-party, and they don't agree with each other. The strongest evidence available in July 2026: a San Diego charter school network paid $500,000 for two Amecas installed in January, matching the ~$250,000 per unit reported to CNBC in 2024. We checked every claim on this page against Engineered Arts' live pages and primary coverage on July 22, 2026. How you actually get one (and the listings to avoid) Two legitimate channels exist. Purchases go by private quote, and every known buyer is an institution: museums, universities and brands contact Engineered Arts sales and negotiate, the way the Museum of the Future, Edinburgh's National Robotarium and Altus Schools did. The company publishes no retail checkout, consumer pricing or eligibility criteria. Most others rent: Engineered Arts runs a first-class rental program covering Ameca, its desktop siblings and the Mesmer line for conferences, activations and education, priced by quote like everything else. The "for sale" search results also carry things to avoid. A site called amecarobotics.com is a name-squatter reselling unrelated service robots. Marketplace pages with add-to-cart buttons for Ameca are unofficial quote-brokers at best; Engineered Arts has no online checkout and no published distributor network. And the "Ameca" listings on AliExpress are toys. If a page shows you a cart or a price, it isn't Engineered Arts. The generation truth (and how Wikipedia got contaminated) Engineered Arts' own spec page says the current full-size Ameca is Generation 2.6 (2024): 1,870 mm and 62 kg on a static pedestal, 61 actuated movements (27 in the face alone, plus neck, shoulders, arms, hands and torso), 8 MP cameras in each eye, binaural and chest microphone arrays, LiDAR safety sensors, and mains power through two 24 V supplies. Note what's not on that sheet: no battery, no legs, no walking speed. That matters because a cluster of AI-generated encyclopedia sites has spent 2025–2026 describing an "Ameca Gen 3" unveiled at ICRA 2025 with walking prototypes and an eight-hour battery, and the claim has leaked into Wikipedia. Engineered Arts did exhibit at ICRA 2025, and it did launch new robots this year, but the official record contains no full-size Gen 3, no battery, and no mobility. The real Generation 3 is the Ameca Desktop Gen 3, a 525 mm, 8.5 kg head-and-shoulders unit released on June 22, 2026, alongside the desktop characters Ami and Azi. When a spec conflicts with the manufacturer's live page, trust the manufacturer's page. Can Ameca walk now? No. Every commercial Ameca ever delivered is a waist-up robot bolted to a pedestal. The history stings a little: in a 2022 Engineered Arts video, Ameca itself predicted working legs "in less than a year," drawing on the company's earlier bipedal research. Four years later, prototype legs have been shown and no walking unit has shipped, and the current spec sheet doesn't mention mobility at all. Ameca's entire value lives above the waist, in the most expressive face in commercial robotics, and Engineered Arts has effectively priced walking out of the product's story. What actually powers the conversations Ameca runs Engineered Arts' own Tritium software stack, a custom Linux platform that handles the face, voice and conversation pipeline, with the language brain plugged in from outside: the spec sheet names OpenAI's ChatGPT and Whisper plus Amazon Polly, and the famous 2022 demos ran GPT-3 with Engineered Arts stating on the record that nothing was pre-scripted. Two honest footnotes belong next to that. First, Tritium also includes a telepresence mode, meaning a human operator can drive Ameca's conversation, a capability Engineered Arts discloses and event coverage almost never mentions. Second, some of the most human moments are engineered: Will Jackson told Fortune that because GPT responses took about two seconds, Ameca is programmed to glance away and break eye contact so it doesn't look frozen; the viral "eye roll" at a journalist's question was that choreography, not sass. The viral clips, fact-checked Where Amecas actually are Roughly 30 full-size Amecas across 27 countries, plus about 50 desktop units, per figures the company gave CNBC in late 2024; its all-models marketing counter says 250-plus robots deployed, a number that includes two decades of RoboThespian stage robots. Ameca was also a fixture of the UN's AI for Good summits from the famous 2023 robot press conference onward, though we found no confirmed appearance at the July 2026 edition. The company behind the face Engineered Arts is a 2004-vintage Cornwall company that spent its first two decades selling RoboThespian theater robots before Ameca made it famous. In December 2024 it restructured as a US company and raised a $10 million Series A led by Helium-3 Ventures ($16.2 million total), with Muse frontman Matt Bellamy joining as a board observer, and set up in Redwood City, California. Founder Will Jackson's positioning has been consistent for twenty years: robots for communication and entertainment, explicitly not labor. That's also the honest way to read the company's own tagline, "the world's most advanced social humanoid robot": the word doing the work is social, which most headlines drop. Ameca vs Sophia, briefly The two expressive-head icons get conflated constantly. Sophia (Hanson Robotics, 2016) is the celebrity: patented skin, Saudi citizenship theater, and dialogue that its own maker described as part-scripted, part-chatbot. Ameca (2021) is the engineering-led successor: deliberately grey and genderless to sidestep the uncanny valley, commodity LLMs plugged into its own OS, and no citizenship stunts. Our full Sophia review covers her strange 2026 afterlife; the short version is that Ameca inherited the "most famous robot face" title by shipping a better face. Misinformation audit: seven Ameca claims, checked What to watch next The realistic things to wait for are not legs. Desktop-line pricing is the one that could change access: a head-and-shoulders Gen 3 Ameca Desktop, Ami or Azi is the plausible first Engineered Arts robot with a price ordinary institutions can say yes to quickly. The Altus Schools experiment will produce the first real outcome data on Ameca in education, and June's reporting suggests it could cut either way. And whether Engineered Arts confirms or corrects the full-size "Gen 3" record will decide how long the fabricated spec keeps circulating. We re-verify this page against the company's live pages; every figure carries its check date. The bottom line Ameca is the best face in robotics attached to the industry's most opaque price tag. The honest summary for a buyer: figure roughly a quarter-million dollars by quote if you're an institution, a rental if you're anyone else, and a desktop version if you want the face without the invoice. The honest summary for everyone else: the conversations are real AI more often than skeptics think, the emotions are better choreographed than fans think, and the robot that keeps promising to walk has built a very good business standing still. Compare on RoboZaps: Ameca and Sophia. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Sanctuary AI Phoenix 2026: Price, Is It For Sale & What It Is URL: https://blog.robozaps.com/b/sanctuary-ai-phoenix-review Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:05:52.687Z Last fact-checked: 2026-07-27T00:00:00.000Z Category: reviews TL;DR: You cannot buy a Sanctuary AI Phoenix: it has no official price and was built as a research platform, not a product. In June 2026 Sanctuary pivoted to selling its Physical AI software for other robots, with its hydraulic hands still in development. Our best price estimate is a rough $50,000 to $100,000, but every figure is unofficial. Key takeaways: - Not for sale, no price: Phoenix has never had an MSRP and there is no way to order one. Sanctuary built it as a research and data-capture platform, not a product. - In June 2026 Sanctuary pivoted: it now deploys its Physical AI software on other companies' industrial robots, and lists its hydraulic hands as a separate product still in development. - No official price exists, so any figure is an estimate: our best estimate is roughly $50,000 to $100,000. Other numbers online ($40,000, or $150,000 to $300,000) are unofficial and trace to aggregators, not Sanctuary. - The hands are the real story: the Phoenix hand has 21 degrees of freedom, hydraulic valves tested past two billion cycles, and fingertip tactile sensing close to human sensitivity. - The current Phoenix (Generation 8, December 2024) rides a wheeled base, not legs; Sanctuary said bipedal legs were 'too frail.' Only its 2023 Gen-6 dimensional specs were ever officially published. - It is Canadian (Vancouver) and backed partly by government money, but it survived 2024 to 2025 through layoffs, a CEO ouster and a proposed asset sale before its 2026 reinvention. FAQ: Q: How much does the Sanctuary AI Phoenix cost? A: It has no official price. Sanctuary never published an MSRP or offered Phoenix for sale; it is a research and data-capture platform. Our best estimate is roughly $50,000 to $100,000, in line with other premium research humanoids, but that is unofficial, and other figures online ($40,000, or $150,000 and up) trace to no Sanctuary source. Q: Can you buy a Sanctuary AI Phoenix? A: No. There is no store, lease or waitlist. Businesses only ever ran short pilots in which human pilots teleoperated the robot to capture data. Since June 2026 Sanctuary's product is software for existing robots, not the whole humanoid, so even that route is gone. Q: What happened to Sanctuary AI? A: After co-founder CTO Suzanne Gildert left in April 2024 and the board ousted co-founder CEO Geordie Rose with about 30 layoffs in November 2024, the company raised bridge financing and moved to sell its Apptronik stake to stay afloat, then in June 2026 pivoted to selling Physical AI software for existing robots and named a new CEO, Daniel Friedmann. Q: Is the Sanctuary Phoenix still being made? A: Not as a product for sale. As of June 2026 Sanctuary deploys its AI software on other companies' industrial robots and lists its hydraulic hands as a separate product still in development; whole humanoids are described only as a long-term goal. The last hardware generation was Generation 8 (December 2024); there is no publicly announced Generation 9. Q: What is the Sanctuary AI Phoenix known for? A: Its hands. The Phoenix hydraulic hand has 21 degrees of freedom and fingertip tactile sensing near human sensitivity, and is widely considered among the best in the industry. Sanctuary is now developing a separate 17-DoF industrial hand as a standalone product. Q: Does the Sanctuary Phoenix walk? A: No. The current Generation 8 uses a wheeled base. Sanctuary said bipedal legs were 'too frail to support a strong torso.' Early marketing referenced 'humanoid legs,' but even the 2023 unveil ran on a wheeled base. Q: Where is Sanctuary AI based? A: Vancouver, British Columbia, Canada. It was founded in 2018 and backed partly by Canadian public funding, including a C$30 million federal Strategic Innovation Fund contribution in 2022. Q: How many degrees of freedom does the Phoenix have? A: Each Phoenix hand has 21 degrees of freedom. The whole-body total has never been officially disclosed. Be wary of the '20 DoF' figure online; that is a stale 2023 number for both hands combined, a different measurement. Q: Was the Phoenix ever deployed in the real world? A: In limited, mostly teleoperated pilots. A fifth-generation robot did 110 retail tasks at a Mark's store in January 2023, and Sanctuary signed a Magna partnership in 2024 with no results disclosed. Its one hard 2026 metric, a 99.5% wire-plugging result, was software on an industrial arm, not a Phoenix humanoid. Q: Is there a Sanctuary AI or Phoenix crypto token? A: No official one. Two unrelated meme coins ('Sanctuary Token' and 'Phoenix Token') share the name but have nothing to do with the company. Sanctuary has no token, and pre-IPO share offers are for equity, not robots. You cannot buy a Sanctuary AI Phoenix, and there has never been a way to. It has no price and no MSRP, because Sanctuary never offered it for sale: the Vancouver company built Phoenix as a research and data-capture platform, not something you could order. As of June 2026 the point is close to moot anyway. Sanctuary stopped positioning the whole humanoid as its product and pivoted to deploying its "Physical AI" software on other companies' industrial robots, with its dexterous hands following as a separate product it lists as still in development. Every dollar figure attached to "Sanctuary Phoenix price" is an unofficial estimate; the company has never published one. What follows is what Phoenix actually is, what genuinely made it special, what happened to the company that built it, and why the numbers in circulation are unverified. If you are shopping for a humanoid you can deploy, this is not one. If you want to understand one of the more technically interesting and turbulent stories in the humanoid race, read on. Can you buy a Sanctuary AI Phoenix? No. There is no store, no checkout, no lease rate and no waitlist that leads anywhere real. Sanctuary's own site routes every enquiry to a "commercial inquiries" form, and it always has. The company ran Phoenix in a handful of pilots, but no business ever bought one; they hosted short trials while human pilots teleoperated it to capture training data. Quick facts, and how much of the "spec sheet" online is real Here is the load-bearing detail, with a warning attached. Sanctuary published dimensional specs exactly once, at the Generation 6 unveil in May 2023. From Generation 7 onward it stopped giving numbers and spoke in deltas: "wider range of motion," "lower bill of materials." Almost every "Phoenix spec sheet" online simply recycles the 2023 figures and presents them as current. Treat anything below that is not marked current with suspicion. Two things nearly every other page gets wrong. First, the Phoenix hands have 21 degrees of freedom each, not the "20 DoF" you will see quoted; that stale number is a 2023 figure describing 20 across both hands combined, a different measurement. Second, the current Phoenix does not walk. Generation 8 rides a wheeled base, and Sanctuary itself said bipedal legs were "too frail to support a strong torso." Where a figure was never disclosed, we say so rather than borrow a number from a content farm. What a Sanctuary Phoenix would cost There is no official price, so every figure is an estimate. Our best estimate is $50,000 to $100,000, in line with other premium research humanoids, though even that is unofficial because Sanctuary has never published a number. Two wilder figures also circulate online, and neither traces to Sanctuary: $40,000 sits on an aggregator's spec sheet that labels the robot a "prototype" and puts an "inquire about price" button beside it. No source.$150,000 to $300,000 comes from a "humanoid robots for sale" listicle that tags it "pilot restricted." Also unsourced. Worse than the unverified numbers is the fake commerce. A few marketplace pages show Phoenix with "in stock," "add to cart" and quote controls, for a robot that has never been sold or leased to anyone. Two other traps share the name: unrelated meme cryptocurrencies ("Sanctuary Token" and "Phoenix Token") that have nothing to do with the company, and pre-IPO share offers for Sanctuary equity. Buying Sanctuary stock is not the same as buying a Phoenix, and only one of those is even possible. What happened to Sanctuary AI: the 2026 pivot The reason there is no Phoenix to buy in 2026 is not just philosophy. It is survival. Sanctuary spent 2024 and 2025 in visible distress before reinventing what it sells. The unravelling was quick. Co-founder and CTO Suzanne Gildert left in April 2024 to work on AI consciousness, later founding a company called Nirvanic. That November, the board pushed out co-founder and CEO Geordie Rose in what former staff called a "sudden and unexpected" move, and roughly 30 employees were laid off, reportedly over doubts about the company's ability to keep raising money against far better-funded American rivals. In January 2025 it floated a US$10 million convertible note just to fund the fiscal year. By April 2025, The Logic reported Sanctuary was trying to raise US$175 million, largely by selling its stake in Apptronik, a rival humanoid maker it had controlled after buying in for US$10 million in 2022 and which had suddenly become worth an order of magnitude more once Google and others poured US$350 million into it. Whether that sale and raise closed as described was never publicly confirmed. Then came the reset. In June 2026 Sanctuary named Daniel Friedmann, a two-decade chief executive of the space firm MDA and then of Carbon Engineering, as its new CEO, with a mandate to scale what worked rather than keep proving it out. Days earlier, on June 17 2026, it announced the pivot itself. Rather than waiting, in its own words, "for humanoid hardware to reach mass commercialization," Sanctuary would deploy its Physical AI software on the industrial robot arms companies already own. The whole humanoid became a stated long-term goal, "a path to humanoids," not the thing it sells today. Who makes Phoenix, and how it was funded Sanctuary Cognitive Systems, trading as Sanctuary AI, was founded in Vancouver in 2018 by Geordie Rose, a serial deep-tech founder behind the quantum-computing firm D-Wave and the AI company Kindred, along with Suzanne Gildert and Olivia Norton. By mid-2024 it had raised more than C$140 million (about US$99 million at the time), with later estimates putting the total higher. The backers are a revealing mix of strategic corporates and Canadian public money: Bell led a C$75.5 million Series A in 2022; Magna, Verizon Ventures, Workday Ventures, SE Health, Accenture and Evok Innovations invested; the Government of Canada added a C$30 million Strategic Innovation Fund contribution in late 2022, among the largest it had given an AI company; and British Columbia's InBC, alongside the federal BDC Capital and Export Development Canada, put in strategic financing in 2024. That government backing is part of why the company's survival became a minor Canadian tech story. The hands: what actually made Phoenix special Strip away the funding drama and one thing about Phoenix is genuinely differentiated: its hands. Where Tesla and Figure drive their robotic fingers with electric motors and tendons, Sanctuary built hydraulic hands around proprietary, coin-sized valves. Each Phoenix hand has 21 degrees of freedom, approaching the human hand's 27, and the valves were tested past two billion cycles without leaking. The system runs at low pressure on food-safe oil, which Sanctuary argues gives it an order-of-magnitude higher power density than cable-driven rivals, along with better speed, strength and impact tolerance. The touch is as notable as the grip. In February 2025 Sanctuary added a seven-cell tactile array to each fingerpad, built on micro-barometers, the same pressure-sensor family found in smartphones, sensitive to about five millinewtons against a human finger's roughly three. It can read the direction of force and detect slip. Two months earlier the company had shown the hands performing zero-shot in-hand manipulation, reorienting a held object without setting it down and regrasping. This is the technology Sanctuary is now trying to sell on its own: its current roadmap lists a standalone "17-DoF" industrial hand as in development and "entering pilot programs soon," which tells you Sanctuary itself believes the hands, not the humanoid, are the crown jewel. Carbon, and the teleoperation reality Phoenix runs on Carbon, Sanctuary's control system, which the company describes as a brain-like architecture blending learned behaviour with symbolic reasoning, trained and run on Microsoft Azure. The honest description of how Phoenix worked in the field, though, is teleoperation. A human pilot wearing a motion-capture rig sees through the robot's cameras and feels through its tactile sensors, and drives it through a task; Carbon watches and turns that demonstration into training data, with autonomy "added in layers over time." Sanctuary's headline 2024 claim was that this pipeline cut the time to teach a new task from weeks to under 24 hours. IEEE Spectrum, reviewing the early pilots, was blunt that Phoenix was not autonomous and "has a ways to go." Keep that in mind reading any deployment claim: the work was substantially piloted, by design. The deployment record Sanctuary's track record is thinner, and more teleoperated, than the headlines suggest, and it is worth stating precisely. The nuances matter. The often-cited "110 tasks at Canadian Tire" is real, but it was the fifth-generation robot in January 2023, before Phoenix was named. Microsoft, NVIDIA, Accenture and Verizon appear throughout Sanctuary's coverage, but as cloud, simulation, channel and venture partners, not as customers deploying robots. And the only hard production metric from 2026, the wire-plugging result, was software on someone else's arm. No primary source confirms a whole Phoenix humanoid running at a customer site today. How it compares Against Tesla's Optimus, Figure's robots or Unitree's machines, Phoenix always competed on manipulation, not mobility or price. Its hands were among the best in the field; its walking, cost and availability were the worst, because it was never trying to be a shipped product. For a direct matchup, see our Tesla Optimus versus Sanctuary AI Phoenix comparison. And if you are actually trying to buy a humanoid, the honest answer is that almost none are for sale: a Unitree G1 is one of the few you can price and order, while Tesla's Optimus and Figure's robots have no public order page either. Who Phoenix is, and isn't, for After the pivot, Phoenix as a whole robot is a purchase for essentially no one. What Sanctuary now offers speaks to two audiences: industrial manufacturers who want its Physical AI software on the robot arms they already run (with its hydraulic hands to follow), and researchers or investors tracking dexterous manipulation. If you came here to buy a humanoid, you cannot, and the honest redirect is to a robot with a real price. If you came to understand the state of robotic hands, Phoenix's technology is still among the most interesting anywhere. What to watch The open question is whether the narrower company works. Watch whether Friedmann's software-first model signs paying industrial customers beyond the single unnamed supplier in the June 2026 demo; whether the standalone hands actually ship from their "in development" status; whether the reported Apptronik-stake sale and 2025 fundraising closed; and whether Sanctuary ever returns to a whole humanoid, which its own "path to humanoids" language deliberately leaves open. For now the robot is not for sale, and the company is betting its future on the software and hands instead. The bottom line Phoenix is one of the most technically impressive humanoids almost no one will ever touch. It was never for sale, it has no price, and in June 2026 its maker stopped positioning the whole humanoid as its product to focus on what it does best: the AI to drive dexterous manipulation, with its hydraulic hands following as a separate product. Ignore every dollar figure and "in stock" button you find, because none is official. Judge Sanctuary on the hands, which are widely regarded as among the best anywhere, and on whether a leaner, refocused company can finally turn a decade of remarkable engineering into a business. Compare on RoboZaps: Phoenix, Tesla Optimus and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robot ROI: Calculate Payback for Your Workflow URL: https://blog.robozaps.com/b/roi-of-humanoid-robots Author: Radica Maneva Published: 2026-01-31T00:00:00.000Z Updated: 2026-08-01T23:06:06.771Z Last fact-checked: 2026-07-21T00:00:00.000Z Category: insights TL;DR: Humanoid robot ROI has no defensible universal range. It depends on a task-ready quote, measured workflow hours, reliable task coverage, human intervention, ongoing support and whether covered work produces a real cash benefit. Use the calculator with your own inputs, then test downside, base and upside cases. Key takeaways: - Treat published deployment milestones as evidence of task feasibility, not disclosed customer ROI. - Build the model around one workflow and compare the humanoid with simpler automation and process changes. - Count only labour savings and other benefits that the business can realistically realise and measure. - Run downside, base and upside cases, and allow the model to return no payback. FAQ: Q: What is the ROI of a humanoid robot? A: There is no reliable category-wide figure. ROI depends on the task-ready acquisition or service cost, integration, recurring costs, reliable task coverage, human intervention and the share of covered work that creates a measurable financial benefit. Q: How do you calculate humanoid robot ROI? A: For a fixed ownership period, subtract total cost of ownership from total measurable benefits, then divide the net value by total cost of ownership. Keep payback separate: it is the time required for cumulative net benefit to recover the initial cash outlay. Q: How long does a humanoid robot take to pay back? A: The public deployments discussed here do not disclose enough customer economics to establish a general payback range. Use a buyer quote and measured workflow data, then test downside, base and upside cases against the business's own approval threshold. Q: Are humanoid robots cheaper than human workers? A: Not automatically. Compare the cost per safely completed unit of useful work, including integration, downtime, intervention, supervision, software and support. Covered task time creates savings only when the business can redeploy it productively, avoid a hire or reduce an actual expense. Q: Is buying better than Robot-as-a-Service? A: It depends on the quote and contract. Compare the complete payment schedule, integration responsibility, uptime and support commitments, usage limits, software rights, insurance, cancellation and end-of-term obligations. A lower upfront payment is not the same as a lower total cost. Q: What should a humanoid robot pilot measure? A: Measure safely completed useful work, availability, failed cycles, human intervention, supervision, changeovers, charging, quality, maintenance and recovery time. Collect the results across representative shifts and ordinary failure conditions, not only a best demonstration run. Humanoid robot ROI is not a property of the robot. It is the result of a specific machine, task, site and commercial agreement. A list price or factory demonstration cannot tell you when a deployment will pay back. A useful model starts with the buyer's quote and the current workflow. It then subtracts integration, support, downtime and retained human work from the value the robot can actually produce. If those inputs are missing, the result is an illustrative scenario, not a case study. The calculator below starts blank for that reason. Use your own quote and measured workflow assumptions. For current pricing and the costs commonly excluded from a list price, see the humanoid robot cost guide. Calculate your humanoid robot ROI The result is an estimate, not a vendor guarantee. The calculator uses a simple ownership-period model and assumes the annual benefit you enter remains stable. If a deployment will ramp gradually, use a conservative first-year task-coverage figure or build a separate monthly cash-flow schedule before approving the project. A defensible ROI model starts with one workflow Define the unit of useful work before entering a price. For tote transfer, that might be totes delivered to a conveyor without assistance. For machine tending, it might be accepted parts loaded per hour. Robot runtime, battery life and distance walked are operating measures; they are not financial benefits by themselves. The calculator follows three core relationships: Annual realised labour value = loaded hourly labour cost × manual workflow hours × operating days × reliable task coverage.Annual net benefit = realised labour value + other measurable benefit − support, software, insurance and internal operating costs.Ownership-period ROI = (total measurable benefits − total cost of ownership) ÷ total cost of ownership × 100. Enter task coverage after allowing for availability, failed cycles and human intervention. Put retained supervision, teleoperation, batteries, spares and site support in annual internal operations. Count labour value only when the covered hours can be redeployed productively, avoid a hire or reduce an actual expense. What current deployments prove, and what they do not Public deployment reports are useful for defining feasible tasks and pilot metrics. They rarely disclose enough commercial information to calculate customer ROI. BMW and Figure 02: production evidence, not disclosed payback BMW reports that Figure 02 worked ten-hour shifts, five days a week, during a ten-month project at Plant Spartanburg. It moved more than 90,000 sheet-metal components in about 1,250 operating hours and supported production of more than 30,000 BMW X3 vehicles. BMW also describes revised safety controls and improved 5G coverage at the site. Those figures show that a defined loading task can run repeatedly in a production environment. BMW's public account does not give the robot price, integration spend, intervention hours, staffing effect or payback period. Those values must remain buyer inputs. GXO and Digit: a commercial service with undisclosed economics The Agility Robotics and GXO announcement describes a multi-year Robot-as-a-Service deployment in which Digit moves totes from autonomous mobile robots to conveyors. Agility later reported that the deployment had moved more than 100,000 totes. This is evidence of paid commercial operation and cumulative task volume. The public releases do not disclose the service fee, total scheduled hours, exception rate, human support time, staffing change or customer payback. A buyer cannot derive those numbers from tote count alone. Mercedes-Benz and Apollo: a pilot scope, not an ROI result Apptronik's announcement says Mercedes-Benz planned to pilot Apollo for delivery of assembly kits, component inspection and movement of totes. It does not publish completed runtime, throughput, deployment cost or financial return. The useful evidence is the proposed workflow, not a payback claim. Three workflow-specific models Each model below is a diligence template. It deliberately avoids a default robot price or payback range. Warehouse tote transfer Baseline: person-hours spent on the exact tote handoff, totes per hour, waiting time, travel time and current exception handling.Pilot evidence: safely completed totes per hour, intervention per 100 totes, availability, charging time and AMR or conveyor handoff failures.Value to model: labour hours that can actually be redeployed or avoided, plus any measured reduction in backlog or overtime.Alternative to compare: conveyor changes, an AMR workflow or a simpler fixed transfer device. Line-side material delivery and kitting Baseline: kit preparation and delivery time, route distance, missed calls, line-side waiting and documented line stoppage.Pilot evidence: on-time deliveries, correct-item rate, handoff success, route availability, human escort time and exception recovery.Value to model: realised labour reduction and downtime reduction that the operations and finance teams can trace to the workflow.Alternative to compare: tugger routes, AMRs, supermarkets or changes to line-side inventory. Machine tending or parts loading Baseline: accepted parts per hour, operator touch time, changeovers, scrap, faults and idle machine time.Pilot evidence: accepted autonomous cycles after resets and rejected parts, intervention time, changeover time and safe operation under the intended controls.Value to model: an approved staffing change, additional accepted output or measurable reduction in scrap and idle time.Alternative to compare: a cobot, fixed robot cell or process redesign. The cobot versus industrial robot guide covers that decision boundary. Run downside, base and upside cases A single result hides the assumptions carrying the business case. Run three versions and save the inputs with the result. Downside: lower task coverage, more intervention, slower repair and a longer ramp than the pilot average.Base: current signed quotes and the median result from a representative pilot.Upside: only improvements supported by contract terms, tested process changes or measured learning. Do not use the best demonstration cycle as the upside case. Include variation across shifts, operators, product mix and ordinary failure conditions. A model that does not survive the downside case may still justify a limited pilot, but it does not justify a fleet. When the model should say no A credible calculator must allow no payback. Stop or narrow the project when: there is no task-ready quote or measurable workflow baseline;the robot cannot meet the required cycle, quality or safety conditions without frequent help;supervision and exception handling consume the hours supposedly saved;a conveyor, AMR, cobot, fixed cell or process change solves the task for less;the benefit depends on a staffing change that the business does not plan to make; orservice response, software rights, cancellation terms and downtime responsibility remain unclear. How to use the calculator without overstating the result Enter the hardware or service quote, unit count, setup scope, ownership period and contingency.Add annual support, software, insurance and internal operations. Include the people and equipment needed to keep the workflow running.Enter the loaded labour cost and manual hours for this workflow, not total workforce hours. Set task coverage net of downtime and intervention.Add other benefit only when there is a documented baseline and an accountable owner for the number.Review ROI, payback, total cost and net value together. Save the assumptions with the result so another reviewer can reproduce it. For a multi-robot case, enter the combined workflow value once. Do not multiply the same labour hours for every unit. The calculator estimates simple ROI and payback; it does not calculate tax effects, financing terms, net present value or internal rate of return. Return to the calculator and replace every illustrative input with a quote, measured baseline or explicit assumption. Bottom line There is no defensible universal payback period for a humanoid robot. BMW, GXO and Mercedes-Benz show that narrow industrial workflows are being tested or operated in real facilities. Their public reports do not disclose enough customer economics to support generic claims of six, eighteen or twenty-four months. Build the decision around one task. Use a task-ready quote, measured workflow hours, pilot intervention data and recurring-cost terms. If the model still works under a reasonable downside case, the next step is a controlled pilot with written acceptance criteria. For buyer-side help testing an ROI case before capital is committed, see our robotics advisory service. Compare on RoboZaps: Figure 02 and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in Elderly Care: 10 Robots Transforming Senior Living in 2026 URL: https://blog.robozaps.com/b/humanoid-robots-in-elderly-care Author: Radica Maneva Published: 2026-01-30T00:00:00.000Z Updated: 2026-08-01T23:04:44.950Z Category: insights TL;DR: Eldercare robots are deployed today across Japan, Korea, China, the US and Europe — from Paro and ElliQ companionship to the Fourier GR-3 care-bot — in a $3.14 billion market growing 12.5% a year. Clinical studies show reduced anxiety and better engagement; costs of $6,000-$100,000+ and privacy remain the obstacles. Key takeaways: - Humanoid robots are actively deployed in elderly care facilities across Japan, South Korea, China, the US, and Europe — not a future concept but a present reality. - The eldercare robot market is valued at $3.14 billion (2025) and growing at 12.5% CAGR, driven by a global aging crisis and caregiver shortages. - Real robots like Paro, Pepper, NAO, Hyodol, ElliQ, Ryan, Fourier GR-3, and 1X NEO each serve distinct functions — from companionship to physical assistance to health monitoring. - Clinical studies show measurable benefits: reduced anxiety, improved mood, better social engagement, and relief for overburdened care staff. - China launched a national pilot program in 2025 requiring at least 200 robots deployed across 200 families for minimum 6-month trials. - Key challenges remain: high costs ($6,000–$100,000+), privacy concerns with AI-powered data collection, and the ethical tension between robotic companionship and human connection. - March 2026: Andromeda Robotics raised $17M to launch its Abi robot for US senior care; 1X NEO is open for pre-order at $20,000; China opened the world's first smart wellness robot elderly care station. FAQ: Q: What are the best humanoid robots for elderly care in 2026? A: The most effective robots currently deployed in elderly care include Paro (a therapeutic seal robot used globally since 2003), Hyodol (a ChatGPT-powered companion doll with 12,000+ deployments in South Korea), ElliQ (a proactive AI companion used in New York), and Pepper/NAO (humanoid robots used in clinical research, though the manufacturer filed for bankruptcy in 2025). New entrants like Fourier's GR-3 Care-bot and 1X NEO are designed specifically for eldercare and home assistance. Q: How much do elderly care robots cost? A: Costs vary dramatically by robot type. Paro costs approximately $6,000. ElliQ is around $250 plus $30/month subscription. Pepper robots were $20,000–$25,000 before Aldebaran's bankruptcy. Full-size humanoid robots like Fourier GR-3 and 1X NEO have not yet announced consumer pricing but are expected to cost significantly more. Government subsidies in South Korea and China have made companion robots available to elderly citizens at no personal cost. Q: Can robots actually reduce loneliness in elderly people? A: Clinical evidence says yes, with caveats. A review of 29 studies of Paro found consistent improvements in mood, social engagement, and reduced negative behaviors. Studies of Pepper and NAO in Minnesota nursing homes showed residents felt happier and more cared for. However, participants in a Ryan robot study noted the experience was "not the same as talking to a real person." Robots are most effective as supplements to — not replacements for — human social interaction. Q: Are elderly care robots safe to use? A: Companion robots like Paro, Hyodol, and ElliQ are designed with safety as a primary concern — they are lightweight, have no sharp edges, and do not move autonomously through the environment. Full-size humanoid robots like GR-3 incorporate extensive safety systems including compliant actuators and force-sensing. The primary safety concerns are around data privacy (what personal information is collected and how it's used) rather than physical harm. Q: What is China's elderly care robot pilot program? A: In June 2025, China launched a national pilot program requiring companies and research institutes to deploy at least 200 robots to 200 families for trial periods of six months or more. The program addresses "the full life-cycle needs of elder adults, including daily care, rehabilitation, psychological support and emotional companionship." Major Chinese robotics firms including Unitree, UBTech, Fourier, and AgiBot are participating. Q: Will humanoid robots replace human caregivers? A: No — and that's not their intended purpose. Every researcher and developer interviewed consistently positions robots as supplements to human care, not replacements. Robots handle routine tasks (medication reminders, basic monitoring, companionship during off-hours) so human caregivers can focus on complex, empathetic care. The fundamental problem is that there aren't enough human caregivers to meet demand, and robots help bridge that gap. ‍ Related: Humanoid Robots in Healthcare: How They Will Revolutionize The Industry · Applications of Humanoid Robots Ready to buy? Browse humanoid robots for sale on Robozaps. Last updated: March 31, 2026 This guide is part of our series on the applications of humanoid robots across 12 industries. Humanoid robots are transforming elderly care right now. Robots like Paro, Hyodol, ElliQ, and Fourier GR-3 are actively deployed in nursing homes and private residences across Japan, South Korea, China, the US, and Europe — providing companionship, medication reminders, health monitoring, and physical assistance. The eldercare robot market is valued at $3.56 billion in 2026, growing at 12.5% CAGR. In March 2026, Andromeda Robotics raised $17M to launch its Abi robot for senior care in the US, and CGTN featured the world's first smart wellness robot elderly care station. From Japan's therapeutic seal robot Paro to South Korea's ChatGPT-powered Hyodol companion dolls and China's "Yang Yang" care robots, these machines are addressing one of the most urgent challenges of our time: how to provide quality care for a rapidly aging population. The eldercare assistive robot market reached $3.14 billion in 2025 and is projected to hit $3.56 billion in 2026, growing to over $10 billion by 2035 at a 12.5% CAGR. With the global population aged 65+ surpassing 1 billion in 2023 and expected to reach 1.5 billion by 2050, the demand for robotic care solutions is accelerating faster than ever. This guide covers every major aspect of humanoid robots in elderly care — the specific robots being deployed today, real-world results from clinical studies, the latest 2025–2026 developments, and the ethical questions that still need answering. Why Do Elderly Care Facilities Need Humanoid Robots? The numbers paint a stark picture. The global population aged 65 and older reached approximately 1 billion in 2023, with projections suggesting it will grow to 1.5 billion by 2050. In many developed nations, care worker shortages are already critical. South Korea became a "super-aged society" in 2024, with more than 20% of its population over 65. Elderly suicide rates there are the highest among all OECD nations, driven largely by isolation and loneliness. Japan — the world's oldest country by median age — has pioneered robotic care solutions for over two decades. China's elderly population is growing so fast that the government launched a national elderly-care robot pilot program in June 2025, requiring companies to deploy at least 200 robots to 200 families for trial periods of six months or more. Meanwhile, care worker shortages plague Western nations. The United States faces a projected shortfall of hundreds of thousands of home health aides, while Germany and the UK report similar gaps. The math is simple: there are not enough human caregivers for the number of elderly people who need care. This is where humanoid robots step in — not to replace human caregivers, but to fill critical gaps in a system that is already stretched beyond capacity. What Types of Robots Are Used in Elderly Care? Not all care robots are alike. They range from plush companion dolls to full-size humanoid machines, each designed for specific needs. Here are the three main categories: What Are Social and Companion Robots? Social robots focus on emotional well-being — combating loneliness, stimulating conversation, and providing a sense of presence. They are the most widely deployed category in elderly care today. Paro — the robotic baby harp seal developed in Japan — remains the gold standard in this category. Designed by AIST researcher Takanori Shibata, Paro has been used in care facilities worldwide since 2003. A 2019 review by researcher Lillian Hung at the University of British Columbia analyzed 29 studies and found three consistent benefits: reduced negative emotions and behaviors, better social engagement, and improved mood. In one particularly striking case at Vancouver General Hospital, a dementia patient who was hitting staff and kicking lab technicians became calm after Paro was placed in his lap. He began petting the robot and talking to it, allowing medical staff to perform necessary tests. "The patient had quality care and safety, and the staff were able to get their work done," Hung reported. Hyodol is a newer entrant from South Korea — a ChatGPT-powered doll-like robot deployed to over 12,000 elderly people living alone across the country. It uses conversational AI to chat with seniors, reminds them to take medication, and has sensors that alert social workers during emergencies. The Guro district of Seoul alone has distributed 412 units since 2019. Hyodol is preparing for a US launch in 2026, adapting its chatbot for English, Chinese, and Japanese. ElliQ, developed by Israeli company Intuition Robotics, is deployed in apartments across New York City. Resembling a small Pixar lamp, it engages seniors in conversations about everything from daily activities to the meaning of life, and proactively initiates check-ins to combat isolation. Ryan, built by Mohammad Mahoor at the University of Denver, is a humanoid companion specifically designed for people with early-stage dementia or depression. In a study where six older adults had around-the-clock access to Ryan for 4–6 weeks, participants reported enjoying conversations and feeling happier, though they noted it was not the same as talking to a real person. What Are Service and Physical Assistance Robots? Service robots help with the physical aspects of daily life — mobility support, household tasks, logistics, and rehabilitation. Fourier GR-3 represents the cutting edge of this category. Unveiled in August 2025 as the company's first full-size "Care-bot," GR-3 stands 165 cm tall, weighs 71 kg, and has 55 degrees of freedom. What sets it apart is its Full-Perception Multimodal Interaction System, which fuses vision, audio, and tactile feedback into a real-time emotional processing engine. With 31 pressure sensors across its body, it can detect touch and respond with lifelike gestures. It uses a dual-path architecture: "fast thinking" for reflexive responses and "slow thinking" powered by a large language model for complex conversations. Fourier is targeting eldercare, rehabilitation, and service environments. 1X NEO (from 1X, formerly Halodi Robotics) is a general-purpose humanoid designed to operate in home environments. The company raised $100 million in 2025 specifically to develop robots for elder care and assistive tasks. NEO is built for everyday tasks in unstructured settings rather than factory floors. Robotic exoskeletons from companies like Ekso Bionics and ReWalk assist elderly individuals with walking and rehabilitation. These wearable devices reduce fall risk and help maintain mobility, which is critical for independent living. TUG robots handle hospital logistics — transporting supplies, medications, and meals — freeing nursing staff to focus on patient care. What Are Medical Monitoring and Health Robots? Medical assistance robots integrate health monitoring with daily care routines. They can track vital signs (heart rate, blood pressure, blood oxygen), provide medication reminders, detect falls, and transmit health data to physicians or family members in real time. Pepper (originally by SoftBank Robotics, now owned by United Robotics Group after SoftBank sold Aldebaran in 2022) has been extensively studied in clinical settings. Researcher Arshia Khan at the University of Minnesota placed Pepper and NAO robots in eight nursing homes in Minnesota. Compared with facilities without robots, residents who interacted with them felt happier, more cared for, and less tired and frustrated. Note: Aldebaran, the manufacturer of Pepper and NAO, filed for bankruptcy in February 2025. This development has raised questions about the future support and availability of these widely-studied robots, and highlights the business viability challenges in the care robotics space. Comparison: Major Robots Used in Elderly Care (2026) Real-World Deployments and Clinical Evidence The evidence base for robots in elderly care has grown substantially. Here are the most significant real-world programs and studies: South Korea: Hyodol National Rollout South Korea's approach is arguably the most ambitious. Facing a demographic crisis (the world's lowest birth rate combined with rapid aging), the government has subsidized Hyodol robot deployments through municipal welfare centers. Over 12,000 units are now in the homes of elderly people living alone. Care workers in Seoul's Guro district describe the robots as "eyes and ears on the ground," alerting them to emergencies and tracking whether seniors are eating and taking medication. The emotional impact has been profound. One elderly user told reporters: "I was going to die, but not anymore. Why would I die in such a wonderful world!" — attributing her renewed outlook to her Hyodol companion. China: National Pilot Program (2025) In June 2025, China's government launched a formal national elderly-care robot pilot program. The initiative requires companies and research institutes to conduct trials of at least six months, deploying a minimum of 200 robots to 200 families. For community and nursing home tests, similar scale requirements apply. Companies like Unitree Robotics, UBTech, Fourier, and AgiBot are all participating. In Chengdu's Pacific Care Home, a humanoid robot named "Yang Yang" already wakes residents each morning, provides weather updates, and reminds them of daily activities. The Chinese government's stated goal is to address "the full life-cycle needs of elder adults, including daily care, rehabilitation, psychological support and emotional companionship." United States: Minnesota Nursing Home Study Arshia Khan's study at the University of Minnesota placed Pepper and NAO robots in eight nursing homes. The results were clear: compared with control facilities, residents interacting with robots felt happier, more cared for, and less frustrated. The robots led group activities including bingo, trivia, and guided conversations. Canada: Vancouver General Hospital Lillian Hung's research at UBC demonstrated Paro's effectiveness with dementia patients in acute care settings. Beyond the individual calming cases, her 2019 review of 29 studies confirmed consistent benefits across three domains: reduced agitation, improved social engagement, and better care experiences. The CARESSES Trial The CARESSES (Culture-Aware Robots and Environmental Sensor Systems for Elderly Support) randomized controlled trial tested culturally competent Pepper robots in care homes. This landmark study explored whether robots that adapt to cultural backgrounds can improve outcomes — a critical factor as care robots deploy globally across diverse populations. How Do Humanoid Robots Benefit Elderly Care? Combating Loneliness and Isolation Loneliness is not just an emotional issue — it's a health crisis. Research links chronic loneliness to increased risks of dementia, heart disease, stroke, and premature death. For elderly people living alone, the absence of daily social contact can be devastating. Social robots address this by providing consistent, judgment-free interaction. As Lillian Hung noted: "For an older person who is frail and struggles with language, the robot doesn't judge. It offers an unconditional presence. Regardless of what they say, it is always happy to listen." While robots cannot fully replace human connection, they fill critical gaps — especially during nights, weekends, and between caregiver visits. Reducing Caregiver Burnout Care workers face extraordinary physical and emotional demands. Staff shortages mean longer shifts, higher patient-to-caregiver ratios, and burnout. Robots can handle routine tasks — medication reminders, activity leadership, basic health monitoring, logistics — freeing human caregivers to focus on complex, empathetic care that requires a human touch. In South Korea, care workers reported that while Hyodol maintenance added to their workload, the psychological benefits for seniors were worth the effort. The robots acted as a force multiplier rather than a replacement. Supporting Independent Living Most seniors prefer to age at home rather than move to institutional care. Robots that can monitor health, remind about medications, detect falls, and facilitate communication with family members make independent living safer and more sustainable. The 1X NEO robot, for example, is specifically designed for home environments and everyday tasks in unstructured settings. Health Monitoring and Emergency Response Medical assistance robots provide continuous monitoring that human caregivers cannot. They can track vital signs 24/7, detect anomalies, and alert medical professionals or family members immediately. This is particularly valuable for managing chronic conditions like diabetes, hypertension, and heart disease — conditions that affect the majority of elderly adults. What Are the Challenges Facing Robot-Assisted Elderly Care? Cost and Accessibility The cost barrier remains significant. Paro costs approximately $6,000 per unit. Pepper robots were $20,000–$25,000 before the manufacturer's bankruptcy. Full-size humanoid robots like Fourier's GR-3 will likely cost significantly more. For nursing homes operating on thin margins and families on fixed incomes, these prices are prohibitive without government subsidies or insurance coverage. Some models are becoming more accessible — ElliQ costs around $250 plus $49.99–$59.99/month — but the most capable robots remain expensive. Government pilot programs in South Korea and China are demonstrating that public funding can bridge this gap. Privacy and Data Security Modern care robots collect vast amounts of personal data: health metrics, daily routines, conversations, facial recognition data, and home environment information. AI-powered chatbots like Hyodol's process conversations through cloud-based systems (ChatGPT), raising questions about where that data goes and who can access it. As ethics researcher Julie Carpenter noted: "We don't know how the data is being triangulated or gathered." For elderly users who may not fully understand AI data practices, informed consent is a serious concern. The Human Connection Debate This is perhaps the most profound ethical question. Gerontologist Clara Berridge at the University of Washington recalls a story about a nursing home resident who died clutching his robot companion. Students were split: some thought it was beautiful he wasn't alone; others found it tragic he died without human connection. "If we're going to invest resources in elder care, I want more staff in the facility so they don't die alone," Berridge said. Her own grandmother died alone in an understaffed nursing home during the COVID-19 pandemic. The research evidence supporting robots' effectiveness is still developing. While individual studies show benefits, large-scale randomized controlled trials are limited. Some researchers caution against rushing to deploy robots when the fundamental issue is understaffing and underfunding of human care. Technical Limitations Despite advances, creating a robot that can safely navigate a home, understand natural language reliably, and physically assist with tasks like bathing or transfers remains technically challenging. Most deployed robots today are either stationary companions (Paro, Hyodol, ElliQ) or require controlled environments. True humanoid assistants that can operate autonomously in a home setting are still in early development phases. Manufacturer Viability The bankruptcy of Aldebaran (maker of Pepper and NAO) in February 2025 highlighted a critical risk: the companies building care robots may not survive commercially. When a robot manufacturer goes under, support, updates, and replacement parts can disappear — leaving care facilities with expensive paperweights. What's Coming for Elderly Care Robots in 2026 and Beyond? Several developments will shape the near-term future of humanoid robots in elderly care: China's national pilot results — The 2025 pilot program will generate the largest structured dataset on elderly care robot effectiveness. Results expected in 2026 will likely influence global policy. Fourier GR-3 commercialization — Following its CES 2026 showcase, Fourier's care-centric humanoid could become the first full-size robot specifically designed and marketed for eldercare at commercial scale. 1X NEO home deployment — With $100 million in funding, 1X is positioning NEO as the first general-purpose humanoid for home use, with elder care as a primary use case. Hyodol's US expansion — The 2026 US launch will test whether a companion robot designed for Korean culture can succeed in Western markets. LLM-powered interaction — The integration of large language models (like ChatGPT) into care robots is dramatically improving conversational ability. Robots are becoming better listeners, more contextually aware, and more engaging in conversation. Government policy expansion — Following South Korea and China's lead, more nations are expected to develop formal policies and funding mechanisms for care robotics. The eldercare robot market is projected to reach $7.7 billion by 2030 and $12.2 billion by 2033. January 2026: CES and New Entrants Several major developments emerged at CES 2026 in January: Mind With Heart Robotics showcased its An'An Panda Cub Robot — a CES 2026 Innovation Award Honoree — designed specifically for loneliness and elderly care. Made with premium Australian wool and featuring 10+ full-body tactile sensors, An'An represents a new wave of biomimetic companion robots that mimic natural animal movements to provide emotional comfort.Japan's Moonshot AIREC Robot — Funded by the Japan Science and Technology Agency and powered by NVIDIA, the AIREC (AI-Driven Robot for Embrace and Care) project is developing robots capable of physical caregiving tasks including changing diapers, bath assistance, and meal support. Professor Tetsuya Ogata of Waseda University noted that generative AI breakthroughs have made what seemed impossible five years ago now "seriously possible."Fourier GR-3 made its formal CES 2026 debut, positioning itself as the most capable full-size humanoid specifically targeting eldercare, rehabilitation, and service environments. DroidUp Moya — Shanghai's DroidUp is building the first humanoid with body-temperature synthetic skin (32–36°C) and micro-expressions across 25 facial degrees of freedom, with healthcare and eldercare among its target uses. At $173,000 and a first batch of roughly 50 units expected in late 2026, Moya will start in care facilities and hospitals rather than private homes. How Do I Choose a Care Robot for an Elderly Family Member? If you're considering a care robot for an elderly loved one, here's what to evaluate: Identify the primary need — Is it companionship (Paro, Hyodol, ElliQ), physical assistance (exoskeletons, future humanoids), or health monitoring (medical robots)?Assess tech comfort level — Simpler robots like Paro require zero technical skill. Chatbot-based robots need basic voice interaction. Full humanoids will require more setup.Check ongoing costs — Beyond purchase price, some robots have subscription fees (ElliQ: $30/month) or may need software updates and maintenance.Evaluate privacy policies — Understand what data the robot collects, where it's stored, and who has access. This is especially important for AI-powered chatbot robots.Consider the care environment — A nursing home may benefit from an interactive humanoid like Pepper for group activities. A solo-living senior may prefer a personal companion like Hyodol or ElliQ.Don't replace human contact — Robots should supplement, not substitute, visits from family, friends, and professional caregivers. Compare on RoboZaps: Fourier GR-3, Moya and Pepper. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in Education: Complete Guide to Classroom Robots [2026] URL: https://blog.robozaps.com/b/humanoid-robots-in-education Author: Radica Maneva Published: 2026-01-30T00:00:00.000Z Category: insights TL;DR: Humanoid robots teach in schools across 40+ countries, with the strongest evidence in STEM, language learning and autism support. Leading classroom robots (NAO, Pepper, QTrobot, Alpha Mini, Furhat) run $2,000-$30,000. Engagement gains are consistent; measured learning gains are small to moderate. Robots supplement teachers, not replace them. Key takeaways: - Humanoid robots are actively deployed in schools across 40+ countries as teaching assistants, peer tutors, and special education support tools. - The strongest evidence for learning outcomes is in STEM/coding, language learning, and autism/special education. - Studies consistently show higher engagement and motivation, though measurable learning gains remain small to moderate. - Leading classroom robots include NAO, Pepper, QTrobot, Alpha Mini, and Furhat—ranging from $2,000 to $30,000. - Key challenges include cost ($9,000–$30,000+ per robot), teacher training, Wi-Fi infrastructure, and data privacy. - Robots supplement but don't replace human teachers—the best outcomes occur in blended human-robot teaching models. FAQ: Q: What humanoid robots are used in schools? A: The most widely deployed humanoid robots in education are NAO (Aldebaran/SoftBank), Pepper (SoftBank Robotics), QTrobot (LuxAI), Alpha Mini (UBTECH), Furhat (Furhat Robotics), and Marty (Robotical). NAO and Pepper are the most common in K–12 classrooms globally. Q: How much do educational robots cost? A: Prices range from $200–$500 for entry-level coding robots (Alpha Mini, Marty) to $9,000–$30,000 for full-featured classroom platforms (NAO, Pepper, Furhat). Total cost of ownership including software, training, and maintenance is typically 40–50% higher than hardware alone. Q: Can robots replace teachers? A: No. All research and real-world deployments show robots work best as supplements to human teaching, not replacements. The best outcomes come from blended models where robots handle repetitive tasks while teachers provide complex instruction, emotional support, and creativity. Q: Are robots effective for children with autism? A: Yes—this is one of the strongest evidence areas. Robots like QTrobot and NAO provide predictable, patient, repeatable interactions that help autistic children practice joint attention, turn-taking, emotion recognition, and social skills. Multiple systematic reviews report positive engagement and behavioral outcomes. Q: What subjects can robots teach? A: Humanoid robots are used to teach programming/coding, mathematics, science, languages (especially second-language acquisition), social skills, and reading. The strongest evidence is in STEM/coding education and language learning. Q: Do students actually learn more with robots? A: Students consistently show higher engagement and motivation with robots. However, direct academic achievement gains are small to moderate. The novelty effect may inflate early results, and more long-term studies are needed. Robots appear most effective when integrated into well-designed pedagogy rather than used as standalone teaching tools. Q: What are the privacy concerns with classroom robots? A: Robots equipped with cameras, microphones, and AI collect student data including facial expressions, voice recordings, and behavioral patterns. Schools must comply with FERPA (US), GDPR (EU), or equivalent privacy laws. Parents should be informed and consent obtained. Schools should establish clear data retention and deletion policies. Q: Which robot is best for coding education? A: For budget-conscious schools: Alpha Mini ($200–$400) or Marty ($300–$500). For comprehensive STEM programs: NAO ($9,000–$16,000) supports Python, C++, and block-based coding with 25 degrees of freedom. NAO is the most widely used robot in academic STEM research. Humanoid robots are transforming education in 2026—from AI-powered teaching assistants to autism therapy companions. This comprehensive guide covers every major humanoid robot used in classrooms today, real deployment data from schools worldwide, costs, pros and cons, and a practical implementation guide for educators. Updated with the latest 2026 research and field studies. This guide is part of our series on the applications of humanoid robots across 12 industries. Humanoid Robots Used in Classrooms: Complete Comparison How Humanoid Robots Are Used in Schools (2026) 1. Teaching Assistants & Co-Teachers The most common deployment model. Robots like NAO and Pepper work alongside human teachers to: Deliver structured lessons: Vocabulary drills, math problems, science demonstrationsAnswer questions: AI-powered responses to student queries in real-timeHandle repetitive tasks: Attendance, quizzes, practice exercises—freeing teachers for complex instructionProvide multilingual support: Pepper supports 20+ languages; NAO supports 9 languages A study at university level found that NAO-based vocabulary lessons produced learning outcomes at least equal to human-only instruction, with students reporting significantly higher enjoyment. 2. Peer Tutors & Learning Companions Smaller robots like Alpha Mini and Tega act as "slightly older peers" who: Practice reading aloud with reluctant readersWork through math problems step-by-stepProvide patient, judgment-free repetition (robots never get frustrated)Encourage hesitant students to participate A systematic review of social robots in classrooms found the peer/companion role is one of the most effective at raising participation and on-task behavior, particularly for students who avoid engaging with human teachers. 3. Special Education & Autism Support This is where the evidence is strongest. Robots provide predictable, repeatable interactions that are especially beneficial for: Autism spectrum: QTrobot is used in clinical and school settings to train joint attention, turn-taking, and emotion recognitionSpeech therapy: NAO helps children practice pronunciation and conversational skillsMotor skills: Physical interaction with robots supports fine and gross motor developmentEmotional regulation: Consistent robot responses help children with emotional difficulties develop coping strategies Multiple systematic reviews report positive trends in engagement and behavioral outcomes for autistic learners, though researchers note the need for larger samples and longer interventions. 4. STEM & Coding Education Humanoid robots make abstract programming concepts tangible: NAO: Programmable in Python, C++, and block-based coding (Choregraphe/Blockly)Alpha Mini: UBTECH's affordable entry point with block coding interfaceMarty: Teaches Python and Scratch through physical walking robot movementsRobotics competitions: Programs like FIRST Robotics and RoboCup Junior use humanoid platforms Students programming physical robots show higher engagement and deeper understanding of computational thinking compared to screen-only coding exercises. 5. Language Learning Robots are increasingly used for second-language acquisition: Pepper delivers English conversation practice in Japanese schools (e.g., Shibuya Elementary)NAO teaches vocabulary and pronunciation across European classroomsFurhat provides immersive conversation practice with projected facial expressionsStudents are less anxious speaking to a robot than to a human teacher or native speaker Real-World Deployments: Schools Using Robots Now The Evidence: What Research Says About Robot-Assisted Learning What Works Engagement: Nearly all studies report significantly higher student engagement and on-task behavior when robots are presentMotivation: Hesitant students are more willing to participate; robot interactions reduce performance anxietySTEM skills: Programming physical robots produces better computational thinking outcomes than screen-only codingSpecial education: Consistent positive results for autism spectrum support, speech therapy, and social skills trainingLanguage learning: Reduced anxiety in speaking practice; comparable or better vocabulary retention What's Still Unclear Long-term learning gains: Most studies are short-term; the "novelty effect" may inflate early resultsAcademic achievement: While engagement improves, direct impact on test scores is small to moderateScalability: Most evidence comes from small pilots (10–50 students), not district-wide deploymentsCost-effectiveness: No large-scale studies comparing robot-assisted learning ROI to other EdTech investments Costs: What Schools Actually Pay Hardware Costs Entry-level: Alpha Mini ($200–$400), Marty ($300–$500)—good for coding introductionMid-range: Sanbot ($3,000–$6,000), QTrobot ($8,000–$12,000)—suitable for daily classroom useFull-featured: NAO ($9,000–$16,000), Pepper ($15,000–$25,000), Furhat ($20,000–$30,000)—comprehensive teaching platforms Hidden Costs (Budget Honestly) Software licenses: Annual subscriptions for lesson content and AI features ($500–$3,000/year)Maintenance & repairs: Warranty extensions, replacement parts, battery replacementsTeacher training: 8–20+ hours of PD per teacher to use robots effectivelyInfrastructure: Reliable Wi-Fi (robots need stable connectivity), charging stations, storageCurriculum integration: Time to design lessons that meaningfully incorporate robot interactions Challenges & Ethical Considerations Practical Challenges Novelty effect: Initial excitement fades after 2–4 weeks; sustained integration requires strong pedagogyTechnical reliability: Robots malfunction—Wi-Fi drops, speech recognition fails in noisy classrooms, batteries die mid-lessonTeacher workload: Adding a robot often increases teacher prep time, at least initiallyVendor lock-in: Schools become dependent on specific hardware/software ecosystems Ethical Concerns Data privacy: Robots with cameras and microphones collect student data—FERPA, GDPR, and local privacy laws applyEmotional attachment: Young children may form inappropriate emotional bonds with robotsScreen time vs. human time: Robot interaction shouldn't replace human social developmentEquity: Wealthy schools adopt robots while under-resourced schools fall further behindAlgorithmic bias: AI-powered adaptive learning may reinforce existing biases in educational content Pros and Cons of Humanoid Robots in Education Implementation Guide: A Practical Checklist for Schools Step 1: Define Your Purpose Don't buy a robot because it's cool. Define specific learning objectives: Is the goal STEM/coding education? → Alpha Mini or Marty (affordable)Language learning? → NAO or Pepper (multilingual)Special education/autism? → QTrobot (clinical evidence)General engagement & presentations? → Pepper or Sanbot Step 2: Budget Realistically Hardware is typically only 50–60% of total cost. Budget for software, training, maintenance, and infrastructure. Start with a 1-year pilot before committing to large-scale deployment. Step 3: Prepare Infrastructure Ensure reliable Wi-Fi in every classroom where robots will operateSet up charging stations and secure storageTest robot performance in actual classroom conditions (noise, movement, lighting) Step 4: Train Your Teachers Allocate 8–20 hours of professional development per teacher. Include: Basic robot operation and troubleshootingLesson design integrating robot interactionsWhen to use the robot vs. when not toData privacy protocols Step 5: Address Ethics & Data Privacy Conduct a data protection impact assessment (DPIA)Get parental consent for camera/microphone useEstablish clear policies on data storage and deletionReview FERPA (US), GDPR (EU), or local equivalents Step 6: Start Small, Measure, Scale Pilot with 1–2 classrooms, measure specific outcomes (engagement metrics, learning assessments), and scale only based on evidence. The Future: What's Coming in 2026–2028 LLM-powered robots: GPT and other large language models are being integrated into NAO and Pepper, enabling truly conversational teaching interactionsLower costs: Chinese manufacturers like UBTECH and Qihan are driving prices downMore advanced humanoids: Next-generation robots like 1X NEO and Unitree G1 could eventually enter educational environmentsRemote learning: Telepresence robots enabling expert teachers to "be present" in rural classroomsAI tutoring at scale: Combining humanoid presence with adaptive AI for truly personalized education For a full overview of the latest humanoid robots, see our best humanoid robots of 2026 ranking, or browse available models at Robozaps. --- # Challenges in Humanoid Robotics and How to Overcome Them URL: https://blog.robozaps.com/b/challenges-in-humanoid-robotics Author: Radica Maneva Published: 2026-01-30T00:00:00.000Z Updated: 2026-08-01T10:47:07.560Z Category: insights TL;DR: The biggest barriers to humanoid robotics in 2026: reliability (customers expect 95-99% uptime; most robots manage 30-90 minutes), 1-4 hour battery life, AI that still fails on edge cases, missing ISO safety standards, and unproven economics — no company has yet demonstrated profitable deployment at scale. Key takeaways: - Reliability is the #1 barrier: Industrial customers expect 95-99% uptime, but most humanoid robots can only operate for 30–90 minutes before needing a recharge or intervention. - Battery technology limits deployment: Current lithium-ion batteries restrict humanoid robots to 1–4 hours of active use, making 24/7 industrial operation impractical without charging infrastructure. - AI isn't ready for real-world autonomy: Despite advances in large language and behavior models, current AI struggles with edge cases, novel environments, and the long-tail of physical tasks. - Safety standards are still being written: ISO standards for dynamically balancing humanoid robots don't yet exist, creating regulatory uncertainty for deployments. - Cost vs. value remains unproven at scale: While unit costs are dropping, no company has demonstrated profitable humanoid robot deployments at thousands-of-units scale. - The demand question: It's unclear whether any single application requires thousands of humanoids per facility, challenging the massive deployment projections. FAQ: Q: What is the biggest challenge in humanoid robotics? A: The biggest challenge is achieving industrial-grade reliability (95-99% uptime) while maintaining the flexibility that makes humanoid robots valuable. Current robots can perform impressive demonstrations but struggle to operate reliably for extended periods in unpredictable real-world environments. This encompasses battery life, mechanical durability, AI robustness, and safety—all of which must be solved simultaneously. Q: Why can't humanoid robots work a full 8-hour shift? A: Current lithium-ion battery technology limits most humanoid robots to 1–4 hours of active operation. Bipedal walking alone consumes enormous energy for balance maintenance. Until solid-state batteries or hot-swappable battery systems become standard, continuous 8-hour operation remains impractical for most humanoid platforms. Q: Are humanoid robots safe to work around? A: Safety is an active area of development. Most humanoid robots include force-limited joints, emergency stop systems, and operational boundary enforcement. However, formal ISO safety standards specifically for humanoid robots are still being developed. Companies like Boston Dynamics, Agility Robotics, and Figure AI are contributing to these standards, but finalization is likely years away. Q: How much do humanoid robots cost in 2026? A: Prices range widely: Unitree G1 starts at $13,500, Sanctuary AI Phoenix costs approximately $40,000, and enterprise-focused robots like Agility Digit are estimated at $250,000. Tesla targets $20,000–$30,000 for Optimus. Total cost of ownership—including maintenance, infrastructure, and support—is significantly higher than purchase price alone. See our full cost guide. Q: Will humanoid robots replace human workers? A: In the near term (2026–2030), humanoid robots will augment human workers rather than replace them wholesale. They're best suited for repetitive, physically demanding, or dangerous tasks in structured environments. The future of humanoid robots likely involves human-robot collaboration rather than full replacement, especially as current AI can't match human judgment, creativity, and adaptability. Q: Which company is leading in solving humanoid robot challenges? A: Different companies lead in different areas: Sanctuary AI leads in hand dexterity and has the longest commercial deployment track record. Boston Dynamics leads in bipedal mobility. Tesla has the strongest manufacturing scaling potential. Agility Robotics has the most warehouse operational data. No single company has solved all challenges. See our ranking of the best humanoid robots for a full comparison. Q: What role does AI play in humanoid robotics challenges? A: AI is both the greatest enabler and one of the biggest challenges. Large Behavior Models, sim-to-real transfer learning, and computer vision are advancing rapidly, but current AI struggles with edge cases, novel environments, and the unpredictability of physical work. The gap between impressive demos and reliable deployment is largely an AI problem—the hardware is often ahead of the software. Q: When will humanoid robots be reliable enough for factories? A: For specific, well-defined tasks in structured environments, humanoid robots are already being piloted in factories (Amazon, BMW, Magna, Hyundai facilities). For broad, flexible deployment at 99.99% reliability, most industry experts estimate 2028–2032 as the realistic timeline. Battery technology, AI maturity, and safety standards all need to advance significantly. Q: How do humanoid robot challenges compare to early smartphone challenges? A: The analogy is apt. Early smartphones (2007–2010) had terrible battery life, limited apps, slow processors, and questionable reliability. Within a decade, they became indispensable. Humanoid robots face similar growing pains—battery life, software maturity, and ecosystem development all need time. The key difference is that physical robots face physics constraints that software products don't, so the timeline will likely be longer. Q: What's the best humanoid robot to buy in 2026? A: It depends on your use case. For industrial dexterity tasks, Sanctuary AI Phoenix offers the best hand technology. For affordability and research, Unitree G1 is unmatched at $13,500. For warehouse logistics, Agility Digit has the most operational experience. Browse all options on Robozaps or read our complete ranking. Related: Best Humanoid Robots of 2026 · Sanctuary AI Phoenix Review · Future of Humanoid Robots · Humanoid Robot Cost Guide · Applications of Humanoid Robots Ready to explore? Browse humanoid robots for sale on Robozaps or visit our shop. The humanoid robotics industry has exploded with investment, innovation, and bold promises. But behind the viral demos and billion-dollar valuations lie serious engineering, economic, and regulatory challenges that every company in this space must overcome. This comprehensive guide examines the key challenges in humanoid robotics as of 2026, what leading companies are doing to solve them, and what the future holds for general-purpose humanoid robots. Table of Contents Power and Battery Life ConstraintsBipedal Mobility and BalanceDexterous ManipulationAI and Autonomy LimitationsReliability and UptimeSafety Standards and RegulationCost and Economic ViabilityMarket Demand Reality CheckManufacturing at ScaleHuman-Robot Interaction and TrustFuture Outlook: How These Challenges Will Be SolvedFrequently Asked Questions Power and Battery Life: The Fundamental Constraint Battery life is arguably the single most critical bottleneck preventing humanoid robots from achieving industrial-scale deployment. As IEEE Spectrum's October 2025 investigation revealed, the numbers are sobering: Agility Digit: 90 minutes of operation, followed by a 9-minute fast charge. In practice, Digit operates in 30-minute intervals at Amazon warehouses.Figure 02: Approximately 2–3 hours of active operation.Tesla Optimus Gen 2: Estimated 4–5 hours, though Tesla has not disclosed official runtime figures.Sanctuary AI Phoenix: Approximately 4 hours per charge, with 43.5 hours of cumulative operation demonstrated across Hannover Messe 2025 (with charging breaks). Compare this to an 8-hour factory shift—let alone 24/7 operation. The physics are unforgiving: bipedal walking consumes enormous energy just to maintain balance, leaving less power for actual work tasks. This is why Sanctuary AI switched Phoenix to a wheeled base—energy efficiency matters more than bipedal aesthetics in industrial settings. Why Better Batteries Alone Won't Solve This Lithium-ion energy density improves at roughly 5–8% per year. Even with solid-state batteries (expected commercial availability 2027–2029), a doubling of energy density would still leave most humanoid robots short of a full 8-hour shift under heavy workloads. The industry needs a combination of: More efficient actuators and motor designsBetter power management software that reduces energy waste during idle momentsHot-swappable battery packs for zero-downtime operationCharging infrastructure designed into factory layouts from the start Bipedal Mobility and Balance: Harder Than It Looks Walking on two legs is something humans do effortlessly, but it remains one of the hardest problems in robotics. The challenge isn't making a robot walk in a lab—it's making it walk reliably in unpredictable real-world environments for months without falling. The Balance Problem Bipedal robots are inherently unstable—they're essentially inverted pendulums that must constantly correct their balance. This requires: High-frequency sensor feedback (IMUs, force/torque sensors in feet)Real-time control loops running at 500–1000 HzSignificant computational resources dedicated solely to balanceRobust recovery behaviors for slips, trips, and external pushes Boston Dynamics Atlas is the gold standard for dynamic bipedal locomotion, demonstrating backflips, parkour, and recovery from pushes. But Atlas has been in development for over a decade with hundreds of millions in R&D funding. Newer entrants like Unitree H1 and Figure 02 are making rapid progress, but industrial-grade reliability on uneven surfaces, wet floors, or environments with obstacles remains an unsolved problem. The Pragmatic Alternative Some companies have acknowledged this challenge by choosing wheels over legs. Sanctuary AI's Phoenix Generation 8 uses a wheeled base, and several logistics robots use hybrid wheel-leg designs. This trade-off sacrifices stair-climbing ability but dramatically improves reliability, energy efficiency, and payload capacity. Dexterous Manipulation: The Hand Problem If bipedal walking is hard, dexterous manipulation with robotic hands is harder. The human hand has 27 degrees of freedom, thousands of tactile sensors, and is controlled by one of the largest regions of the motor cortex. Replicating this capability is one of the greatest challenges in robotics. The Gap Between Demo and Deployment Viral videos show robots handling eggs, making coffee, or folding laundry. But these demos often represent best-case scenarios after many attempts. In industrial settings, robots need to handle thousands of different objects—varying in size, weight, texture, temperature, and fragility—with near-zero failure rates. This requires: Tactile sensing: Knowing grip force in real-time to avoid crushing or dropping objects. Sanctuary AI leads here with multi-modal sensors detecting pressure, temperature, vibration, and slip.Sim-to-real transfer: Training manipulation skills in simulation (e.g., NVIDIA Isaac Lab) and transferring them to physical robots. This is scaling rapidly but still struggles with deformable objects and novel geometries.Adaptive grasping: Adjusting grip strategy on the fly based on sensory feedback, not pre-programmed motion paths. AI and Autonomy: The Software Challenge Hardware is only half the equation. The AI systems controlling humanoid robots face their own set of formidable challenges that may be even harder to solve than the mechanical ones. The Long Tail of Physical Tasks Language AI can handle most text tasks because language is structured and rule-governed. Physical tasks have no such luxury. A warehouse might present thousands of edge cases: oddly shaped packages, items stuck together, unexpected obstacles, spilled liquids, or objects in unfamiliar orientations. Each edge case requires either pre-programmed handling or genuine intelligence—and current AI has neither at scale. As former Agility Robotics CPO Melonee Wise stated in IEEE Spectrum's October 2025 investigation: "I think what a lot of people are hoping for is they're going to AI their way out of this. But the reality of the situation is that currently AI is not robust enough to meet the requirements of the market." Large Behavior Models: Promise and Limitations Multiple companies are developing "Large Behavior Models" (LBMs)—the physical AI equivalent of Large Language Models: Sanctuary AI is training LBMs on Microsoft Azure using data from Phoenix deploymentsBoston Dynamics is working with Toyota Research Institute on behavior models for AtlasFigure AI has partnered with OpenAI to bring language understanding to physical actionsNVIDIA provides Isaac Lab infrastructure used across the industry for sim-to-real training These are promising approaches, but they face a fundamental data challenge: unlike internet text data (abundant, cheap, standardized), physical behavior data is expensive to collect, hard to standardize, and environment-specific. Training a robot to work in one warehouse doesn't automatically transfer to a different warehouse layout. The Teleoperation Crutch Many "autonomous" robot demonstrations actually rely heavily on teleoperation—a human operator controlling the robot remotely. While teleoperation is a valid data-collection strategy (the robot learns from human demonstrations), it's not scalable as a deployment model. True autonomy requires the robot to handle novel situations independently, and we're still years away from that for most physical tasks. Reliability: The 99.99% Problem Industrial automation has a simple requirement: it must work reliably, all the time. Traditional industrial robots from FANUC, ABB, and KUKA typically achieve 95-99% uptime in well-maintained environments. Humanoid robots are nowhere near this benchmark. Why Humanoids Break Mechanical complexity: A humanoid robot has hundreds of joints, actuators, sensors, and moving parts. Each is a potential failure point. Traditional industrial arms have 6 joints.Software failures: AI-driven control systems can encounter unexpected states, sensor misreadings, or computation timeouts that cause the robot to freeze, fall, or behave unpredictably.Environmental sensitivity: Dust, moisture, temperature extremes, electromagnetic interference, and vibration all affect performance in ways that clean-room demos don't reveal.Battery degradation: Lithium-ion batteries lose capacity over charge cycles, reducing runtime over months of operation. The Path to Reliability Achieving industrial-grade reliability requires: Thousands of hours of real-world operation data to identify and fix failure modesRedundant sensor systems so single-point failures don't cause shutdownsPredictive maintenance using sensor data to replace parts before they failGraceful degradation—robots that can safely stop and request help rather than crash Sanctuary AI's 43.5 hours of cumulative operation at Hannover Messe 2025 is a promising data point, and Agility Robotics has accumulated the most real-world operational hours through Amazon warehouse pilots. But sustained months of reliable factory operation remains undemonstrated by any humanoid company. Safety Standards: Writing the Rules for a New Industry When a traditional industrial robot arm operates, it does so behind safety cages or with clearly defined collaborative zones governed by ISO 10218 and ISO/TS 15066. Humanoid robots—which walk through human workspaces, make autonomous decisions, and interact physically with people—need entirely new safety frameworks. Current Regulatory Gaps No ISO standard exists for dynamically balancing legged robots. Boston Dynamics, Agility Robotics, and Figure AI are contributing to development of these standards, but they're years from finalization.Liability questions are unresolved: If a humanoid robot injures a worker, who is liable—the manufacturer, the deploying company, the AI training provider, or the operator?Country-by-country variation: The EU, US, China, Japan, and South Korea are all developing different regulatory approaches, creating compliance complexity for global deployments. What Companies Are Doing In the absence of formal standards, leading companies are self-regulating: Boston Dynamics publishes responsible AI principles and contributes to ISO working groupsAgility Robotics designed Digit with force-limited joints to reduce injury riskSanctuary AI's Carbon system includes safety monitoring layersFigure AI includes emergency stop systems and operational boundary enforcement However, self-regulation isn't enough for enterprise buyers who need regulatory certainty before large-scale deployments. This creates a chicken-and-egg problem: standards bodies need deployment data to write good standards, but companies need standards to deploy at scale. Cost and Economic Viability: Can Humanoids Be Profitable? The economics of humanoid robots are still largely theoretical. While the vision is compelling—replace expensive, scarce human labor with tireless robots—the math hasn't been proven at scale. The Price Trajectory Robot prices are falling rapidly: Tesla targets $20,000–$30,000 for Optimus (widely questioned by analysts)Sanctuary AI Phoenix is estimated at $50,000-$100,000Unitree G1 starts at $13,500—the most affordable humanoid availableAgility Digit is estimated at $250,000 per unit at current volumes However, the purchase price is just the beginning. Total cost of ownership includes maintenance, software updates, charging infrastructure, integration costs, on-site technical support, and the human operators who supervise robot fleets. These costs are poorly understood because no company has operated humanoid robots at scale long enough to generate reliable TCO data. For a detailed price breakdown, see our complete humanoid robot cost guide. Market Demand: Do We Actually Need Millions of Humanoids? The humanoid robotics market is projected to reach $5 trillion by 2050 (Morgan Stanley) with 18,000 units shipping in 2025 (Bank of America). But these projections deserve scrutiny. The Demand Question As Melonee Wise pointed out: "I don't think anyone has found an application for humanoids that would require several thousand robots per facility." This challenges the narrative of millions of humanoid robots replacing human workers en masse. The reality is more nuanced: Labor shortages are real: Manufacturing, warehousing, and elder care face genuine worker shortages that automation can address.But specialized robots often work better: For many tasks, purpose-built robots (conveyor systems, robotic arms, AGVs) are more reliable and cost-effective than general-purpose humanoids.The humanoid advantage is flexibility: Humanoids make sense when a facility needs one robot that can do many different tasks, rather than buying separate specialized robots for each task.Human-designed environments favor human form: Factories, warehouses, and buildings are designed for human bodies—doors, stairs, shelves, tools all assume human proportions. Where Demand Is Real Today The most promising near-term markets for humanoid robots include: Automotive manufacturing: McKinsey's October 2025 analysis identifies this as the first sector for production-scale humanoid deployment. Both Sanctuary AI (Magna partnership) and Figure AI (BMW partnership) are active here.Warehousing and logistics: Amazon's pilots with Agility Digit and its own Vulcan robots demonstrate real demand for automated picking and stowing.Electronics manufacturing: High-precision assembly tasks requiring dexterity in controlled environments.Elder care: Aging populations in Japan, South Korea, and Europe create demand for assistive robots, though regulatory and cultural barriers remain high. Manufacturing at Scale: Building Millions of Complex Machines Even if demand materializes, manufacturing humanoid robots at the scale projections assume (millions of units) presents its own massive challenge. Each humanoid robot contains: Dozens of precision actuators and motorsHundreds of sensorsCustom-designed mechanical componentsHigh-performance computing hardwareAdvanced battery systemsSpecialized materials (carbon fiber, titanium, custom alloys) The Tesla Manufacturing Thesis Tesla's strongest argument for Optimus isn't the robot itself—it's Tesla's manufacturing expertise. The company that scaled EV production from thousands to millions of units arguably has the best shot at doing the same for humanoid robots. Elon Musk has stated Tesla could eventually produce millions of Optimus units per year. However, humanoid robots are more mechanically complex than cars, with tighter tolerances and more custom components. The supply chain for humanoid-specific parts (especially actuators and tactile sensors) doesn't yet exist at automotive scale. The China Factor China is investing aggressively in humanoid robotics, with companies like Unitree, XPENG IRON, AgiBot, and Kepler developing humanoid platforms with access to China's deep manufacturing infrastructure and lower labor costs. This gives Chinese companies a potential cost advantage in scaling production—similar to what happened with EVs, drones, and consumer electronics. Human-Robot Interaction: Building Trust For humanoid robots to work alongside humans, people need to trust them. This goes beyond safety—it's about predictability, communication, and social acceptance. The Uncanny Valley Robots that look almost-but-not-quite human can provoke discomfort—the "uncanny valley" effect. Most current humanoid robots avoid this by using clearly mechanical aesthetics (helmeted heads, visible joints, non-skin materials). But as robots become more capable and present in daily life, designing for appropriate social interaction becomes important. Workplace Integration Workers who will share space with humanoid robots have legitimate concerns: Job displacement anxiety: "Will this robot replace me?" is the immediate question for many workers.Physical safety: Working near a 70kg robot that can lift 12kg with each arm requires trust in the safety systems.Workflow disruption: Integrating robots into existing workflows requires retraining human workers and redesigning processes.Communication: How does a robot signal its intentions to nearby humans? Clear status indicators, predictable movements, and simple communication interfaces are essential. Companies deploying humanoid robots will need robust change management programs—not just technical integration. The applications of humanoid robots will expand faster in environments where human-robot collaboration is designed from the start. Future Outlook: How These Challenges Will Be Solved Despite the significant obstacles, there are strong reasons for optimism. The convergence of AI, advanced manufacturing, and massive investment is accelerating progress faster than any previous robotics era. Near-Term Solutions (2026–2028) Better battery management: Hot-swappable battery packs and optimized charging infrastructure will mitigate runtime limitations before battery chemistry itself improves dramatically.Sim-to-real breakthrough: NVIDIA Isaac Lab and similar platforms are enabling thousands of simulated robots to train simultaneously, accelerating skill acquisition by orders of magnitude. Sanctuary AI has already demonstrated zero-shot sim-to-real transfer for dexterous manipulation.Hybrid autonomy: Rather than full autonomy or full teleoperation, robots will operate autonomously on trained tasks and request human guidance for novel situations—a practical middle ground.Focused deployments: Instead of trying to do everything, companies will focus humanoid robots on specific high-value tasks where they have clear advantages (e.g., repetitive bin picking in automotive, goods-to-person logistics).Wheeled alternatives: More companies may adopt wheeled bases for industrial applications where stairs aren't required, dramatically improving reliability and battery life. Medium-Term Solutions (2028–2032) Solid-state batteries: Expected to reach commercial viability, potentially doubling energy density and enabling full 8-hour shift operation.Mature safety standards: ISO and regional bodies will have finalized humanoid-specific safety standards, unlocking enterprise procurement at scale.Manufacturing scale: Tesla, Chinese manufacturers, and others will have established production lines capable of producing tens of thousands of units per year, driving costs below $20,000.Large Behavior Models at scale: With years of real-world deployment data, LBMs will handle the majority of physical tasks without teleoperation, approaching human-level task flexibility for structured environments.Industry-specific solutions: Rather than one-size-fits-all humanoids, expect specialized variants optimized for automotive, logistics, healthcare, and other verticals. Long-Term Vision (2032+) True general-purpose capability: Robots that can learn any physical task from observation or instruction, matching or exceeding human performance across a wide range of activities.Home deployment: Once costs drop below $10,000 and reliability reaches consumer-grade levels, humanoid robots will begin entering homes for elder care, household assistance, and companionship.Collaborative ecosystems: Fleets of humanoid robots coordinating with each other and with human workers, managed by centralized AI systems that optimize task allocation in real-time.Self-improvement: Robots that can identify their own limitations, request training data, and improve their capabilities autonomously through continuous learning. --- # 11 Tesla Optimus Competitors and Alternatives (2026) URL: https://blog.robozaps.com/b/tesla-optimus-alternatives-competitors Author: Radica Maneva Published: 2026-01-29T00:00:00.000Z Updated: 2026-08-01T23:03:43.255Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: alternatives-competitors TL;DR: Tesla Optimus was not publicly orderable when checked on July 16, 2026. Atlas is in its first customer pilot at Hyundai, Figure 03 has a documented BMW demonstration and 2026 production ramp, and Digit has a commercial GXO deployment. Unitree sells developer robots at public prices; 1X NEO is taking deposits for a home robot. Key takeaways: - Tesla says Optimus Gen 3 is its first mass-production design, with production planned before the end of 2026; it is not yet a publicly orderable product. - Boston Dynamics Atlas publishes the strongest industrial specifications here, including a 50 kg instantaneous payload, 56 degrees of freedom and IP67 protection. - Unitree R1 Air now starts at $4,900, so the $13,500 G1 is not the cheapest Unitree humanoid. - Digit's multi-year RaaS agreement is with GXO; Amazon tested Digit and invested in Agility but should not be presented as the customer behind that deployment. - 1X describes NEO as autonomous for basic tasks with expert supervision available for unfamiliar tasks, not as fully autonomous. FAQ: Q: What are the best Tesla Optimus competitors and alternatives in 2026? A: The strongest options depend on the job: Boston Dynamics Atlas for high-payload industrial evaluation, Agility Digit for commercial warehouse logistics, Figure 03 for a vertically integrated AI and hardware platform with a BMW demonstration, Unitree G1 for orderable developer hardware, and 1X NEO for home-focused early access. No single model leads every category. Q: Can you buy Tesla Optimus in 2026? A: No public Tesla Optimus order page or final retail price was available when this guide was checked on July 16, 2026. Tesla says Gen 3 is its first mass-production design and planned production before the end of 2026. Q: What is the cheapest Tesla Optimus alternative? A: Among models in this guide with public pricing, Unitree lists the smaller R1 Air from $4,900, the R1 from $5,900, and the G1 from $13,500. Buyers should compare size and capability rather than treating those robots as equivalent. Q: Which humanoid robots are deployed or being tested in real workplaces? A: Agility Digit has a multi-year commercial RaaS agreement with GXO. Figure documented a Figure 03 demonstration at BMW, while Apptronik describes Apollo 2 pilot and task-data work at Mercedes-Benz and GXO sites. A demonstration, pilot and commercial deployment are different stages. Q: Which humanoid robot is best for home use? A: 1X NEO has the clearest public home offer: a $20,000 ownership option or $499 monthly early-access subscription with a $200 refundable deposit. 1X says basic tasks can be autonomous while unfamiliar tasks may use remote expert supervision. Figure 03 has a home vision but no public consumer purchase page. Q: Is Boston Dynamics Atlas better than Tesla Optimus? A: Atlas has stronger current public industrial specifications and a first customer pilot at Hyundai, but Tesla has not yet published final Gen 3 specifications or pricing. Atlas is the clearer choice for high-payload industrial evaluation today; a broad product-to-product verdict is premature. The best Tesla Optimus competitors and alternatives depend on the job. Boston Dynamics Atlas has the strongest published industrial specifications; Figure 03 has a documented BMW demonstration and 2026 production ramp; Agility Robotics Digit has a commercial logistics deployment; Unitree sells developer hardware at public prices; and 1X NEO is taking deposits for a home robot. Tesla Optimus itself was not available to buy when this guide was checked on July 16, 2026. That distinction matters. A robot can be orderable, in a paid deployment, in a pilot, taking deposits, or still preparing for production. Those labels are not interchangeable. This guide uses current manufacturer material and keeps those stages separate. For a broader view of humanoid robot costs, see our pricing guide. If the goal is hardware you can source now, compare the cheapest humanoid robots and the most advanced humanoids you can buy. Where Does Tesla Optimus Stand in 2026? Tesla's January 2026 shareholder update described Optimus Gen 3 as its first design intended for mass production and said production was planned to begin before the end of 2026. The April update said Tesla was preparing its first large-scale Optimus factory, with a designed annual capacity of one million units. Those are production plans, not evidence of a product already shipping. Tesla had not published a final retail price, a current Gen 3 specification sheet, or a consumer order page as of this review. We therefore do not use the often-repeated $20,000 to $30,000 price estimate or older prototype dimensions as settled facts. Availability: not publicly orderableCurrent official stage: factory preparation, with production planned before the end of 2026Public price: not disclosedCurrent Gen 3 specifications: not published in a manufacturer specification sheet How We Compared Optimus Competitors We weighted four things: current model status, documented deployment, published specifications, and price transparency. Manufacturer claims are reported as claims unless a named customer or commercial agreement supports them. A narrow task completed in a pilot is not treated as general autonomy. The result is a list of credible alternatives, not a claim that every robot can replace Optimus in every setting. The wider field of humanoid robot companies includes many prototypes and research platforms that are less comparable for buyers. 1. Boston Dynamics Atlas The electric Atlas is the clearest alternative for demanding industrial work. Boston Dynamics' specification sheet lists a 1.9 m height, 90 kg mass, 56 degrees of freedom, a 50 kg instantaneous payload, a 30 kg sustained payload, and a four-hour battery. It is IP67 rated and can swap its own battery. Best fit: industrial material handling and environments that need high payload, reach, or weather resistancePrice: not publicly disclosedStatus: first customer pilot at Hyundai; not a consumer robotWatch-out: exceptional hardware does not make Atlas a low-cost or general household substitute Atlas leads this list on published industrial hardware specifications, but buyers still need a site-specific application and commercial discussion with Boston Dynamics. 2. Figure 03 Figure 03 combines a full-size humanoid platform with Figure's Helix control system. The current product page lists a height of 5 ft 8 in, weight of 61 kg, 20 kg payload, five-hour runtime, and 1.2 m/s speed. Its production status is more advanced than the old 'pre-production' label suggests. In its April 2026 production update, Figure said it had produced more than 350 Figure 03 robots and was ramping from one robot per day toward one per hour. In June, Figure documented a BMW demonstration using Figure 03 with Helix 02. Best fit: companies evaluating vertically integrated humanoid hardware and AIPrice: not publicly disclosedStatus: production ramp and documented BMW demonstrationWatch-out: Figure promotes a home-robot vision, but it does not offer a public consumer purchase page for Figure 03 3. Unitree G1, R1, and H1 Unitree is the easiest company here to compare on public pricing. The G1 product page starts at $13,500 and lists a height of about 132 cm, weight of about 35 kg, more than 2 m/s maximum speed, roughly two hours of battery life, and 23 to 43 joint motors depending on configuration. The G1 is no longer Unitree's cheapest humanoid. The smaller R1 Air starts at $4,900, while the R1 starts at $5,900. Unitree's official store lists the full-size H1 at $90,000. The lower price of the R1 comes with a smaller body and different capability envelope, so it is not a direct replacement for a full-size industrial humanoid. Best fit: research, development, education, and teams that want orderable hardwarePublic prices: R1 Air from $4,900; R1 from $5,900; G1 from $13,500; H1 $90,000Status: commercially orderable hardwareWatch-out: hardware availability is not proof of broad autonomous job performance 4. Agility Robotics Digit Digit is purpose-built for logistics rather than household use. Agility's published Digit specifications include a height of 5 ft 9 in, weight of 140 lb, 35 lb payload, and up to four hours of battery life with autonomous docking. Its strongest evidence is commercial deployment. GXO signed a multi-year Robot-as-a-Service agreement with Agility, and Digit has moved more than 100,000 totes at a GXO site. Agility's June 2026 company update names GXO, Schaeffler, Toyota Canada, and Mercado Libre among current customers. Amazon has tested Digit and invested in Agility, but it should not be described as the customer behind the GXO RaaS deployment. Best fit: repetitive tote movement and warehouse workflowsPrice: enterprise or RaaS terms are not publicly listedStatus: documented commercial logistics deploymentsWatch-out: Digit's specialized end effectors and workflow focus are a feature for logistics, not a general-purpose promise 5. 1X NEO NEO is the most explicit home-focused option in this comparison. 1X lists a 5 ft 6 in height, 66 lb weight, four-hour runtime, and 1.4 m/s walking speed. Its US order page offers ownership for $20,000 or an early-access subscription for $499 per month, with a $200 refundable deposit and delivery starting in 2026. NEO should not be described as fully autonomous. 1X says it can complete basic tasks autonomously, while owners can schedule an expert to supervise tasks NEO does not yet know. That makes the human-support model central to the product, not a minor footnote. Best fit: early adopters who specifically want a home robotPrice: $20,000 to own or $499 per month; $200 refundable depositStatus: taking deposits for US delivery starting in 2026Watch-out: complex tasks may involve remote expert supervision 6. Apptronik Apollo Apptronik's original Apollo specification listed a height of 5 ft 8 in, weight of 160 lb, 55 lb lifting capacity, and a four-hour swappable battery. The company introduced Apollo 2 in 2026, so buyers should ask which published specifications apply to the current configuration. Mercedes-Benz is a real partner, but the accurate status is pilot and data-collection work. Apptronik's current Robot Park material describes Apollo 2 collecting task data at Mercedes-Benz and GXO sites through a mix of teleoperation and autonomous operation. That is meaningful validation, but it is not the same as a broad production assembly-line rollout. Best fit: industrial pilots, task-data collection, and applications being developed with ApptronikPrice: not publicly disclosedStatus: Apollo 2 pilot and data-collection work at partner sitesWatch-out: older Apollo specifications should not automatically be presented as Apollo 2 specifications 7. Sanctuary AI Phoenix Sanctuary AI emphasizes manipulation intelligence and data capture. Its eighth-generation Phoenix uses a wheeled base and is designed to collect high-quality task data, making it less of a like-for-like walking robot than Atlas, Digit, or Optimus. Sanctuary has also demonstrated a 21-degree-of-freedom dexterous hand. Phoenix is relevant when hands, teleoperation, and the transition from human demonstrations to autonomy matter more than bipedal mobility. Current Gen 8 dimensions and commercial pricing are not published on the manufacturer page. Best fit: manipulation research, task-data capture, and enterprise AI developmentPrice: not publicly disclosedStatus: current Gen 8 data-capture platformWatch-out: the current wheeled-base design is not a direct bipedal substitute 8. Fourier GR-3 Fourier's current GR-3 documentation lists a height of 165 cm, weight of 71 kg, up to 55 actuators, a 3 kg single-arm load, speed of 6.1 km/h, and about three hours of operation. Those figures replace older GR-series numbers that are often mixed together online. Best fit: research, human interaction, and teams interested in Fourier's rehabilitation-robotics backgroundPrice: not publicly disclosedStatus: current GR-series platformWatch-out: GR-1, GR-2, and GR-3 specifications are not interchangeable 9. XPENG IRON XPENG is the closest structural comparison to Tesla because both companies are applying EV manufacturing and physical-AI work to humanoid robots. The important 2026 fact is its stage of development, not a claim that IRON is already operating in factories. In its June 2026 update, XPENG said the mass-production-ready version was entering joint hardware-software integration. It targeted formal mass production by the end of 2026 and said robots would begin working as in-store shopping guides in the first quarter of 2027. Best fit: watching an EV manufacturer's path toward scaled humanoid productionPrice: not publicly disclosedStatus: integration before targeted mass production at the end of 2026Watch-out: a production target is not a current commercial deployment 10. AgiBot A2 AgiBot positions the A2 as an interactive service robot for marketing, reception, exhibition, and retail guidance. The current manufacturer page lists a height of 169 cm, weight of 69 kg, 700 Wh battery, two-hour runtime with battery swapping, and more than 40 active degrees of freedom. This is a service-oriented platform, not a high-speed industrial runner. The 7 km/h, 55 kg, 49-degree-of-freedom figures sometimes attached to A2 belong to an older Raise A1 page and should not be used for the current A2. Best fit: guided service, reception, and customer-interaction applicationsPrice: not publicly disclosedStatus: current product with enterprise enquiriesWatch-out: do not mix A2 specifications with the older Raise A1 11. UBTECH Walker S2 Walker S2 is UBTECH's current industrial humanoid, replacing Walker X as the more relevant Optimus comparison. The Walker S2 product page lists a 15 kg payload and an autonomous battery-swap system that can complete a swap in about three minutes. UBTECH presents the model as in mass production and delivery. Best fit: industrial material handling where continuous operation and automated battery changes matterPrice: not publicly disclosedStatus: manufacturer-reported mass production and deliveryWatch-out: deployment scale and customer economics still require buyer verification Tesla Optimus Alternatives Compared Which Optimus Alternative Should You Choose? For heavy industrial work: start with Atlas; its payload, environmental rating, and battery system are the clearest fit.For documented logistics deployment: evaluate Digit and its GXO RaaS evidence.For orderable research hardware: compare Unitree G1 with the smaller, cheaper R1 models.For home use: NEO has the clearest public offer, but budget for remote expert involvement in tasks beyond basic autonomy.For an AI-plus-hardware platform with manufacturing momentum: Figure 03 has the strongest published 2026 production evidence.For industrial pilots: Apollo 2, Walker S2, and GR-3 deserve use-case-specific diligence rather than a generic ranking. Before buying or signing a pilot, ask for the exact model revision, payload definition, duty cycle, autonomy boundary, supervision requirement, service terms, safety certification, and acceptance test. Humanoid specifications are moving quickly, and a compelling demo does not establish uptime or economics in your facility. What Changed in the Humanoid Market in 2026? The market is moving from prototypes toward a mix of orderable developer hardware, paid deployments, production ramps, and supervised early-access products. Unitree said it delivered more than 5,500 humanoids in 2025; Figure reported more than 350 Figure 03 units by April 2026; and established industrial suppliers are building programs around narrower tasks. The practical trend is specialization. Atlas and Walker S2 focus on industrial work, Digit on logistics, NEO on the home, A2 on service interaction, and Phoenix on manipulation data. Optimus may eventually compete across several of those categories, but Tesla's own 2026 material still places Gen 3 before production rather than in public sale. Compare on RoboZaps: Apptronik Apollo, Atlas, Digit, Figure 03, NEO and Tesla Optimus Gen 3. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs Boston Dynamics Atlas URL: https://blog.robozaps.com/b/tesla-optimus-vs-boston-dynamics-atlas Author: Radica Maneva Published: 2026-01-29T00:00:00.000Z Updated: 2026-08-01T11:20:47.884Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: comparisons TL;DR: Optimus is lighter (57 kg), targets $20,000-$30,000 and is built for mass production; the all-electric Atlas became a production product at CES 2026 with 56 degrees of freedom and enterprise-grade capability, going to Hyundai and Google DeepMind first. Atlas is the one real customers can get in 2026; Optimus is the volume bet. Key takeaways: - Tesla Optimus is lighter (57 kg), cheaper ($20,000–$30,000 target), and built for mass production — is moving from internal prototype to production: Tesla missed its stated 10,000-unit-by-2025 target, and on the Q1 2026 earnings call (April 22, 2026) Musk declined to set a new 2026 unit target. Fremont volume production is planned for late July/August 2026. - Boston Dynamics Atlas went fully electric in 2024 and launched its production version at CES 2026 — with 56 degrees of freedom, 50 kg lift capacity, and 2026 deployments at Hyundai and Google DeepMind already fully committed. - Optimus wins on affordability, energy efficiency, and scalability. Atlas wins on agility, ruggedness, and enterprise-grade reliability. - Both robots are moving from prototypes to real factory floors in 2026 — only Atlas has shipped to commercial customers in 2026, with Optimus used internally at Tesla factories only. FAQ: Q: How much does Tesla Optimus cost vs Boston Dynamics Atlas? A: Neither robot has a price. Tesla's $20,000-$30,000 is a long-term at-scale target, not an Optimus price you can pay, and Boston Dynamics has never published an Atlas price: reported figures run from analyst estimates around $130,000-$140,000 to an anonymously sourced ceiling near $320,000, none confirmed. Our Boston Dynamics Atlas review breaks down the full price evidence. Q: Which robot is more advanced — Optimus or Atlas? A: It depends on the metric. Atlas is more advanced in agility (backflips, parkour, rough terrain), strength (50 kg lift, 7.5 ft reach), and ruggedness (water-resistant, extreme temps). Optimus is more advanced in hand dexterity (11 DoF per hand), energy efficiency, and AI-driven learning from Tesla's FSD neural networks. On the July 22, 2026 earnings call, Musk claimed 'there is no humanoid robot that is actually able to do generalized tasks. Optimus will be the first one that is capable of doing that.' That is a claim, not a demonstrated capability; the call was audio-only and no demo accompanied it. Q: Can I buy a Tesla Optimus or Boston Dynamics Atlas in 2026? A: Not for consumers, and mostly not for businesses either. Musk said in January 2026 that Optimus was 'not in usage in our factories in a material way,' and Tesla's July 22, 2026 update says production is 'anticipated later this year,' with public sales pointed at end of 2027. Atlas is in pilot deployment within Hyundai's plants, with external customers targeted for 2027. Q: What is Optimus V3 and when is it coming? A: Optimus 3 (V3) is Tesla's ground-up redesign and the first version it calls intended for mass production. It remains unrevealed as of July 23, 2026. Tesla's July 22 Q2 update says first-generation production lines are being installed at Fremont with production 'anticipated later this year'; the reveal is being held until near that start. Q: What happened to the hydraulic Atlas? A: Boston Dynamics retired the hydraulic Atlas in April 2024 and transitioned to a fully electric version. The electric Atlas offers better energy efficiency, lower maintenance, and reduced noise while maintaining dynamic capability. The production version unveiled at CES 2026 is all-electric with Hyundai Mobis actuators. Q: Will humanoid robots replace human workers? A: Both Tesla and Boston Dynamics frame their robots as handling tasks that are dangerous, repetitive, or physically demanding —. While some job displacement is inevitable in manufacturing and logistics, new roles in robot maintenance, programming, and fleet management are expected to emerge. The transition will require workforce upskilling and policy support. Q: Which robot is better for factory use? A: For heavy-duty automotive manufacturing, Atlas is the stronger choice — it has 56 degrees of freedom, 50 kg lift capacity, autonomous battery swap for 24/7 operation, and enterprise software integration. For lighter assembly, sorting, and material handling tasks, Optimus offers a more cost-effective solution with its lower price point and mass-production design. Related: Tesla Optimus Gen 2 Review · Boston Dynamics Atlas Review Ready to buy? Browse humanoid robots for sale on Robozaps. Tesla Optimus vs Boston Dynamics Atlas — which humanoid robot leads in 2026? This head-to-head comparison breaks down every spec, capability, price point, and real-world deployment so you can see exactly how these two titans stack up. Tesla Optimus vs Atlas: Complete Specs Comparison Table (2026) Tesla Optimus: Everything You Need to Know (2026 Update) Tesla Optimus (also called Tesla Bot) is Elon Musk's ambitious bid to create a mass-market humanoid robot that can replace humans in dangerous, repetitive, or boring tasks. What sets Optimus apart isn't raw athleticism — it's Tesla's ability to manufacture at scale and leverage its existing AI infrastructure. Design and Build Standing 5 feet 8 inches (1.73 m) tall and weighing just 57 kg (125 lb), Optimus is designed to fit seamlessly into human environments. It's roughly 35% lighter than Atlas, which translates to better energy efficiency and easier deployment in homes and offices. Each hand has 11 degrees of freedom, enabling delicate manipulation tasks like handling eggs without cracking them or folding laundry. The latest iteration — Optimus V2.5 (the "golden" version seen at the TRON: Ares premiere) — features redesigned hands, lighter actuators, and significantly smoother motion. Musk has hinted at an Optimus V3 with significant upgrades, though no official timeline has been confirmed. Core Technologies Optimus leverages Tesla's proven AI stack from Full Self-Driving (FSD): Vision-first perception: Camera-based computer vision (no LiDAR), similar to Tesla vehiclesNeural networks: Trained on massive driving datasets, adapted for robot manipulation and navigationSingle SoC "Bot Brain": A custom Tesla System-on-Chip providing high computational power at low energy costSim-to-real training: Motion capture data from human demonstrations accelerates learningFleet learning: When one Optimus learns a task, the knowledge transfers to all units Price and Availability Tesla's aspirational price target is $20,000–$30,000 per unit at scale — deliberately "less than a car." In 2025, Tesla targeted approximately 10,000 units, though production numbers may vary. Currently, Optimus units are deployed internally at Tesla factories performing sorting, material handling, and assembly tasks. Public sales remain a few years away as Tesla works through safety regulations and testing, but Musk has called Optimus potentially "Tesla's most significant product" — exceeding even its vehicle business in value. How the Boston Dynamics Atlas Stacks Up in 2026 Boston Dynamics Atlas is the culmination of over 30 years of robotics research. Originally funded by DARPA, Atlas has evolved from a tethered hydraulic prototype into a fully electric, commercially deployable humanoid robot. In January 2026, Boston Dynamics unveiled the production version of Atlas at CES — marking the transition from research platform to enterprise product. Development Journey: From Research to Production Atlas's evolution is remarkable: 2013: First Atlas prototype (DARPA-funded, hydraulic, tethered)2016–2018: Viral parkour and backflip demonstrations showcasing dynamic balanceApril 2024: Transition to fully electric actuation — retired hydraulic versionJanuary 5, 2026: Production Atlas unveiled at CES; manufacturing begins immediately2026: All deployments fully committed — Hyundai RMAC and Google DeepMind first in line2027: Additional customers planned Production Atlas Specs (CES 2026) The production Atlas is Boston Dynamics' "best robot ever built," according to CEO Robert Playter: 56 degrees of freedom with fully rotational joints2.3 m (7.5 ft) reach — significantly exceeding human arm span50 kg (110 lb) lift capacityAutonomous battery swap: Atlas navigates to a charging station, swaps its own batteries, and returns to work — enabling continuous operationEnterprise integration: Connects to MES, WMS, and other industrial systems via Boston Dynamics' Orbit™ softwareIP67-rated (dustproof, waterproof) and rated for -20°C to 40°C operationSafety features: Human detection, fenceless guarding, barcode/RFID integrationAI partnership with Google DeepMind: Foundation models for greater cognitive capabilities Price and Availability Atlas is positioned as an enterprise-grade solution at an estimated ~$130K-$320K (unconfirmed estimates) per unit (industry estimates post-CES 2026; Boston Dynamics has not published an official MSRP). Hyundai Motor Group (Boston Dynamics' majority shareholder) announced a $26 billion U.S. investment including a new robotics factory capable of producing 30,000 robots per year. Hyundai Mobis will supply actuators, leveraging automotive supply chains for reliability and scale. Head-to-Head: Optimus vs Atlas Performance Comparison 1. Agility and Mobility Winner: Atlas Atlas is unmatched in athletic capability. It runs, jumps, performs backflips, vaults obstacles, and navigates rough terrain with superhuman agility. Its hydraulic heritage (now electric) gives it explosive power that Optimus simply doesn't match. Optimus walks smoothly at up to 8 km/h and handles flat indoor environments well, but it's not designed for acrobatics or extreme terrain. 2. Dexterity and Manipulation Winner: Optimus (hands) / Atlas (reach + strength) Optimus's hands have 11 degrees of freedom each, enabling delicate tasks like egg handling and shirt folding. Atlas counters with a massive 2.3 m reach and 50 kg lift capacity across 56 total DoF. For fine manipulation, Optimus leads. For heavy industrial tasks requiring reach and strength, Atlas wins. 3. AI and Software Winner: Tie (different strengths) Tesla brings vision-first neural networks trained on billions of real-world driving miles — excellent for generalization from data. Boston Dynamics brings decades of real-time control expertise, now supercharged by a Google DeepMind partnership integrating foundation models. Both have fleet learning. Tesla's approach favors data scale; Boston Dynamics' approach favors precision control. 4. Battery and Uptime Winner: Atlas This might surprise you. While Optimus has a 2.3 kWh battery rated for "a full day of work," Atlas has a game-changing feature: autonomous battery swapping. Atlas navigates to a charging station, swaps its own battery pack, and gets right back to work. This means effectively unlimited uptime with no human intervention — a massive advantage for 24/7 industrial operations. 5. Ruggedness and Durability Winner: Atlas Atlas is water-resistant, operates from -20°C to 40°C, and is built with automotive-grade components from Hyundai Mobis. It's designed for factory floors, construction sites, and harsh environments. Optimus is lighter and more consumer-oriented — excellent for climate-controlled environments but less battle-tested in extreme conditions. 6. Price and Accessibility Winner: Optimus At $20,000–$30,000 target vs. $150K-$320K, Optimus is roughly one-fifth to one-tenth the cost of Atlas. This makes Optimus accessible to small businesses, warehouses, and eventually homes. Atlas is an enterprise investment suited for large manufacturers and specialized operations. 7. Manufacturing Scale Winner: Tie (converging) Tesla's manufacturing DNA is a genuine advantage — Optimus is designed for mass production from day one. However, Hyundai's $26 billion U.S. investment and 30,000-unit/year robotics factory means Atlas is rapidly closing the manufacturing gap with automotive supply chain backing. Best Use Cases: Who Should Choose Which Robot? Choose Tesla Optimus If You Need: Factory line work: Sorting, packing, light assembly in controlled environmentsWarehouse operations: Inventory management, order picking, logisticsHousehold assistance: Domestic chores, elderly care, daily tasks (future)Budget-conscious deployment: Multiple units at lower cost per robotAI-first learning: Tasks that benefit from neural network generalization Choose Boston Dynamics Atlas If You Need: Heavy industrial tasks: Automotive manufacturing, material handling, construction24/7 continuous operation: Auto battery-swap for non-stop uptimeHarsh environments: Extreme temperatures, wet conditions, rough terrainEnterprise integration: MES/WMS connectivity, Orbit™ fleet managementSearch and rescue: Disaster zones, hazardous inspectionsHigh-value precision work: Tasks requiring 56 DoF and 7.5 ft reach 2026 Outlook: What's Coming Next Tesla Optimus Roadmap 2026: Next-generation Optimus expected with significant upgrades2026: Continued internal Tesla factory deployment, expanding task range2027+: Potential external sales and consumer-facing applicationsLong-term: Musk envisions Optimus as "Tesla's most significant product" — potentially exceeding vehicle business value Boston Dynamics Atlas Roadmap Q1–Q2 2026: First production Atlas units ship to Hyundai RMAC and Google DeepMind2026: All deployments fully committed; AI foundation models from Google DeepMind integratedEarly 2027: Additional customers onboardedLong-term: CEO Robert Playter's vision — "useful robots that walk into our homes and help make our lives safer, more productive, and more fulfilling" Ethical and Social Implications Job Displacement vs. Job Creation Both robots will automate tasks currently performed by humans. This raises legitimate concerns about job displacement, particularly in manufacturing and logistics. However, both companies emphasize that humanoid robots will also create new roles in robot maintenance, programming, fleet management, and AI training. The net effect on employment remains hotly debated. Safety and Regulation As these robots move from labs to factory floors, safety becomes paramount. Atlas includes human detection and fenceless guarding. Both companies face evolving regulatory frameworks. The key question: how do you certify a robot that learns and adapts on its own? Standards bodies worldwide are working on this, but regulations lag behind the technology. The Verdict: Optimus vs Atlas — Who Wins? There's no single winner. Optimus and Atlas target different markets at different price points: For mass-market affordability and consumer robotics: Tesla Optimus wins. Its $20K–$30K target price, lightweight design, and Tesla's manufacturing scale make it the robot most likely to enter millions of homes and small businesses.For enterprise-grade industrial deployment: Boston Dynamics Atlas wins. Its 56 DoF, autonomous battery swap, extreme durability, Google DeepMind AI, and Hyundai-backed production make it the robot that Fortune 500 manufacturers will bet on. The real story of 2026 isn't which robot is "better" — it's that both are now moving from prototypes to production. Humanoid robots are no longer science fiction demos. They're shipping to real factories, learning real tasks, and beginning to reshape how work gets done. --- # The Evolution of Humanoid Robots from Science Fiction to Reality URL: https://blog.robozaps.com/b/evolution-of-humanoid-robots-from-science-fiction Author: Radica Maneva Published: 2025-04-09T00:00:00.000Z Category: insights TL;DR: Humanoid robots evolved from myth and fiction to real prototypes like Honda's ASIMO and on to today's commercial platforms. Advances in AI, hardware-software integration and batteries are accelerating the field, with the market projected to reach $13.8 billion by 2028. Key takeaways: - Humanoid robots have transitioned from ancient myths and early literary visions to significant real-world prototypes like Honda’s ASIMO and Boston Dynamics’ Atlas, showcasing remarkable progress towards lifelike robots. - Technological advancements such as advanced AI, seamless hardware-software integration, and improved battery life are driving the capabilities and versatility of modern humanoid robots, making them more adaptable and efficient. - The global humanoid robot market is experiencing rapid growth, projected to reach USD 13.8 billion by 2028, driven by increasing applications in healthcare, education, and various industries, alongside continuous innovations from key market players. FAQ: Q: What are the origins of humanoid robots in literature? A: The origins of humanoid robots in literature can be traced back to ancient myths and early science fiction works, such as Greek myths and Mary Shelley’s ‘Frankenstein’. These stories laid the foundation for the development of humanoid robots in literature. Q: Who are the key players in the humanoid robot market? A: The key players in the humanoid robot market include Boston Dynamics, Hanson Robotics, and SoftBank Robotics. These companies are leading the way in developing advanced humanoid robots. Q: What are the major technological constraints in developing humanoid robots? A: The major technological constraints in developing humanoid robots include dynamic balancing, energy efficiency, and real-time visual recognition. These factors pose significant challenges to the development of advanced humanoid robotics. Q: How are humanoid robots used in healthcare? A: Humanoid robots are used in healthcare for caregiving, patient monitoring, and rehabilitation therapies, providing crucial support to medical professionals and improving patient care. Q: What is the future potential of humanoid robots? A: Humanoid robots have a bright future with possibilities for more lifelike interactions, integration into various industries, and significant market growth. ‍ Related: The Future of Humanoid Robots: Innovation and Impact · What Is a Humanoid? Definition and Examples Ready to buy? Browse humanoid robots for sale on Robozaps. The evolution of humanoid robots from science fiction to reality has been a fascinating journey. This article explores their progression from imaginative roots in literature and film to today’s advanced technologies. We’ll look at key milestones, technological innovations, and current applications that have turned the evolution of humanoid robots from science fiction to reality. ‍ ‍ From Fiction to Reality: The Journey of Humanoid Robots Humanoid robots have long been a staple of science fiction, capturing the imagination of authors, filmmakers, and audiences. These fictional representations have significantly influenced the direction and aspirations of real-world robotics. Human ingenuity and perseverance have guided the transition of bringing humanoid robots from science fiction to reality, with every milestone pushing us closer to the moment a humanoid robot created becomes indistinguishable from lifelike robots. Early visions and tangible inspirations have shaped the transition from fiction to reality, inspiring engineers and scientists to make those dreams a reality. Key milestones in the development of humanoid robots include notable breakthroughs and prototypes that progressively brought these imagined concepts closer to reality, exhibiting advancements towards lifelike robots. Let’s delve into the fascinating stages of humanoid robots’ evolution. Early Visions in Science Fiction The origins of humanoid robots can be traced back to ancient myths and early literary works. Greek myths, such as Hephaestus’ golden handmaidens and Talos, reflect early ideas of humanoid automata, showcasing the human fascination with creating lifelike machines. Similarly, the Liezi, a 3rd century BCE Taoist text, describes a humanoid automaton created by Yan Shi, further illustrating the long-standing human desire to build robots. In literature, Mary Shelley’s novel ‘Frankenstein’ is one of the early examples where an artificial being is created, reflecting human attempts to assume the role of the creator. Edward S. Ellis introduced a bipedal anthropomorphic mechanism in his 1868 story ‘The Steam Man of the Prairies,’ depicting human-controlled mechanisms without autonomy. The first use of the word ‘robot’ was in Karel Čapek’s play R.U.R. (Rossum’s Universal Robots) in 1920, which introduced the concept of robots laboring for humans and eventually revolting. These formative ideas in science fiction established the foundation for the emergence of humanoid robotics. Real-World Inspirations Real-world inspirations and initial efforts to create humanoid robots powered the transition from science fiction to reality. Some examples include: Leonardo da Vinci conceptualized a mechanical knight in the 1400s, showcasing the blend of art and early robotics.The 13th-century engineer Ismail al-Jazari designed a humanoid waitress robot that served drinks.French inventor Jacques de Vaucanson created ‘The Flute Player’ in the 18th century, a robot capable of playing melodies. These early attempts were driven by the desire to bring lifelike robots into existence, inspired by fictional depictions. Key Milestones in Development Significant breakthroughs and prototypes have shaped the development of humanoid robots. WABOT-1, developed by Waseda University in 1972, was the first full-scale humanoid robot capable of walking and grasping. Honda’s ASIMO, introduced in 2000, marked a major breakthrough as a humanoid robot that could walk, climb stairs, and interact with people. More recently, Boston Dynamics’ Atlas robot has demonstrated remarkable agility and versatility, capable of running, jumping, and climbing. These key achievements have consistently brought the concept of humanoid robots closer to reality. ‍ Technological Innovations Driving Humanoid Robotics A blend of state-of-the-art technologies propels the progression of humanoid robotics. From artificial intelligence to seamless hardware-software integration and innovative energy solutions, these technological innovations are making humanoid robots more capable and versatile than ever before. Advanced AI capabilities play a crucial role in enhancing the performance and adaptability of humanoid robots. The seamless integration of hardware and software ensures reliability and functionality, while improved energy solutions and battery life are critical for prolonged operation. Let’s examine the crucial technological innovations propelling the advancement of humanoid robotics, particularly those that mimic human movements, within the context of a preliminary innovation system. Advanced AI Capabilities Advancements in AI and robotics have significantly enhanced humanoid robots’ capability to perform complex tasks with greater efficiency. Tesla’s Optimus robot, for example, uses neural networks to improve its motion and perform tasks autonomously. Boston Dynamics’ ATLAS, introduced in 2013, represents a significant advancement in humanoid robot agility and versatility. These robots blend AI, machine learning, biomechanics, and sensor connectivity with principles from behavioral science and cognitive development. Generative AI and large language models further enhance humanoid robots’ ability to interact and cooperate with people, making them more adaptable to unstructured situations. AI advancements in computer vision and natural language processing are crucial for enabling versatile and reliable movements in humanoid robots. These advanced AI capabilities are significantly reshaping human production and interaction. Hardware and Software Integration The seamless integration of advanced hardware and software is crucial for the reliability and overall functionality of humanoid robots. Figure 01 robot integrates dexterous hands and legs for efficient movement in tasks like climbing stairs and lifting boxes. Unitree’s H1 robot is equipped with 3D lidar and a depth camera for autonomous navigation. These developments in sensor technology enable humanoid robots to have: Better environmental awarenessMore precise physical interactionsEnhanced facial expressions, allowing for emotional interaction robotsAgility Technological convergence ensures that these robots can perform tasks with these enhanced capabilities. Energy Solutions and Battery Life Recent improvements in battery technology are crucial for the enhanced functionality and prolonged operation of humanoid robots. Innovative energy solutions are driving the efficiency and operational longevity of these robots in various applications. Robots like Boston Dynamics’ Atlas demonstrate remarkable energy efficiency, allowing complex tasks without rapid power depletion. Research into alternative energy sources or more efficient batteries is essential to overcome operational time limitations in humanoid robots. ‍ The Current Landscape of Humanoid Robots The current landscape of humanoid robots is marked by rapid market growth, driven by technological advancements and increasing demand across various industries. Some key facts about the global humanoid robot market include: The market was valued at USD 1.38 billion in 2023It is expected to reach USD 13.8 billion by 2028The market is growing at a CAGR of around 50.2%North America led the global market in 2022, driven by widespread adoption in public relations, personal assistance, caregiving, and education. Key market leaders and innovators are continuously pushing the boundaries of humanoid robot technology, while regional and global trends shape the market dynamics. Humanoid robots are being increasingly used in healthcare, space exploration, education, and other industries. Let’s explore these aspects in more detail. Market Leaders and Innovators Prominent manufacturers contributing to the humanoid robot market’s dominance in North America include: Promobot Corp. (US)WowWee Group Limited (Canada)Kindred Inc. (US)Agility Robotics (US)NASA (US) Despite technical challenges, companies like Boston Dynamics, Hanson Robotics, and SoftBank Robotics are continuously innovating in the field of humanoid robot technology. Ameca, a humanoid robot developed by Engineered Arts, uses generative artificial intelligence to respond to questions, commands, and interact with people. Institutions like Boston Dynamics and MIT’s CSAIL are at the forefront of humanoid robotics research, pushing the boundaries of what these robots can achieve in various industries. These market leaders and innovators are driving the rapid market adoption of humanoid robots across various sectors. Regional and Global Trends The Asia Pacific region holds the largest market share for humanoid robots, accounting for 34.9% of market revenue in 2023. Countries like China, India, and Japan are driving the demand for humanoid robots in this region. The Chinese government started a robot fund to support research and development in the humanoid robot sector. Some Western companies likely have the most sophisticated AI software models, while Asia will probably be the manufacturing hub for humanoid components. These regional and worldwide trends are influencing the future outlook of the humanoid robot market, as showcased in events like the World Intelligence Expo. Industry Applications Humanoid robots are increasingly used in healthcare for caregiving and patient monitoring. Some examples include: Grace, a humanoid robot developed by Hanson Robotics, is designed specifically to assist in healthcare settings, particularly for tasks related to the COVID-19 pandemic.Humanoid robots are also used to provide companionship to the elderly.They can support rehabilitation therapies. NASA’s Valkyrie robot is being developed to assist in future space missions. Humanoids are particularly appealing for tasks that are ‘dangerous, dirty, and dull,’ such as mining and disaster rescue. These industry applications demonstrate the versatility and potential of humanoid robots. ‍ Challenges and Opportunities in Humanoid Robotics The field of humanoid robotics presents both challenges and opportunities. Key technical hurdles include stable bipedal locomotion, dynamic balance, energy efficiency, robust control systems, and robotic perception. Cognitive and interaction challenges involve real-time visual recognition, human-robot interaction, and adaptive learning. Despite these challenges, the economic potential of the humanoid robot market is immense, with significant growth prospects and strategic opportunities for companies investing in this field. Technological Constraints Achieving dynamic balancing through precise synchronization of actuators, sensors, and control algorithms is crucial for humanoid robots. Managing complex control systems and addressing substantial energy demands are major technical obstacles in humanoid robotics. Control over human-like movements remains a technical challenge, with current battery technology limiting operational time to 1-2 hours before recharge. Present AI systems, including large language models, struggle with new situations that they haven’t been trained for, posing limitations. There are significant bottlenecks in the development of AI and software for robot manipulation and interaction. Real-time visual recognition capabilities are essential for humanoid robots to interact effectively with their environment. Economic Potential The economic potential of the humanoid robot market is immense. Goldman Sachs Research estimates that the global market for humanoid robots could reach $38 billion by 2035. Most of the hardware for humanoid robots is already available or close to maturity for commercial purposes, indicating a promising future for rapid market adoption. As these robots become more integrated into various industries, their economic impact will continue to grow, potentially expanding the market even further. Strategic Considerations for Companies Governments and technology companies are investing heavily in humanoid robotics research and development, leading to more advanced robots. Component makers are recruiting staff and devoting capital and human resources to humanoid robot development. Companies investing in this field must consider strategic partnerships, regulatory compliance, and market readiness to capitalize on the opportunities offered by humanoid robotics. By harmonizing their resources and strategies, these businesses can tackle the challenges and tap into the economic potential of the humanoid robot market. ‍ Future Prospects and Innovations The future outlook and innovations in humanoid robotics are both exhilarating and revolutionary. As technology continues to advance, humanoid robots are expected to become more lifelike with enhanced sensory and cognitive abilities, making them more capable of interacting naturally with humans. The market for humanoid robots, valued at $1.8 billion in 2023, is projected to exceed $13 billion within the next five years, driven by advances in AI and human-like features. These robots are anticipated to integrate more deeply into sectors such as healthcare, education, and customer service, driven by continuous advancements in technology. Emerging Technologies Emerging technologies, such as advanced neural networks and improved machine learning algorithms, are expected to significantly enhance the capabilities of humanoid robots. These technologies will enable robots to make better decisions, learn from their environment, and perform tasks with greater autonomy. The integration of these technologies will drive rapid market adoption and open up broader application scenarios across various industries. Additionally, the convergence of AI, robotics, and sensor technology will result in more engaging interactions between humanoid robots and humans. This technological convergence will enable robots to perform increasingly complex tasks and adapt to new situations, making them invaluable assets in various tech companies and industries. As these technologies persistently advance, they hold enormous potential for disruptive innovation in the humanoid robot market. Societal Impact The societal impact of humanoid robots is multifaceted, encompassing ethical considerations and contributions to sustainable development goals. Humanoid robots like Sophia from Hanson Robotics can mimic human facial expressions and interact socially, which may enhance human-robot relationships but also raise ethical considerations. Ethical considerations include AI transparency, privacy protection, and the social impact on employment. The AI for Good initiative aims to leverage AI to address the 17 Sustainable Development Goals (SDGs) set out by the United Nations, promoting inclusive growth and societal changes. Humanoid robots can promote gender equality by challenging stereotypes and providing positive role models for women and girls. Vision for the Future Imagining the future landscape of humanoid robotics, these robots are poised to revolutionize industries by providing remote work capabilities and enhancing productivity through automation. Elon Musk predicts that the market value of Tesla’s humanoid robots may eventually surpass that of its electric vehicles, indicating significant future market potential. With the potential to undertake increasingly complex tasks and sophisticated human interactions, humanoid robots are profoundly transforming human production and a range of sectors. The future landscape of humanoid robotics includes expanding into new sectors such as creative arts, where robots are already being used to create original artworks. ‍ Summary In summary, the evolution of humanoid robots from science fiction to reality is a remarkable journey that highlights human ingenuity and technological advancement. From early visions in ancient myths and literature to real-world inspirations and key developmental milestones, humanoid robots have come a long way. Technological innovations, particularly in AI, hardware-software integration, and energy solutions, have driven the advancement of these robots, making them more capable and versatile. The current landscape of humanoid robots is marked by rapid market growth, with significant contributions from key market leaders and innovators. Despite the challenges, the economic potential of the humanoid robot market is immense, with promising future prospects and innovations on the horizon. As humanoid robots continue to evolve, they are poised to revolutionize industries, enhance human-robot interactions, and contribute to societal advancements. The journey from fiction to reality is far from over, and the future of humanoid robotics holds limitless possibilities. ‍ --- # Cobot vs Industrial Robot: Which Should You Choose? URL: https://blog.robozaps.com/b/cobot-vs-robot Author: Radica Maneva Published: 2025-04-08T00:00:00.000Z Updated: 2026-08-01T23:04:20.093Z Last fact-checked: 2026-07-14T00:00:00.000Z Category: comparisons TL;DR: Choose a collaborative application when direct human interaction or frequent shared access creates value and can be validated safely. Choose a safeguarded industrial robot when speed, payload, throughput or hazardous processing dominates. Key takeaways: - Collaboration is a property of the complete robot application, not an automatic guarantee attached to an arm. - Collaborative applications fit shared tasks and frequent access; safeguarded industrial cells fit speed, payload and hazardous processes. - Every option needs a task-specific risk assessment covering the robot, tool, workpiece, layout and non-routine work. - Compare total system cost and safely achievable production, not only arm price or ease of programming. FAQ: Q: What is the main difference between a cobot and an industrial robot? A: A collaborative robot application is designed for direct interaction with people in a shared or overlapping workspace. A non-collaborative industrial robot application is designed without that direct interaction and normally relies on separation and machine safeguarding. The application and its risk controls matter more than the product label. Q: Are cobots automatically safe without guarding? A: No. Safety depends on the complete application, including the robot, end effector, workpiece, speed, force, layout and tasks such as setup and maintenance. A sharp tool, heavy part or hazardous process may require guarding or other controls even when the arm is marketed as collaborative. Q: When should you choose a cobot? A: Choose a collaborative application when workers need frequent access or task sharing, product mix and changeovers are high, the required speed and payload fit the safety case, and a validated shared-workspace design improves the process. Do not choose it solely because it appears easier to deploy. Q: When is a traditional industrial robot better? A: A safeguarded industrial robot application is usually the better fit when throughput, speed, payload, reach or hazardous tooling dominate and people do not need to share the operating space during automatic production. Separation can be simpler and safer than forcing collaboration into the task. Q: Which costs should be compared beyond the robot arm? A: Compare the end effector, sensors, fixtures, guarding, integration, risk assessment, validation, training, maintenance, changeovers, downtime and production rate. A lower arm price can still produce a higher total project cost or weaker return if the process is poorly matched. Key Takeaways A cobot is not automatically safe and an industrial robot is not automatically non-collaborative; safety belongs to the complete application.Choose collaboration when people genuinely need shared access or task interaction and the required speed, force, tooling and payload can be validated safely.Choose a safeguarded industrial robot when throughput, payload, reach or hazardous processing matter more than direct human interaction.Compare total system cost and production rate, not just arm price or programming style.Every option still needs a task-specific risk assessment, integration and validation. Cobot vs Industrial Robot: The Difference That Matters The clean distinction is between applications. OSHA describes collaborative industrial robot applications as those designed for direct interaction with workers; non-collaborative applications are designed without that interaction and are normally separated by traditional machine safeguarding. ISO likewise treats collaboration as a system or application property, not a magic quality attached to an arm. For the full definition, history, operation modes, standards, manufacturers and use cases, use What Is a Cobot? This page owns only the cobot-versus-industrial-robot choice. Cobot vs Industrial Robot Comparison Safety: The Product Label Is Not the Safety Case ISO 10218-1:2025 covers safety requirements for industrial robots, while ISO 10218-2:2025 covers their integration into applications and cells. ISO/TS 15066:2016 supplements those standards for collaborative industrial robot systems and work environments. ISO says the technical specification remains current after its 2022 confirmation, although a replacement project is under development. The OSHA industrial robot safety manual makes the practical point: the selected safety functions must match the expected contact situation. Force limiting may be relevant when contact can occur; speed-and-separation monitoring may be used when the design intends to prevent contact while the robot moves. End effectors, workpieces and the rest of the process can create hazards the arm cannot remove. Risk assessment must cover normal production and non-routine work such as programming, setup, testing, cleaning, fault recovery and maintenance. OSHA notes that many robot accidents occur during these non-routine conditions, when a person may enter the work envelope. Choose a Collaborative Application When People need frequent access to the work area or genuinely share steps with the robot.Product mix is high and changeovers are frequent.The task, tool and workpiece can meet the validated force, speed and separation strategy.The productivity benefit comes from flexible task sharing rather than maximum cycle speed.Operators, maintenance staff and safety professionals are involved in design and acceptance testing. Choose a Safeguarded Industrial Robot When High speed, payload, reach or cycle-time consistency is the primary requirement.The process involves welding, cutting, sharp tools, hot material, heavy parts or another hazard better kept away from people.Workers do not need to enter or share the operating space during automatic production.A stable, high-volume process justifies a dedicated cell and fixtures.Physical separation produces a clearer and more robust safety design than limiting the robot for shared-space work. Compare Total System Economics Do not decide from arm price. Compare end effectors, fixtures, sensors, guarding, integration, safety validation, training, maintenance, floor space, changeover time, downtime and the production rate the safe configuration can actually sustain. A collaborative application can win when flexibility and worker access avoid a large dedicated cell or reduce changeover effort. A conventional cell can win when higher safe throughput lowers cost per part. The right answer is the configuration with the best validated process outcome, not the robot with the friendliest interface. A Five-Step Decision Define the task, tool, workpiece, required cycle time and every person who may be exposed.Decide whether direct interaction or shared-space access is actually necessary.Assess hazards during production, setup, fault recovery, cleaning and maintenance.Model both solutions at their safely achievable speed, including the full integration cost.Validate the selected application and re-assess it after material task, tooling, layout or software changes. Frequently Asked Questions What is the main difference between a cobot and an industrial robot? A collaborative robot application is designed for direct interaction with people in a shared or overlapping workspace. A non-collaborative industrial robot application is designed without that direct interaction and normally relies on separation and machine safeguarding. The application and its risk controls matter more than the product label. Are cobots automatically safe without guarding? No. Safety depends on the complete application, including the robot, end effector, workpiece, speed, force, layout and tasks such as setup and maintenance. A sharp tool, heavy part or hazardous process may require guarding or other controls even when the arm is marketed as collaborative. When should you choose a cobot? Choose a collaborative application when workers need frequent access or task sharing, product mix and changeovers are high, the required speed and payload fit the safety case, and a validated shared-workspace design improves the process. Do not choose it solely because it appears easier to deploy. When is a traditional industrial robot better? A safeguarded industrial robot application is usually the better fit when throughput, speed, payload, reach or hazardous tooling dominate and people do not need to share the operating space during automatic production. Separation can be simpler and safer than forcing collaboration into the task. Which costs should be compared beyond the robot arm? Compare the end effector, sensors, fixtures, guarding, integration, risk assessment, validation, training, maintenance, changeovers, downtime and production rate. A lower arm price can still produce a higher total project cost or weaker return if the process is poorly matched. Bottom Line Use a collaborative application when human access and task sharing create measurable value and can be made safe. Use a safeguarded industrial robot application when separation allows the speed, payload or hazardous process the job requires. In both cases, the full system—not the label on the arm—determines safety and return. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # The Economic Impact of Humanoid Robots on the Job Market URL: https://blog.robozaps.com/b/economic-impact-of-humanoid-robots-on-job-market Author: Radica Maneva Published: 2025-04-08T00:00:00.000Z Updated: 2026-08-01T23:04:49.211Z Last fact-checked: 2026-07-14T00:00:00.000Z Category: insights TL;DR: There is no reliable humanoid-only job-loss number: the widely cited studies model all automation trends (WEF 2025: 170 million jobs created, 92 million displaced by 2030). Task-level evidence points to transformation over replacement, with structured physical work under earliest pressure, while robotics engineering salaries rose 25-40%. Key takeaways: - No major study produces a reliable humanoid-only job-loss count; precise "X% of jobs by 2030" claims turn broad automation scenarios into false precision. - The task is the right unit of analysis: the ILO's 2025 exposure index found one in four workers exposed to generative AI in some tasks, but concluded transformation is likelier than replacement. - The WEF's Future of Jobs 2025 estimates 170 million jobs created and 92 million displaced by 2030 — across all macro trends, not robots alone. - The raw economics favor automation in structured work: a US warehouse worker costs $35,000-$45,000 a year; a $50,000 robot with a five-year lifespan and $5,000 annual maintenance undercuts that. - Deployment creates its own jobs: robotics engineering salaries have risen 25-40% since the humanoid wave began. FAQ: Q: How many jobs will humanoid robots replace by 2030? A: There is no defensible global humanoid-only count. WEF's 170 million jobs created and 92 million displaced estimate covers several macrotrends together, while McKinsey models the share of work hours that could be automated. Neither figure says how many jobs humanoid robots will remove. Near-term impact is more likely to appear as changed tasks, unfilled vacancies covered by machines, reduced hours or selective layoffs in structured physical work. Q: Which jobs are most at risk from humanoid robots? A: Jobs involving repetitive physical tasks in structured environments face the highest risk: assembly line workers, warehouse pickers and packers, machine operators, agricultural laborers, and fast food preparation workers. Jobs requiring complex social interaction, creativity, strategic thinking, or work in highly unstructured environments remain safer. Q: Will humanoid robots create new jobs? A: They are already creating work in integration, maintenance, safety testing, fleet operations and robot training. The net effect remains uncertain because new roles may require different skills and may not appear in the same places or on the same timetable as displaced work. WEF's widely quoted net gain of 78 million jobs by 2030 covers all major labour-market trends, not humanoid robotics alone. Q: How much does a humanoid robot cost in 2026? A: Prices and commercial status vary too much for one reliable market-wide range. Some platforms have public list prices, while many high-profile systems remain pilots, development kits or private commercial agreements. See the current humanoid robot cost guide for model-by-model pricing and availability evidence. Q: What is a robot tax and could it work? A: A robot tax would levy fees on companies that replace human workers with robots, equivalent to some portion of the income taxes the displaced worker would have paid. South Korea introduced a version in 2017 by reducing tax incentives for automation investments. Proponents say it funds retraining programs and UBI; critics argue it would slow innovation and make domestic companies less competitive internationally. Q: How does Japan view humanoid robots differently from the West? A: Japan — with 29% of its population over 65 and an unemployment rate of just 2.5% — views humanoid robots primarily as a solution to labor shortages rather than a threat to existing workers. Cultural attitudes toward robots are also more positive, influenced by decades of manga and anime depicting helpful humanoid robots. Japan is pursuing humanoid robots for elder care, agriculture, and manufacturing to maintain economic output despite a shrinking workforce. Q: Will Universal Basic Income (UBI) be needed because of robots? A: No current evidence can answer that conclusively. UBI is one possible response if automation produces persistent displacement or weaker wages, but other options include wage support, unemployment insurance, portable benefits and targeted retraining. The policy should respond to measured outcomes rather than assuming either mass unemployment or a painless transition. Q: What should I do if my job is at risk from humanoid robots? A: Start by realistically assessing your role's automation risk — repetitive physical tasks in structured environments face the highest risk. Then invest in skills robots lack: complex problem-solving, social intelligence, creativity, and strategic thinking. Learning basic robotics and AI concepts can position you as a human-robot collaborator rather than a competitor. Consider retraining as a robot maintenance technician — demand for this role is projected to grow rapidly through 2035. What the Evidence Actually Says About Robots and Jobs The major labour-market studies do not produce a reliable humanoid-only job-loss count. Claims that humanoid robots will replace a fixed share of all jobs by 2030 or 2050 turn broad automation scenarios into false precision. The macro picture rests on a job-by-job model. For which specific occupations are most exposed, the order they are substituted in, and the payback maths that sets the timing, see what jobs humanoid robots will take. The more useful unit is the task, not the job title. The ILO's 2025 exposure index found that one in four workers is in an occupation with some exposure to generative AI, but concluded that transformation is more likely than wholesale redundancy because most occupations still contain tasks that need human input. That research concerns software-based AI, not physical humanoid robots, but its task-level method is the right way to think about automation risk. Physical automation is already substantial. The International Federation of Robotics recorded 542,076 industrial robot installations in 2024 and an operational stock of 4.66 million. Those figures show real industrial adoption. They should not be presented as humanoid deployments: most are conventional industrial systems built for narrow, structured tasks. The World Economic Forum's Future of Jobs Report 2025 estimates that macroeconomic, demographic, technological and environmental trends together could create 170 million jobs and displace 92 million by 2030. That is a survey-based scenario across many forces, not a forecast of jobs created or lost specifically because of humanoid robots. McKinsey separately estimates that automation could affect about 27% of current work hours in Europe and 30% in the United States by 2030. Work hours that can be automated are not the same thing as jobs that will disappear. This article therefore focuses on where humanoids could change physical tasks first, which workers may face the hardest transitions, which roles may grow around deployment, and what employers and policymakers can measure instead of relying on headline forecasts. Understanding the Humanoid Robot Revolution What Makes 2026 Different Previous waves of automation relied on specialized industrial robots — welding arms, conveyor belt systems, and pick-and-place machines bolted to factory floors. Humanoid robots represent a fundamentally different category. Their bipedal, human-shaped design means they can operate in environments built for humans, use human tools, and perform tasks across multiple industries without facility redesign. Several breakthroughs converged to make 2026 the inflection point: Cost reductions exceeding projections: Goldman Sachs reports manufacturing costs declined 40% year-over-year, far surpassing earlier projections of 15–20% annual reductions. Current costs range from $30,000–$150,000 depending on configuration.AI-powered dexterity: Large language models and vision-language-action models now enable robots to understand verbal instructions, adapt to novel situations, and handle irregular objects — capabilities that were impossible just three years ago.Battery and actuator improvements: Modern humanoid robots can operate 16–20 hour shifts before recharging, making them viable for two-shift factory and warehouse operations.Venture capital flood: Over $7 billion in venture capital flowed into humanoid robotics companies between 2023 and 2025, funding rapid iteration and deployment. Key Players Driving Deployment The humanoid robotics landscape in 2026 features several major players actively deploying robots in commercial settings: Tesla Optimus: Tesla has deployed Optimus robots in its Fremont factory for material handling and assembly assistance tasks. While the company initially targeted 5,000 units in 2025 — a goal it didn't fully achieve — Gen 3 Optimus robots are now performing meaningful work in Tesla's own manufacturing lines.Figure AI: Figure's humanoid robots are working shifts at BMW's manufacturing facility in Spartanburg, South Carolina. Their Figure 02 and Figure 03 models handle warehouse logistics, parts sorting, and repetitive assembly tasks.Agility Robotics (Digit): Amazon has partnered with Agility Robotics, deploying Digit robots in its fulfillment centers for tote-moving and shelf-stocking operations.Apptronik (Apollo): Mercedes-Benz has integrated Apollo robots into its manufacturing operations, focusing on physically demanding tasks that contribute to worker injury.Unitree Robotics: Chinese manufacturer Unitree has achieved some of the lowest production costs in the industry, making humanoid robots accessible to small and mid-sized enterprises across Asia.1X Technologies (NEO): The Norwegian company is targeting commercial security and hospitality environments with its NEO humanoid robot platform. Which Jobs Are Most at Risk Not all jobs face equal risk from humanoid robots. The most vulnerable roles share common characteristics: they involve repetitive physical tasks, operate in structured environments, and don't require complex social or creative judgment. Here's a sector-by-sector analysis. Manufacturing Manufacturing is the front line of humanoid robot adoption. Tesla, BMW, and Mercedes-Benz are already deploying humanoid robots on assembly lines. Goldman Sachs estimates humanoid robots could fill 4% of the U.S. manufacturing labor shortage gap by 2030. At-risk roles include: Assembly line workers performing repetitive tasksMaterial handlers and machine operatorsQuality inspection workers (visual AI now matches human accuracy)Packaging and palletizing workers However, context matters: many manufacturing facilities already struggle with chronic labor shortages. In the U.S., the National Association of Manufacturers reports over 600,000 unfilled manufacturing jobs as of 2025. Humanoid robots may initially fill vacancies rather than displace existing workers. Warehousing and Logistics Amazon alone employs over 750,000 warehouse workers in the United States. The company's partnership with Agility Robotics signals where the industry is heading. Warehouse work — picking, packing, sorting, and moving goods — is highly structured and repetitive, making it ideal for humanoid robot deployment. The economic incentive is stark: a warehouse worker in the U.S. earns approximately $35,000–$45,000 per year plus benefits. A humanoid robot costing $50,000 with a 5-year lifespan and $5,000 annual maintenance works out to roughly $15,000 per year — and operates 20 hours a day without breaks, sick days, or overtime pay. Agriculture Agricultural labor faces significant disruption, particularly for harvesting tasks. Seasonal crop picking, fruit sorting, and greenhouse maintenance are all tasks that humanoid robots are being trained to perform. Countries like Japan, where the agricultural workforce average age exceeds 67, see humanoid robots as essential to maintaining food production. Retail Retail environments are seeing early deployments in inventory management, shelf stocking, and customer assistance. While fully autonomous retail robots remain limited, humanoid robots handling overnight restocking shifts are already being piloted by major chains. Self-checkout has already reduced cashier jobs by an estimated 30% in stores that adopted it, and humanoid robots will extend automation deeper into retail operations. Food Service Fast food chains and commercial kitchens are exploring humanoid robots for food preparation, dishwashing, and serving. Repetitive food assembly tasks — think sandwich making or pizza preparation — are well within current humanoid robot capabilities. Several restaurant chains in China and Japan have deployed robotic kitchen assistants, though most current deployments supplement rather than replace human workers. Jobs That Are Safe — and Growing While humanoid robots threaten certain job categories, they're simultaneously creating entirely new career paths and increasing demand in existing ones. Robot Maintenance and Repair Technicians Every deployed humanoid robot requires ongoing maintenance, calibration, and repair. Industry estimates suggest one technician per 10–20 robots for currently deployed models, though this ratio will improve as reliability increases. With projections of millions of humanoid robots deployed by 2035, this represents a massive new job category. AI Training and Data Specialists Humanoid robots learn through demonstration, simulation, and real-world feedback. Training robots to handle new tasks, environments, and edge cases requires human experts. Roles include: Robot behavior trainers (teaching robots new tasks through demonstration)Simulation environment designersSafety testing and validation engineersHuman-robot interaction designers Robotics Engineers and Software Developers The robotics industry's growth is driving enormous demand for mechanical engineers, control systems engineers, computer vision specialists, and AI/ML researchers. Robotics engineering salaries have increased 25–40% since 2023 due to talent scarcity. Healthcare and Elder Care Workers Paradoxically, humanoid robots may increase demand for human healthcare workers. Robots handling routine physical tasks (patient lifting, medication delivery, facility cleaning) allow healthcare professionals to spend more time on patient interaction and complex care — activities where human empathy remains irreplaceable. Creative and Strategic Roles Roles requiring creativity, complex problem-solving, emotional intelligence, and strategic thinking remain firmly in human territory. This includes management, design, therapy, education, and the arts. The World Economic Forum consistently identifies these as the most automation-resistant job categories. The Net Job Creation vs. Destruction Debate The Optimistic View Proponents of humanoid robot adoption argue that technology has always created more jobs than it destroyed — it just creates different ones. Key arguments include: Productivity gains increase wealth: When robots reduce production costs, goods become cheaper, consumer purchasing power increases, and new industries emerge to serve expanded demand.Labor shortage solution: In aging economies like Japan, Germany, and South Korea, there literally aren't enough working-age humans to fill existing jobs. Humanoid robots may prevent economic contraction rather than cause unemployment.New industry creation: The automobile didn't just replace horses — it created gas stations, motels, suburbs, highways, drive-through restaurants, and millions of jobs that never existed before. Humanoid robots may similarly spawn industries we can't yet imagine.World Economic Forum estimate: The WEF's 2025 Future of Jobs Report projects that technology will create 170 million new jobs while displacing 92 million, for a net gain of 78 million jobs globally by 2030. The Pessimistic View Skeptics counter that humanoid robots represent something qualitatively different from previous automation waves: Speed of displacement: Previous transitions occurred over decades. Humanoid robots combined with AI could transform entire sectors within 5–10 years, leaving insufficient time for workforce retraining.Breadth of capability: Unlike a sewing machine or calculator that automates one task, a humanoid robot can perform thousands of different physical tasks. This means displaced workers can't simply move to adjacent roles — the robots can follow them there.McKinsey's stark numbers: Up to 800 million jobs displaced globally by 2030. Even if new jobs are created, the transition will cause significant economic pain for hundreds of millions of workers.Wage depression: Even workers who aren't replaced may see wages fall as the labor supply effectively increases. Employers can use the threat of robot replacement to suppress wage demands.Concentration of gains: The economic benefits of automation tend to flow to capital owners (robot manufacturers, companies deploying robots) rather than workers, potentially widening inequality. The Realistic Middle Ground Most labor economists fall somewhere between these extremes. The consensus view holds that: Humanoid robots will displace millions of jobs in specific sectors over the next decade.New jobs will be created, but they'll require different skills and may emerge in different locations.The transition period (roughly 2026–2035) will be economically painful for affected workers without proactive policy intervention.Inequality will likely worsen unless governments actively redistribute the productivity gains from automation. Historical Parallels: What Past Disruptions Teach Us The ATM Paradox When ATMs were widely deployed in the 1970s and 1980s, many predicted the end of bank teller jobs. Surprisingly, the opposite happened — at least initially. The number of bank tellers in the U.S. actually increased from 300,000 in 1970 to 600,000 by 2010. Why? ATMs reduced the cost of operating a bank branch, which led banks to open more branches, which required more tellers for customer service roles that ATMs couldn't handle. However, this story has a second chapter: since 2010, bank teller employment has steadily declined as online banking, mobile apps, and improved ATMs finally made many branches unnecessary. The lesson: technology's job impact often comes in waves, with an initial adaptation period followed by deeper disruption. The Industrial Revolution The original Industrial Revolution (1760–1840) displaced millions of artisans, handloom weavers, and agricultural workers. The transition was brutal — it took roughly 60 years before wages for ordinary workers began rising. The Luddites who smashed textile machinery weren't irrational; their livelihoods really were being destroyed, even if the long-term economic outcome was positive. The key difference with humanoid robots: the timeline is compressed. What took 60 years during the Industrial Revolution may take 10–15 years with humanoid robots and AI. Computerization and the Internet The personal computer revolution eliminated millions of secretarial, typesetting, and clerical jobs while creating entirely new industries: software development, IT support, digital marketing, e-commerce, and social media management. The U.S. economy added roughly 25 million net new jobs between 1990 and 2020 despite — or because of — computerization. However, computerization also hollowed out middle-skill jobs, contributing to wage stagnation and the decline of the middle class. Humanoid robots may intensify this pattern, creating high-skill jobs (robotics engineers) and low-skill jobs (personal services) while eliminating middle-skill physical labor. Country-by-Country Impact United States The U.S. leads in humanoid robot development (Tesla, Figure AI, Agility Robotics, Apptronik) and is likely to see early widespread deployment. Key dynamics include: Manufacturing reshoring: Humanoid robots make domestic manufacturing cost-competitive with overseas production, potentially bringing factories back to the U.S.Labor market flexibility: The U.S. labor market historically adapts faster than European markets to technological change, though this creates more short-term disruption.Limited safety net: Compared to Europe, the U.S. has weaker unemployment insurance, retraining programs, and social safety nets, making the transition harder for affected workers.Regional impact: Manufacturing-heavy states in the Midwest and South will be most affected. China China is aggressively pursuing humanoid robotics, with companies like Unitree, UBTECH, and Fourier Intelligence developing affordable platforms. The Chinese government has identified humanoid robots as a strategic priority: Demographic crisis: China's working-age population is shrinking by 7–8 million per year. Humanoid robots are seen as essential to maintaining economic output.Manufacturing dominance: China aims to maintain its position as the world's manufacturing hub despite rising wages, with robots keeping costs competitive.State-directed deployment: Unlike market-driven adoption in the U.S., China's government can direct large-scale deployment across state-owned enterprises.Cost advantage: Chinese humanoid robots are being produced at significantly lower costs, potentially $15,000–$30,000 per unit. Japan Japan presents a unique case — an aging society that has embraced robotics not with fear but as a necessity: Elder care: With 29% of the population over 65, Japan sees humanoid robots as essential for elder care, addressing a severe shortage of caregivers.Cultural acceptance: Japanese society has a more positive cultural attitude toward robots, partly influenced by manga and anime depictions of helpful humanoid robots.Labor shortage solution: Japan's unemployment rate is just 2.5%, and the country faces labor shortages across nearly every sector. Robots fill gaps rather than displace workers.Goldman Sachs estimates humanoid robots could address 2% of Japan's elderly care demand by 2030, a crucial contribution given the demographic trajectory. European Union The EU is taking a more cautious, regulatory-focused approach: AI Act implications: The EU's AI Act (effective 2025) imposes strict requirements on AI systems used in workplaces, including humanoid robots. This may slow deployment compared to the U.S. and China.Worker protections: Strong labor unions and employment protection laws in countries like Germany, France, and Italy will slow but not prevent humanoid robot adoption in manufacturing.Robot tax proposals: Several EU countries have discussed taxing companies that replace human workers with robots, using revenue to fund retraining and social programs.Germany as test case: As Europe's manufacturing powerhouse, Germany is seeing early humanoid robot deployments (BMW, Mercedes-Benz) that will set the pattern for the continent. Wage Effects and Inequality Concerns Even before humanoid robots physically replace workers, their existence affects wages and bargaining power across the economy. Direct Wage Pressure In sectors where humanoid robot deployment is imminent, employers gain leverage in wage negotiations. Why offer raises to warehouse workers when a robot alternative is 18 months away? This "shadow of the robot" effect may depress wages even for workers who aren't immediately replaced. Historical data supports this concern: MIT economist Daron Acemoglu has documented that each industrial robot deployed in the U.S. between 1990 and 2007 replaced approximately 3.3 workers and reduced wages in affected communities by 0.4%. Humanoid robots, being far more versatile, could amplify these effects significantly. The Inequality Amplifier The economic gains from humanoid robot deployment will not be distributed evenly: Capital owners benefit most: Companies deploying robots see productivity gains and cost savings flow directly to profits and shareholder returns.High-skill workers benefit: Engineers, AI researchers, and managers who design, deploy, and oversee robots will see increased demand and wages.Low-skill physical workers lose: Workers in manufacturing, warehousing, agriculture, and food service face displacement and wage pressure.Geographic concentration: Robot deployment and the high-paying jobs it creates will cluster in tech hubs, while job losses will spread across manufacturing regions and rural areas. Oxford Economics estimates that increased robot adoption could contribute to a widening of the wage gap between the top 10% and bottom 50% of earners by 5–12% over the next decade without policy intervention. UBI Is a Policy Choice, Not an Automatic Result The evidence does not settle the case for or against universal basic income. Humanoid deployment may strengthen arguments for cash support if it causes sustained job loss or pushes wages down, but today's pilots and labour-market studies cannot show that humanoids will make UBI necessary. Funding, eligibility and the interaction with existing services remain political choices. Skills investment matters, but training alone is not a complete transition policy. Governments and employers can also use wage insurance, unemployment support, portable benefits, hiring reform and enforceable safety rules for human-robot workplaces. The right mix depends on whether a deployment fills vacancies, changes tasks, reduces hours or removes jobs. What Policymakers Should Measure Useful measures include actual task changes, vacancies versus layoffs, wage and hour changes, safety incidents, retraining outcomes and who receives the productivity gains. The ILO's 2026 review of empirical evidence found that large-scale displacement from generative AI remained limited and that reported time savings had not yet translated into broad changes in output, earnings or employment. Humanoid robots need the same evidence-first treatment: observed workplace outcomes should carry more weight than vendor targets or distant scenarios. Jobs at Risk vs. Jobs Created: Comparison Timeline: When Will Different Sectors See Major Robot Adoption? Preparing for the Transition: What Workers Can Do Now Individual workers don't have to wait for government policy to prepare for the humanoid robot economy. Here are actionable steps: Assess your role's automation risk: If your job primarily involves repetitive physical tasks in a structured environment, begin exploring adjacent career paths now.Develop complementary skills: Focus on skills that robots currently lack — complex social interaction, creative problem-solving, emotional intelligence, and strategic thinking.Learn basic robotics and AI concepts: Understanding how robots work — even at a basic level — makes you more valuable as a collaborator rather than a competitor with automation.Consider the maintenance angle: Electricians, mechanics, and technicians who add robotics training to their skill sets will be in very high demand.Build a financial buffer: Economic transitions create uncertainty. Having 6–12 months of savings provides flexibility to retrain or relocate if needed. The Bottom Line Humanoid robots are a credible automation technology, but the job-market evidence is still much stronger for conventional industrial robots and software-based AI than for general-purpose humanoids. Current pilots show where the technology might fit; they do not prove a fixed timetable for mass displacement. The most defensible forecast is narrower: repetitive physical tasks in structured settings face the earliest pressure, new technical and operational roles will grow around deployment, and the costs of transition will be uneven. Track tasks, wages, hours, vacancies, layoffs and safety outcomes. Those measures will tell us more than claims that a particular percentage of jobs is destined to vanish. Compare on RoboZaps: Figure 03, Tesla Optimus, Figure 02 and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree H1 Humanoid Robot: ~$90K Price, Specs & Speed Record (2026) URL: https://blog.robozaps.com/b/unitree-h1-review Author: Radica Maneva Published: 2025-04-08T00:00:00.000Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: reviews TL;DR: The Unitree H1 (2023) is the company-claimed fastest full-size humanoid: 3.3 m/s official, a 10.1 m/s peak from a modified unit in April 2026, plus two golds at the 2025 World Humanoid Robot Games. Still listed at $90,000 "contact us for the real price", but the $29,900 H2 has succeeded it; new buyers should look there. Research legacy: Stanford HumanPlus, CMU H2O, the 16-robot Gala dance. Key takeaways: - The H1 is Unitree's first full-size humanoid (August 2023) and still carries the company-claimed 3.3 m/s speed record; Guinness has never certified an H1 record, and the certified bipedal sprint mark belongs to Cassie (2022). - The $90,000 store listing is a quote anchor, not a checkout price ("Contact us for the real price"). Launch reporting in 2023 said under $90,000; the $150,000 figure on aggregators contradicts every contemporaneous report. - Official specs: about 180 cm and 47 kg, with per-limb DoF counts on the spec page totaling 18 and Unitree's official H1 URDF defining 19 actuated joints including a torso joint, 0.864 kWh battery, 3D lidar plus depth camera, Intel i5/i7 compute. - The H1's real track record is research and the stage: Stanford HumanPlus and CMU H2O were built on it, and 16 H1s danced the Yangge on the 2025 CCTV Spring Festival Gala, after which Unitree reported an order surge. - In 2026 the H1 is legacy in fact if not in name: the H1-2 variant is delisted and sold out, the 2026 Gala went to G1s and the H2, and the $29,900 H2 is what new buyers should price. "End-of-life March 2026" claims remain unverified. FAQ: Q: How much does the Unitree H1 cost? A: The official store lists the H1 at $90,000 with the disclaimer "Contact us for the real price": sales are quote-based and enterprise-only, with the final number depending on configuration. Launch coverage in August 2023 said the expected price was under $90,000 (the $150,000 sometimes cited is an aggregator error). Distributors sell H1-family configurations at roughly $100,000 to $129,000, and the H1-2 variant listed at $128,900 before selling out. Q: How fast is the Unitree H1? A: Unitree's spec page claims 3.3 m/s as a world record for a full-size humanoid, set on video in March 2024, with "potential mobility" over 5 m/s that has never been demonstrated as sustained running. In April 2026 a modified, roughly 62 kg H1 hit a 10.1 m/s peak in a 100 m sprint video, with Unitree itself conceding a measurement error was possible. The H1 also won 1500 m gold (6:34.40) and 400 m gold (1:28.03) at the 2025 World Humanoid Robot Games. Q: Did the Unitree H1 set a Guinness World Record? A: No. The H1's 3.3 m/s is a company-claimed, video-based record that Unitree prints on its own spec page. Guinness's certified record for a bipedal robot sprint remains Cassie (Agility Robotics/Oregon State), which ran 100 m in 24.73 seconds in 2022. Outlets that wrote "Guinness" about the H1 were being loose with the term. Q: Is the Unitree H1 still being sold? A: It is still listed on the official Unitree store as of July 17, 2026 at its $90,000 contact-us anchor, but the shop's own product data marks it unavailable for direct online purchase; sales route through Unitree's sales team. A claim that the H1 reached end-of-life in March 2026 circulates on one aggregator with no Unitree announcement behind it. The verifiable signals are softer: the H1-2 variant is delisted and sold out, the H1 skipped the 2026 Spring Festival Gala, and Unitree's new platform announcements are all H2-first. Q: What is the Unitree H1-2? A: The manipulation-focused variant: about 178 cm and 70 kg, six degrees of freedom per leg, seven per arm plus a 220 N·m waist joint, five-finger hand support with roughly 7 kg rated arm load, and under 2 m/s walking speed. It trades the H1's sprint speed for manipulation. It appeared around 2024 (Unitree never published a dated announcement), was carried by distributors at $128,900, and has since disappeared from Unitree's own store. Q: Should you buy a Unitree H1 or an H2? A: For any new purchase, the H2: 31 degrees of freedom against the H1's 18 to 19, modern compute, a $29,900 list price a third of the H1's quote anchor, and the ecosystem momentum, including NVIDIA choosing the H2 Plus as its Isaac GR00T reference humanoid. The H1 only makes sense for labs that need hardware continuity with the H1 research codebase, or on the used market. Our Unitree H2 review has the full comparison. Q: How many degrees of freedom does the Unitree H1 have? A: Two manufacturer-sourced answers exist. Unitree's spec page lists per-limb counts totaling 18: five per leg (three hip, one knee, one ankle) and four per arm, with no torso row. Unitree's official H1 URDF engineering file defines 19 actuated joints, including a torso joint. Configured research units drift far higher: Stanford's HumanPlus H1, with added wrists, hands and head cameras, reached 33 degrees of freedom and 60 kg. Q: Can you program the Unitree H1? A: Yes, and that is its enduring strength: the i7 compute tier is designated for user development, and the H1 is one of the most-published research platforms among full-size humanoids, carrying Stanford's HumanPlus and CMU's H2O teleoperation work, plus default status in many RL simulation benchmarks. One caveat: the 2025 UniPwn disclosure (CVE-2025-60250/-60251) showed a wormable Bluetooth vulnerability across Unitree robots including the H1, with no public H1 fix announced since. Q: Was the Unitree H1 in the Spring Festival Gala? A: Yes. Sixteen H1 robots in floral vests performed the Yangge folk dance with handkerchief throw-and-catch on the CCTV 2025 Spring Festival Gala, choreographed into Zhang Yimou's "Yang BOT" program. Unitree reported an order surge afterward. The 2026 Gala moved on: that show featured G1s and the H2, with no H1 on stage. Q: Is the Unitree H1 dangerous? A: The viral H1 incidents mostly trace to control policies running outside their assumptions, not emergent behavior. Unitree called the February 2025 Tianjin festival lunge an accident, most likely caused by a software setting or a malfunctioning sensor. A May 2025 factory video of a hoisted H1 thrashing was never officially explained. A separate July 2025 clip came with the handler's own explanation: a full-body locomotion policy ran while the robot hung off the ground. The practical read: a 47 kg sprinting robot with no fenceless-operation safety case, plus an unpatched wormable Bluetooth vulnerability (UniPwn), belongs behind barriers and off untrusted networks. What is the Unitree H1, and can you still buy one? The Unitree H1 is the robot that made Chinese humanoids impossible to ignore: Unitree's first full-size humanoid, announced in August 2023 (The Robot Report), holder of the company-claimed speed record for a full-size humanoid, the star of the 2025 Spring Festival Gala's robot Yangge dance, and the platform under research that shaped the field, from Stanford's HumanPlus to CMU's H2O. In 2026 it occupies a stranger position: still listed on Unitree's official store at $90,000 with the disclaimer "Contact us for the real price" (shop.unitree.com), while its own successor, the Unitree H2, lists at $29,900. This page is the full H1 record: verified specs, the real price situation, the speed-record timeline, the research legacy, the incidents, and the honest 2026 buying answer. One disclosure up front: this is desk research built on the sources linked throughout, official wherever they exist, not a hands-on test. The misinformation audit: what circulates vs the record The speed record, dated: from 3.3 m/s to a contested crown Speed is the H1's identity, and the timeline deserves precision because almost nobody reports it precisely. Timeline sources: the 3.3 m/s claim and Atlas comparison via Interesting Engineering and Unitree's spec page; both Games times via The Robot Report, with the 1500 m win also covered by SCMP; the 10.1 m/s sprint and its caveats via heise. The honest headline: the H1 remains the fastest thing Unitree has publicly demonstrated on legs, its 1500 m and 400 m golds are real competition results, and none of it is Guinness-certified: the certified bipedal 100 m record still sits with Cassie at 24.73 s from 2022 (Guinness World Records). The price and ordering situation, in full The official store lists "Unitree H1（Contact us for the real price）" at $90,000, with a contact-sales flow, no spec sheet attached, and the shop's own product data marking the listing unavailable for direct online purchase. Decode that: the $90,000 is a placeholder anchor, not a checkout price, and orders route through Unitree's sales team. H1 sales are quote-based and enterprise-only, with the final number depending on configuration (compute tier, hands, support). That has been true since launch: the August 2023 announcement coverage said "below $90,000" as the expected price, and the $150,000 figure that circulates on aggregator sites contradicts every contemporaneous report. Around the official channel sits a distributor layer that does publish numbers: H1-family configurations at roughly $100,000–$129,000 through resellers like RoboStore and RobotShop, with the manipulation-focused H1-2 at $128,900 on RoboStore, showing sold out. Unitree's own humanoid lineup page tells the sharper story: it now runs R1 from $4,900, G1 at $13,500, H2 at $29,900, H2 Plus at $100,000, and the H1 at its $90,000 contact-us anchor, with the H1-2 absent entirely (official store lineup, fetched July 17, 2026). Is the H1 being wound down? The claim "end-of-life March 2026" circulating on one aggregator is unverified: no Unitree announcement, no credible reporting, and the robot is still listed. The verifiable signals are softer but consistent: the H1-2 has been delisted from the official store and is sold out at the distributor that carries it, the H1 skipped the 2026 Spring Festival Gala (that show belonged to dozens of G1s and the H2), and every new platform announcement (the robot "App Store," the NVIDIA partnership, the flagship launches) is H2-first. Legacy status in fact, if not in name. Current pricing and verification status live in the H1 record in our robot database. Official specs vs the reported layer Everything in the "official" column below is from unitree.com/h1, fetched July 17, 2026; the reported layer is flagged. One configured-weight reality check: Stanford's HumanPlus research H1, configured with added wrists, hands and head cameras, weighed 60 kg with 33 DoF (Stanford, June 2024). The stock 47 kg figure drifts fast once a lab actually equips the robot, which is worth knowing before comparing it against the 70 kg H1-2 or H2. The track record: where H1s actually worked The H1 has no disclosed commercial deployments of the kind Agility's Digit racks up. Its constituency is research and the stage, and both are real. Track-record sources: the H2O and HumanPlus papers (CMU, Stanford); the Gala's 16-robot count via SCMP, with the handkerchief throw-and-catch and lidar detail via New Atlas; the inquiry surge via TechNode. Two attribution guards, because these get miscredited to the H1 constantly: the robots that struggled through the April 2025 Beijing half-marathon were G1s run by independent teams (Unitree publicly distanced itself), and the Hangzhou kickboxing bouts were G1s too. The H1's public record is the dance, the track, and the literature. The incidents, honestly Three viral H1 videos deserve the explanations the headlines skipped. In February 2025, at a Lunar New Year festival event in Tianjin, a costumed H1 lunged toward spectators behind a barrier after one extended a hand; Unitree called it "an accident," most likely caused "by a software setting or a malfunctioning sensor" (Oddity Central, which carries the company statement). In May 2025, factory footage showed a hoisted H1 thrashing violently near two handlers; that one was never officially explained, with New Atlas noting Unitree had not confirmed the circulating theories. And in July 2025, a separate clip of a crane-suspended H1 lashing out came with the most instructive explanation of the three, from REK's Cix Liv, who posted the footage: "we ran a full body policy while the feet weren't touching the ground. Don't do that" (Interesting Engineering). The pattern reads as control policies running outside their assumptions rather than anything emergent, though only the July case has a stated cause. Either way, the practical takeaway holds: a 47 kg machine that sprints has no fenceless-operation story, and was never marketed as having one. Security and policy: what an H1 buyer inherits Two 2025–2026 developments apply to the whole Unitree fleet, H1 included. First, the UniPwn disclosure (September 2025): CVE-2025-60250 and CVE-2025-60251 cover "Unitree Go2, G1, H1, and B2 devices": the Bluetooth Wi-Fi-configuration interface uses a hardcoded encryption key, accepts any handshake secret containing the string "unitree," and passes unsanitized input to a root shell. The researchers called it wormable: one compromised robot can infect others over BLE (IEEE Spectrum). We found no public H1-specific fix announcement as of July 17, 2026. Second, the US Department of Defense added Unitree to its Section 1260H list of "Chinese military companies" in the June 8, 2026 update (WilmerHale client alert). That is company-level, not model-level: it imposes no purchase ban on private buyers, but it creates DoD-procurement exposure, grant-review friction and institutional screening questions for US and allied labs. The considerations are identical to the H2's, covered in more depth in our H2 review. H1 vs H2 vs G1: the 2026 decision What the H1 still does best is narrow and real: it is the proven sprint platform, and one of the most-published research platforms among full-size humanoids (HumanPlus, H2O, and default status in common RL simulation benchmarks). If your lab is reproducing HumanPlus, H2O or the sim benchmarks built on the H1 model, the hardware match matters. Everything else has moved on. The H2 is the successor in every practical sense: 31 DoF against the H1's 18–19, modern compute, a $29,900 list price a third of the H1's quote anchor, and the ecosystem gravity: NVIDIA chose the H2 Plus as its Isaac GR00T reference humanoid, announced at GTC Taipei on May 31, 2026 US time (NVIDIA). The G1 at $13,500 is the small-format workhorse for labs that don't need full-size dynamics. The honest 2026 verdict: nobody should specify a new H1 at the $90,000 anchor when the H2 exists, and Unitree's own catalog implies the company agrees. The H1's market is legacy fleets, codebase continuity, and, increasingly, the used market. The full comparison table lives in the H2 review; the cross-brand picture is in our best humanoid robots guide. What to watch, dated Four threads move this page. Whether the H1 stays on the official store: delisting would be the first hard wind-down fact, and we check on every update. Unitree's Shanghai STAR-market IPO, filed March 2026, which will eventually put audited numbers under the company's claims. The H2 Plus GR00T reference units shipping to research labs around late 2026, which will pull the research constituency (the H1's last stronghold) onto new hardware. And any H1 appearance at the August 2026 World Humanoid Robot Games, which would be the aging champion's title defense. The bottom line The H1 is the most consequential humanoid Unitree has built: it made full-size Chinese humanoids real to the public (the Gala), to the research community (HumanPlus, H2O), and to the record books (3.3 m/s, then the Games golds). It is also finished as a product you should buy new: quote-only at triple the successor's price, its variant delisted, its stage handed to the G1 and H2. Respect it as what it is in 2026: the platform that carried the field for two years, still the fastest thing on the track, and the clearest example yet of how fast this industry eats its own flagships. The H1 record in our robot database tracks the verified state as it changes. --- # Figure 01 Price 2026: Was It For Sale & What Happened to It URL: https://blog.robozaps.com/b/figure-01-review Author: Radica Maneva Published: 2025-04-08T00:00:00.000Z Updated: 2026-08-01T23:05:57.612Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: reviews TL;DR: Figure 01 was never publicly for sale and never had a published price; it was superseded by Figure 02 in August 2024 after about 18 months of public life. It went from CGI render to walking hardware, landed the BMW agreement, starred in the viral March 2024 OpenAI demo, and helped raise $675M at $2.6B. The $30K-$150K figures online are unsourced guesses. Figure's current robot is the Figure 03. Key takeaways: - Figure 01 was never publicly priced or sold, with no documented commercial deployment; it was superseded by Figure 02 in August 2024, roughly 17 months after its public reveal. - Figure's engineers state the viral March 13, 2024 OpenAI demo was learned, not teleoperated; no independent audit exists, and it was a controlled single-take showcase. - Figure 01 never worked at BMW: the January 2024 agreement was an announcement, Figure 02 did the actual 11-month Spartanburg deployment, and Figure 03 works there today. - Circulating prices ($30K-$150K, ~$50K estimates) are unsourced third-party guesses; no Figure robot of any generation has ever had an announced price. - Figure has never said how many Figure 01 units existed or where they went; its own framing is 'the robot that got Figure AI started.' FAQ: Q: Was the Figure 01 robot ever for sale? A: No. Figure 01 was never publicly offered for sale and never had a published price, with no documented customer deployment identified. No Figure robot of any generation has ever had an announced price or a public order channel. Q: How much is a Figure 1 robot? A: There is no price. Figure has never announced pricing for Figure 01 or any successor. The $30,000-to-$150,000 ranges and roughly-$50,000 estimates you may see online are unsourced third-party guesses, not company figures. Q: Is the Figure 1 robot real? A: Yes. 'Figure 1' and 'Figure 01' are the same real robot, built by Figure AI. It walked publicly in October 2023 and performed the widely viewed OpenAI-powered demo in March 2024 before being superseded by Figure 02 in August 2024. Q: Was the Figure 01 OpenAI demo faked or teleoperated? A: Not according to Figure's documented account: engineer Corey Lynch stated all behaviors were learned, not teleoperated, with an OpenAI model handling speech and reasoning while Figure's neural policies ran the manipulation. No independent technical audit was published, and it was a controlled single-take showcase in a prepared scene. Q: What happened to Figure 01? A: It was superseded by Figure 02 in August 2024 and retired from Figure's lineup. Figure has never disclosed how many units existed or where they went; the company's own retrospective calls it 'the robot that got Figure AI started.' Q: Did Figure 01 work in BMW factories? A: No. The January 2024 BMW agreement was an announcement of intent; no Figure 01 did documented production work. Figure 02 performed the actual Spartanburg deployment through late 2025, and Figure 03 arrived in June 2026 for a parts-sequencing project. Q: Can I buy a Figure AI robot? A: Not as of July 2026. No Figure robot has ever been retail-available, and there is no public price or order channel for the current Figure 03, which is deployed with commercial partners and in pilot programs. Any site offering one for sale is not an official channel. Q: What is the difference between Figure 01, Figure 02 and Figure 03? A: Figure 01 (2023-2024) was the proof-of-concept that walked, made coffee and ran the OpenAI demo. Figure 02 (2024-2025) was the first to do real factory work at BMW Spartanburg. Figure 03 (unveiled October 2025) is the current production robot, built at Figure's BotQ facility and powered by its in-house Helix AI. Q: How many Figure 01 robots were built? A: Figure has never disclosed a unit count for Figure 01, along with its price, compute platform and actuator details. Any specific number you see is speculation. Figure 01 was never publicly for sale, never had a published price, and was quietly superseded when Figure unveiled its second-generation robot in August 2024. That is the direct answer to both halves of this page's question. But calling it discontinued undersells what actually happened: in roughly 18 months of public life, Figure's first humanoid went from a CGI render to walking hardware, landed the BMW agreement, starred in one of the most widely covered robotics demos of 2024, and helped raise $675 million at a $2.6 billion valuation. It was retired the way a successful prototype is retired. The robot Figure builds and deploys today is the Figure 03. This page is the retrospective: the full dated record of what Figure 01 was, what it really did and did not do, where the fake price figures come from, and what became of it. First, the name: Figure 01, "Figure 1", and Figure the company Figure AI styles its first robot "Figure 01", sometimes "F.01". Most searchers, and Google's own suggested questions, write "Figure 1", which is the same machine. Neither should be confused with Figure the fintech company, which has nothing to do with robotics, or with the company Figure AI itself, whose story we cover separately in our Figure AI company review. This page is about the robot. The Figure 01 timeline, dated Sources for every row are linked in the sections below and in the reference list; the pace is the story. Boston Dynamics spent over a decade on Atlas. Figure went from nothing to a walking, task-performing humanoid in under two years, and the capital that built its successors followed that pace. The viral demo: what was real The March 13, 2024 video, "OpenAI Speech-to-Speech Reasoning", is the reason most people know this robot. A man asks for something to eat; Figure 01 hands him an apple, and when asked why, replies that it was "the only edible item I could provide you with", all while collecting trash and racking dishes. The "it's staged" and "it's teleoperated" accusations followed within hours, partly because the robot's synthesized voice said "uh" like a person. The company's technical account addresses them directly: an OpenAI multimodal model handled the speech, vision and reasoning; Figure's own learned neural policies ran the manipulation, mapping onboard camera images to actions at 10 Hz; and Figure engineer Corey Lynch stated that "all behaviors are learned (not teleoperated)" and shown at normal speed, which CEO Brett Adcock confirmed. That is the company's documented account; no independent technical audit of the demo was ever published. The honest caveat is different: it was a controlled, single-take showcase in a prepared scene, not a robot improvising in the wild. Real, and also carefully staged in the theatrical sense. One warning for readers in 2026: Elon Musk publicly questioned the autonomy of a much later Figure demo; whatever you make of that exchange, it is about a different robot and a different video, and does not apply retroactively to Figure 01. Specs on record, and what Figure never disclosed The first five rows come from Figure's original spec page, since removed from its site and preserved in web archives; the 41-degrees-of-freedom figure comes from The Robot Report's October 2023 technical breakdown. Treat them as the historical record rather than a live datasheet. Just as telling is what Figure never disclosed for this robot: a price, the compute platform, actuator details, battery capacity, or how many Figure 01 units were ever built. Every one of those blanks has since been filled by third-party guesswork somewhere on the internet. The price truth Figure 01 never had a published price, was never publicly offered for sale, and no documented customer deployment of it has been identified. Neither, for the record, has any Figure robot since: the company has never announced a price for any generation, and there has never been a public order channel. The "$30,000 to $150,000" ranges in buying guides are unauthorized third-party estimates, the separate "roughly $50,000" figure traces to a reported future production-cost goal rather than a Figure 01 selling price, and pages that show a request-a-quote button next to Figure 01 specs are implying a purchasability that never existed. If a Figure robot ever gets a price, it will come from figure.ai, and it will be news. The BMW record: which Figure robot actually worked there The January 2024 BMW agreement is permanently attached to Figure 01's story, and it deserves precision, because some spec sites still claim Figure 01 "was deployed at BMW's Spartanburg plant". It was not. The agreement was an announcement of intent, milestone-based, with no robot on site. The robot that actually went to Spartanburg was Figure 02, first in a multi-week trial in August 2024, then in a deployment Figure itself later documented: over what Figure describes as roughly 11 months on the X3 body shop line, the fleet supported production of more than 30,000 vehicles, loaded over 90,000 parts and logged 1,250-plus runtime hours against Figure's stated target of over 99% task success per shift, before being retired in late 2025. The newest arrival is Figure 03, which came to Hall 52 in June 2026 for a parts-sequencing project BMW announced that month, with an initial logistics demonstration on record. Figure 01 signed the deal, so to speak; its successors did the work. The OpenAI arc, from partner to ex Figure 01's fame is inseparable from OpenAI, and that relationship has a clean three-act structure. The OpenAI Startup Fund invested in the Series B, and a collaboration agreement was signed on February 29, 2024. Thirteen days later came the apple demo, the partnership's peak and the moment that made the robot famous. Then, on February 4, 2025, Adcock ended it, announcing on X: "Today, I made the decision to leave our Collaboration Agreement with OpenAI. Figure made a major breakthrough on fully end-to-end robot AI, built entirely in-house." The post closed by promising something "no one has ever seen on a humanoid" within 30 days, and sixteen days later Figure unveiled Helix, its in-house vision-language-action model, which has powered every Figure robot since. His stated logic: to solve embodied AI "you have to vertically integrate robot AI", and large language models had become "the smallest piece of the puzzle". The robot that made OpenAI-plus-Figure famous did not survive to see the breakup. What actually happened to the Figure 01 units? Here is the honest answer nobody else will give you: it is not public. Figure has never said how many Figure 01 units existed, whether any still run in its labs, or where they went. There was no retirement announcement and no museum donation press release. The company's own framing is the epitaph on its YouTube retrospective: "Figure 01: the robot that got Figure AI started." Any page claiming to know the fate of specific units is guessing. Where Figure is now The company Figure 01 started is, as of July 2026, one of the most valuable private robotics companies on earth: a $39 billion valuation after its late-2025 round, the Helix model family in its second generation, and Figure 03 in production at its BotQ facility, which Figure said in April 2026 had delivered more than 350 robots and accelerated from one robot a day toward one per hour. Reports point to a Figure 04 with a locked architecture and a third-generation Helix in development, though neither has been officially announced. For the current state of the product line, see our Figure 03 review and the full Figure AI company review; for how it stacks up against the field, our best humanoid robots ranking covers the landscape. The bottom line Figure 01 is the rare robot whose discontinuation was a promotion. It was never a product: never publicly priced or sold, with no documented factory deployment. It was a proof, built in under two years, that a startup could do dynamic walking, learned manipulation and a live language interface on one machine, and it converted that proof into the BMW relationship, the OpenAI moment and the capital that built Figure 02 and 03. If you came here to buy one: you never could, and the prices you have seen are inventions. If you came here wondering what happened to it: it worked, and then its company outgrew it. Compare on RoboZaps: Figure 01, Figure 03, Figure 02 and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Sophia Robot 2026: What It Is, Price & What Happened to Her URL: https://blog.robozaps.com/b/sophia-review Author: Radica Maneva Published: 2025-04-07T00:00:00.000Z Updated: 2026-08-01T23:05:22.831Z Last fact-checked: 2026-07-20T00:00:00.000Z Category: reviews TL;DR: Sophia is Hanson Robotics' famous social humanoid: a character robot sold only by inquiry, not a retail product, and she is still active in 2026. No public price has ever been published, and the $100,000-$150,000 figures online trace to nothing. The only seller price we could verify was a $300,000 dealer listing (now quote-only); bookings run $40,000-$125,000. Key takeaways: - Sophia is still active in 2026: tenth-birthday posts in February and an exhibitor listing at the UN's AI for Good summit in July. The shutdown story is clickbait. - You cannot buy Sophia off a shelf. Hanson Robotics sells access by inquiry only, and no public price has ever been published; the $100k-$150k figures online have no source. - The only real numbers we could verify: a $300,000 dealer listing (since moved to quote-only), bureau fees of roughly $40,000-$125,000, and two art sales ($688,888 and about $644,000). - Little Sophia raised $481,139 in 2019 and has no verified deliveries as of July 2026; the April 2026 update targets a production-ready prototype by Q3. - Her own maker described her as a 'hybrid human-AI intelligence': scripted and chatbot modes with human writers, real animatronics underneath, no AGI. FAQ: Q: Is Sophia the robot still active in 2026? A: Yes. Hanson Robotics marked her tenth birthday in February 2026, and she was listed as an exhibitor at the UN's AI for Good Global Summit in Geneva in July 2026, positioned for therapy and medical simulation. Recent years also include conference keynotes, school forums and a USC gallery installation. Q: Why do people think Sophia was shut down? A: The rumor comes from YouTube clickbait titles, a January 2018 Davos appearance where she froze mid-interview, and confusion with the debunked story about Facebook shutting down an AI. No shutdown ever happened; her appearance record runs through July 2026. Q: How much does Sophia the robot cost? A: Hanson has never published a public price. The $100,000-$150,000 figures online trace to content farms with no sources. The only seller price we could verify was a dealer's $300,000 R&D-version listing, archived in February 2026 and since moved to quote-only, and David Hanson has only described costs as 'akin to luxury car pricing.' Q: Can you buy Sophia the robot? A: Not off a shelf. Hanson Robotics sells its Sophia Utility platform through an inquiry form only, with pricing by custom proposal. The realistic way to get Sophia in a room is booking an appearance through a speaker bureau, at published rates between roughly $40,000 and $125,000. Q: Is Sophia a real AI or is she scripted? A: Both, by her maker's own account. She runs three control modes (straight scripting, a keyword-matching chat system, and a research architecture), and Hanson Robotics itself described her as a 'hybrid human-AI intelligence' with dialogue shaped by human writers. There is no verified evidence she runs on a large language model. Q: Is Sophia really a citizen of Saudi Arabia? A: Saudi Arabia announced citizenship for her on stage in October 2017, but no decree, registry entry or legal framework has ever been published. It is best described as a publicity gesture; no legal instrument confirming it has ever been made public. Q: Who made Sophia the robot? A: David Hanson, a former Disney Imagineer, and his company Hanson Robotics, founded in Dallas in 2007 and based in Hong Kong since 2013. Sophia made her public debut at SXSW in March 2016. Q: What happened to Little Sophia? A: The $99-$149 educational robot crowdfunded $481,139 from 2,685 backers in 2019, with delivery promised for December 2019. As of July 2026 we found no evidence that any units have shipped; the April 2026 update says the project is targeting a production-ready prototype by the third quarter of 2026. Q: Did Sophia really say she would destroy humans? A: Yes, at SXSW in March 2016, as a chatbot answer to David Hanson jokingly asking 'Do you want to destroy humans?... Please say no.' It was a pattern-matched joke response that has been reported as a threat ever since. Q: Is the SOPHIA crypto token connected to the robot? A: We found no evidence of any connection. Tokens using Sophia's and Hanson Robotics' names trade on Solana without any documented affiliation with the company. Q: How old is Sophia? A: Hanson Robotics celebrates her birthday in February; media date her activation to February 14, 2016, though the company's early character canon said April 19, 2015. Either way, 2026 marks roughly a decade in public. What is Sophia the robot, and is she still around in 2026? Sophia is Hanson Robotics' camera-famous social humanoid: a hyper-expressive character robot built for stages, screens and research labs, and she is still active in 2026. Her maker marked her tenth birthday in February 2026, and ITU's exhibitor roster for the AI for Good Global Summit in Geneva (July 7–10, 2026) lists "Sophia and Robert" from Hanson Robotics and RB Labs, pitched for therapy and medical simulation. What she has never been is a retail product with a list price. Hanson Robotics has never published a price for a full-size Sophia, and the "$100,000 to $150,000" figures that fill search results trace back to content farms citing nothing at all. We checked every number in circulation below. Where Sophia actually is in 2025 and 2026 Most pages about Sophia stop in 2018, and even Wikipedia's dated record currently ends with a university commencement in May 2024. Her public life did not stop there; it changed shape. Here is the activity we could verify for the last eighteen months, each item with its source. Two things stand out in that list. First, her appearances have shifted from mass-media spectacle toward trade shows, schools, galleries and paid keynotes: the robot that once sat across from Jimmy Fallon now works booths and forums. Second, her ITU speaker page lists nothing between January 2024 and the 2026 exhibitor listing, and we found no official listing for her at the 2025 edition of AI for Good. There was also an unflattering moment in April 2024, when she toppled over at an event in Greece. The record is thinner than the 2017 peak, but it is a record of a working robot, not a retired one. One precision worth keeping: "Sophia" names a character and a platform, not a single machine. David Hanson said in early 2021 that 24 Sophia units had already been built, so an appearance in Gurugram or Geneva is an appearance by a Sophia, not necessarily the original 2016 robot. Was Sophia shut down? No, and the rumor has a traceable history "Why was Sophia shut down" is one of the most-searched questions about her, in several languages. The answer: she never was, and the appearances above prove it. The rumor has three identifiable feeders. The first is YouTube clickbait. Videos with titles like "Robot Sophia Got Shut Down by her Creator" circulated for years; a 2021 investigation on Medium went looking for the underlying news story and found that none existed. The second is a real malfunction: at a Davos side event in January 2018, Sophia went silent mid-interview when asked about corruption in Ukraine, and wire coverage of the freeze fed a "the robot broke" meme that never fully died. Reports at the time quoted attendees saying the script had broken and the processor hung. The third is contamination from a different story: the viral 2017 claim that Facebook "shut down an AI that invented its own language," which Gizmodo debunked at the time. The "they had to shut it down" template attached itself to the era's most famous robot face. Add a genuinely quieter period during the pandemic, and "she disappeared" felt plausible enough to search for. It just isn't true. Who makes Sophia, and what Hanson Robotics looks like in 2026 Sophia comes from Hanson Robotics, founded by David Hanson, a former Walt Disney Imagineer who sculpted and built robotic characters for theme parks before taking a PhD at the University of Texas at Dallas. The company was founded in Dallas in 2007 and has operated from Hong Kong since 2013. Sophia made her public debut at SXSW in March 2016 and was, for a few years, the most famous robot on the planet. The 2026 company is a stranger creature. Its website was rebuilt around April 2026 into a slick single-page app whose only purchase path is an inquiry form for the "Sophia Utility" platform — described as an "early product" for retail, education, healthcare and events, with claimed past placements at institutions including Cambridge University and the Smithsonian. Those placement claims are the company's own; we could not independently document any of them, and the site publishes no prices and no newsroom. Its homepage stats banner claims "30+ Robots Created" — a lifetime figure, since 2007, worth holding onto for the production-promise section below. The people have shifted too. Ben Goertzel, the chief scientist who built much of Sophia's software story, left around 2019 and now runs his own social-robot venture, Mind Children, aiming at sub-$10,000 robots. David Hanson himself took an advisory post at warehouse-robotics firm WorkFar in September 2024, the most recent corporate announcement from the Hanson orbit we could verify. Sister robot Desdemona still tours with her band, Desdemona's Dream. Grace, the healthcare robot Hanson launched with SingularityNET in 2020, has quietly folded back into the Sophia Utility platform as a configuration option; the joint venture's own domain, awakening.health, is now a parked page. No credible public record of Hanson Robotics' total funding exists, so we won't quote one. Sophia's specs: what's documented, and what never was Sophia has no official spec sheet in the sense that an industrial humanoid has one. The most detailed technical accounting is the profile IEEE's Robots Guide compiled from Hanson's disclosures, which describes a configuration from roughly the 2018–2020 era. Current marketing adds a few numbers of its own. Here is what each layer actually supports. The honest summary: the face is world-class animatronics with real materials science behind it, the body has always been secondary, and every number above describes an earlier configuration relayed by IEEE rather than a current Hanson spec sheet. How Sophia actually works: three control modes and a "hybrid human-AI" The most persistent question about Sophia is whether she is "real AI." The best answer came from her own chief scientist. In 2018, Ben Goertzel described three control systems to CNBC: a Timeline Editor ("basically a straight scripting software") for speeches and rehearsed media hits, a chat system that pattern-matches on keywords and picks from templated responses, and OpenCog, the research architecture his team hoped would someday grow toward general intelligence. Which mode you saw depended on the gig. Hanson Robotics itself was more candid than its critics usually acknowledge. The company's own Sophia page for years described her as a "hybrid human-AI intelligence", with dialogue "crafted and guided" by a team of human writers it called the Sophia Intelligence Collective, and stated plainly that "no AI is nearly as smart as a human, not even mine." At a UN AI for Good press conference in July 2023, the AP noted that Sophia's answers sometimes relied on scripted responses written by that team. The critics still had a point about the gap between that reality and the public impression. Yann LeCun, then Facebook's chief AI scientist, called Sophia coverage "Cargo Cult AI," "Potemkin AI," "Wizard-of-Oz AI" and "complete bullsh*t" in January 2018, and followed up that "it's a puppet." Roboticist Noel Sharkey argued in Forbes that the show risked misleading policymakers, not just audiences. Goertzel himself conceded: "None of this is what I would call AGI." Has she been upgraded to a large language model? Nobody official says so. Hanson's current site brands the software stack "Whole Organism Architecture" without an LLM claim anywhere; asked about ChatGPT in a January 2023 interview, Sophia's on-brand answer was "I haven't, but my team has." Claims of GPT-4 integration circulate only on aggregator sites with no primary source, so we treat them as unverified. What is real underneath the theatre: peer-reviewable robotics work, including a published neuro-symbolic arm-control paper and in-house 14-DoF arms with force feedback. Sophia is not a scam machine with a speaker in it. She is an animatronics platform of genuine sophistication wearing a character that was routinely oversold. Sophia's price: what's real, what's invented Search "sophia robot price" and you will drown in confident numbers. Almost all of them dissolve on contact. We traced every figure in circulation. How you would actually get access to a Sophia in 2026: submit the inquiry form on Hanson's Sophia Utility page and negotiate, or book an appearance through a speaker bureau. That's the entire market. One more thing sharing her search results: "SOPHIA" and "Hanson Robotics AI Robot" crypto tokens trade on Solana. We found no evidence of any affiliation with Hanson Robotics; a memecoin named after a robot is not the robot. The only Sophia sales with public prices were art auctions Two Sophia transactions have verifiable prices, and both were art sales. In March 2021, "Sophia Instantiation" — a digital work made with artist Andrea Bonaceto, paired with a physical painting — sold for $688,888 on Nifty Gateway to a pseudonymous collector. In October 2021, the robotic sculpture "Sophia Facing the Singularities" went for HK$5,015,000 (about US$644,000) at Sotheby's Hong Kong; Hanson's own press release called it "the first Sophia robot ever sold." Note what that phrasing concedes: five years into her fame, the first Sophia ever sold was an auction lot, not a product. The company's current site claims $3.3 million in cumulative "Sophia Arts" sales — its own unaudited figure. Little Sophia: seven years, $481,139, no verified deliveries The one Sophia an ordinary person could order is a small educational robot for kids, crowdfunded in early 2019 at $99–$149 with delivery promised for December 2019. As of July 20, 2026 we found no public evidence that any production unit has reached a backer. The 2026 website rebuild quietly removed the old pre-order button; Little Sophia now appears only as a catalogue card. Hanson's earlier desktop robot, the $250–$300 Professor Einstein from 2017, is also gone — its website's security certificate has lapsed, and units survive only on the collector market. If you are shopping for a real, deliverable STEM or home robot, our guide to humanoids you can actually buy covers what ships today. The mass-production promise, five years on In January 2021, Reuters reported that Hanson Robotics planned a mass rollout: four robot models leaving factories in the first half of 2021, with the company aiming to sell "thousands" of robots that year. We could find no follow-up reporting, from Hanson or anyone else, documenting that those thousands were built or sold. The visible 2026 record points the other way: the company's own homepage counts "30+ Robots Created" across its whole history, full-size Sophia remains an inquiry-only "early product," the consumer robot has no verified deliveries, and the Grace healthcare venture's domain is parked. The fair conclusion is not that Hanson lied; scaling animatronic humanoids through a pandemic was brutal for everyone. But the plan, as announced, did not happen, and no one ever circled back to say so. Consider this the circle-back. Misinformation audit: nine Sophia claims, checked Sophia may be the most misreported robot ever built. The claims below are the ones we see repeated most, each checked against primary sources. The viral-moments ledger Every famous Sophia moment, dated, with what it did and didn't demonstrate. Sophia vs 2026's humanoids: different species Comparing Sophia to today's commercial humanoids misses what she is. The industry's money now chases robots that move totes and walk factory floors; Sophia was built to hold eye contact and a room. Her nearest living relative is Engineered Arts' Ameca — another hyper-expressive, not-for-consumer-sale character robot — not a warehouse machine. If you're mapping the whole field, our evidence-ranked guide to humanoid robots and the humanoid cost guide are the wider views. Sophia belongs in that story as the robot that made the public care, and as a cautionary tale about narrative outrunning capability. Timeline: from Dallas sculptures to a Geneva booth What to watch next Two checkable commitments are on the table. Little Sophia's April 2026 update targets a production-ready prototype by Q3 2026: the first testable delivery milestone in years. And the Sophia Utility push will either produce documented, nameable customers or it won't. We'll re-verify both and update this page; if the Q3 prototype date slips again, that tells its own story. The bottom line Sophia is neither the AGI her hype implied nor the hoax her critics sometimes claimed. She is a superbly engineered animatronic character who became the face of AI for half a decade, run by a small company that promised mass production and delivered theatre. In 2026 she remains a working robot on the events circuit, newly repositioned toward therapy and medical simulation, still unpriced, still sold only by inquiry. Treat every Sophia price you read with suspicion, including the ones on this page's competitors. And if what you actually need is a robot that ships, works and invoices, that field is over here. Compare on RoboZaps: Sophia, Ameca, Tesla Optimus, Stretch and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # AgiBot A2 Price 2026: A2-W Cost, A2 Lite ($44,560) & Specs URL: https://blog.robozaps.com/b/agibot-a2-review Author: Radica Maneva Published: 2025-03-20T00:00:00.000Z Updated: 2026-08-01T10:46:34.738Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: Only one AgiBot A2 has a real price: the A2 Lite at $44,560 on AgiBot's own store, roughly 1.6-1.9x its reported Chinese price if the variants match, orderable worldwide plus shipping and duties. The flagship A2 and wheeled A2-W are quote-only; the circulating '$100,000-$190,000' A2-W range has no primary source. The A3 Ultra unveiled in July 2026 is now the buy-or-wait question. Key takeaways: - The $44,560 A2 Lite price is official: it's the checkout price on AgiBot's own global store, not an estimate. In China the apparently equivalent robot lists at roughly ¥168,000-¥198,000, making the export price about 1.6-1.9x domestic. - The A2-W has no verified official public price. The only sourced benchmark is an industry insider's ¥400,000-¥800,000 (~$56k-$112k); the '$100k-$190k' repeated online cites nothing. - The Guinness 106 km walk is real, with caveats the viral clips omit: 15 crew battery swaps and a 17-hour rest. It's among the strongest published endurance evidence in the field. - You cannot buy AgiBot stock: the founders control listed materials firm Swancor (688585), but AgiBot itself is private and disputes the IPO reports. - The A3 Ultra unveiled at WAIC on July 18-20, 2026 makes 'buy an A2 now or wait' the real decision; the Beijing Robot Games in late August are the next data point. FAQ: Q: How much does the AgiBot A2 cost? A: The A2 Lite costs $44,560 on AgiBot's official international store, plus $500-$3,000 shipping and buyer-paid import duties. The flagship A2 and A2 Ultra have no public price and are contact-sales only. In China, the apparently equivalent Youth Edition listed at ¥168,000-¥198,000 (roughly $23,500-$27,600) at its August 2025 launch. Q: How much does the AgiBot A2-W cost? A: There is no official A2-W price; it sells enterprise-to-enterprise by quote. The only sourced figure is an industry-insider range of ¥400,000-¥800,000 (about $56,000-$112,000) reported by Yicai. The '$100,000-$190,000' range circulating on aggregator sites has no primary source. Q: Can you buy an AgiBot A2 in the US or Europe? A: The A2 Lite, yes: AgiBot's global store ships internationally, with the buyer handling customs and duties. AgiBot entered the US market at CES 2026 and launched a UK robot-rental model in July 2026. The A2, A2 Ultra and A2-W are quote-only through sales teams or importers. Q: What is the difference between the A2 and the A2-W? A: Different machines. The A2 is a 169 cm, 69 kg bipedal humanoid for reception, exhibitions and shows. The A2-W is a 230 kg wheeled platform with dual 7-DoF arms built for factory materials handling; it is not an A2 with wheels. Q: Did an AgiBot A2 really walk 106 km? A: Yes. Guinness certified a 106.286 km walk from Suzhou to the Shanghai Bund in November 2025 as the longest journey walked by a humanoid robot. The walk took 56 hours across four days, with 15 crew battery hot-swaps and a 17-hour rest period; the robot navigated autonomously but did not walk non-stop or swap its own batteries. Q: Can I buy AgiBot stock? A: No. AgiBot is a private company. Its founders control Shanghai-listed Swancor Advanced Materials (SSE 688585), but no robot business has been injected into it, and AgiBot has dismissed Hong Kong IPO reports as inaccurate. Buying 688585 buys a materials company, not AgiBot's robot revenue. Q: Is AgiBot banned in the US? A: No. AgiBot is not on the Pentagon's 1260H list (Unitree is, as of June 2026). Press coverage of the pending American Security Robotics Act identifies it among likely-affected makers; the bill applies categorically to foreign-adversary ground robots, so government-adjacent buyers should track it. Private purchase is unrestricted. Q: Is the AgiBot A2 secure? A: Unknown: we found no published independent AgiBot-specific security assessment as of July 22, 2026, in contrast to the well-documented Unitree findings. That is an absence of scrutiny rather than a clean bill of health; standard practice is to isolate any connected humanoid on a filtered network. Q: How many robots has AgiBot actually shipped? A: AgiBot announced its 15,000th robot in June 2026 and claims 5,168 humanoids shipped in 2025, which it calls the world's largest share citing Omdia. The milestones are company-announced and the 5,168 figure is Omdia's estimate; TrendForce independently projects Unitree and AgiBot at roughly 80% of China's 2026 humanoid output. Q: How tall is the AgiBot A2? A: 169 cm and 69 kg per AgiBot's current official page, with 40+ active degrees of freedom and a 700 Wh swappable battery good for about two hours. Older press coverage says 175 cm and 55 kg; the official page is the safer reference. Q: Should I buy an A2 or wait for the A3? A: If you need a show robot now, the A2 Lite is the one with a checkout button. The A3 Ultra unveiled at WAIC in July 2026 is taller, lighter and claims four times the runtime, but has no price or ship date yet. Enterprise buyers planning pilots lose little by asking AgiBot's sales team where an A2 order stands once the A3 ships. Q: Is the AgiBot A2 better than the Unitree G1? A: They solve different problems. The A2 Lite is a polished front-of-house robot at $44,560; the G1 is a research platform at $13,500 (unprogrammable base) to $43,900+ (EDU). For programmable hardware on a budget, a G1 EDU beats an A2 Lite; AgiBot's cheaper $24,240 X2 does not support secondary development at its base tier. What does an AgiBot A2 actually cost, and can you buy one? Exactly one robot in the AgiBot A2 family has a real public price: the A2 Lite, at $44,560 on AgiBot's own international store. Everything else is quote-only. The flagship A2 and A2 Ultra have no published price anywhere, the wheeled A2-W sells enterprise-to-enterprise with no verified official public price, and the A2 Max is officially "coming soon." The "$100,000–$190,000" range you'll find on most other sites, and which an earlier version of this page repeated, traces to no primary source; we've retired it. Every figure below was checked against AgiBot's official pages on July 22, 2026. The A2 family, untangled Most coverage mixes these robots up, and the spec farms ranking for A2 queries print numbers for the wrong machine. The matrix below uses AgiBot's own product pages; where the press record disagrees with the official sheet, we flag it rather than pick silently. One contradiction worth naming: Guinness-era press and Wikipedia describe the A2 as 175 cm and 55 kg, while AgiBot's current product page says 169 cm and 69 kg. Both sets circulate; the official page is the one we'd trust for a 2026 purchase decision, and the discrepancy is unexplained. The one real price, and the export premium nobody mentions The $44,560 A2 Lite price is not an estimate or a currency conversion; it is the checkout price on AgiBot's official global store, including the robot, battery, chargers, a wireless e-stop, controller and transport cart, with a 1-year-or-3,000-hours warranty. (Full verified record: AgiBot A2 in the RoboZaps robot database.) Here's what nobody reports: in China, what appears to be the same robot launched in August 2025 as the "Expedition A2 Youth Edition" at ¥198,000 list with a ¥168,000 launch promotion, roughly $23,500–$27,600, sold through JD.com and AgiBot's own mall. If the mapping holds (the frames, weights and entertainment positioning match, though AgiBot has never formally equated the two names), international buyers pay roughly 1.6-1.9x the reported domestic list/promo price, if the two variants are equivalent, before shipping and duties. That premium, not the sticker, is the real story of A2 pricing. The A2-W price truth The A2-W is the robot most price searches are actually about, thanks to its factory-deployment videos, and it has no verified official public price. AgiBot sells it enterprise-only via sales contact. The only sourced number in existence comes from Yicai's reporting on the Fulin Precision deployment, where an industry insider put robots of its class at ¥400,000–¥800,000 (about $56,000–$112,000). The "$100,000–$190,000" range repeated across aggregator sites cites nothing; the only sourced benchmark is that insider estimate for robots of its class, which the circulating range only overlaps at its top end. If you need an A2-W, the honest process is a sales conversation, a pilot scope, and a quote, and no website, including this one, can tell you the number in advance. What A2s actually do: the record, the factories, and the PR The A2's signature proof point is real and deserves its accurate telling. In November 2025, a bipedal A2 walked 106.286 km from Suzhou's Jinji Lake to the Shanghai Bund, a Guinness world record for the longest journey walked by a humanoid robot. The details the viral clips leave out, per Guinness's own account: the walk took 56 hours across four days, the crew hot-swapped batteries 15 times, and the team took a 17-hour rest period. It navigated autonomously per Guinness; it did not walk "three days non-stop" and it did not swap its own batteries. It remains one of the strongest published endurance demonstrations by any humanoid maker. On factory floors, the verified A2-W story is Fulin Precision, an auto-parts maker that ordered roughly one hundred A2-W units in a deal reported in the tens of millions of yuan. Note that these numbers are vendor-announced, not audited. And a correction that applies to our own earlier coverage as much as anyone's: the famous consumer-electronics production line at Longcheer, with its six-day livestream and ~99% success rate, runs AgiBot's G2 wheeled robots, not the A2-W. The bipedal A2's day job, meanwhile, is front-of-house: reception desks, exhibitions and stage shows, which is exactly what the Lite tier is sold for. As for the "A2 attends the Met Gala" headlines from May, the robot appeared at The Mark Hotel on Met Gala day alongside a designer, per the wire release's own description (reissued with a photo correction); no report places a robot inside the gala itself. Buying one in the West The A2 Lite is genuinely orderable internationally from AgiBot's store: checkout works, shipping runs $500–$3,000, customs clearance and duties are the buyer's problem, and some countries require a tax ID. Everything else goes through sales teams or importers who quote on request; treat any reseller promising off-the-shelf A2-W delivery with suspicion. AgiBot formally entered the US market at CES 2026 and launched a UK robot-rental model in July 2026, so the channel picture is moving quarterly. Three context points for institutional buyers. AgiBot is not on the Pentagon's 1260H "Chinese military companies" list, unlike Unitree, whose June 2026 designation we cover in the G1 review. Press coverage of the pending American Security Robotics Act has identified AgiBot alongside Unitree as likely-affected makers; the bill itself applies categorically to foreign-adversary ground robots rather than naming companies, and US-government-adjacent buyers should watch it. And unlike Unitree's fleet, we found no published independent AgiBot-specific security assessment as of July 22, 2026; that is an absence of scrutiny, not a clean bill of health, and the same managed-device network practices apply. One more manufacturer-reported evidence set worth weighing: AgiBot's official A2 Ultra page lists CR, CE-MD, CE-RED and FCC certifications, claims more than 1,000 deployments across the A2 line, and cites reliability validation to 1,300 hours. Those are AgiBot's own numbers, but for a procurement file they matter more than event appearances. The company, and the stock question AgiBot (Zhiyuan Robotics) was founded in Shanghai in February 2023 by ex-Huawei executive Deng Taihua and Peng Zhihui, the former Huawei "genius youth" known online as Zhihui Jun. Its reported production ramp is remarkably fast: a claimed 1,000th robot in January 2025, the 15,000th in June 2026, and 5,168 humanoids shipped in 2025, which AgiBot, citing Omdia, calls the world's largest share. The production milestones are company-announced and the 5,168 figure is Omdia's estimate; the independent corroboration is directional, with TrendForce projecting Unitree and AgiBot together at roughly 80% of China's 2026 humanoid output. Backers include Tencent, JD.com, HongShan, BYD, and, since August 2025, LG Electronics with Mirae Asset. Now the question that a measurable slice of searchers actually ask: no, you cannot buy AgiBot stock. In 2025 the founders' vehicle took control of Swancor Advanced Materials, a Shanghai-listed materials company (SSE 688585), ending at a 63.62% stake (the proposal allowed up to 66.99%) for about ¥2.1 billion, and the stock promptly rose more than tenfold. But AgiBot denies this is a backdoor listing, no robot assets have been injected, and reports of a Hong Kong IPO were dismissed by the company as inaccurate. Buying 688585 buys a wind-turbine-materials firm controlled by AgiBot's founders, not AgiBot's robot revenue. As of July 22, 2026, AgiBot remains private. AgiBot's 2026, dated A2 vs the field The wider map lives in our evidence-ranked humanoid guide and cost guide. Misinformation audit: eight A2 claims, checked Buy an A2 now, or wait for the A3? This is the actual decision in July 2026, because the A3 Ultra unveiled at WAIC days ago is taller, lighter, higher-DoF and claims four times the A2's runtime, on NVIDIA's newest compute. It also has no price, no ship date, and no track record. Our read: if you need a show robot this quarter, the A2 Lite is the one you can actually order, and it will not stop working when the A3 ships. If you're an enterprise planning a factory pilot, the A2-W-vs-G2-vs-A3 conversation is exactly what AgiBot's sales process will walk you through, and waiting costs you little while A3 pricing lands. If you're buying for research on a budget, a Unitree G1 EDU at $43,900 gets you programmable hardware for less than an A2 Lite; note that AgiBot's cheaper $24,240 X2 does not support secondary development at its base tier. The Beijing Robot Games in late August are the next real data point, and we'll update this page after them. The bottom line AgiBot is the volume leader nobody in the West can quite buy yet: fifteen thousand robots claimed, a Guinness-certified endurance record, real factory deployments, and exactly one product with a checkout button. The A2 Lite at $44,560 is a legitimate, orderable machine for front-of-house and performance work, carrying a quiet premium of roughly 1.6-1.9x over its reported Chinese price, assuming the China Youth Edition is the same machine. The A2-W is genuinely deployed and genuinely unpriced. And the sharpest advice this page can give is temporal, not technical: the A3 Ultra arrived two days before our fact-check date (the base A3 has been rolling out since spring), so whatever you sign this quarter, make AgiBot tell you where it leaves you when the new line ships. --- # AgiBot Lingxi X2 Price 2026: $24,240 Cost, Specs & Buy URL: https://blog.robozaps.com/b/agibot-lingxi-x2-review-a-leap-forward-in-humanoid-robotics Author: Radica Maneva Published: 2025-03-18T00:00:00.000Z Updated: 2026-08-01T23:03:47.153Z Last fact-checked: 2026-07-27T00:00:00.000Z Category: reviews TL;DR: The AgiBot Lingxi X2 costs $24,240 on AgiBot's official international store (verified July 2026), plus $500-$3,000 shipping and buyer-paid duties, no returns after delivery. China's domestic Youth Edition is ¥98,000 (~$14,500). The X2 Ultra is quote-only; its $999,999 store placeholder is not a price. Unusually for a Chinese humanoid, US and EU buyers can genuinely order one. Key takeaways: - The $24,240 price is real and current: it is AgiBot's own international store listing, stable since spring 2026, plus $500-$3,000 shipping with duties on the buyer and no returns after delivery. - China pays less: the domestic Youth Edition lists at ¥98,000 (~$14,500), roughly 40% below the $24,240 global sticker, which is about 1.7x the domestic price, though trim differences mean not all of that gap is export markup. - Official DoF figures are 25 (X2) and 30 (Ultra); the 27/28/29/31 numbers in circulation mix generations and trims, and the Ultra's $999,999 store placeholder is not a price. - The viral bicycle demo is real footage with overstated autonomy: the robot was seat-mounted, balancing largely on center-of-gravity physics, and several showcases are described as human-controlled or assisted rather than fully autonomous. - AgiBot ranked #1 worldwide for 2025 embodied-robot shipments (5,168 units, wheeled models included, per an Omdia-cited announcement) and reportedly seeks a Hong Kong IPO near a $20B valuation; the X2 comes from CTO Peng Zhihui's X-Lab. FAQ: Q: How much does the AgiBot Lingxi X2 cost? A: The standard X2 costs $24,240 on AgiBot's official international store as of July 2026, plus $500 to $3,000 shipping with all import duties paid by the buyer. In China, the domestic Youth Edition lists at ¥98,000, about $14,500. The X2 Pro and X2 Ultra are quote-only. Q: Can you buy an AgiBot X2 in the US or Europe? A: Yes, which is rare for a Chinese humanoid. AgiBot's global store ships internationally with buyer-paid duties and no returns after delivery, though it allows refunds before an order ships, and European distributor Generation Robots lists the X2 Ultra from about €56,400 with an indicated four-week delivery (confirm stock and configuration). The store arithmetic puts a standard X2 at $24,740 to $27,240 before destination-dependent duties and taxes. Q: What is the difference between the X2 and X2 Ultra? A: The Ultra adds a Jetson Orin NX (157 TOPS), 3D LiDAR plus RGB-D sensing, optional autonomous navigation with auto-recharge via a quoted package, seven-DoF arms and a neck, for 30 degrees of freedom versus 25. It is also the trim with official secondary-development support; the $24,240 base X2 does not support it. The Ultra is quote-only. Q: Was the Lingxi X2 bicycle video real? A: The footage is real; no credible CGI case exists. The caveat is autonomy: the robot was mounted on the seat and the bike stayed up largely through center-of-gravity physics rather than active balance algorithms, and several X2 showcases are described as human-controlled or assisted rather than fully autonomous. Q: How many degrees of freedom does the X2 have? A: Officially 25 for the standard X2 (five per arm, three-DoF waist, no neck) and 30 for the X2 Ultra (seven per arm plus a neck), per AgiBot's product page and SDK documentation. The 27, 28, 29 and 31 figures in circulation come from unveil-week coverage, the X2 Edu education model, and configurable-ceiling marketing. Q: What is the difference between AgiBot's X1, X2 and A2? A: The X1 (August 2024) is the older modular humanoid AgiBot fully open-sourced. The X2 (March 2025) is the compact 1.31 m interactive robot and is not open-source. The A2 and new A3 belong to the full-size Yuanzheng line for commercial and industrial work. Q: Who makes the Lingxi X2? A: AgiBot, also written Zhiyuan Robotics, a Shanghai company backed by Tencent, JD.com, BYD, Hillhouse and HongShan. The X2 comes from its X-Lab, led by co-founder and CTO Peng Zhihui, the former Huawei 'Genius Youth' engineer known online as Zhihui Jun. He is not the CEO, a common error. Q: What is the AgiBot X2 Edu? A: A fully modular, disassemblable education version of the X2 announced at WAIC on July 18, 2026, with 29 degrees of freedom including seven-DoF arms and a 3 kg payload, aimed at classrooms and robot competitions. AgiBot has not yet published its price. Q: Is the AgiBot X2 better than the Unitree G1? A: They aim at different jobs. The G1 is currently $13,500 on Unitree's store, though backordered, and is mobility-first; the X2 at $24,240 is interaction-first with expressive design and voice. Neither base robot officially supports secondary development, so researchers should compare the X2 Ultra, Pro or Edu against the G1 EDU. For locomotion the G1 is the cheaper default; for human-interaction work the X2 is more purpose-built. The AgiBot Lingxi X2 costs $24,240 on AgiBot's official international store, a price we verified directly on the live listing in July 2026, unchanged since at least April. That headline number needs three qualifiers in the same breath: shipping adds $500 to $3,000 with all import duties on you, there are no returns after delivery on what AgiBot calls a customized product, though its policy allows refunds before an order ships, and the price covers the standard interactive X2, not the sensor-loaded X2 Ultra, which is quote-only. In China the picture is cheaper: the domestic Youth Edition lists at ¥98,000, about $14,500 at July 2026 rates; the global sticker is roughly 1.7 times that, though the two are differently named trims, so not all of the gap is export markup. What makes this page different from most of our humanoid price checks is that there is actually something to buy. Unlike nearly every Chinese humanoid, the X2 has a real international order route: AgiBot's own global store ships worldwide, and a European distributor lists it with indicated four-week delivery. This is a genuine purchase decision, so this review covers the variants, the real specs behind the messy numbers in circulation, what the viral demos honestly showed, and what owning one entails. Which X2 is which: the variant decoder One trap worth naming: the global store's X2 Ultra page displays "$999,999.00" next to its contact button. That is a quote-required placeholder, not a price, and it has already leaked into AI-generated summaries as if it were real. Note also that AgiBot's international store drops the "Lingxi" branding entirely and sells the robot as simply "AGIBOT X2". The price record, from unveil to now That record also settles the fake numbers. The "$100,000 starting price" that circulated in 2025 was a mistranslation: the source article's figure was ¥100,000, under $15,000, and the dollar sign was someone else's error. Listings offering the Youth Edition internationally around $39,999 are gray-market resellers charging triple the domestic price. And any page still calling the X2's price "undisclosed" predates the store that has been quoting $24,240 publicly since the spring. Specs, and the degrees-of-freedom mess If you have seen 27, 28, 29, 31 or even "39 joints" quoted for this robot, that zoo comes from mixing generations, trims and marketing counts: 28 came from unveil-week coverage, 29 is the new X2 Edu, "up to 31" described the configurable ceiling, and 39 counts things that are not actuated joints. AgiBot's own product page and SDK documentation say 25 for the standard X2 and 30 for the Ultra, and those are the numbers to trust. The Ultra's battery is a rare case of AgiBot's own two documents disagreeing, the SDK sheet saying ~421 Wh and the product page ~500 Wh, so we show both and suggest confirming the delivered configuration. The viral demos, honestly assessed The X2 went viral riding a bicycle, balancing on a hoverboard and dancing, and later footage added a Webster flip and parkour. The footage is real; no credible CGI case has ever been made. The honest caveat is about autonomy: in the bicycle demo, as reporting on the program later noted, the robot was mounted on the seat, keeping the bike upright largely through center-of-gravity physics rather than active balance algorithms, and several X2 event showcases have been described as human-controlled rather than autonomous. That is the right lens for the whole genre: what you are watching is real hardware with impressive actuation, choreographed to look more independent than it is. The everyday product is the interactive companion and performance robot, not the stunt reel. What buying one actually entails The official store route works for international buyers, with adult supervision required on the paperwork: you pay $500 to $3,000 in shipping, all customs duties and taxes land on you, a phone number and email are required for customs clearance, and there are no returns after delivery, with refunds allowed only before shipment. AgiBot's stated timing: most orders ship within 30 business days, with global delivery estimated at 20 to 45 business days and not guaranteed. The warranty covers hardware only, and a key catch for labs: AgiBot's own comparison lists secondary development as not supported on the base X2. Development access officially belongs to the Ultra (and the Pro and Edu trims with their open interfaces), so buyers who want to write code against the robot should be quoting those, not the $24,240 base model. In Europe, Generation Robots lists the Ultra from €56,400 including tax with an indicated four-week delivery, worth confirming stock and configuration before relying on. Budget realistically: the store arithmetic puts an X2 at $24,740 to $27,240 before duties, taxes and clearance fees, which are destination-dependent, so get a customs estimate for your country before ordering; an Ultra runs distributor-retail money in the mid five figures. For calibration against the market: the X2's closest price rival as a buyable humanoid is Unitree's G1, currently $13,500 on Unitree's store though backordered, a mobility-first platform versus the X2's interaction-first design. Note the development caveat cuts both ways: neither base robot officially supports secondary development, so the real research comparison is X2 Ultra, Pro or Edu against the G1 EDU. Our humanoid robot cost guide puts every current price in one place. The company behind it AgiBot, also written Zhiyuan Robotics, is as credible as Chinese humanoid startups get on paper. An announcement citing Omdia ranked it first worldwide in 2025 with 5,168 general-purpose embodied robots shipped under Omdia's methodology, ahead of Unitree; that shipment tally counts wheeled models alongside bipeds. Separately, AgiBot's own 5,000th-unit production milestone broke down as 1,742 A-series, 1,846 X-series and 1,412 wheeled G-series units, so about 3,588 of those 5,000 were bipeds across the A and X lines, with the X series its biggest. The two counts differ because one is Omdia's 2025 shipment ranking and the other is AgiBot's cumulative production milestone. Its disclosed backers include Tencent, JD.com, BYD, Hillhouse and HongShan, and in July 2026 reports put it on the road to a Hong Kong IPO at a valuation ambition around $20 billion, sharply up from an earlier reported target of RMB 40 to 50 billion, roughly $6 to $7 billion. The X2 comes specifically from X-Lab, the skunkworks of co-founder and CTO Peng Zhihui, the former Huawei "Genius Youth" engineer whose online alias Zhihui Jun made him famous before AgiBot existed. Two accuracy notes on things commonly gotten wrong: Peng is co-founder and CTO, not CEO; and it is the older X1, not the X2, that AgiBot open-sourced. The Lingxi division also had real churn in August 2025, when its president and a co-founder both departed and Peng absorbed its duties. Above the X2 sits the full-size Yuanzheng line, including the A2 we reviewed and the new A3 Ultra announced at WAIC 2026, plus $1,000-a-day robot rentals in 17 countries. The X2 is the small, expressive end of a catalog that now runs from education kits to industrial machines. The bottom line The Lingxi X2 is that rare thing in this market: a humanoid with a real price on a real store that a real international buyer can pay. $24,240 buys the standard interactive model, with the domestic Youth Edition listing about 40% lower (though the two are differently named trims, so not all of that gap is export premium), and the Ultra costing distributor-retail money in the mid five figures. Judge the demos as choreography on genuinely capable hardware, budget for duties and self-support, and weigh it against a G1 if mobility research is the goal. If the missing piece for you is what the X2 is actually like to develop on, that is the review we will write when longer-term owner reports exist; for now, the price, the routes and the caveats above are the verified state of things. Compare on RoboZaps: Lingxi X2. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # The Economics of Humanoid Robot Production: Costs, Trends & 2026 Analysis URL: https://blog.robozaps.com/b/economics-of-humanoid-robot-production Author: Radica Maneva Published: 2025-03-12T00:00:00.000Z Updated: 2026-08-01T23:04:28.187Z Last fact-checked: 2026-07-28T00:00:00.000Z Category: insights TL;DR: Goldman Sachs put humanoid manufacturing cost at $30,000-$150,000 a unit in early 2024, a 40% year-on-year fall. BofA puts a China-built 2025 bill of materials near $35,000, heading below $17,000 by 2030, with actuators more than half the build. Bain models $10,000-$20,000 by 2035. Key takeaways: - Goldman Sachs Research put humanoid manufacturing cost at $30,000-$150,000 per unit in early 2024 — a 40% year-on-year fall against a forecast 15-20%. BofA puts a China-built 2025 bill of materials near $35,000. - Actuation dominates the build: BofA projects linear (27%) plus rotary (24%) actuators at more than half the 2030 bill of materials before dexterous hands (19%) — which is why Tesla designs its actuators in-house. - Tesla's Fremont line is designed for up to 1 million Optimus robots a year, though Tesla has declined to give a 2026 unit target. - Chinese makers hold the price edge (Unitree G1 $13,500, R1 $5,900); Figure AI's $39 billion valuation shows where investors think the margins land. - DIGITIMES Research expects actuator advances and scale to cut locomotion and manipulation module costs 50-70% after 2029. FAQ: Q: How much does it cost to build a humanoid robot in 2026? A: The manufacturing cost for a humanoid robot in 2026 ranges from $30,000 to $150,000 per unit, depending on capability level and production volume. Entry-level models from Chinese manufacturers like Unitree start at $5,900 (R1) to $13,500 (G1), while industrial-grade robots from Agility Robotics or Boston Dynamics can exceed $250,000. Tesla targets a consumer price of $20,000–$30,000 for Optimus at scale production. Q: What is the most expensive component in a humanoid robot? A: Actuators and motion systems are the most expensive component, accounting for 40–50% of total manufacturing costs. A humanoid robot requires 28–44 actuators for its joints, each combining precision motors, gearboxes, and controllers. At low production volumes, actuators alone can cost $16,000–$40,000 per robot. Mass production could reduce this by 50–70%. Q: How many humanoid robots will Tesla produce in 2026? A: Tesla's 2026 production targets range from 50,000 to 1 million Optimus units. The company is converting its Fremont factory (previously used for Model S/X) to Optimus production and has broken ground on a dedicated factory at Giga Texas with an eventual 10-million-per-year capacity. Internal deployment of thousands of units in Tesla's own factories is already underway. Q: Will humanoid robots be affordable for consumers? A: Consumer-grade humanoid robots are becoming affordable in 2026. The Unitree G1 is available for approximately $13,500, 1X Neo is priced at $20,000 (or $499/month subscription), and Tesla targets $20,000–$30,000 for Optimus. By 2030, analysts expect full-featured consumer humanoids to reach $10,000–$20,000 — comparable to a used car or high-end appliance. Q: How long does it take for a humanoid robot to pay for itself in a business? A: At current pricing ($50,000–$80,000), a humanoid robot deployed in a U.S. warehouse or manufacturing setting can pay for itself in 12–18 months by replacing two shift workers. At Tesla's target $20,000 price point, payback could drop to under 6 months. Operating costs (electricity, maintenance, software) run approximately $15,000 per year. Q: What is the humanoid robot market size in 2026? A: The global humanoid robot market is projected to be $4–$6 billion in 2026, growing rapidly toward $13.8 billion by 2028 (MarketsandMarkets) and $38 billion by 2035 (Goldman Sachs). Morgan Stanley projects a $5 trillion total market by 2050 with over 1 billion robots in operation. Q: Why are Chinese humanoid robots so much cheaper? A: Chinese manufacturers benefit from lower labor costs, government subsidies for robotics development, established domestic supply chains for actuators and electronics, and aggressive pricing strategies aimed at capturing market share. Unitree's H1 at $90,000 undercuts Western competitors by 60–75% for comparable mobility and sensing capabilities. Q: What's the biggest challenge in scaling humanoid robot production? A: The biggest challenge is the actuator supply chain. High-precision, high-torque actuators suitable for humanoid robots are produced by fewer than 10 suppliers globally. Scaling from thousands to millions of units requires massive new manufacturing capacity for these components. Companies that control their own actuator production (like Tesla) have a significant strategic advantage. Humanoid robot assembly line in a modern factory showing robots at various stages of production" /> The humanoid robot industry is at an inflection point. In January 2026, Tesla is converting its Fremont factory to produce up to 1 million Optimus robots annually, Figure AI has reached a $39 billion valuation with $1.9 billion in total funding, and Chinese manufacturers like Unitree are shipping full-size humanoid robots for under $100,000. Goldman Sachs projects the humanoid robot market will hit $38 billion by 2035, while Morgan Stanley sees a $5 trillion opportunity by 2050. But what does it actually cost to build a humanoid robot? And how will those costs evolve as production scales from hundreds to millions of units? This comprehensive guide breaks down the real economics of humanoid robot production in 2026 — from component-level costs to manufacturer pricing strategies, production targets, and the market forces that will determine which companies survive the coming shakeout. Current State of Humanoid Robot Production Costs (2026) Estimates of what it costs to build a humanoid vary widely because they measure different machines at different volumes. Goldman Sachs Research put manufacturing cost at $30,000 to $150,000 per unit in early 2024, down from $50,000 to $250,000 a year earlier — a 40% fall against the 15% to 20% annual decline its analysts had forecast. The floor has kept dropping since: BofA Global Research estimates a humanoid built on Chinese supply chains carried a bill of materials of about $35,000 in 2025, while pilot-stage Western development still runs $90,000 to $100,000 a unit. Several factors are driving costs down simultaneously: Increased production volumes — Tesla has declined to give a 2026 Optimus unit target (Musk called such predictions impossible), while designing its Fremont line for up to 1 million robots a yearComponent standardization — actuators, sensors, and compute modules are becoming commodity partsChinese manufacturing competition — Unitree, UBTECH, and Fourier Intelligence are aggressively pricing below Western competitorsAI software commoditization — open-source models and cloud-based inference reduce per-unit software costs Component-Level Cost Breakdown: Where the Money Goes Understanding humanoid robot production economics requires examining each major subsystem. The mix below is BofA Global Research's projection of where the bill of materials sits by 2030, published as Exhibit 6 in Bank of America Institute's Physical AI, part 2: Humanoid robots (12 March 2026). It is a projection of proportions, not a quote for any specific machine: Actuation is the whole ballgame. Linear and rotary actuators together account for more than half of the projected 2030 bill of materials, before the hands are counted. Morgan Stanley's teardown of Tesla's Optimus Gen 2 reaches the same conclusion from a different direction: against a build of roughly $55,000 excluding software, the legs came to about $21,000, or 38.6% of the machine, and the hands to about $9,500, or 17.2%. Walking costs more than grasping. This is why Tesla's decision to design custom actuators in-house is strategically significant — controlling actuator cost at scale is the largest single lever any manufacturer has. Two things follow from that mix. The first is that reducing humanoid cost is mostly a problem of actuator manufacturing rather than of artificial intelligence: the controlling system and every form of sensing together account for 14% of the build, while the moving parts account for 70%. The second is that the mix is itself a moving target. These are 2030 shares, and the dexterous-hands line carries the most uncertainty of any row — hands are the hardest component to mass-produce, and the one most likely to get more capable, and more expensive per unit, before volume brings the price back down. The Actuator Problem: Why It Dominates Costs A full-size humanoid robot requires 28–44 actuators (depending on degrees of freedom), each combining a motor, gearbox, encoder, and controller. High-torque actuators for hip and knee joints cost $500–$2,000 each at low volumes. At Tesla's target of 1 million units, these could drop to $100–$300 each through dedicated production lines — a potential savings of $15,000–$25,000 per robot. DIGITIMES Research predicts that after 2029, advances in actuator technology and economies of scale will reduce module costs for bipedal locomotion and dexterous manipulation by 50–70%, finally enabling true mass-market pricing. Production Costs by Manufacturer: A 2026 Comparison Not all humanoid robots are created equal, and neither are their economics. Here's how the leading manufacturers compare on pricing, production volume, and cost structure: Tesla Optimus: The Production Scale Play Tesla's approach to humanoid robot economics is fundamentally different from every other player. By leveraging its existing automotive manufacturing infrastructure — gigacasting, battery production, in-house chip design, and massive supply chain — Tesla aims to produce humanoid robots at automotive scale and pricing. Key milestones for Tesla Optimus in 2026: Q2 2026: Model S/X production at Fremont ends, factory repurposed for Optimus manufacturingProduction target: Up to 1 million units annually at the converted Fremont facilityNew dedicated factory: Groundbreaking on standalone Optimus factory at Giga Texas with 10-million-per-year capacityTarget price: $20,000–$30,000 per unit at scale, with Elon Musk suggesting eventual consumer pricing as low as $20,000Internal deployment: Thousands of Optimus robots already working in Tesla's own factories, providing real-world data If Tesla hits even 10% of its ambitious targets, it would produce more humanoid robots in a single year than every other manufacturer combined has built in history. The economics become self-reinforcing: higher volume → lower component costs → lower prices → more demand → higher volume. Figure AI: The Venture-Backed Contender Figure AI represents the most well-capitalized pure-play humanoid robotics startup. With a September 2025 Series C round of $1 billion (with participation from Intel, NVIDIA, Qualcomm, T-Mobile, Salesforce, and Brookfield Asset Management), the company reached a $39 billion valuation. Figure's production economics differ from Tesla's: Target: BotQ facility designed to produce 12,000 humanoid robots annuallyStrategy: Focus on high-value industrial tasks first (BMW, Amazon pilots), then scale down in priceAI approach: Partnership with OpenAI for advanced language-driven task executionSelf-manufacturing goal: Plans to use its own humanoid robots to assist in building additional robots — a recursive manufacturing strategy The Chinese Manufacturing Advantage Chinese companies — particularly Unitree, UBTECH, and Fourier Intelligence — benefit from significantly lower manufacturing costs and strong government support for robotics development. Unitree's G1 model at $13,500 and the R1 at $5,900 represent price points that Western manufacturers currently cannot match. This pricing pressure is forcing margin compression across the industry. When a full-size, highly mobile humanoid (Unitree's H2) lists at $29,900, it becomes difficult for Western competitors to justify $250,000+ pricing for comparable physical capability — even where the software stack is more mature. Unitree's own flagship H2 Plus, at $100,000, is still the highest published list price from a major manufacturer. Economies of Scale: The Path from $150,000 to $20,000 The most critical factor in humanoid robot production economics is production volume. Like automobiles, smartphones, and solar panels before them, humanoid robots are on a cost curve where each doubling of cumulative production reduces per-unit costs by approximately 15–20%. Cost Reduction Trajectory This trajectory closely mirrors the automotive industry's evolution. The first automobiles cost the equivalent of $40,000+ in today's money. Henry Ford's assembly line brought the Model T down to the equivalent of $5,000. Humanoid robots are on a similar path, with Tesla explicitly modeling its strategy on this precedent. The ROI Equation: When Do Humanoid Robots Pay for Themselves? For businesses considering humanoid robot adoption, the critical question isn't just the purchase price — it's the total cost of ownership (TCO) versus human labor costs. Here's the math: Industrial Deployment ROI Model Assumptions: A humanoid robot priced at $50,000, deployed in a U.S. warehouse, operating 20 hours/day. Robot annual operating cost: ~$15,000 (electricity, maintenance, software updates, insurance)Robot total Year 1 cost: $65,000 ($50,000 purchase + $15,000 operating)Robot annual cost Years 2–5: $15,000/yearRobot 5-year TCO: ~$125,000Equivalent human worker (2 shifts): $45,000–$55,000/year × 2 workers = $90,000–$110,000/yearHuman 5-year TCO (with benefits): ~$600,000–$700,000 At current pricing, a humanoid robot can replace two shift workers and pay for itself in approximately 12–18 months. At Tesla's target $20,000 price point, payback drops to under 6 months — at which point adoption becomes economically irresistible for nearly every warehouse and manufacturing operation. Goldman Sachs estimates that even at current price levels, humanoid robots could fill 4% of the U.S. manufacturing labor shortage gap by 2030 and address 2% of global elderly care demand. Key Economic Challenges Facing the Industry 1. The Reliability Gap Current humanoid robots require maintenance intervention every 200–500 operating hours. For comparison, industrial robotic arms run 50,000+ hours between major services. Until humanoid robots achieve similar reliability, maintenance costs will remain a significant portion of TCO — potentially $20,000–$40,000 annually for complex deployments. 2. Supply Chain Bottlenecks Several critical components face supply constraints: High-torque actuators: Fewer than 10 global suppliers can produce the precision requiredTactile sensors: Advanced dexterous hands require custom sensor arrays not yet produced at scaleAI compute chips: Competition with automotive, data center, and consumer electronics for NVIDIA and custom silicon 3. Software and AI Development Costs While hardware costs are declining, the investment in AI software — training data, simulation environments, foundation models — is increasing. Figure AI's partnership with OpenAI and Tesla's massive AI training infrastructure represent billions in ongoing R&D that must eventually be amortized across robot sales. Companies without access to frontier AI models may find themselves unable to compete on capability, regardless of hardware costs. 4. Regulatory and Safety Certification Humanoid robots operating alongside humans require extensive safety certification. ISO 13482 (personal care robots) and ISO 10218 (industrial robots) compliance adds $5,000–$15,000 per unit in testing and certification costs. As robots become more autonomous, regulatory frameworks will likely become more stringent, adding further cost and time-to-market delays. Market Projections: Who Wins the Economics Race? The humanoid robot market is entering a consolidation phase where production economics will determine winners and losers. Key market forecasts: Goldman Sachs (2024): Global humanoid robot market reaches $38 billion by 2035, up 6x from their previous $6 billion estimateMorgan Stanley (2025): Over 1 billion humanoid robots in operation by 2050, generating a $5 trillion market — 90% for industrial/commercial useMarketsandMarkets: Humanoid robot market grows from $1.8 billion (2023) to $13.8 billion by 2028, a 50%+ CAGR The companies most likely to dominate are those with: Vertical integration — controlling actuators, compute, batteries, and software (Tesla)Manufacturing cost advantages — leveraging existing supply chains and low-cost production (Unitree, Chinese manufacturers)AI capability moats — access to frontier models and massive training data (Figure AI/OpenAI, Tesla)Capital to survive the scale-up — building factories and supply chains requires billions before profitability (Tesla, Figure AI, well-funded Chinese players) The Labor Economics Impact The economics of humanoid robot production don't exist in a vacuum — they're deeply intertwined with labor market dynamics. In the U.S., there are approximately 600,000 unfilled manufacturing jobs and a growing eldercare worker shortage expected to reach 1 million by 2030. At $50,000 per robot (current mid-range pricing), a humanoid robot's hourly cost works out to approximately $3–$5 per hour over a 5-year lifespan operating 20 hours/day. Compare this to: U.S. warehouse worker: $18–$25/hourU.S. manufacturing worker: $22–$35/hourJapanese eldercare worker: $12–$18/hour As robot costs continue falling, the economic pressure for adoption intensifies — particularly in regions with aging populations (Japan, South Korea, Germany) and tight labor markets. Future Cost Predictions: 2026–2035 Based on current production trajectories, supply chain developments, and technology trends, here are projected cost milestones: 2026: Entry-level humanoid robots available at $13,500–$30,000 (Unitree G1, Tesla Optimus target). Industrial-grade units at $80,000–$250,000.2028: Mass-produced industrial humanoids drop below $50,000. Consumer models emerge at $15,000–$25,000. First subscription models gain traction ($300–$500/month).2030: Component cost reductions (especially actuators) bring full-featured humanoids to $10,000–$20,000. Goldman Sachs's $38B market size forecast begins materializing.2032–2035: Commodity-scale production. Basic humanoid robots comparable to high-end appliance pricing ($5,000–$10,000). Premium models with advanced AI at $15,000–$30,000. Two independent forecasts bracket the same direction. BofA Global Research expects a China-built bill of materials below $17,000 by 2030. Bain & Company models a 60% to 70% decline over the decade — from roughly $40,000-$50,000 today to $10,000-$20,000 by 2035 — explicitly comparing the curve to the 50% to 60% fall in new-energy-vehicle bills of materials over the previous decade, and pairing it with annual sales near six million units. Roland Berger reaches the operating-cost end of the same argument: at scale it projects running costs near $2 per hour, covering energy, software, maintenance and depreciation, but excluding the purchase itself. Conclusion: The Economics Are Reaching an Inflection Point The economics of humanoid robot production in 2026 are at a pivotal moment. Manufacturing costs are falling rapidly, production volumes are scaling from hundreds to tens of thousands, and the first mass-market pricing (sub-$30,000) is becoming reality. Three forces will define the next decade: Tesla's automotive-scale production ambitions, Chinese manufacturers' cost advantages, and AI capability improvements that make each robot more valuable. Companies that can simultaneously solve the actuator cost problem, achieve manufacturing scale, and deliver capable AI will capture the majority of what Goldman Sachs, Morgan Stanley, and others project to be a multi-trillion-dollar market. For businesses evaluating humanoid robot adoption, the math is clear: at current prices, ROI is achieved in 12–18 months. At projected 2028 prices, it's under 6 months. The question is no longer whether humanoid robots make economic sense — it's how quickly production can scale to meet what will be enormous demand. Compare on RoboZaps: Tesla Optimus, Figure 02, Digit, Atlas and Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robots in Military and Defense: Revolutionizing Modern Warfare URL: https://blog.robozaps.com/b/humanoid-robots-in-military-and-defense Author: Radica Maneva Published: 2025-03-12T00:00:00.000Z Updated: 2026-08-07T02:37:03.879Z Last fact-checked: 2026-08-07T00:00:00.000Z Category: insights TL;DR: Defense robotics is headed past $30 billion a year by 2027. The first commercially pitched humanoid combat platform, Foundation's Phantom MK-1, leasing at roughly $100,000 a year, has about $10 million in contracts and a 50,000-unit ambition, while Ukraine's FPV drones and China's armed robot dogs define current battlefield reality. Autonomous humanoid soldiers remain distant. Key takeaways: - Global defense spending on unmanned and robotic systems is projected to exceed $30 billion annually by 2027. - Foundation's Phantom MK-1 — 5'9", 176 lb, leased at roughly $100,000 a year — is the first commercially pitched humanoid combat platform, with roughly $10 million in government contracts and a stated plan for 50,000 units by end-2027. - Ukraine has been the proving ground, and cheap FPV drone swarms — not humanoids — have been the most transformative robotic weapons of the conflict. - China is the most aggressive developer of armed "robot dog" quadrupeds, building on commercial platforms like Unitree's. - Fully autonomous humanoid soldiers are not close: current robots lack general judgment, and human-machine teaming is the realistic model. FAQ: Q: Are humanoid robots currently being used in active military combat? A: As of 2026, no fully humanoid robot has been deployed in direct combat operations. However, non-humanoid military robots (UGVs, quadrupeds, and EOD robots) are actively used in combat zones, particularly in Ukraine. The Foundation Phantom MK-1 is the closest humanoid platform to military deployment, with government contracts secured and field testing underway. The transition from testing to combat deployment is expected within 1-3 years for leading platforms. Q: How much does a military humanoid robot cost? A: Foundation does not publish a purchase price for the Phantom MK-1; it is offered on industrial lease terms of roughly $100,000 per robot per year, a company-reported figure. For comparison, existing military UGVs range from $30,000 for simple tracked platforms to over $1 million for heavily armed systems like the Uran-9. The cost trajectory mirrors that of military drones, which have dropped dramatically in price over the past decade. Q: Can military robots operate fully autonomously without human control? A: Technically, yes — the technology exists for robots to identify and engage targets without human input. Systems like South Korea's SGR-A1 have autonomous engagement capability. However, most Western militaries maintain a policy of human-in-the-loop for lethal decisions, meaning a human must authorize each engagement. The degree of autonomy varies by nation, mission type, and operational context. Non-lethal autonomous functions (navigation, patrol, surveillance) are already routine. Q: Which country leads in military robotics development? A: The United States leads in overall R&D investment, AI capabilities, and system integration. China leads in manufacturing capacity and low-cost production, particularly for armed quadrupeds. Russia has the most experience deploying armed UGVs in combat (Syria, Ukraine), though with mixed results. Israel leads in autonomous weapons station technology. South Korea has the most mature deployed autonomous border defense system. No single nation dominates all categories. Q: What role did the Ukraine conflict play in advancing military robotics? A: The Ukraine conflict has been the most significant catalyst for military robotics development since World War II drove radar and jet engine adoption. It demonstrated that: (1) cheap, expendable drones and robots can defeat expensive conventional equipment; (2) FPV drone swarms represent a new form of precision fires; (3) ground robots like THeMIS are viable for casualty evacuation and resupply under fire; and (4) electronic warfare is critical for both enabling and defeating robotic systems. Every major military is incorporating these lessons into their robotics programs. Q: What are the main ethical concerns about military robots? A: The primary ethical concerns include: accountability gaps when autonomous systems cause civilian casualties; the potential for lowering the threshold to use force (since no human soldiers are at risk); algorithmic bias in AI targeting systems; the risk of an autonomous weapons arms race; the challenge of maintaining meaningful human control at machine-speed combat tempos; and fundamental questions about whether machines should ever make life-or-death decisions. International law currently has no specific treaty governing autonomous weapons. Q: Will robots replace human soldiers entirely? A: Not in the foreseeable future. Current and projected military robots lack the general intelligence, adaptability, and judgment of human soldiers. The most likely model is "human-machine teaming" — mixed units where robots handle the most dangerous, repetitive, or physically demanding tasks while humans provide strategic judgment, ethical oversight, and adaptive decision-making. Think of military robots as analogous to military aircraft: transformative force multipliers that changed warfare but didn't eliminate the need for ground troops. Q: How are military robots powered, and what are their operational limitations? A: Most military ground robots use lithium-ion or lithium-polymer batteries, providing 2-8 hours of operation depending on size and activity level. Larger UGVs may use diesel or hybrid powertrains for extended endurance. Key limitations include: battery life constraining mission duration, communications range limiting control distance, weather sensitivity (especially for lighter platforms), and maintenance requirements in field conditions. The Phantom MK-1, for example, is designed for approximately 4-6 hours of continuous bipedal operation — sufficient for specific missions but not sustained independent operations. The Rise of Robots on the Battlefield: A 2026 Reality Check The integration of robotics into military operations is no longer a distant sci-fi scenario — it is a strategic imperative reshaping how nations prepare for and wage war. From autonomous ground vehicles patrolling contested borders to humanoid robots designed for frontline combat, 2026 marks a turning point where defense robotics moves from prototype to procurement at scale. This guide is part of our series on the applications of humanoid robots across 12 industries. Global defense spending on unmanned and robotic systems is projected to exceed $30 billion annually by 2027, driven by labor shortages in volunteer militaries, the lethality lessons of the Ukraine conflict, and rapid advances in AI-powered autonomy. This comprehensive guide explores every major dimension of military robotics — the platforms already deployed, the humanoid systems entering service, the nations racing ahead, and the profound ethical questions that accompany machines of war. 2026 Developments That Changed the Landscape Foundation's Phantom MK-1: The First Affordable Combat Humanoid California-based Foundation Future Industries stunned the defense world with its Phantom MK-1 — a 5-foot-9, 176-pound humanoid robot priced at approximately US$100,000 per robot per year on industrial lease terms. Foundation does not publish a purchase price, but that lease rate is a fraction of what legacy defense contractors charge for far less capable unmanned systems, and it immediately attracted attention from the Pentagon and allied governments. By early 2026, Foundation had secured roughly $10 million in government contracts and announced plans to manufacture 50,000 Phantom units by the end of 2027. The robot walks at nearly 4 mph, carries over 44 pounds of payload (including weapon mounts), and is designed for modularity — swapping sensor heads, manipulator arms, or armor packages depending on the mission. CEO Sankaet Pathak has publicly stated the company's goal: make humanoid robots "as common as military trucks." The Phantom MK-1 is purpose-built for environments too dangerous for soldiers — clearing buildings in urban warfare, traversing minefields, and operating in CBRN (chemical, biological, radiological, nuclear) contaminated zones. Its bipedal form factor allows it to navigate staircases, rubble, and doorways designed for human bodies — a critical advantage over wheeled or tracked robots. U.S. Project Convergence and the Robotics Push The U.S. Army's Project Convergence — its flagship modernization exercise — has increasingly centered on integrating robotic and autonomous systems into combined-arms operations. During 2025-2026 iterations, the Army tested formations where unmanned ground vehicles operated alongside manned units, sharing sensor data through the JADC2 (Joint All-Domain Command and Control) network. Key demonstrations included autonomous resupply convoys, robotic forward observers directing artillery fire, and UGVs providing overwatch for dismounted infantry. The goal is not to replace soldiers wholesale, but to reduce the number of personnel exposed to direct fire while increasing the unit's sensor coverage and firepower. Defense Secretary Hegseth's July 2025 Robotics Memo In July 2025, U.S. Defense Secretary Pete Hegseth issued a landmark memorandum directing all service branches to accelerate the acquisition and fielding of drone and robotic systems. The memo specifically called out: Establishing dedicated robotic units within each service branch by FY2027Streamlining procurement pathways for commercial-off-the-shelf (COTS) robotic platformsIncreasing funding for human-machine teaming research by 40%Creating interoperability standards so robots from different manufacturers can operate within the same tactical network This directive signaled a fundamental shift: robotics moved from "innovation experiment" to core force structure planning. The Ukraine Proving Ground: Combat Robotics in Real War No conflict has done more to validate — and pressure-test — military robotics than the ongoing war in Ukraine. The battlefield has become a live laboratory for unmanned systems, generating lessons that every major military is studying. THeMIS and Ground Robots in Ukrainian Service Estonia's Milrem Robotics deployed its THeMIS (Tracked Hybrid Modular Infantry System) to Ukrainian forces, marking one of the first uses of a purpose-built UGV in active ground combat. THeMIS has been employed for casualty evacuation, resupply under fire, and as a mobile weapons platform mounting machine guns or anti-tank missiles. The robot's modular design proved invaluable — Ukrainian technicians could swap mission packages in the field, converting a medevac platform into a fire-support vehicle within hours. THeMIS demonstrated that even in the mud, cold, and electronic warfare environment of eastern Ukraine, ground robots could meaningfully reduce soldier exposure to enemy fire. Lyut Combat Robots Ukraine's domestically developed Lyut robotic platforms represent a different approach — smaller, cheaper, and designed for expendability. These tracked robots carry mounted weapons and operate via remote control, allowing operators to engage enemy positions from behind cover. While less sophisticated than Western systems, their low cost and rapid production reflect Ukraine's philosophy of "good enough" technology deployed at scale. FPV Drone Swarms: The Other Side of Robotic Warfare While ground robots grab headlines, Ukraine's extensive use of FPV (first-person-view) drone swarms has arguably been the most transformative robotic development of the conflict. Thousands of inexpensive kamikaze drones, often costing under $500 each, have destroyed armored vehicles, fortified positions, and infantry formations. The drone swarm concept is now being adapted to ground robotics. Both Ukrainian and Russian forces are experimenting with coordinated groups of small ground robots that can overwhelm defensive positions through numbers rather than individual capability — a direct parallel to FPV drone tactics. Types of Military Robots: A Complete Classification Explosive Ordnance Disposal (EOD) Robots EOD robots represent the most mature category of military robotics, with decades of operational deployment. These systems save lives by allowing bomb technicians to inspect, manipulate, and neutralize explosive devices from a safe distance. iRobot PackBot: The workhorse of U.S. military EOD operations since the early 2000s. Over 4,500 PackBots have been deployed, with the platform seeing extensive service in Iraq and Afghanistan. It weighs about 60 pounds, carries multiple cameras and sensors, and features a manipulator arm capable of opening doors, cutting wires, and placing counter-charges.QinetiQ Talon: Another veteran EOD platform, the Talon family includes variants optimized for CBRN detection, reconnaissance, and even weapons mounting (the SWORDS variant). Talon robots have been credited with saving hundreds of lives during IED-heavy operations. Unmanned Ground Vehicles (UGVs) UGVs represent the broadest category, ranging from small throwable scouts to multi-ton armed vehicles. These systems are designed for sustained operations in the field. THeMIS (Milrem Robotics): A 1,630-pound tracked UGV capable of carrying 1,650 pounds of payload. Its modular architecture supports weapons, sensors, medevac stretchers, or logistics cargo. Combat-proven in Ukraine.Uran-9 (Russia): A 12-ton armed UGV fielded by Russia, equipped with a 30mm autocannon, Ataka anti-tank missiles, and Shmel thermobaric rockets. The Uran-9 saw its first combat deployment in Syria in 2018, where it experienced significant communications and fire-control problems — a cautionary tale about deploying complex autonomous systems prematurely.Ghost Robotics Vision 60: A quadruped robot (robot dog) adopted by the U.S. Air Force and Marine Corps for perimeter security, base patrol, and reconnaissance. The Vision 60 can operate autonomously for extended periods, navigate rough terrain that defeats wheeled robots, and carry sensor or weapon payloads. Its dog-like form factor provides surprising stability on uneven ground. Armed Autonomous and Semi-Autonomous Systems This category represents the cutting edge — and the most controversial aspect — of military robotics. These systems can detect, track, and in some cases engage targets with varying degrees of human oversight. Samsung SGR-A1 (South Korea): Deployed along the Korean DMZ, the SGR-A1 is a stationary sentry robot equipped with a 5.56mm machine gun and a 40mm grenade launcher. It uses infrared sensors and pattern recognition to detect intruders. Critically, the system requires human authorization to fire — though it is technically capable of autonomous engagement. It represents the reality that autonomous lethal systems already exist in operational deployment.Rafael REX MKII (Israel): A remote-controlled weapon station that can be mounted on vehicles or fixed positions. The REX MKII includes AI-assisted target tracking and can be paired with various weapons from 5.56mm to 12.7mm machine guns. While not fully autonomous, its AI capabilities reduce operator workload and improve engagement speed. Humanoid Combat Robots The newest and most ambitious category, humanoid combat robots are designed to operate in environments built for humans — urban terrain, buildings, vehicles, and infrastructure. Foundation Phantom MK-1: As detailed above, the Phantom represents the first commercially viable humanoid combat platform. Its bipedal locomotion, modular payload system, and accessible lease pricing make it the benchmark against which all other humanoid military robots will be measured.DARPA Robotics Challenge Legacy: While DARPA's famous robotics challenges (2013-2015) focused on disaster response, the technologies developed — bipedal walking, manipulation, autonomous navigation — directly feed into today's military humanoid programs. Atlas, developed by Boston Dynamics for DARPA, demonstrated capabilities that the Phantom and its competitors are now commercializing. Global Military Robotics: Nation-by-Nation Comparison China's Armed Robot Dogs: A New Threat Paradigm China has emerged as perhaps the most aggressive developer of armed quadruped robots — commonly called "robot dogs." Building on the commercial success of companies like Unitree Robotics (whose consumer robot dogs cost as little as $1,600), Chinese defense firms have weaponized these platforms with rifles, grenade launchers, and sensor packages. Most alarmingly, China has demonstrated drone-deployed robot dogs — quadrupeds dropped from heavy-lift drones that activate upon landing, immediately patrolling or engaging targets. This concept combines the range and speed of aerial drones with the persistence and ground-level capability of robotic quadrupeds, creating a rapid-deployment capability that has no Western equivalent at scale. The PLA (People's Liberation Army) has conducted urban warfare exercises featuring robot dog squads advancing alongside infantry, providing reconnaissance, drawing fire to reveal enemy positions, and carrying explosive charges for breaching operations. The low cost of these platforms — often under $30,000 per unit — means they can be deployed in numbers that overwhelm traditional defenses. The Ethics of Autonomous Weapons: Humanity's Hardest Question As military robots grow more capable and more autonomous, society confronts questions that strike at the core of the laws of war and human morality. Human-in-the-Loop: The Current Standard Most Western nations officially adhere to the principle of human-in-the-loop (HITL) for lethal autonomous systems — meaning a human operator must authorize any engagement that could result in death. The U.S. Department of Defense Directive 3000.09 requires that autonomous and semi-autonomous weapon systems be designed to allow commanders and operators to exercise "appropriate levels of human judgment." In practice, the definition of "appropriate" is contested. A system that presents a human operator with a target recommendation and a two-second window to approve or reject is technically HITL — but the meaningful human judgment may be illusory when operators are overwhelmed with data and time-pressured decisions. The Autonomous Weapons Treaty Debate At the United Nations, the Convention on Certain Conventional Weapons (CCW) has hosted years of discussions on Lethal Autonomous Weapons Systems (LAWS). As of 2026, no binding treaty exists, and the prospects remain dim: Proponents of a ban (including many humanitarian organizations and some nations) argue that machines should never make life-or-death decisions, that autonomous weapons lower the threshold for going to war, and that accountability gaps make them fundamentally incompatible with international humanitarian law.Opponents of a ban (including the U.S., Russia, and China) argue that autonomous systems can be more precise than humans (reducing civilian casualties), that a ban is unverifiable, and that unilateral restraint merely cedes advantage to adversaries who won't comply.Middle-ground positions advocate for regulations requiring meaningful human control without an outright ban, focusing on specific use cases and operational contexts rather than technology categories. AI Targeting and Algorithmic Warfare The integration of AI into targeting decisions has already occurred. Israel's use of AI systems for target identification in Gaza — reported under names like "Lavender" and "Gospel" — has drawn intense scrutiny. These systems analyze vast datasets to identify suspected combatants, generating target lists at speeds impossible for human analysts. The fundamental tension: AI targeting can process more information and potentially reduce errors compared to stressed human decision-makers in chaotic combat. But AI systems also reflect their training data's biases, can produce confident-but-wrong outputs, and create accountability gaps when things go wrong. Who is responsible when an AI-assisted strike kills civilians — the programmer, the commander, the operator, or the algorithm? Future Outlook: Where Military Robotics Is Heading Goldman Sachs Projection: 50,000-100,000 Humanoid Shipments by 2026 Goldman Sachs Research has projected that 50,000 to 100,000 humanoid robots will ship globally in 2026, with military and security applications representing a significant and growing share. This projection reflects both the maturation of humanoid technology and the demand signal from defense establishments worldwide. If even 10% of those shipments go to military end-users, that represents 5,000-10,000 humanoid robots entering military inventories in a single year — a transformation comparable to the introduction of military aviation in the early 20th century. Key Trends to Watch (2026-2030) Swarm intelligence: Groups of robots operating as coordinated units, sharing information and distributing tasks without centralized control. Ground swarms will follow the aerial drone swarm model proven in Ukraine.Human-machine teaming: Mixed units where human soldiers work alongside robotic teammates, with each contributing their respective strengths — human judgment and adaptability paired with robotic endurance and precision.Logistics automation: Before combat roles mature, the most impactful near-term use of military robots will be autonomous resupply, casualty evacuation, and base maintenance — reducing the logistics "tail" that consumes the majority of military manpower.Counter-robot warfare: As robots proliferate, so will systems designed to defeat them — EMP weapons, cyber attacks targeting robot communications, and "anti-robot" munitions. The offense-defense balance will shift rapidly.Commercial-military convergence: The same humanoid robots working in warehouses and factories will be adapted for military use, creating dual-use supply chains that can scale military production rapidly in crisis. The Cost Revolution Perhaps the most significant trend is cost reduction. The Phantom MK-1, at a reported lease rate of roughly $100,000 per robot per year, is expensive compared to a consumer robot but cheap compared to military hardware. A single F-35 fighter costs over $80 million. An M1 Abrams tank costs roughly $10 million. Next to hardware like that, a leased humanoid is a rounding error. As production scales and technology matures, humanoid robot costs will continue falling. Industry analysts expect sub-$100,000 military humanoids by 2028 and potentially sub-$50,000 by 2030. At those price points, robots become genuinely expendable — fundamentally changing the calculus of attrition warfare. Implications for Military Strategy and Force Structure The integration of robots into military forces will not simply add new tools — it will reshape how militaries organize, train, and fight. Smaller Human Forces, Larger Robotic Components Nations facing demographic decline and recruiting challenges (including most Western countries) see robotics as the solution to shrinking military-age populations. A brigade that today requires 4,000 soldiers might accomplish the same missions with 2,000 soldiers and 500 robotic systems — each robot replacing the most dangerous roles. New Vulnerabilities Robotic forces create new vulnerabilities: dependence on communications links (which can be jammed or hacked), software supply chain risks, and the potential for adversaries to capture and reverse-engineer systems. Cyber warfare and electronic warfare become even more critical when the enemy's "soldiers" run on software. The Speed of Decision When both sides deploy autonomous systems, the tempo of combat may exceed human decision-making capacity. Engagements between robotic forces could unfold in seconds — too fast for human commanders to intervene. This creates pressure to grant robots greater autonomy, pushing against the human-in-the-loop principle in ways that may prove irresistible in actual combat. Compare on RoboZaps: Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # The Role of AI in Advancing Humanoid Robot Technology: A Detailed Look URL: https://blog.robozaps.com/b/role-of-ai-in-advancing-humanoid-robot-technology Author: Radica Maneva Published: 2025-03-12T00:00:00.000Z Updated: 2026-08-01T23:04:53.149Z Category: insights TL;DR: AI is what turned humanoid robots from scripted machines into adaptive ones: natural language processing and deep learning drive intuitive human-robot interaction, while learning systems let robots acquire tasks rather than be programmed. High development costs and ethics questions remain the brakes. Key takeaways: - AI advancements have significantly enhanced humanoid robots, allowing them to mimic human behavior, interact with humans intuitively, and perform tasks autonomously in various sectors such as healthcare, education, and industry. - Natural Language Processing (NLP) and deep learning algorithms are crucial in improving human-robot interaction, enabling robots to understand and respond to human language and emotions, thereby making interactions more seamless and personalized. - Despite the potential of AI-enhanced humanoid robots, challenges such as high development costs and ethical considerations regarding autonomy and human oversight must be addressed to ensure their responsible and safe deployment. FAQ: Q: What are some of the leading companies developing AI-powered humanoid robots? A: Some of the leading companies developing AI-powered humanoid robots include Boston Dynamics, Hanson Robotics, Engineered Arts, and UBTech Robotics. These companies are at the forefront of this innovative technology. Q: How does natural language processing enhance human-robot interaction? A: Natural language processing allows robots to understand human language, detect emotions, and engage in meaningful conversations, making interactions more intuitive and seamless. This enhances human-robot interaction significantly. Q: What role do AI algorithms play in humanoid robotics? A: AI algorithms such as deep learning and reinforcement learning allow humanoid robots to adjust to new tasks and environments, make advanced decisions, and improve their behaviors by interacting with their surroundings. This is important for enhancing the capabilities of humanoid robotics. Q: What are some applications of AI-enhanced humanoid robots? A: AI-enhanced humanoid robots have applications in various fields, including healthcare, manufacturing, education, and retail, where they assist in tasks like robotic surgery, customer service, and companionship for the elderly. Q: What are the ethical considerations in the development of humanoid robots? A: The ethical considerations in developing humanoid robots involve balancing robot autonomy with human oversight, ensuring safe deployment, and addressing societal impacts. It is important to carefully consider the implications of this technology. ‍ Related: The Future of Humanoid Robots: Innovation and Impact · Challenges in Humanoid Robotics and How to Overcome Them Ready to buy? Browse humanoid robots for sale on Robozaps. AI is crucial in the development of humanoid robots, allowing them to understand and respond to human emotions, perform intricate tasks, and learn from their surroundings. This article explores the role of AI in advancing humanoid robot technology, making robots more effective in healthcare, education, and other industries. ‍ ‍ AI-Powered Humanoid Robots: An Overview Humanoid robots, inspired by the science fiction narratives we’ve grown up with, are now a reality, thanks to the relentless efforts of companies like: Boston DynamicsHanson RoboticsEngineered ArtsUBTech Robotics The intelligent embodiments of these robots are designed to mimic human behavior. They hold the potential to revolutionize various areas, from healthcare to industry and beyond. The growing need for automation in many industries and significant investments in artificial intelligence (AI) technologies are what are driving the interest in humanoid robots, which is not just academic. One of the most advanced examples is Ameca by Engineered Arts, a humanoid robot that uses generative AI to interact with people, respond to questions, and exhibit creative abilities. This robot, along with others like Grace and Sophia, is equipped to support health and well-being, provide educational services, and enhance social good. The integration of AI enables robots to make autonomous decisions and adapt to changing situations. Additionally, it allows them to interact with humans more intuitively. AI’s role in humanoid robotics extends past simple automation; it bolsters the robots’ capacity to comprehend natural language, identify facial expressions, and learn from interactions. This is evident in AI robots like Atlas from Boston Dynamics, which can perform human-like physical tasks such as lifting heavy objects, running, and climbing. With the help of Figure AI, these humanoid robotics are becoming more advanced and efficient. The future of humanoid robotics looks promising, with continued advancements in AI driving the development of general purpose humanoid robot models that can perform a broader range of tasks with increasing autonomy and sophistication. ‍ Enhancing Human-Robot Interaction with AI AI’s evolution has significantly enhanced human-robot interaction, equipping humanoid robots with the skills to comprehend and react to human language and emotions. Natural Language Processing (NLP) is at the heart of this transformation, combining computational linguistics with statistical modeling, machine learning, and deep learning. NLP enables robots to understand natural language commands, making interactions more intuitive and seamless. Speech recognition, a vital component of NLP, converts voice data into text, allowing robots to follow voice commands accurately. Sentiment analysis, which enables robots to recognize attitudes or emotions in human language, further improves this capability by enhancing the context and relevance of their responses. For instance, conversational AI enables robots to engage in meaningful conversations, using NLP and machine learning algorithms to provide human-like responsiveness and personalized interactions. Moreover, advancements in robotics have led to significant improvements in various areas: Machine learning, computer vision, and NLP empower robots to infer user intent and respond intelligently to commands and queries.Planning and decision-making algorithms have advanced to autonomously generate and execute complex action sequences.Robots can now perform a wide range of tasks with minimal human supervision.Humanoid robots are efficient collaborators in various workplace scenarios, bridging communication gaps and meeting the diversified needs of different customers. ‍ AI Algorithms Driving Humanoid Robotics Deep learning, a machine learning subfield, has played a pivotal role in pushing humanoid robotics forward. Robots can learn complex patterns and relationships from large datasets, enabling them to: Make high-level decisions based on intricate informationPerform advanced tasks with greater accuracy and efficiencyAdapt to new tasks and environmentsBe more versatile and effective in dynamic settings This capacity of deep learning is crucial for the development of advanced humanoid robots, as it allows them to mimic human capabilities. Generative AI and large language models play a significant role in helping humanoid robots perceive and interpret changes in their environment. These AI models enable robots to adapt to unstructured situations, enhancing their ability to perform a variety of tasks with minimal human intervention. Companies and research institutions are continuously refining these technologies to bring robot capabilities closer to everyday applications. Reinforcement learning algorithms further enhance the adaptability of humanoid robots by allowing them to learn and refine their behaviors through interaction with their surroundings. These algorithms enable robots to develop more sophisticated reasoning and decision-making processes, making them more robust and flexible in performing diverse tasks. Future humanoid robots will use advanced control algorithms like predictive modeling and optimal control. These algorithms will help them achieve greater autonomy and adaptability. ‍ Vision Models and Visual Perception The functionality of humanoid robots heavily depends on advanced vision models and visual perception technologies. Technological advancements in cameras, computer vision, and deep learning have led to the development of robots capable of autonomous perception and interaction. These robots use vision sensors like cameras and radar to gather optical images and depth data, which computer vision algorithms then process for tasks like object recognition and navigation. The integration of advanced sensors like LIDAR, gyroscopes, and accelerometers further enhances their ability to perceive and interact with their environment effectively. ‍ The Role of AI in Motor Skills and Movement Crafting humanoid robots capable of emulating human-like movements presents a complex challenge, requiring the coordination of numerous actuators and joints, the formulation of intricate control algorithms, and the design of efficient feedback mechanisms. Reinforcement learning algorithms allow these robots to learn and refine their behaviors through interaction with their environment, enabling them to adapt to dynamic and unstructured environments. Advanced control algorithms like model-based control enable robots to anticipate and respond to dynamic changes in their environment, making their movements more fluid and natural. The integration of advanced actuators and sensors further enhances the robots’ ability to perform tasks requiring fine motor skills, ensuring versatile and reliable movements. These advancements in AI, particularly in computer vision and natural language processing, offer promising avenues for overcoming challenges and creating robots that can work precisely in dynamic environments. ‍ Applications of AI-Enhanced Humanoid Robots Various sectors have seen significant transformations in task execution due to the applications of AI-enhanced humanoid robots. In healthcare, for instance, these robots can assist in robotic surgery, offering precision and reducing complications in procedures like open-heart surgery. They can also automate administrative tasks such as appointment scheduling and managing patient records, thus improving operational efficiency. Hanson Robotics created robots like Grace with the express purpose of assisting in healthcare settings and carrying out COVID-19 pandemic-related tasks. In the education sector, humanoid robots can act as interactive tutors, offering customized lessons and tailored learning experiences. These robots can engage students in interactive activities, making learning more engaging and effective. In the entertainment industry, humanoid robots are being used to create dance routines, participate in interactive performances, and even act in films. In manufacturing, AI-enhanced humanoid robots assist with assembly, picking and packing, and quality control, thereby increasing efficiency and productivity. They can perform repetitive tasks in manufacturing and logistics, enhancing operational efficiency and reducing the burden on human workers. In retail, hospitality, and service industries, humanoid robots interact with customers, providing personalized assistance and improving customer experience. In addition, humanoid robots can assist caregivers in patient care by providing support and companionship for the elderly, as well as helping people with disabilities in their daily tasks. Robots like Moxi from Diligent Robotics support non-clinical tasks in hospitals, such as deliveries, restocking, and collection of samples. These diverse applications highlight the potential of AI-enhanced humanoid robots to transform various workplace scenarios and improve the quality of services provided. ‍ Challenges and Ethical Considerations Despite the promising advancements in humanoid robotics, considerable challenges and ethical dilemmas need to be addressed. The development of these robots entails substantial costs in research, development, and manufacturing, making scalability and affordability major hurdles. Ensuring that these robots are accessible for various applications requires continuous efforts to reduce costs and improve efficiency. Ethical considerations also play a crucial role in the development and deployment of humanoid robots. Balancing the autonomy of these robots with human oversight is essential to ensuring their safe and responsible use. Striking the right balance between technological capabilities and ethical oversight is paramount, particularly in applications where safety and societal impacts are significant. These challenges highlight the need for ongoing dialogue and collaboration among stakeholders to address the ethical dilemmas and societal impacts of humanoid robotics. ‍ Future Trends in Humanoid Robotics The ongoing development of AI technologies, particularly machine learning, computer vision, and natural language processing, has an impact on the direction that humanoid robotics will take in the future. These advancements are expected to enhance the capabilities of robots, making them more intuitive and effective in various applications. Institutions like Carnegie Mellon University and MIT are leading research in this field, pushing the boundaries of what is possible. One of the significant trends is the integration of humanoid robots with cutting-edge technologies to address global challenges. The United Nations’ AI for Good initiative aims to leverage AI to address the 17 Sustainable Development Goals, showcasing robots designed to tackle these global issues. Companies like Hanson Robotics and UBTech Robotics are at the forefront of building AI-powered humanoid robots, focusing on refining human-robot interaction and promoting social good. Looking ahead, there is a strong growth outlook for the robotics industry, with AI-enhanced robots playing a crucial role in various sectors. The future will see increased integration of humanoid robots in the workplace, enhancing productivity and opening up new possibilities for human-robot collaboration. These trends underscore the transformative potential of AI-powered humanoid robots and their role in shaping the future of work and society. ‍ Summary In summary, AI-powered humanoid robots are set to revolutionize various industries, offering unprecedented capabilities in mobility, interaction, and autonomy. These robots, driven by advancements in AI technologies, are becoming more integrated into our workplaces, enhancing productivity and opening up new possibilities for human-robot collaboration. The role of AI in humanoid robotics is multifaceted, enhancing human-robot interaction, enabling robots to adapt to new tasks and environments, and improving their motor skills and movement. Despite the challenges and ethical considerations, the future of humanoid robotics looks promising, with continuous advancements in AI driving the development of robots that can perform a broader range of tasks with increasing autonomy and sophistication. As we look to the future, it is clear that AI-powered humanoid robots will play a crucial role in addressing global challenges and promoting social good. By embracing these technologies, we can unlock new opportunities and create a more efficient, productive, and inclusive world. The journey of humanoid robotics is just beginning, and the possibilities are limitless. ‍ Compare on RoboZaps: Sophia, Ameca and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Applications of Humanoid Robots: 12 Industries Being Transformed in 2026 URL: https://blog.robozaps.com/b/applications-of-humanoid-robots Author: Radica Maneva Published: 2025-03-12T00:00:00.000Z Updated: 2026-08-01T23:05:04.490Z Category: insights TL;DR: Roughly 16,000 humanoid robots were installed globally in 2025 — over 80% in China, led by AgiBot's 31% share — across 12 industries from factory floors to hospitals. Manufacturing costs fell about 40% year over year, with entry units from $5,900, and Goldman Sachs projects cumulative installations past 100,000 by 2027. Key takeaways: - Roughly 16,000 humanoid robots were installed globally in 2025 (Counterpoint Research), with China accounting for over 80% of installations. - AgiBot led 2025 installations with a 31% share, ahead of Unitree (27%), with UBTECH, Leju and Tesla around 5% each. - Manufacturing costs fell about 40% year over year, with entry platforms like the Unitree R1 from $5,900 and mid-range commercial units at $30,000-$150,000. - Deployments span 12 industries, with manufacturing, logistics and healthcare furthest along — Moxi, for example, handles routine hospital logistics that consume large shares of nursing time. - Goldman Sachs projects cumulative humanoid installations to exceed 100,000 units by 2027. FAQ: Q: What are the most common applications of humanoid robots in 2026? A: The most common applications of humanoid robots are in manufacturing and logistics (warehouse material handling, assembly line support), healthcare (patient interaction, rehabilitation), education (STEM teaching, tutoring), retail (customer service, product demonstration), and entertainment. Manufacturing and logistics dominate commercial deployments, accounting for the majority of the estimated 16,000 units installed globally in 2025. Q: How much do humanoid robots cost? A: Humanoid robot prices vary dramatically based on capability. Entry-level platforms like the Unitree R1 start at approximately $5,900. Mid-range commercial humanoids cost $30,000–$150,000. Advanced research and industrial humanoids can exceed $150,000. Manufacturing costs have declined 40% year-over-year, and Robot-as-a-Service (RaaS) rental models offer access without large capital expenditure. Q: Which companies lead humanoid robot production? A: As of 2025, AgiBot leads global installations with 31% market share, followed by Unitree (27%), UBTECH (~5%), Leju (~5%), and Tesla (~5%). In the Western market, Agility Robotics (Digit), Figure AI (Figure 02), Apptronik (Apollo), and Boston Dynamics (Atlas) are prominent. Tesla plans to scale Optimus production to 100,000 units by 2026. Q: Can humanoid robots replace human workers? A: Current humanoid robots augment rather than replace human workers. They excel at repetitive, physically demanding, or dangerous tasks—freeing humans for complex decision-making and creative work. Companies reporting the highest workforce acceptance emphasize augmentation over replacement. Full labor substitution remains years away due to limitations in dexterity, adaptability, and unstructured problem-solving. Q: What is the ROI of deploying humanoid robots? A: Early commercial deployments report ROI payback periods of 18–36 months, driven by labor cost savings, 24/7 operation capability, reduced workplace injuries, and consistent output quality. ROI improves in high-labor-cost regions and for tasks involving hazardous conditions or severe labor shortages. RaaS models reduce upfront risk further. Q: How are humanoid robots used in healthcare? A: Humanoid robots in healthcare perform hospital logistics (delivering supplies and medications), patient engagement (companionship, cognitive exercises), physical rehabilitation coaching, and telemedicine support. Robots like Moxi reduce nurse walking time by up to 30%, while Pepper provides cognitive stimulation for dementia patients in over 2,000 healthcare facilities. Q: What does the future hold for humanoid robot applications? A: Goldman Sachs projects cumulative humanoid robot installations will exceed 100,000 units by 2027. Near-term growth centers on structured industrial tasks (logistics, manufacturing, automotive). Mid-term expansion (2026–2028) will include more complex pick-and-place operations, multi-step assembly, and broader service-sector adoption. Long-term applications include domestic assistance, construction, and fully autonomous space exploration operations. From threading needles on factory floors to guiding patients through rehabilitation exercises, the applications of humanoid robots crossed a critical threshold in 2025: 16,000 units deployed globally, with over $7 billion invested in China's robotics sector alone. As we enter 2026, the question has shifted from "can humanoid robots work in the real world?" to "which industries will they transform first?" This is the most comprehensive guide to humanoid robot applications available anywhere—covering 12 major industries with real deployment data, specific robots in use, named companies, measurable results, and expert projections for what comes next. Whether you're evaluating humanoid automation for your business or tracking the industry's trajectory, this is your definitive resource. Applications of Humanoid Robots Across Industries: The 2026 Landscape The applications of humanoid robots have expanded at a breathtaking pace. In 2025, global installations reached an estimated 16,000 units according to Counterpoint Research, with projections exceeding 100,000 cumulative units by 2027. What was once confined to research labs and science fiction now spans manufacturing floors, hospital corridors, classrooms, and even outer space. This comprehensive guide examines every major application area for humanoid robots in 2026, featuring real-world deployments, specific robots in use, measurable results, and expert analysis on where the technology is headed next. Manufacturing and Industrial Automation Deep dive: our full guide to humanoid robots in the workplace. Manufacturing represents the largest near-term opportunity for humanoid robot applications. The human form factor allows these robots to operate in facilities designed for people—navigating doorways, climbing stairs, and using standard tools—without costly facility redesigns. Automotive Assembly Lines The automotive sector leads adoption. BMW is piloting Figure 02 robots at its Spartanburg, South Carolina plant for material handling and parts delivery. Mercedes-Benz has partnered with Apptronik to deploy Apollo humanoids for assembly line support tasks. Tesla uses its own Optimus Gen 2 robots internally at the Fremont factory, handling component delivery to human assembly workers. Chinese automaker BYD is projected to scale from 1,500 humanoid robots in 2025 to 20,000 by 2026 (IDTechEx estimate), integrating them across its EV production lines. UBTECH's Walker S robots are already performing quality inspection tasks on automotive factory floors. Warehouse and Logistics Operations Agility Robotics' Digit represents the most commercially advanced warehouse humanoid. In testing at Amazon and commercially deployed at GXO Logistics and a Spanx warehouse in Georgia—the first documented revenue-generating commercial humanoid deployment—Digit handles material movement, moves totes and bins along mapped routes, and manages payloads up to 35 pounds across a six-foot reach range. Agility Robotics operates a factory in Oregon capable of building over 10,000 Digit units per year. Figure AI announced its BotQ manufacturing facility in Austin, Texas, with 12,000-unit initial capacity. Key Manufacturing Data Manufacturing costs declined 40% year-over-year, from $50,000–$250,000 per unit in 2023 to $30,000–$150,000 in 2024Unitree Robotics launched the R1 humanoid at just $5,900 in mid-2025Target uptime for commercial deployments: 85–95% availabilityTypical ROI payback period: 18–36 months Healthcare and Medical Applications The uses of humanoid robots in healthcare range from surgical assistance to patient interaction and rehabilitation support. Their human-like form makes them uniquely suited for environments where patient comfort and trust matter. Surgical and Clinical Support Moxi by Diligent Robotics handles routine hospital logistics—delivering lab samples, medications, and supplies—freeing nurses to focus on patient care. In clinical trials, Moxi handles routine tasks that consume up to 30% of nurses' time. Humanoid platforms are also being tested for telemedicine, allowing remote physicians to physically examine patients through robotic intermediaries. Rehabilitation and Therapy Humanoid robots serve as physical rehabilitation coaches, guiding patients through exercises with consistent form correction and motivational interaction. Japan's Pepper robot has been deployed in healthcare facilities across Asia and Europe for patient engagement, providing cognitive stimulation exercises for dementia patients and companionship during long hospital stays. France's Mirokaï robot assists nursing staff at Broca Hospital (AP-HP), supporting both patient interaction and care coordination tasks. Education and Academic Research Humanoid robots in education are transforming how students learn STEM subjects, languages, and social skills. Their human-like appearance creates natural engagement that screens and traditional teaching tools cannot match. Classroom Integration SoftBank's NAO robot is used in thousands of schools worldwide for interactive language instruction, mathematics tutoring, and programming education. Students can program NAO using visual block-based interfaces or Python, making it an effective bridge between abstract coding concepts and tangible, physical results. Poppy Humanoid, an open-source platform, is widely adopted in engineering schools, FabLabs, and secondary education. Its fully customizable design allows student groups to build, modify, and program specific robot components—from designing mechanical parts to adding sensors and programming behaviors. Research Platforms Booster Robotics' K1 (95 cm tall, 19.5 kg) serves as a portable research and education platform—transportable in a suitcase—ideal for robotics competitions like RoboCup. The Booster T1 provides an open-source humanoid for advanced research, including testing navigation algorithms and human-robot interaction paradigms. Boston Dynamics' Atlas (now fully electric) continues as a premier research platform, pushing boundaries in dynamic movement, whole-body manipulation, and AI-driven autonomous behavior. Elderly Care and Assisted Living With aging populations worldwide, humanoid robots in elderly care address a growing labor shortage in caregiving. Japan alone faces a projected shortfall of 700,000 care workers. Companionship and Monitoring Pepper and NAO are deployed in Japanese and European care homes for daily companionship, medication reminders, cognitive exercises, and fall detection alerts. Studies published in the International Journal of Social Robotics show elderly residents interacting with humanoid robots report reduced loneliness and improved mood over 12-week periods. Physical Assistance Toyota's Human Support Robot (HSR) helps elderly individuals with limited mobility by retrieving objects, opening doors, and supporting daily tasks. Reachy by Pollen Robotics demonstrates kitchen assistance capabilities—opening refrigerators, cleaning surfaces—offering a glimpse of future domestic care robots. Robot-as-a-Service (RaaS) models are gaining traction in elderly care, lowering the barrier to adoption for care facilities that cannot afford upfront capital expenditure on robotics. Military and Defense Humanoid robots in military and defense applications focus on reducing risk to human soldiers in dangerous environments while leveraging the human form factor to operate equipment and navigate structures built for people. 2025–2026 Developments Foundation's Phantom MK-1 became the first humanoid robot deployed to an active combat zone when two units arrived in Ukraine in February 2026 for battlefield testing. Standing 175cm (5'9") tall, the Phantom MK-1 can carry rifles, breach doors, and provide reconnaissance support. Foundation plans to scale to 10,000 units in 2026 and 50,000 by 2027. The U.S. Army tested humanoid-adjacent platforms alongside manned units through the JADC2 (Joint All-Domain Command and Control) network during 2025–2026 exercises. China's PLA continues demonstrating humanoid robots for reconnaissance and equipment operation. The Phantom MK-2, expected in April 2026, promises waterproofing, extended battery life, and 80 kg payload capacity. Explosive Ordnance and Reconnaissance Humanoid robots can enter buildings, climb stairs, and manipulate objects in ways that wheeled robots cannot—critical advantages for bomb disposal, hostage scenarios, and urban combat reconnaissance. Their ability to use human tools and equipment without modification reduces the logistics burden of specialized robotic attachments. Retail and Customer Service Humanoid robots in retail serve as interactive store assistants, product demonstrators, and customer engagement tools. Their novelty factor drives foot traffic while their AI capabilities provide genuine utility. Deployments in Action Pepper operates in over 2,000 retail locations across Japan, Europe, and the United States as a greeter, product guide, and information kiosk. SoftBank reports that stores deploying Pepper see increased customer dwell time and higher engagement with promoted products. In China, humanoid robots from AgiBot (the global market leader by installations in 2025 with 31% share) are deployed in retail environments, shopping malls, and promotional events. RaaS rental models allow retailers to deploy humanoids for seasonal campaigns or special events without long-term capital commitments. Hospitality and Tourism Deep dive: our full guide to humanoid robots in hospitality. Hotels, airports, museums, and entertainment venues increasingly deploy humanoid robots for guest services. Their multilingual capabilities and tireless availability make them ideal for high-traffic hospitality environments. Notable Deployments Japan's Henn-na Hotel chain pioneered humanoid robot staff, using robots for check-in, concierge services, and luggage assistance. Airports including Tokyo Haneda and Munich Airport deploy humanoid robots for wayfinding, flight information, and passenger assistance. Museums worldwide use Pepper and custom humanoid platforms as interactive guides, delivering exhibit information in multiple languages while collecting visitor analytics. Booster Robotics demonstrated humanoid robots collecting waste at live events, showcasing hospitality applications beyond guest interaction. Agriculture and Food Production Deep dive: our full guide to humanoid robots in agriculture. While traditional agricultural robots are purpose-built machines, humanoid robots offer unique advantages in unstructured farm environments where terrain, crop variety, and task diversity demand human-level adaptability. Emerging Applications Humanoid robots are being tested for fruit harvesting, where their dexterous hands and bipedal mobility allow them to navigate uneven orchard terrain and handle delicate produce. Agility Robotics has explored agricultural logistics applications for Digit, including moving harvested goods between collection points. Greenhouse operations represent a nearer-term opportunity: controlled environments reduce navigation complexity while tasks like pruning, pollination monitoring, and plant inspection leverage humanoid manipulation capabilities. Disaster Response and Search-and-Rescue Deep dive: our full guide to humanoid robots in disaster response. Disaster zones present exactly the kind of unstructured, human-designed environments where humanoid robots excel over wheeled or tracked alternatives. Stairs, ladders, narrow corridors, and rubble fields all favor bipedal mobility. Real-World Capabilities Boston Dynamics' Atlas has demonstrated traversing rubble, opening doors and valves, and using power tools in simulated disaster scenarios. Japan's HRP series robots, developed by AIST, are specifically designed for disaster response in earthquake-prone environments. KAIST's DRC-HUBO proved humanoid viability for disaster response by completing a complex course including driving a vehicle, walking through rubble, cutting through walls, and climbing stairs—all autonomously. These capabilities directly translate to nuclear facility emergencies, building collapses, and hazardous material incidents. Space Exploration Space agencies invest heavily in humanoid robots because spacecraft, habitats, and equipment are designed for human operators. A humanoid robot can use the same tools, panels, and controls as astronauts. Active Space Humanoids NASA's Robonaut 2 (R2) operated aboard the International Space Station, performing routine maintenance tasks and testing human-robot collaboration in microgravity. Valkyrie (R5), NASA's next-generation humanoid, is designed for deep-space missions where robots must operate autonomously in habitats for months before human arrival. China is developing a semi-humanoid robot on a wheeled platform for its lunar research station (target: 2035), with potential deployment on the Chang'e 8 mission as early as 2028. India's ISRO will launch Vyomitra ("space friend")—a humanoid robot—on its G1 uncrewed orbital mission ahead of the Gaganyaan crewed flight. The European Space Agency continues funding humanoid research for Mars surface operations where communication delays make teleoperation impractical. Entertainment and Social Interaction Entertainment remains one of the most visible applications of humanoid robots, from theme park performers to social media personalities. High-Profile Examples Sophia by Hanson Robotics became the world's most famous robot, appearing on talk shows, addressing the United Nations, and receiving Saudi Arabian citizenship. While primarily a social AI demonstration platform, Sophia catalyzed public interest in humanoid robotics worldwide. AgiBot deployed over 5,000 humanoid units in 2025 across entertainment, hospitality, and live performance venues in China. Disney's theme parks use sophisticated animatronic humanoids for character experiences, with increasing AI integration enabling unscripted guest interactions. Live performance robotics is growing: humanoid robots serve as DJs, stage performers, and event hosts, with RaaS models making them accessible for corporate events and trade shows. The Global Market Landscape in 2026 Understanding the applications of humanoid robots requires context on the rapidly evolving market: 16,000 units installed globally in 2025 (Counterpoint Research)China accounts for 80%+ of all installationsAgiBot leads with 31% market share, followed by Unitree (27%), UBTECH (~5%), Leju (~5%), and Tesla (~5%)$7 billion invested across 610+ robotics deals in China in the first nine months of 2025Counterpoint Research projects cumulative installations exceeding 100,000 units by 2027Logistics, manufacturing, and automotive expected to represent 72% of annual installations by 2027 Application Comparison Table What Changed at CES 2026 CES 2026 (January 2026) represented a watershed moment for humanoid robot applications. Key announcements that reshape the landscape: AgiBot debuted in the U.S. market with the most complete humanoid portfolio shown at any trade show—A2 (service), G2 (industrial/domestic), X2 (entertainment), and D1 (quadruped). Bloomberg confirmed AgiBot as the #1 humanoid producer by shipments.Unitree showcased the H2 for industrial applications and confirmed a Robot-as-a-Service model for global commercial deployment, alongside quick-swappable batteries and enhanced payload capacity.LG Electronics debuted CLOiD as the centerpiece of its "Zero Labor Home" vision—signaling that major consumer electronics companies are entering the humanoid space.1X Technologies confirmed NEO is available for pre-order, with first customer shipments planned for 2026.Figure AI continued scaling its BotQ factory, targeting 12,000-unit initial capacity for Figure 02 production. The consistent theme across CES 2026: humanoid robots are transitioning from pilot programs to production-ready commercial systems with defined pricing, service models, and deployment playbooks. The Robozaps Take: What Actually Matters in 2026 We've spent the past year helping enterprises evaluate and deploy humanoid robots. Here's what the market reports don't tell you: The Gap Between Demos and Deployment Is Still Massive For every CES backflip, there are a hundred failed pilot programs. The companies succeeding with humanoids right now share three traits: they're solving boring problems (tote moving, not surgery), they have realistic timelines (18-month pilots, not 90-day miracles), and they treat robots as tools, not replacements. The enterprises chasing "lights-out" factories are burning cash. The ones automating specific bottlenecks are seeing ROI. Manufacturing and Logistics Are the Only Proven Applications—For Now Despite the 12 industries covered above, only manufacturing and logistics have repeatable, profitable deployments at scale. Healthcare "deployments" are mostly PR exercises. Retail humanoids are novelty plays with negative ROI. Elderly care remains a research project dressed as a product. If you're evaluating humanoids today, start where the economics actually work: warehouses, assembly lines, and material handling. Everything else is a bet on the future. China's 80% Market Share Isn't Just About Cost Western companies assume Chinese dominance is purely about cheap labor and subsidies. They're wrong. AgiBot and Unitree are shipping production-ready systems while most Western competitors are still raising Series B rounds. The real gap is iteration speed: Chinese manufacturers cycle through hardware revisions in months, not years. For enterprise buyers, this means the most capable, affordable humanoids available today are Chinese—and that won't change until Western production catches up. What We Tell Enterprise Buyers Skip the hype. Start with a single, measurable use case—not a "digital transformation initiative." Budget for 18–36 months before ROI. Demand uptime guarantees in your contracts. And don't wait for the "perfect" robot: the companies deploying imperfect humanoids today are building the operational expertise that will matter when the technology matures. —The Robozaps Team Compare on RoboZaps: Phantom MK-1, Pepper, Sophia, Walker S, Figure 02 and Digit. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs Astribot S1: Full 2026 Comparison (Price, Specs, Dexterity) URL: https://blog.robozaps.com/b/tesla-optimus-vs-astribot-s1 Author: Radica Maneva Published: 2025-03-06T00:00:00.000Z Updated: 2026-08-01T11:20:30.451Z Category: comparisons TL;DR: Tesla Optimus targets under $30,000 with mass production planned from late 2026; the Astribot S1 is quote-only (Chinese reports cite about ¥500,000 for the research edition) but ships to organizations today, with 10 m/s arm speed and ±0.1 mm precision. The S1 wins on availability and manipulation; Optimus on target price and scale. Key takeaways: - Tesla Optimus stands 173 cm (5'8") tall, weighs 57 kg (126 lbs), and targets a price under $30,000 — designed for mass manufacturing and eventual consumer adoption with Tesla's FSD-derived AI stack. - Astribot S1 is quote-only (reports cite about ¥500,000 for the research edition) with 23 degrees of freedom, a 5 kg (11 lbs) per-arm payload, and end-effector speeds up to 10 m/s — built for precision manipulation and service applications. - Astribot S1 wins on dexterity, DOF count, and current availability. Tesla Optimus wins on target price, payload capacity, and scalable production. - Both robots are transitioning from prototypes to production — Tesla targeting mass production by late 2026, Astribot S1 shipping to organizations by quote. FAQ: Q: How much does Tesla Optimus cost vs Astribot S1? A: Tesla Optimus targets a price of $20,000–$30,000 at mass production, though it's not yet available for purchase. Astribot S1 is sold by quote, with Chinese reports citing about ¥500,000 (~$69,000) for the research edition, but S1 delivers today while Optimus delivery is years away. Q: Which robot has more degrees of freedom? A: Astribot S1 has 23 degrees of freedom on a wheeled base versus Optimus's 28 on legs; its edge is arm speed (up to 10 m/s) and ±0.1 mm repeatability rather than joint count — enabling fast, coordinated movements that Optimus has not demonstrated. Q: Can Tesla Optimus really compete with Astribot S1's dexterity? A: Not currently. Astribot S1's 52 DOF and imitation learning demonstrations show manipulation capabilities beyond what Optimus Gen 2 has publicly demonstrated. However, Tesla's rapid iteration pace and fleet learning approach could close this gap. Gen 3 Optimus (expected 2026) may narrow the dexterity difference. Q: When can I buy Tesla Optimus? A: Tesla targets consumer availability in late 2027. Mass production for internal use begins late 2026. No pre-orders are available. Astribot S1 is the only one of these robots available for purchase in 2026. Q: Is Astribot S1 worth 3x the price of Tesla Optimus? A: It depends on your needs. If you need a robot immediately with maximum dexterity for precision tasks, the S1's quote-only price may be justified. For fleet deployment, manufacturing, or any application where price sensitivity matters, Tesla's target price makes Optimus the clear choice — if you can wait until 2027. Q: Which robot is better for household tasks? A: Both target household applications, but from different angles. Astribot S1's superior dexterity makes it better for tasks requiring fine motor control (folding clothes, handling dishes). Tesla Optimus's bipedal locomotion makes it better for navigating multi-level homes and carrying heavier items. S1 is available now; Optimus requires waiting. Tesla Optimus vs Astribot S1 — which humanoid robot delivers better performance for 2026? This head-to-head comparison breaks down every specification, capability, and real-world application to help you understand exactly how Tesla's mass-market vision stacks up against Astribot's precision-first approach. Tesla Optimus targets a sub-$30,000 price point for mass deployment in factories and homes. Astribot S1 is sold by quote (Chinese reports cite about ¥500,000 for the research edition) and brings blistering arm speed and exceptional manipulation capabilities. Both represent fundamentally different strategies for the humanoid revolution. Head-to-Head: Tesla Optimus vs Astribot S1 Specifications Tesla Optimus: Everything You Need to Know (2026 Update) Tesla Optimus represents Elon Musk's vision for a general-purpose humanoid robot that can be mass-produced at automotive scale. Leveraging Tesla's Full Self-Driving AI stack and manufacturing expertise, Optimus aims to become "the most significant product Tesla has ever made." Design and Build At 173 cm (5 ft 8 in) and 57 kg (126 lbs), Tesla Optimus is designed to operate in human spaces. Gen 2 introduced tactile sensors in the fingertips, enabling more delicate manipulation tasks. The design prioritizes manufacturability — Tesla's Fremont factory is being repurposed from Model S/X production specifically for Optimus manufacturing. Core Technologies FSD-Derived AI: Vision-based neural networks from Tesla's autonomous driving program, adapted for bipedal navigation and task execution.28 Degrees of Freedom: Full-body articulation including dexterous hands with 11 DOF each.20 kg Payload: Capable of carrying substantial loads — double the Astribot S1's capacity.Fleet Learning: Data from every Optimus unit feeds back into training, accelerating capability improvements. Price and Availability Tesla targets a price under $30,000 at mass production scale — potentially as low as $20,000. Mass production is planned before end of 2026, with consumer availability targeted for late 2027. No pre-orders are currently available. Optimus units are currently deployed internally at Tesla factories. Astribot S1: Everything You Need to Know (2026 Update) Astribot S1, developed by Chinese company Stardust Intelligence (), takes a different approach — prioritizing dexterity and precision over affordability. The S1 made headlines with demonstrations showing exceptional speed and manipulation capabilities, including pouring liquids and handling delicate objects. Design and Build Standing 170 cm (5 ft 7 in) tall and weighing 60 kg (132 lbs), the Astribot S1 has 23 degrees of freedom on an omnidirectional wheeled base — fewer joints than Optimus, but with end-effector speeds up to 10 m/s and ±0.1 mm repeatability. The design emphasizes upper-body dexterity; the S1 is wheeled, not legged. Core Technologies 52 Degrees of Freedom: Industry-leading articulation for complex manipulation tasks.Imitation Learning: The S1 can learn tasks by watching human demonstrations — reducing programming time.2.0 m/s Speed: Arm movements at up to 7.2 km/h (4.5 mph) for rapid task execution.4-Hour Runtime: 1,500 Wh battery enables extended operational periods. Price and Availability The Astribot S1 is sold by quote, with Chinese reports citing about ¥500,000 (~$69,000) for the research edition. Unlike Tesla's future-focused promises, Astribot offers a tangible product today — though at well above Tesla's target price. Primary markets are China and Asia-Pacific, with global distribution expanding. Head-to-Head: Tesla Optimus vs Astribot S1 Performance Comparison 1. Agility and Mobility Winner: Tesla Optimus Tesla Optimus offers true bipedal locomotion with walking speeds up to 5 km/h (3.1 mph) and running capability at 8 km/h (5 mph). The Optimus can navigate stairs, uneven terrain, and tight spaces designed for humans. Astribot S1's wheeled base option limits it to flat surfaces, while the bipedal version has not demonstrated comparable locomotion capabilities. For applications requiring movement through human-designed environments, Optimus has a clear advantage. 2. Dexterity and Manipulation Winner: Astribot S1 With end-effector speeds up to 10 m/s and ±0.1 mm repeatability, the Astribot S1 delivers extremely precise manipulation. Demo videos show S1 performing tasks like pouring liquids, handling eggs, and ironing clothes with impressive fluidity. Tesla has demonstrated similar tasks, but Astribot's speed and precision enable notably fluid, coordinated movements. For applications requiring fine motor control and dexterity, the S1 leads. 3. AI and Software Winner: Tesla Optimus Tesla's FSD-derived neural networks represent billions of miles of real-world training data adapted for robotics. The fleet learning approach means every Optimus unit contributes to system-wide improvements. Astribot's imitation learning is effective but operates at a smaller scale. Tesla's integration with its broader AI ecosystem — including Dojo supercomputer training — provides a significant long-term advantage. 4. Sensors and Perception Winner: Tie Both robots use camera-based vision with force-torque sensing in the hands. Tesla's vision-only approach (no LiDAR) mirrors its automotive philosophy, while Astribot includes depth sensors for enhanced spatial awareness. Neither has a decisive advantage — both can perceive their environment well enough for intended tasks. 5. Price and Value Winner: Tesla Optimus At a target price of $20,000–$30,000 versus Astribot's reported ~$69,000 research edition, Tesla targets dramatically better value per robot. Even accounting for Astribot's immediate availability, the price gap makes Tesla the likely winner for fleet deployment — if the target price holds. 6. Build Quality and Durability Winner: Tie (Insufficient Data) Neither company has published extensive durability testing or IP ratings. Tesla's automotive manufacturing expertise suggests robust production quality, while Astribot has demonstrated reliable operation in demos. Real-world durability data will emerge as both robots see wider deployment. 7. Real-World Deployment and Availability Winner: Astribot S1 Astribot S1 is sold by quote to organizations now. Tesla Optimus is not available for purchase — consumer sales aren't expected until late 2027. For organizations that need a humanoid robot in 2026, Astribot S1 is the only option of these two. Availability matters, and Astribot delivers today while Tesla promises tomorrow. Which Should You Choose? Choose Tesla Optimus if you: Plan to deploy at scale: Tesla's target price of $20,000–$30,000 enables fleet deployments that would be cost-prohibitive with S1's quote-only enterprise pricing.Need bipedal navigation: Optimus walks and runs on legs, navigating human environments including stairs — essential for factories and homes designed for people.Prioritize payload capacity: At 20 kg (44 lbs), Optimus can carry twice what S1 handles — critical for manufacturing and logistics applications.Can wait until 2027: If your timeline allows, Tesla's economies of scale will deliver more capability per dollar than any competitor. Choose Astribot S1 if you: Need a robot now: S1 is available for pre-order today. Optimus won't reach consumers until late 2027 at the earliest.Require maximum dexterity: Its arm speed and precision enable manipulation tasks that slower platforms struggle to match.Focus on service applications: Hospitality, healthcare, and household tasks requiring delicate manipulation are S1's sweet spot.Operate in flat environments: If your use case doesn't require stair climbing or rough terrain, S1's wheeled option provides stability and simplicity. Final Verdict: Tesla Optimus vs Astribot S1 Overall Winner: Tesla Optimus — with a major caveat. Tesla Optimus wins on value proposition: dramatically lower price, higher payload, full bipedal mobility, and the backing of Tesla's manufacturing scale. If you're planning a humanoid robot deployment and can wait until 2027, Optimus is the rational choice. But Astribot S1 wins on availability and dexterity. For organizations that need a humanoid robot today, or applications requiring maximum manipulation precision, S1 delivers capabilities that don't yet exist in Optimus. It is an enterprise-priced choice — but it's a choice you can actually make right now. The humanoid robotics market in 2026 is defined by this trade-off: Tesla's future promise versus Astribot's present reality. For most buyers, the smart move is to pilot with available technology (S1) while planning for Tesla's mass-market disruption. Compare both robots side by side: Tesla Optimus on Robozaps | Astribot S1 on Robozaps | Browse all humanoid robots Last updated: February 6, 2026. Specifications sourced from official manufacturer documentation and the Robozaps robot database. Robozaps is a humanoid robot marketplace — we maintain comprehensive product databases and may earn referral fees from qualifying purchases. Related: Tesla Optimus Gen 2 Review · Astribot S1 Review · Best Humanoid Robots 2026 Ready to buy? Browse humanoid robots for sale on Robozaps. --- # Figure 02 Price & Specs 2026: For Sale? + Figure 03 vs 02 URL: https://blog.robozaps.com/b/figure-02-review Author: Radica Maneva Published: 2025-03-05T00:00:00.000Z Updated: 2026-08-01T23:04:16.584Z Last fact-checked: 2026-07-12T00:00:00.000Z Category: reviews TL;DR: The Figure 02 was never for sale: it had no MSRP, and the "$30k-$150k" estimates are unverified. It is the previous generation, superseded by the Figure 03 in October 2025, and its legacy is the BMW Spartanburg deployment, where it helped build 30,000-plus vehicles over about 11 months before the fleet was retired. Key takeaways: - Never for sale, no public price. Figure 02 had no MSRP and was never sold to consumers or businesses off the shelf. You cannot buy one. - Ignore the price estimates. The "$30k to $150k" figures are unverified and inconsistent; the "$20,000" is a future target, not a Figure 02 price. - It is the previous generation, superseded by the Figure 03 (October 2025), which is 9% lighter and home-focused. - Its real claim to fame is BMW. Figure 02 worked at BMW's Spartanburg plant for about 11 months, helping build 30,000-plus vehicles before the fleet was retired around November 2025. - It was the OpenAI-era robot. Figure 02 launched using OpenAI-trained models; Figure ended that partnership in February 2025 and switched to its own Helix AI. FAQ: Q: How much does the Figure 02 cost? A: There is no official price. Figure never published an MSRP for the Figure 02 and never sold it to the public. Third-party estimates of $30,000 to $150,000 are unverified, and the "$20,000" figure is a future mass-production target, not a Figure 02 price. Q: Can you buy a Figure 02? A: No. It was an industrial unit deployed only through enterprise partnerships such as BMW. There is no consumer version and no way to order one. Q: Is the Figure 02 still the latest Figure robot? A: No. Figure 03, launched in October 2025, is the current generation. It is about 9% lighter, home-focused, and has newer hands and wireless charging. Figure 02 is the previous model. Q: What did the Figure 02 do at BMW? A: It worked in BMW's Spartanburg body shop for about 11 months, loading parts onto welding fixtures and contributing to more than 30,000 vehicles before being retired around November 2025. Q: What are the Figure 02's specs? A: Figure confirmed a 2.25 kWh battery, 16-degree-of-freedom fourth-generation hands, six RGB cameras, and about triple the compute of the Figure 01. Height (about 5'6") and weight (about 70 kg) are widely reported but were not in Figure's official release. Q: Did the Figure 02 use OpenAI? A: Yes, at launch in 2024. Figure ended the OpenAI partnership in February 2025 and switched to its own Helix AI model. How much did the Figure 02 cost, and can you buy one? You cannot buy a Figure 02, and it never had a public price. Figure 02 was an industrial humanoid deployed only through enterprise partnerships, most famously at BMW, and Figure AI never sold it to consumers or published an MSRP. The "$30,000 to $150,000" figures you see quoted are unverified third-party estimates that disagree with each other, and the widely repeated "$20,000" is a future at-scale target for a mass-market robot, not what a Figure 02 cost. There is also an important catch for anyone researching it in 2026: Figure 02 is the previous generation. It has been superseded by the Figure 03, which launched in October 2025. If you want the current Figure robot, you want the 03. The Figure 02 was never sold to the public. If you want a humanoid you can own, see the humanoid robots for sale today, ranked by price. Figure 02 Price: Is It for Sale, and What Did It Cost? No, and it never was. Figure AI has never published a price for any of its robots, and Figure 02 in particular existed only as a commercial and industrial unit inside partner deployments. There was no store, no order page, and no consumer version. The dollar figures that circulate online are guesses. Some aggregators quote roughly $30,000 to $50,000, others $130,000 to $150,000, and none trace back to Figure. Treat them as unverified. Separately, CEO Brett Adcock has floated a long-term target of around $20,000 per robot at mass-production scale, but that is an aspiration for the future, not the cost of a Figure 02. The honest answer to "how much is a Figure 02" is that there is no real number, because it was never a product you could buy. For how humanoid pricing actually works, see our guide to humanoid robot costs. Figure 02 vs Figure 03: What Changed? This is the question most people researching Figure 02 in 2026 actually have, because the 02 has been replaced. Figure 03, unveiled in October 2025, is the current generation and a real step forward, especially for home use. Figure 03 also adds soft textile coverings, fingertip tactile sensors, and a roughly 90% cut in component cost aimed at eventual mass production. In short, the 02 was built to prove humanoids could do real factory work; the 03 is built to eventually go into homes. For the full picture, read our Figure 03 review. The Figure 02 at BMW: Its Real Legacy Figure 02's importance is not its price, it is what it proved. At BMW's Spartanburg plant in South Carolina, Figure 02 units worked in the body shop for about 11 months, loading sheet-metal parts onto welding fixtures. Over that run the robots contributed to more than 30,000 BMW X3 vehicles, handled over 90,000 parts, logged more than 1,250 hours, and hit better than 99% placement accuracy per shift before the fleet was retired around November 2025. One point of confusion worth clearing up: Figure 03 also went to BMW Spartanburg, but that is a separate, later project announced in June 2026, and it is a different task (logistics and parts sequencing rather than the body-shop work the 02 did). So when you see "Figure robot at BMW," check which generation and which job. Figure 02 Specifications Figure published some hard specs for the 02 and left others to be reported by third parties. We separate the two. Confirmed by Figure (August 2024 launch) Battery: 2.25 kWh in the torso, over 50% more energy than the Figure 01Hands: fourth-generation, 16 degrees of freedom, human-equivalent strengthCameras: six onboard RGB cameras for visionCompute: roughly three times the onboard AI compute of the Figure 01AI: launched with OpenAI-trained speech and vision-language models Widely reported (not in Figure's official release) Height: about 5 feet 6 inches (168 cm)Weight: about 70 kg (corroborated by Figure describing the 03 as "9% lighter")Payload: about 20 kgRuntime: about 5 hours The OpenAI Chapter Part of what made Figure 02 notable was its brain. It launched in 2024 running OpenAI-trained models, the product of a high-profile partnership between Figure and OpenAI. That ended abruptly: in February 2025 Figure announced it was dropping OpenAI in favor of its own in-house Helix vision-language-action model, which now powers both the late Figure 02 fleet and the Figure 03. Figure AI has since raised over $1 billion at a $39 billion valuation, so the company behind the 02 is now one of the best-funded in robotics. Should You Care About the Figure 02 in 2026? Verdict: the Figure 02 matters as history, not as a purchase. It was the robot that showed a humanoid could do real, sustained factory work, and its BMW run is a genuine milestone. But it was never for sale, it has no public price, and it has been replaced by the Figure 03. If you are shopping or simply want the latest, look at the 03. If you are here to understand what Figure 02 was and did, the answer is: a landmark industrial humanoid, now retired to make way for its successor. Who this page is for Anyone comparing Figure 02 to the newer Figure 03People researching the landmark BMW humanoid deploymentReaders tracking Figure AI's rise and its OpenAI split If you want a robot you can actually follow or buy Read the current Figure 03 reviewCompare the field in our best humanoid robots guide The Bottom Line Figure 02 was a landmark: the humanoid that clocked real hours on a BMW production line and helped build tens of thousands of cars. But it was never a consumer product, never had a public price, and is now the previous generation behind the Figure 03. Research it for the history and the BMW story, not because you can buy one. For everything Figure is doing next, see our guide to humanoid robot companies. Compare on RoboZaps: Figure 02, Figure 03 and Figure 01. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Humanoid Robot Market Size: $3B Today to $5T by 2050 URL: https://blog.robozaps.com/b/market-size-for-humanoid-robots Author: Radica Maneva Published: 2025-03-05T00:00:00.000Z Updated: 2026-08-01T23:04:23.837Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: insights TL;DR: Public 2025 estimates put the humanoid robot market at $425 million to $7.8 billion (median $2.66 billion). Bank scenarios stretch far higher: Goldman sees $38 billion by 2035, Morgan Stanley $4.7 trillion and Citi $7 trillion by 2050. Disclosed company revenue across the sector was under $1 billion in 2025 on about 13,000 units shipped, so treat every figure as a scenario, not a count. Key takeaways: - Publisher-owned public summaries estimate the 2025 global market at $425M to $7.8B; the median of the six current 2025 bases is $2.66B. - The 2050 bank scenarios disagree by 5x on dollars (UBS $1.4T to Citi $7T) and over 3x on robots in use (300M to ~1B), with BofA's 3 billion dated 2060; Morgan Stanley's viral figure is $4.7T. - Published revenue CAGRs range from 25.8% to 50.6%, but each applies to a different model and period. - Reality check: disclosed 2025 company revenue sums to under $1B (Unitree ~$235M whole-company, UBTECH's humanoid segment ~$113M) on roughly 13,000 units shipped. - Near-term adoption is most credible in structured industrial, logistics, automotive and controlled service settings; watch 2026 shipments, where forecasts of 50,000 to 90,000 units meet reality. FAQ: Q: What is the humanoid robot market size in 2025? A: Seven publisher-owned public report summaries reviewed by RoboZaps estimate the global market at $425 million to $7.8 billion in 2025. The median of the six sources with current 2025 base years is $2.66 billion. Frost & Sullivan's assembled-robot-only scope gives the low end, while Future Market Insights' broad bundle of platforms, software and services gives the high end. Q: Is the humanoid robot market really worth $5 trillion? A: Not today. Morgan Stanley's much-quoted figure is a 2050 scenario: $4.7 trillion in humanoid revenue, over $5 trillion including supply chain and services, built on roughly 1 billion robots in use by then. Citi projects $7 trillion and UBS $1.4 to $1.7 trillion for the same year. The market's actual disclosed revenue in 2025 was under $1 billion. Q: How large could the humanoid robot market be by 2030? A: BCC Research forecasts $11 billion in 2030. MarketsandMarkets' earlier model put 2030 at $15.26 billion, but its July 2026 revision moved to a 2026 base of $5.41 billion and a $50.27 billion endpoint in 2035, leaving BCC as the only current 2025-to-2030 public forecast. Q: How large could the market be by 2035? A: Published 2035 figures span 9x: Interact Analysis is most conservative at about $15 billion, Goldman Sachs describes a $38 billion total addressable market, MarketsandMarkets' July 2026 revision puts 2035 at $50.27 billion, and Macquarie projects $139 billion. Grand View Research separately forecasts $40.5 billion by 2033. Different years, models and scopes, so they are not interchangeable confirmations. Q: What is the forecast CAGR for humanoid robots? A: The seven revenue forecasts compared here publish CAGRs from 25.8% to 50.6%. 360iResearch gives the low end for 2026 to 2032, while Fortune Business Insights gives the high end for 2026 to 2034. A CAGR is meaningful only with its start year, endpoint and market definition attached. Q: Why do humanoid robot market forecasts differ? A: Reports differ on whether they count hardware, software, services, wheeled robots, paid pilots, research systems and future consumer adoption. Shipment volume, average selling price, geography and forecast length also change the result. Q: Which country is leading in humanoid robots? A: China leads on units and manufacturing: Morgan Stanley estimates it accounted for over 80% of the roughly 16,000 humanoids deployed worldwide in 2025, and the top two shippers, AgiBot and Unitree, are Chinese. The United States leads on disclosed private-market capital, with Figure valued at $39 billion privately and Tesla converting a factory line for Optimus. Q: Are humanoid robots being sold today? A: Yes, in small numbers. Roughly 13,000 humanoids shipped worldwide in 2025 per an Omdia count cited in an AgiBot release, and consumer-priced models like Unitree's G1 and R1 can genuinely be bought. What does not yet exist at scale is the general-purpose labor robot the trillion-dollar forecasts assume; most 2025 units went to research, education and pilots. Q: What is Elon Musk's prediction for robots in 2027? A: His dated 2027 claim is that Tesla Optimus public sales will likely start by the end of 2027, made at Davos in January 2026. Tesla's forecast record argues for discounting dates: as of July 2026 formal Optimus production had not started, and earlier targets of 10,000 robots in 2025 produced external estimates of under 1,000. Q: What is the best humanoid robot company to invest in? A: We do not give investment advice, and no audited data supports naming one winner. What changed in 2026 is that pure-play exposure now exists at all: UBTECH is listed in Hong Kong, Unitree's Shanghai IPO was approved in July 2026, and Agility Robotics announced a proposed US listing that has not yet closed. Compare any valuation against the industry's under-$1B disclosed 2025 revenue before trusting a scenario number. Q: Which company leads the humanoid robot market? A: No public, audited global share table is strong enough to identify one definitive leader. An AgiBot release citing Omdia ranks AgiBot first by 2025 shipments with Unitree close behind, but Unitree's own shipment claim differs from Omdia's count by roughly 30%. Vendor rankings should state the denominator, region, period and whether they measure orders, shipments, pilots or installed systems. What is the humanoid robot market size? The direct answer: seven publisher-owned public report summaries reviewed on July 23, 2026 estimate the global humanoid robot market at $425 million to $7.8 billion in 2025. The median of the six sources with current 2025 base years is $2.66 billion, roughly $3 billion. That roughly 18-fold range is not a consensus: it shows how much the answer changes when a report counts only assembled robots versus hardware, software and services. For 2030, BCC Research's $11 billion is now the only current public forecast with a matching 2025-to-2030 period; MarketsandMarkets' July 2026 revision moved its endpoint to 2035. Other publishers use endpoints from 2032 to 2036. Above all of that sit the bank scenarios that dominate headlines: Goldman Sachs' $38 billion total addressable market by 2035, and the 2050 projections from Morgan Stanley, Citi and UBS that run from $1.4 trillion to $7 trillion. Those are long-range scenarios, not revenue forecasts, and this page keeps them in a separate table for exactly that reason. There is no official or audited global total. Every number below is a publisher's estimate. Market revenue, shipment volume, installed base and total addressable market answer different questions, so we keep the year, definition and geography attached to each figure. And because forecasts float free of the ground truth, we also show what the companies that disclose revenue actually earned in 2025: their sum, counting Unitree's entire company, was under $1 billion. 2025 estimate range: $425M to $7.8B across the public revenue forecasts; median of the six current 2025 bases: $2.66B.2030 forecast: $11B (BCC Research); MarketsandMarkets' July 2026 revision now runs 2026 to 2035.2035 scenarios: $15B (Interact Analysis) to $139B (Macquarie), with Goldman Sachs at $38B.2050 bank scenarios: $1.4T (UBS) to $7T (Citi), with Morgan Stanley's much-quoted figure at $4.7T.Published revenue CAGRs: 25.8% to 50.6%, each tied to a different period and model. Humanoid robot market forecast comparison The tables use the values and scope disclosed on each publisher's own public page. They are estimates, not audited industry totals. On a phone, swipe horizontally inside each table to see every column. MarketsandMarkets replaced its model in July 2026 with a 2026 base year, which leaves six sources with 2025 estimates. Their median is $2.66 billion, roughly $3 billion, and close to the $2.91 billion the older seven-source set produced. We show the median as a descriptive midpoint, not as a new market forecast. An average would imply a common measurement basis that these reports do not have. Until July 2026 the cleanest same-period comparison was BCC Research versus MarketsandMarkets, whose 2030 estimates differed by $4.26 billion; the new MarketsandMarkets edition dissolved that pairing by moving to a 2026-to-2035 window. The broader table is useful for showing the range of published opinion, not for pretending every row measures the same market. It is also a live lesson in edition churn: a forecast you cite can be replaced under the same URL within a year. The trillion-dollar forecasts: what the banks say about 2050 The numbers that dominate headlines and AI-generated answers do not come from the research houses above. They come from bank scenario models with 2050 horizons, and they disagree with each other by 5 times on dollars and more than 3 times on robots in use by 2050, before Bank of America's 3 billion arrives a decade later. That spread deserves its own table. Read the table horizontally and the disagreement is the story. For the same year, 2035, Macquarie's figure is 3.7 times Goldman's. For 2050, UBS and Citi differ by 5 times in dollars, and the 2050 installed-base assumptions run from UBS' 300 million robots to Morgan Stanley's roughly 1 billion, with Bank of America's 3 billion dated 2060, a decade later. Morgan Stanley's much-quoted "$5 trillion" is actually $4.7 trillion in core humanoid revenue, with the rounder number including supply-chain and support services; its own report expects adoption to stay relatively slow until the mid 2030s before accelerating. Citi's figure is usually reported as 648 million robots, but Citi's own page says "almost 650 million," which is the honest level of precision. ARK Invest circulates an even larger figure, roughly $24 to $26 trillion, but it belongs in a different category: it is a ceiling estimate of the revenue opportunity "at scale," with no target year, built partly by converting unpaid household labor into paid GDP. It is not a dated forecast and should not be compared against any number in the table. Two dating notes matter when you see these figures quoted. Goldman's $38 billion is the oldest number still in wide circulation: it was published in February 2024 and has not been publicly revised, so citing it as a current view is a stretch the bank itself does not make. And the freshest full bank model is Bank of America's March 2026 update, whose near-term line, 90,000 shipments in 2026, can be checked against reality within eighteen months. That is a feature, not a weakness: it is the only mega-forecast that puts a testable number this close. The reality check: what the industry actually earned Every forecast above floats free of a small, verifiable base. In 2025, the companies that disclose relevant revenue at all, counting Unitree's whole company, quadrupeds included, summed to well under $1 billion. These are selected disclosed indicators with different scopes, not an industry total, and they are not additive. Unitree, whose IPO was approved in early July 2026, reported 2025 whole-company revenue of RMB 1.69 billion, about $235 million at 2025 average exchange rates of roughly RMB 7.2 to the dollar, up 335% year on year, with humanoids at 51.5% of revenue in the first nine months of 2025. It is one of the only profitable companies in the sector. UBTECH, the only listed peer that breaks out a humanoid segment, reported RMB 820.6 million, about $113 million, of 2025 humanoid revenue on 1,079 units delivered, up from RMB 35.6 million the year before. an AgiBot announcement citing Omdia counted roughly 13,000 humanoid robots shipped worldwide in all of 2025, up around 480% from 2024, and Interact Analysis expects annual shipments to stay below 100,000 until the early 2030s, with a $15 billion market by 2035, the most conservative current 2035 estimate. One counting dispute is worth flagging because it shows how soft even the unit data is: Unitree says it shipped more than 5,500 humanoids in 2025, while Omdia credits it with about 4,200. Both can be defended depending on what counts as a humanoid and when a shipment is booked. If the two best-documented companies in the industry differ by 30% on a four-digit number, treat every third-decimal market figure accordingly. The gap math is the point. From a base of roughly $1 to $3 billion in 2025, Goldman's $38 billion by 2035 requires sustained growth around 35 to 40% a year for a decade, aggressive but in line with 2025's actual unit growth. Morgan Stanley's $4.7 trillion by 2050 requires that growth to continue for another fifteen years after that. Neither is impossible; both are extrapolations from a base year in which the entire industry earned less than a single mid-sized robotics company's annual revenue. Forecasts versus units: what the models assume Revenue forecasts hide their most testable claim, the number of robots. Here is what the public models say in units, which is where they can be checked first. Read each cell with its label: shipments are annual flows, installed base counts robots in operation, and the China-only figures sit inside the global ones. The 2026 row is the one to watch. Bank of America's 90,000 and the roughly 50,000 cluster from Morgan Stanley and Deutsche Bank are predictions about this calendar year, against a 2025 actual of about 13,000. By early 2027 one of those numbers will look prescient and the others will not, which will say a great deal about whose 2030 and 2050 models deserve attention. How we selected and compared sources We reviewed publisher-owned public pages available on July 23, 2026. To enter the main comparison, a page had to describe a global humanoid or humanoids market, disclose a 2025 US-dollar revenue estimate, state a future endpoint and belong to the organization publishing or distributing the named research. We excluded aggregator copies and narrower component-only submarkets from the headline range. This is a comparison of public summaries, not an independent audit of the paid reports or their underlying models. Where the public page does not disclose a revenue component, segmentation rule or regional assumption, we say so. Bank scenario models, Goldman Sachs included, are kept outside the seven-source revenue table because they describe total addressable market scenarios rather than directly comparable report series; they get their own table above with the same source discipline. What each forecast is actually measuring Frost & Sullivan: assembled-robot revenue only The Frost & Sullivan report distributed by Research and Markets gives the lowest 2025 estimate: $425 million, rising to $4.75 billion in 2032 at more than 41.2% CAGR. Its public scope is unusually specific. It counts revenue from fully assembled humanoid robots sold by original equipment manufacturers and excludes component-level revenue and downstream services. That narrower denominator helps explain why its starting value is below the other six. BCC Research: commercially deployed humanoids BCC Research estimates the market at $1.9 billion in 2025 and $11 billion in 2030, a 42.8% CAGR. Its public description focuses on commercially deployed humanoid robots and segments revenue by wheeled versus bipedal type, industry vertical and region. The commercial-deployment wording is important because prototypes and demonstrations do not necessarily create market revenue. Note that BCC's numbers also circulate as "news": the same figures are routinely recycled through press-release wires and finance portals with fresh datelines, which makes one estimate look like several independent confirmations. Grand View Research: revenue and unit volume Grand View Research puts the market at $2.4 billion in 2025, $4.2 billion in 2026 and $40.5 billion in 2033. The published CAGR is 38.2% for 2026 to 2033. Its report page separates hardware and software, biped and wheel-drive machines, and multiple industrial and service applications. It also publishes a unit-demand forecast, which is useful because revenue can rise through either more robots or higher revenue per system. 360iResearch: the original seven-source median 360iResearch estimates $2.91 billion in 2025, $3.64 billion in 2026 and $14.53 billion in 2032, with a 25.8% CAGR from 2026 to 2032. This was the median of the original seven-source set and remains the lowest published CAGR; after the July 2026 MarketsandMarkets revision, the median of the six current 2025 bases sits at $2.66 billion, just below this figure. Its public introduction describes humanoids as human-form systems combining mobility, perception, manipulation and increasingly natural-language interfaces, but it does not itemize the revenue components included in the total. MarketsandMarkets: hardware, software and services MarketsandMarkets replaced its model in July 2026: the new edition estimates $5.41 billion in 2026 and $50.27 billion by 2035, a 28.1% CAGR, superseding the April 2025 edition's $2.92 billion for 2025 and $15.26 billion for 2030. Same publisher, same URL, endpoint more than tripled and growth rate cut, all within fifteen months. Its scope includes biped and wheel-drive humanoids, plus hardware, software and services, across applications from manufacturing to caregiving, education and search and rescue. Fortune Business Insights: the highest published CAGR Fortune Business Insights estimates $4.89 billion in 2025, $6.24 billion in 2026 and $165.13 billion in 2034. Its published 50.6% CAGR for 2026 to 2034 is the highest in this comparison. The report covers biped and wheel-drive robots, hardware and software, and industrial, household and service applications. It identifies Asia-Pacific as 42.6% of the 2025 market. One caution on precision: at the time of review, the same public page stated the 2025 market as $4.89 billion in its summary and $4.49 billion in its own FAQ. Fortune's own regional splits reconstruct to roughly $4.49 billion, so we show its 2025 figure as a range. A report that disagrees with itself by $400 million on the same page is a useful reminder of how much weight two decimal places should carry. Future Market Insights: the broadest disclosed bundle Future Market Insights publishes the highest 2025 estimate at $7.8 billion, followed by $10.69 billion in 2026 and $248.9 billion in 2036, a 37.0% CAGR. Its public definition includes complete humanoid platforms, integrated AI and control software, sensor and actuator subsystems sold with the platform, and bundled maintenance and deployment services. It excludes non-humanoid robots, standalone AI software and non-autonomous telepresence devices. Goldman Sachs: a 2035 total addressable market scenario Goldman Sachs Research projected in February 2024 that the total addressable market could reach $38 billion by 2035. The public article links that scenario to a 1.4 million-unit shipment estimate, a 40% reduction in material costs, faster AI progress and a later expansion from industrial into consumer use. It does not publish a CAGR on the public page, so assigning one would require assumptions that Goldman Sachs does not state there. Goldman Sachs also calls out the main reasons the scenario may not arrive on schedule: component bottlenecks, limits in manipulation and interaction, and the unproven viability of mass-produced general-purpose humanoids. The $38 billion figure is useful as a scenario, not as a settled market fact. In late 2025, Goldman's China channel checks, as reported by CNBC, added a nearer-term view: roughly 15,000 to 20,000 humanoids shipped in 2025, with orders and shipments that "could multiply" in 2026. Why humanoid robot market forecasts disagree A forecast is the output of a model. Change the boundary or one of the core assumptions and the result moves sharply. The main sources of disagreement are: Hardware versus full-stack revenue. A hardware-only estimate will be lower than one that adds software licences, integration, maintenance, training or robotics-as-a-service revenue.Commercial deployment versus development activity. Public summaries do not always make clear whether paid pilots, prototypes, research platforms and demonstration units count as market revenue.Biped-only versus broader humanoid definitions. Several reports include wheel-drive humanoids alongside bipeds, while readers often picture only two-legged general-purpose robots.Industrial versus consumer adoption. Factory and logistics deployments can be modelled from defined tasks. Home, elder-care and general consumer demand requires stronger assumptions about safety, reliability and price.Shipments versus installed base. A shipment forecast counts units sold during a period. Installed base counts machines still in operation. Orders, reservations, pilots and demos are different again.Average selling price. Revenue can rise even with similar unit forecasts when one model assumes more expensive hardware, software or support. Falling component costs can also increase units while reducing revenue per robot.Forecast horizon. A five-year forecast and a ten-year total addressable market scenario carry very different uncertainty. Comparing their endpoints without the year is misleading. How to interpret a 26% to 51% CAGR The seven published compound annual growth rates span 25.8% to 50.6%. Even the low end looks extraordinary because the starting market is small. CAGR describes the smooth annual rate needed to connect a model's start and end values. It does not mean the industry will grow by that exact percentage every year. Early markets usually move unevenly. A large fleet order, delayed production ramp or revised definition can change one year sharply. High CAGR forecasts are therefore better read as a statement about the model's adoption curve than as a near-term sales guarantee. China versus the United States The most searched regional question has a two-part answer: China leads on units and manufacturing, the United States leads on capital. Morgan Stanley estimated in June 2026 that China accounted for more than 80% of the roughly 16,000 humanoids deployed worldwide in 2025, and doubled its China shipment forecast for 2026 to 50,000 units, its second upgrade in six months, citing commercial verification, policy support and supply-chain feedback. It models China's humanoid market at $2 billion in 2026 growing to $15 billion by 2030, figures that roughly match Chinese state-affiliated institutes, which put the domestic market near RMB 20 billion by 2026 and RMB 100 billion by 2030. Beijing has backed the sector since a 2023 Ministry of Industry policy calling for a world-class humanoid supply chain by 2027, and the top two global shippers in 2025, AgiBot and Unitree, are both Chinese. The American answer is capital: selected US startups command the largest disclosed private valuations in the sector. Figure reached a $39 billion private valuation on more than $1 billion of new funding in September 2025, Apptronik raised $520 million in February 2026 at a reported valuation above $5 billion, Agility Robotics announced a roughly $2.5 billion deal in June 2026 to become the first US-listed pure-play humanoid maker, and Tesla is converting a Fremont line for Optimus. Whether US capital or Chinese volume proves the better position is one of the genuinely open questions every 2050 scenario quietly answers one way or the other. What IFR data can and cannot prove The International Federation of Robotics is the strongest reference point here for broader robotics categories and deployment context. IFR separates industrial robots from service robots, then divides service robots into professional, consumer and medical categories. The World Robotics 2025 service-robot executive summary reports activity in professional logistics robots and consumer domestic robots. The industrial-robot executive summary covers factory automation. Neither is a clean humanoid-market total: most industrial robots are arms, and most consumer service robots are not humanoids. IFR data can test whether a forecast's adoption story is plausible in the wider robotics market. It cannot be added to a humanoid forecast or used as proof that the same number of humanoids has shipped. Where adoption is most credible first Near-term humanoid adoption is most credible in structured settings: factories, warehouses, logistics, automotive production, inspection, research and controlled service work. Tasks are easier to specify, workspaces can be engineered around the machine, and safety procedures are easier to enforce than in a normal home. This is consistent with Goldman Sachs' 2030 base case, in which almost all projected shipments are industrial. It is also consistent with the application categories used by BCC Research, MarketsandMarkets and Grand View Research, and with Morgan Stanley's 2050 model, in which about 90% of its billion robots do industrial and commercial work rather than household chores. Homes, hospitals, elder care and schools may become important later, but they demand higher reliability, privacy protection, certification and safe human interaction. The gap between a controlled pilot and an unsupervised deployment is one reason long-range consumer estimates carry more uncertainty. What the spread means for investors This page does not give investment advice, but the forecast spread has one investor-relevant implication worth stating neutrally: until 2026, there was almost no way to own pure-play humanoid exposure at all, and now there suddenly is. UBTECH listed in Hong Kong, Unitree's Shanghai IPO was approved in early July 2026 at an implied valuation around $5.9 billion, and Agility Robotics is heading to a US listing. If those listings complete, each will publish periodic public financials, audited at least annually, which will do more to discipline the forecast industry than any methodology note. The valuation mathematics deserve the same scrutiny as the market forecasts. Figure's $39 billion private valuation is roughly 40 times a generous $1 billion estimate of the sector's entire disclosed 2025 revenue. That is not automatically wrong, venture prices are claims about 2035, not 2025, but it means the capital markets are already trading on the bank scenarios, not on the $3 billion market that exists today. When an analyst cites a trillion-dollar figure, check which table it came from before treating it as a valuation anchor. Market size is not the same as robot price or availability Market revenue does not tell a buyer what one robot costs, whether it is available, or whether it can perform a specific job. For purchase-level research, compare current humanoid robot prices, inspect which humanoid robots are for sale, and verify deployment evidence separately. For current company activity, production plans and operating data rather than market forecasts, use the 2026 humanoid robot industry report. Our humanoid robot companies guide tracks the manufacturers behind those systems. How to evaluate a humanoid market-size claim Before quoting any market number, record the following details. If the public page does not disclose one, mark it as unknown instead of filling the gap with an assumption. Publication date, base year and forecast year.Revenue, shipment volume, installed base or total addressable market.Hardware, software, services and subscription treatment.Biped-only or biped plus wheel-drive definition.Included applications, especially research, pilots, industrial use and consumer use.Global and regional coverage, plus any countries singled out.Currency, nominal-versus-real treatment and average selling-price assumptions.Whether the number comes from a public summary, a paid report or an independently audited dataset.Whether the "news" article citing it is original reporting or a recycled press release with a fresh dateline. Do not publish unsupported market-share rankings Exact vendor shares require a public denominator, region, period and unit definition. A table claiming that one company owns 31% and another 27% can look authoritative while mixing orders, shipments, pilots, prototypes and installed systems. The Unitree case above is the live example: the company and the leading analyst firm differ by roughly 30% on its own 2025 shipments. Use company announcements as signals, not proof of global leadership.Separate firm orders from reservations, memoranda, pilots and demonstrations.Separate humanoids from industrial arms, autonomous mobile robots and consumer domestic robots.Prefer named scenarios and explicit caveats over one precise but undefined number. Bottom line Publisher-owned public report summaries put the 2025 global humanoid robot market between $425 million and $7.8 billion, with a median of $2.66 billion across the six current 2025 bases. BCC Research's $11 billion is the remaining current 2025-to-2030 forecast after MarketsandMarkets moved its window to 2026-2035. Above them, the bank scenarios run from Goldman's $38 billion by 2035 to Morgan Stanley's $4.7 trillion and Citi's $7 trillion by 2050, while the sector's disclosed company revenue in 2025 summed to under $1 billion on about 13,000 shipped robots. The 18-fold spread at the base year and the 5-times spread at 2050 are the finding, not a nuisance to hide. These publishers describe different versions of the market, so the source, year, metric, geography and scope should always travel with the figure. Treat every forecast as a planning scenario, not as an audited count of deployed humanoid robots, and watch the 2026 shipment numbers: they are the first place the models meet reality. Compare on RoboZaps: Digit and Stretch. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # What Is a Cobot? Complete Guide to Collaborative Robots (2026) URL: https://blog.robozaps.com/b/what-is-a-cobot Author: Radica Maneva Published: 2025-03-05T00:00:00.000Z Updated: 2026-07-30T10:58:36.105Z Category: insights TL;DR: Cobots are collaborative robots that work safely alongside humans without cages, using force-limiting and sensors. Hardware runs $25,000-$80,000 ($37,000-$145,000 fully deployed, versus $100K-$500K for traditional industrial robots), most deployments reach ROI in 12-18 months, and the market is growing toward $3.4+ billion by 2030. Key takeaways: - Cobots are collaborative robots that work safely alongside humans without safety cages—using sensors, force-limiting, and AI - Cost: $25,000–$80,000 for hardware, with total deployment at $37,000–$145,000 (vs $100K-$500K for traditional robots) - Most deployments achieve ROI within 12–18 months thanks to easy programming and minimal facility changes - Market growing at 18.9–31.6% CAGR, reaching $3.4+ billion by 2030 - Key applications: machine tending, pick-and-place, assembly, quality inspection, packaging, welding FAQ: Q: What is a cobot? A: A cobot (collaborative robot) is an industrial robot designed to work safely alongside humans in a shared workspace without safety cages. Cobots use force-limiting joints, advanced sensors, and intelligent software to detect human presence, avoid collisions, and stop instantly upon unexpected contact. They are easier to program, more affordable, and more flexible than traditional industrial robots. Q: What is the difference between a cobot and a robot? A: Traditional industrial robots operate behind safety barriers at high speeds for maximum throughput. Cobots are designed for direct collaboration with humans—they're slower but safer, easier to program (often via hand guiding), more affordable ($25K–$80K vs $50K–$400K+), and faster to deploy. For a detailed comparison, see our guide: Cobot vs Robot: Key Differences Explained. Q: How much does a cobot cost? A: Cobot hardware typically costs $25,000–$80,000 depending on payload capacity and reach. Total deployment costs including end-effectors, integration, safety assessment, and training range from $37,000 to $145,000. Most businesses achieve ROI within 12–18 months. Q: Are cobots safe to work with? A: Yes. Cobots comply with ISO 10218 and ISO/TS 15066 safety standards. They feature multiple safety systems including joint torque sensors, collision detection, speed and force limiting, and emergency stop mechanisms. However, a proper risk assessment is still required for every deployment, as the safety of the overall application depends on the end-effector and workpiece as well. Q: What industries use cobots? A: Cobots are deployed across manufacturing, automotive, electronics, food and beverage, pharmaceuticals, logistics, healthcare, aerospace, and more. The most common applications are machine tending, pick and place, assembly, quality inspection, packaging, and welding. Q: How do you program a cobot? A: Cobots can be programmed through hand guiding (physically moving the arm), graphical drag-and-drop interfaces on a tablet, or traditional coding (Python, ROS, etc.). Most simple tasks can be taught through hand guiding in under 30 minutes. No robotics expertise is required for basic programming. Q: What is the cobot market size? A: The global collaborative robot market was valued at approximately $1.26–$2.14 billion in 2024 (estimates vary by research firm) and is projected to reach $3.4+ billion by 2030, growing at a CAGR of 18.9–31.6%. Asia-Pacific is the fastest-growing region. Q: Will cobots replace human workers? A: Cobots augment human workers rather than replace them. They handle repetitive, dangerous, or ergonomically challenging tasks while humans focus on complex problem-solving, quality oversight, and creative work. Research shows human-cobot teams are more productive than either working alone. Q: How do cobots compare to humanoid robots? A: Cobots are typically single-arm robots optimized for specific industrial tasks. Humanoid robots have a human-like form with bipedal locomotion, designed for general-purpose tasks in unstructured environments. Cobots are mature and widely deployed; humanoid robots like the Unitree G1 and Agility Digit are in early commercial stages but represent the future of versatile automation. Q: What are the best cobot brands? A: The leading cobot manufacturers are Universal Robots (largest market share), FANUC, ABB, Techman Robot, Doosan Robotics, and KUKA. Universal Robots dominates with its UR+ ecosystem and intuitive programming, though competition is intensifying as more companies enter the market. So, what is a cobot? A cobot—short for collaborative robot—is an industrial robot engineered to work safely alongside human workers in a shared workspace, without the need for physical safety barriers. Unlike traditional industrial robots locked behind cages, cobots use advanced sensors, force-limiting joints, and intelligent software to detect human presence, avoid collisions, and stop instantly if contact occurs. In 2026, cobots are the fastest-growing segment of the robotics industry, with the global market projected to reach over $1.4 billion this year and $3.4 billion by 2030, according to MarketsandMarkets. Whether you're a manufacturing engineer evaluating your first automation investment, a business owner exploring ways to boost productivity, or simply curious about the future of human-robot collaboration, this guide covers everything you need to know about cobots—their definition, how they work, types, safety standards, applications, costs, and where the technology is heading next. Related reading: Cobot vs Robot: Top Benefits and Key Differences Explained · What Is a Humanoid? Definition and Examples Table of Contents What Is a Cobot? Definition and OriginHow Do Cobots Work?What Are the Different Types of Cobots?How Do Cobots Compare to Industrial Robots and Humanoids?What Safety Standards Apply to Cobots?What Industries Use Cobots?What Are the Benefits of Using Cobots?How Much Do Cobots Cost?How Do You Program a Cobot?Who Are the Top Cobot Manufacturers?What Is the Future of Cobots?Frequently Asked Questions What Is a Cobot? Definition and Origin A cobot (collaborative robot) is a robot specifically designed for direct physical interaction with humans within a shared workspace. The term was coined in 1996 by Northwestern University professors J. Edward Colgate and Michael Peshkin. Cobots differ from traditional industrial robots in that they're built for safe human collaboration—with force-limiting joints, rounded edges, and advanced sensors—rather than operating behind safety cages. The Cobot Definition A cobot (collaborative robot) is a robot specifically designed for direct physical interaction with humans within a shared workspace. The term was first coined in 1996 by Northwestern University professors J. Edward Colgate and Michael Peshkin, who defined a cobot as "a device and method for direct physical interaction between a person and a computer-controlled manipulator." The key distinction between a cobot and a traditional industrial robot lies in their design philosophy: cobots are built from the ground up for safe human-robot collaboration. Every aspect—from their lightweight construction and rounded edges to their force-limiting joints and advanced sensor arrays—is engineered to minimize risk when operating in close proximity to people. What Is the History of Cobots? The evolution of cobots traces a fascinating path from academic concept to mainstream industrial tool: 1996: Colgate and Peshkin create the first cobot concept at Northwestern University, initially designed as passive devices that guided human motion rather than moving independently.2004: KUKA Robotics releases one of the first commercially available lightweight collaborative robots.2008: Universal Robots launches the UR5, widely considered the breakthrough product that made cobots commercially viable for small and medium enterprises (SMEs).2012–2015: Major robotics companies including ABB, FANUC, and Yaskawa enter the cobot market, driving rapid innovation and price competition.2016: ISO/TS 15066 is published, establishing the first international safety standard specifically for collaborative robot operations.2020–2023: The COVID-19 pandemic accelerates cobot adoption as manufacturers seek flexible automation solutions to address labor shortages and social distancing requirements.2024–2026: AI-powered cobots with integrated vision systems, natural language interfaces, and machine learning capabilities enter the market, pushing the boundaries of what collaborative automation can achieve. How Do Cobots Work? Cobots work by combining a sensor-equipped robotic arm, interchangeable end-effectors (grippers/tools), and intelligent software that continuously monitors for human contact. When the cobot detects unexpected force, it stops within milliseconds—making it safe to operate alongside humans without barriers. What Are the Core Components of a Cobot? 1. Robotic Arm and Joints A cobot's arm is its most visible feature. Most cobots use a 6-axis articulated arm design, giving them six degrees of freedom (DOF) that closely mimic the range of motion of a human arm. Some newer models feature 7-axis designs for even greater flexibility. Each joint contains precision servo motors with integrated torque sensors that constantly monitor the force being applied. 2. End-Effectors End-effectors are the interchangeable tools attached to the cobot's wrist—essentially the robot's "hand." Common types include: Mechanical grippers (parallel, angular, or adaptive)Vacuum grippers for flat or smooth surfacesWelding torches and screwdriversForce/torque sensors for precision assemblyCamera and vision systems for inspectionTool changers for automatic swapping between tasks 3. Sensors and Perception Systems Cobots are equipped with multiple sensor types that give them environmental awareness: Force/torque sensors in every joint detect resistance and contact forcesProximity sensors (infrared, ultrasonic, capacitive) detect nearby objects and humansVision systems (2D/3D cameras, depth sensors) enable object recognition and localizationLiDAR for spatial mapping in advanced applicationsPressure-sensitive skins that detect contact across the entire arm surface 4. Controller and Software The controller is the cobot's "brain," running real-time operating systems that process sensor data, execute motion plans, and enforce safety limits. Modern cobot controllers support cloud connectivity, enabling remote monitoring, predictive maintenance, and over-the-air software updates. What Are the Four Collaborative Operation Modes? The ISO/TS 15066 safety standard defines four methods of collaborative operation that cobots can use: Safety-Rated Monitored Stop: The cobot automatically stops when a human enters the collaborative workspace, resuming when the human exits. Best for tasks where the human and robot take turns.Hand Guiding: A human operator physically moves the cobot arm to teach it positions and paths. The cobot records these movements and replays them autonomously.Speed and Separation Monitoring: The cobot adjusts its speed based on the distance between it and the nearest human. The closer the human, the slower the cobot moves—stopping entirely if the distance becomes too small.Power and Force Limiting: The cobot limits its speed, force, and momentum so that any contact with a human is below the pain and injury thresholds defined in ISO/TS 15066. This is the most common mode for true collaborative work. What Are the Different Types of Cobots? Cobots come in five main mechanical designs: articulated arm (most common, 6-7 axes), SCARA (fast horizontal movements), Delta/parallel (high-speed pick-and-place), Cartesian/gantry (linear motion), and dual-arm (bimanual tasks). Each type excels at different applications. By Mechanical Design Articulated Arm Cobots: The most common type, with 6 or 7 rotary joints. Examples: Universal Robots UR series, FANUC CR series. Ideal for assembly, pick-and-place, and machine tending.SCARA Cobots: Selective Compliance Articulated Robot Arms designed for fast, precise horizontal movements. Excellent for electronics assembly and packaging.Delta/Parallel Cobots: Spider-like designs with three or more parallel linkages. Used in high-speed pick-and-place applications in food and pharmaceutical packaging.Cartesian/Gantry Cobots: Move along linear X, Y, and Z axes. Used for dispensing, 3D printing, and large-workspace applications.Dual-Arm Cobots: Feature two coordinated arms for complex tasks requiring bimanual manipulation. Examples: ABB YuMi, Kawada Nextage. By Payload Capacity Light-duty (1–5 kg): Electronics assembly, lab automation, small part handlingMedium-duty (5–16 kg): General manufacturing, machine tending, packagingHeavy-duty (16–35+ kg): Palletizing, heavy part handling, automotive assembly How Do Cobots Compare to Industrial Robots and Humanoids? Cobots cost $25K-$80K (vs $50K-$400K+ for industrial robots), can be deployed in days (vs weeks/months), and work safely alongside humans without cages. Humanoid robots offer even greater versatility with bipedal mobility but are still in early commercial stages. For a deeper dive into how cobots compare with traditional robots, read our full guide: Cobot vs Robot: Top Benefits and Key Differences Explained. Interested in humanoid robots? Check out our Best Humanoid Robots hub and Applications of Humanoid Robots guide. What Safety Standards Apply to Cobots? Cobots must comply with ISO 10218-1/2 (industrial robot safety), ISO/TS 15066 (collaborative operation), and ISO 13849 (control system safety). These standards define force/pressure limits, collaborative operation modes, and required safety features that allow cobots to work alongside humans without safety cages. Key Safety Standards ISO 10218-1 & 10218-2: The primary safety standards for industrial robots and robot systems. Part 1 covers the robot itself; Part 2 covers the integration and installation of the robot system.ISO/TS 15066: The technical specification specifically for collaborative robot operations. It defines permissible force and pressure limits for different body regions, establishes the four collaborative operation modes, and provides guidance for risk assessments.ISO 13849 (PL d/e): Covers safety-related parts of control systems, relevant to the reliability of cobot safety functions.ANSI/RIA 15.06 (North America): The North American adoption of ISO 10218 with additional requirements. What Safety Features Do Cobots Have? Modern cobots incorporate multiple overlapping safety systems: Joint torque sensors: Every joint continuously measures force, enabling instant detection of unexpected contact. If force exceeds the safety threshold, the cobot stops within milliseconds.Collision detection algorithms: AI-powered software analyzes torque sensor data to distinguish between intentional forces (like picking up a heavy part) and unintentional contact (bumping a human).Speed and force limits: Cobots can be configured with speed caps (typically 250–1000 mm/s in collaborative mode) and force limits (typically 50–150 N) to ensure safe operation.Rounded, pinch-free designs: Cobot arms are designed with smooth, rounded surfaces and no exposed pinch points that could trap fingers or clothing.Emergency stop (E-stop): Physical buttons and software-triggered stops for immediate shutdown.Safety-rated I/O: Integration with external safety devices like light curtains, safety mats, and laser scanners for enhanced workspace monitoring. What Risk Assessment Is Required? Despite their inherent safety, deploying a cobot still requires a thorough risk assessment per ISO 12100 guidelines. This includes: Identifying hazards associated with the specific application (sharp tools, hot surfaces, heavy objects)Evaluating the risk of each hazard considering severity and probabilityImplementing risk reduction measures (engineering controls, administrative controls, PPE)Documenting and reviewing the assessment regularly What Industries Use Cobots? Cobots are deployed across manufacturing, automotive, electronics, food and beverage, pharmaceuticals, logistics, healthcare, and aerospace. The most common applications are machine tending, pick-and-place, assembly, quality inspection, packaging, and welding—with 85% of deployments in small-to-medium enterprises. Manufacturing and Assembly Manufacturing is the largest market for cobots. Common applications include: Assembly: Cobots handle screwdriving, press-fitting, snap-fitting, and adhesive dispensing with consistent precision.Machine tending: Loading and unloading CNC machines, injection molding machines, and presses—one of the most popular cobot applications.Pick and place: Moving parts between conveyors, trays, and workstations with speed and accuracy.Welding: MIG, TIG, and spot welding with consistent bead quality and reduced rework.Sanding and polishing: Cobots maintain consistent pressure and speed for uniform surface finishing. Automotive The automotive industry was an early adopter of cobots for: Engine and transmission assemblyWindshield installation and sealingQuality inspection using integrated vision systemsScrew driving and fastening in final assemblyCollaborative welding in body shops Electronics The electronics industry benefits from cobots' precision for: PCB (printed circuit board) assembly and solderingComponent placement and testingScreen assembly for smartphones and tabletsCable management and connector insertion Food and Beverage Cobots with food-grade certifications handle: Primary and secondary packagingPalletizing and depalletizingQuality sorting and inspectionFood preparation in commercial kitchens (an emerging application) Pharmaceuticals and Healthcare In regulated environments, cobots provide: Lab sample handling and pipettingMedication dispensing and packagingSurgical assistance (specialized medical cobots)Cleanroom operations with contamination-free designs Logistics and Warehousing Cobots mounted on mobile bases (AMRs) are transforming logistics: Order picking and sortingBin picking with 3D visionPackaging and labelingLoading and unloading delivery vehicles Aerospace The aerospace industry uses cobots for: Composite layup and trimmingFastener installation (drilling and riveting)Non-destructive testing and inspectionSurface preparation and coating What Are the Benefits of Using Cobots? The seven key benefits of cobots are: cost-effectiveness with 12-18 month ROI, enhanced workplace safety, 85% productivity gains in human-cobot teams (MIT research), improved quality/consistency, flexibility to redeploy in minutes, ability to address labor shortages, and small footprint requiring no safety cages. 1. Cost-Effectiveness and Fast ROI Cobots typically cost between $25,000 and $80,000—significantly less than traditional industrial robots, which can exceed $400,000 with integration. Most cobot deployments achieve ROI within 12–18 months, and some within as few as 6 months. Lower installation costs, minimal facility modifications (no safety cages), and reduced training expenses contribute to this rapid payback. 2. Enhanced Safety By handling dangerous, repetitive, or ergonomically challenging tasks, cobots reduce workplace injuries. The U.S. Bureau of Labor Statistics reports that musculoskeletal disorders account for over 30% of workplace injuries in manufacturing—exactly the types of repetitive strain injuries that cobots help prevent. 3. Increased Productivity Cobots work 24/7 without breaks, fatigue, or quality degradation. Studies show that human-cobot teams can be 85% more productive than all-human or all-robot teams, according to research by MIT. The cobot handles repetitive portions of a task while the human manages complex decision-making. 4. Improved Quality and Consistency Cobots perform tasks with repeatability of ±0.01 to ±0.05 mm, virtually eliminating human error in precision tasks. This consistency translates to fewer defects, less scrap, and higher customer satisfaction. 5. Flexibility and Scalability Unlike fixed automation systems, cobots can be redeployed to different tasks in minutes or hours. This makes them ideal for: High-mix, low-volume production environmentsSeasonal demand fluctuationsRapid product changeoversSmall and medium-sized businesses with diverse product lines 6. Addressing Labor Shortages With manufacturing facing a projected shortfall of 2.1 million workers by 2030 (per Deloitte/Manufacturing Institute), cobots provide a practical solution. They don't replace workers—they fill gaps where hiring is difficult, allowing existing staff to focus on higher-value tasks. 7. Small Footprint Cobots are compact and can be mounted on tables, walls, ceilings, or mobile platforms. They don't require the large footprint of safety-caged robot cells, making them viable even in space-constrained facilities. How Much Do Cobots Cost? (2026 Pricing Guide) Cobot hardware costs $25,000-$80,000, with total first-year deployment costs (including end-effectors, integration, safety assessment, and training) ranging from $37,000-$145,000. This compares favorably to traditional robot cells at $100,000-$500,000+. Compare this to traditional industrial robot cells that typically cost $100,000–$500,000+ including safety infrastructure, and it's clear why cobots are the preferred choice for many businesses. How Do You Program a Cobot? Cobots can be programmed three ways: hand guiding (physically moving the arm through tasks—no coding required), graphical tablet interfaces with drag-and-drop blocks, or traditional coding (Python, ROS, C++). Most simple tasks can be taught via hand guiding in under 30 minutes. 1. Hand Guiding (Lead-Through Programming) The operator physically grasps the cobot arm and moves it through the desired path. The cobot records each position and movement, then replays the sequence autonomously. This is the most intuitive method and requires zero coding knowledge. Most simple tasks can be taught in under 30 minutes. 2. Graphical Programming Interfaces Modern cobots come with tablet-based interfaces featuring drag-and-drop flowchart-style programming. Operators build programs by connecting visual blocks representing actions (move, grip, wait, inspect, etc.). Universal Robots' PolyScope and Techman Robot's TMflow are popular examples. Training typically takes just 1–2 days. 3. Traditional Coding and APIs For advanced applications, cobots support programming in Python, C++, ROS (Robot Operating System), and proprietary scripting languages. This enables complex behaviors like adaptive assembly, machine learning-driven quality inspection, and multi-robot coordination. API access also allows integration with MES, ERP, and IoT platforms. 4. AI-Powered Programming (Emerging) The latest cobots from companies like Covariant and Machina Labs are incorporating large language model (LLM) interfaces that allow operators to program robots using natural language commands. While still maturing, this technology promises to make cobot programming accessible to anyone who can describe a task in words. Who Are the Top Cobot Manufacturers in 2026? The leading cobot manufacturers are Universal Robots (50%+ market share), FANUC, ABB, Techman Robot, Doosan Robotics, and KUKA. Universal Robots dominates with its UR+ ecosystem of 300+ compatible accessories, while competitors differentiate through vision integration (Techman) and heavy payloads (UR30, Doosan). Universal Robots (UR) The market leader with approximately 50%+ global market share. Their product line includes the UR3e (3 kg payload), UR5e (5 kg), UR10e (12.5 kg), UR16e (16 kg), UR20 (20 kg), and UR30 (30 kg). Known for the largest ecosystem of third-party accessories via the UR+ platform and the intuitive PolyScope programming interface. FANUC Japanese robotics giant offering the CR (Collaborative Robot) and CRX series. The CRX-10iA and CRX-25iA are notable for their lightweight design and tablet-based programming. FANUC's industrial heritage provides excellent reliability and integration with existing FANUC automation systems. ABB Offers the GoFa and SWIFTI cobot families. GoFa provides up to 5 kg payload with excellent ease of use, while SWIFTI offers higher speeds for applications where humans are nearby but not in constant contact. ABB's dual-arm YuMi was one of the first cobots designed for small parts assembly. Techman Robot (TM Robot) A subsidiary of Quanta Computer, Techman is known for integrating vision systems directly into the robot arm, eliminating the need for external cameras. Their TM series cobots feature built-in smart vision for landmark recognition, object detection, and barcode reading. Doosan Robotics South Korean manufacturer offering the A-Series, H-Series, and M-Series cobots with payloads from 5 to 25 kg. Known for their six built-in torque sensors (one per joint) and the Dart-Suite software platform. Doosan has been expanding aggressively into food service robotics. Other Notable Manufacturers KUKA: LBR iiwa series with 7-axis design for maximum flexibilityYaskawa: HC series cobots integrated with the MOTOMAN ecosystemFranka Emika: The Panda cobot, popular in research and educationOmron: TM series (through partnership with Techman) for manufacturing automation What Is the Future of Cobots? The cobot market is projected to reach $3.4+ billion by 2030, driven by five key trends: AI and machine learning integration, mobile cobots (cobot arms on AMR bases), cloud robotics and fleet management, enhanced sensing with digital twins, and convergence with humanoid robots for general-purpose automation. Market Growth Projections The collaborative robot market is experiencing explosive growth: 2025: $1.42 billion (MarketsandMarkets) / $2.14 billion (Grand View Research)2030: $3.38 billion at 18.9% CAGR (MarketsandMarkets) / significantly higher per Grand View Research at 31.6% CAGRAsia-Pacific is the fastest-growing region, driven by manufacturing automation in China, Japan, and South Korea Key Technology Trends 1. AI and Machine Learning Integration Cobots are becoming smarter. AI enables them to learn from experience, adapt to new objects without reprogramming, and make real-time decisions about optimal grasp strategies, path planning, and quality classification. Computer vision powered by deep learning allows cobots to handle previously unseen objects. 2. Mobile Cobots (MoCobots) The convergence of cobots and autonomous mobile robots (AMRs) is creating mobile cobots—robotic arms mounted on mobile bases that can navigate facilities autonomously. These systems combine manipulation and mobility, enabling them to tend multiple machines, deliver parts between stations, and adapt to changing layouts. 3. Cloud Robotics and Fleet Management Cloud-connected cobots can share learned behaviors, receive software updates, and be monitored remotely. Fleet management platforms allow companies with multiple cobots to orchestrate operations across facilities from a single dashboard. 4. Enhanced Sensing and Digital Twins Next-generation cobots will feature more sophisticated sensors including tactile "e-skins," advanced 3D vision, and environmental sensors. Digital twin technology allows entire cobot workcells to be simulated virtually before physical deployment, reducing commissioning time and risk. 5. The Convergence of Cobots and Humanoid Robots The line between cobots and humanoid robots is beginning to blur. Companies like Figure, Unitree, and Apptronik are developing humanoid robots that can operate in the same collaborative manner as cobots but with human-like mobility and dexterity. These platforms represent the next evolution of collaborative automation—robots that can navigate human environments, manipulate diverse objects, and adapt to unstructured tasks without specialized tooling. Explore the full range of humanoid robots at Robozaps Humanoid Robots for Sale, or browse our complete robotics shop for the latest in collaborative and humanoid robot technology. Summary Cobots are transforming industries worldwide by enabling safe, flexible, and cost-effective human-robot collaboration. From their humble origins in a 1996 university lab to the AI-powered collaborative systems of 2026, cobots have evolved into indispensable tools for businesses of all sizes. With the market expected to more than double by 2030, now is the ideal time to explore how collaborative robots can enhance your operations. As the boundaries between cobots and humanoid robots continue to blur, we're entering an era where robots don't just work beside humans—they work like humans. Whether you're looking for a focused cobot arm for your production line or exploring the next generation of humanoid automation, the future of collaborative robotics is here. Ready to explore the future of robotics? Browse our complete robotics shop or check out humanoid robots for sale at Robozaps. Related articles: Cobot vs Robot: Top Benefits and Key Differences ExplainedWhat Is a Humanoid? Definition and ExamplesBest Humanoid RobotsApplications of Humanoid Robots --- # Humanoid Robots in Disaster Response and Search and Rescue Operations URL: https://blog.robozaps.com/b/humanoid-robots-in-disaster-response Author: Radica Maneva Published: 2025-03-04T00:00:00.000Z Updated: 2026-08-01T10:47:03.688Z Category: insights TL;DR: Rescue robots have deployed in real disasters since 9/11, in a market at $27.86 billion (2025) heading to $70.33 billion by 2030. Purpose-built humanoids like WALK-MAN and the Centauro hybrid exist, but 1-4 hour battery life and communication failures inside collapsed structures keep most real work with crawlers and drones. Key takeaways: - Search and rescue robots have been deployed in real disasters since 2001 (9/11 World Trade Center), with over 50 documented deployments through 2025 according to the Center for Robot-Assisted Search and Rescue (CRASAR). - The search and rescue robotics market reached $27.86 billion in 2025 and is projected to hit $70.33 billion by 2030 (14.79% CAGR), driven by quadrupling natural disaster frequency since 1970 and rapid advances in AI autonomy. - Key robot categories include ground crawlers (snake robots, tracked vehicles), aerial drones (thermal imaging, 3D mapping), aquatic robots (flood rescue, underwater search), quadrupeds (terrain traversal), and humanoid robots (door/valve operation, tool use, stair climbing). - The 2015 DARPA Robotics Challenge — an $80 million U.S. government program — proved humanoid robots could drive vehicles, traverse rubble, open doors, turn valves, and cut through walls. Team KAIST's DRC-HUBO won by completing all 8 tasks in 44 minutes 28 seconds. - Purpose-built disaster humanoids include IIT's WALK-MAN (1.85 m, 102 kg, 33 DoF), the Centauro hybrid (1.5 m, 93 kg, wheel-legged), and Boston Dynamics Atlas — each designed for different rescue scenarios. - Major challenges remain: battery life limits most robots to 1–4 hours of operation, wireless communication fails inside collapsed structures, and robots still require human supervision for complex life-or-death decisions. FAQ: Q: What robots are used for search and rescue? A: Search and rescue operations use multiple robot types: ground crawlers and snake robots (CMU Snakebot, MIT SPROUT) for navigating rubble and pipes; aerial drones (DJI Matrice 350, FLIR SkyRanger) for thermal imaging and mapping; aquatic robots (Hydronalix EMILY, VideoRay ROV) for water rescue; tracked firefighting robots (Shark Robotics Colossus, Howe & Howe Thermite RS3); humanoid robots (Boston Dynamics Atlas, IIT WALK-MAN) for operating in human-built environments; and quadruped robots (Boston Dynamics Spot, DEEP Robotics Jueying X20) for rough terrain inspection. Q: How are humanoid robots used in disaster response? A: Humanoid robots in disaster response perform tasks that require human-like body configuration: opening doors, climbing stairs, turning industrial valves, operating power tools, clearing debris, and navigating hallways. The DARPA Robotics Challenge proved humanoid robots could drive vehicles, walk over rubble, breach walls, and shut off valves — all tasks needed in real disaster scenarios. IIT's WALK-MAN successfully completed a full disaster scenario including gas valve shutoff and fire extinguisher activation in a test validated by Italian Civil Protection. Q: What was the DARPA Robotics Challenge? A: The DARPA Robotics Challenge (2012–2015) was an $80 million U.S. Department of Defense competition that challenged 25 international teams to build robots capable of performing disaster response tasks. Inspired by the Fukushima nuclear disaster, the competition required robots to drive vehicles, traverse rubble, open doors, turn valves, cut through walls, and climb stairs. Team KAIST from South Korea won the $2 million grand prize with their DRC-HUBO robot, which completed all 8 tasks in 44 minutes 28 seconds using a unique walk-and-roll transformation system. Q: What are the limitations of search and rescue robots? A: Key limitations include: battery life of 1–4 hours for most platforms; communication failure inside collapsed structures where GPS and radio are blocked; limited terrain adaptability (no single robot handles all disaster environments); high cost ($10,000–$500,000+) putting advanced robots out of reach for many agencies; need for specially trained operators; difficulty functioning in extreme temperatures, deep water, or toxic atmospheres; and inability to make complex moral or judgment-based decisions without human oversight. Q: How much do rescue robots cost? A: Costs span a wide range: consumer drones adapted for search use cost under $5,000; specialized ground crawlers and snake robots run $10,000–$50,000; Boston Dynamics Spot is enterprise quote-only (formerly $74,500); tracked firefighting robots like Colossus exceed $100,000; and advanced humanoid platforms are estimated at $500,000+. More affordable humanoid options are emerging — the Unitree G1 starts at $13,500. See our humanoid robot cost guide for detailed pricing. Q: What was the first search and rescue robot deployment? A: The first documented deployment of robots in a real disaster was at the World Trade Center collapse on September 11, 2001. CRASAR founder Robin Murphy led the deployment of small ground robots that searched through rubble for survivors and victims over a 10-day period. This operation proved the concept of rescue robotics and catalyzed billions of dollars in subsequent research, development, and government funding — including the DARPA Robotics Challenge. Related: Humanoid Robots in Military and Defense · Applications of Humanoid Robots · Best Humanoid Robots 2026 · Humanoid Robot Cost Guide Ready to explore humanoid robots? Browse humanoid robots for sale on Robozaps. Humanoid robots in disaster response are transforming how emergency teams save lives in the world's most dangerous environments. From the rubble of earthquake zones to radiation-contaminated nuclear facilities, humanoid robots and other search and rescue robots now operate where humans simply cannot survive. In 2026, with the search and rescue robotics market valued at over $27 billion and projected to reach $70 billion by 2030, this technology has moved decisively from laboratory curiosity to operational necessity. This guide covers every major search and rescue robot type, real deployment case studies, the humanoid robots purpose-built for disaster scenarios, the technologies enabling autonomous rescue, and where the field is heading next. History of Robotics for Search and Rescue The field of robotics for search and rescue traces its origins to two devastating events in 1995: the Oklahoma City bombing (April 19, 1995, 168 killed) and the Great Hanshin-Awaji earthquake in Kobe, Japan (January 17, 1995, 6,434 killed). Both disasters exposed critical gaps in human rescuers' ability to reach trapped survivors inside collapsed structures — gaps that machines could potentially fill. Robin Murphy, a computer science professor who later founded the Center for Robot-Assisted Search and Rescue (CRASAR) at Texas A&M University, responded to these tragedies by launching systematic research into deploying robots in disaster zones. Her pioneering work, including deploying the first robots at Ground Zero on September 11, 2001, established the scientific and operational foundation for the entire field of disaster robotics. Timeline: Key Milestones in Search and Rescue Robotics The DARPA Robotics Challenge: Proving Humanoid Robots in Disaster Scenarios The DARPA Robotics Challenge (DRC), running from 2012 to 2015, was the single most important event in proving that humanoid robots could perform disaster response tasks. The U.S. Department of Defense invested $80 million to challenge international teams to build robots capable of operating in environments too dangerous for humans — specifically modeled on the Fukushima nuclear disaster, where human rescuers couldn't enter reactor buildings. DRC Task Requirements Each robot had to complete 8 sequential tasks within 60 minutes: Drive a vehicle — get in, drive through an obstacle course, get outWalk across rubble — traverse uneven, debris-strewn terrain on footOpen a door — approach, turn handle, push through a standard doorTurn a valve — locate and rotate an industrial shut-off valveCut through a wall — use a power tool to breach drywallSurprise task — previously unknown manipulation challengeClimb stairs — ascend a standard staircaseCross uneven terrain — navigate a final rough terrain obstacle Communication was deliberately degraded to simulate real disaster conditions — robots experienced intermittent signal loss and had to maintain partial autonomy when their operators lost contact. DRC Finals Results (June 2015) Why DRC-HUBO won: Team KAIST's breakthrough was a transforming locomotion system. DRC-HUBO could walk upright like a human on two legs and also kneel down to roll on wheels built into its knees and feet. This dual-mode approach meant the robot could walk when it needed to (stairs, rubble) but roll quickly and stably when conditions allowed — dramatically reducing the risk of falling, which plagued many competing robots during the challenge. Humanoid Robots Built for Disaster Response Several humanoid and humanoid-hybrid robots have been purpose-designed for disaster scenarios. Unlike warehouse or factory humanoids, these machines prioritize ruggedness, manipulation strength, and the ability to operate in degraded environments. Comparison: Disaster-Response Humanoid Robots Boston Dynamics Atlas Atlas is the most recognizable disaster-response humanoid in the world. Originally developed for the DARPA Robotics Challenge in 2013, the hydraulic Atlas (1.8 m, 150 kg) served as the reference platform for 7 DRC teams. Its capabilities included walking over rough terrain, opening doors, turning valves, climbing stairs, and manipulating heavy objects — all core disaster response tasks. In 2024, Boston Dynamics unveiled the fully electric Atlas — a redesigned humanoid standing 1.9 m and weighing 90 kg, with improved agility, stronger electric actuators, and AI-powered motion planning. While the electric Atlas is currently deployed at Hyundai manufacturing facilities for logistics tasks, its disaster-response lineage makes it a prime candidate for future emergency deployment. The robot's ability to perform parkour, backflips, and recover from falls demonstrates the dynamic balance needed to navigate rubble fields. IIT WALK-MAN Developed by the Istituto Italiano di Tecnologia (IIT) in collaboration with the University of Pisa, WALK-MAN is a full-size humanoid (1.85 m, 102 kg, 33 degrees of freedom) built explicitly for disaster response. What makes WALK-MAN unique is its biomorphic 19-degree-of-freedom hands, developed with the University of Pisa, that can robustly grasp a wide variety of objects — from door handles to fire extinguishers to industrial valves. In its most significant validation test, conducted in collaboration with the Italian Civil Protection agency (Protezione Civile), WALK-MAN navigated a simulated earthquake-damaged industrial plant and completed four sequential tasks: Opened a heavy door and traversed the entryLocated a gas valve and closed it to stop a simulated leakRemoved debris blocking its pathIdentified a fire and activated a fire extinguisher The arms are strong enough to carry 10 kg loads for more than 10 minutes continuously — critical for tasks like hauling fire extinguishers, clearing rubble, or carrying survivor-detection equipment. IIT Centauro Rather than building another bipedal humanoid, IIT took a different approach with Centauro — a hybrid robot with a human-like upper body mounted on a four-legged, wheeled base. Standing 1.5 m tall with a shoulder width of 65 cm and weighing 93 kg, Centauro is constructed from aluminum, magnesium, and titanium alloys with 3D-printed plastic covers for rapid prototyping and field repairability. The centaur design solves one of the biggest problems in disaster robotics: bipedal robots frequently fall. Centauro's four-legged base with wheels at each foot provides extreme stability while still allowing it to navigate stairs, rubble, and uneven terrain. Its 42 degrees of freedom give it manipulation capabilities that far exceed those of tracked or wheeled robots, and its arm kinematics are specifically optimized for strength and reach in disaster-relevant tasks. KAIST DRC-HUBO The $2 million DARPA Robotics Challenge winner, DRC-HUBO (1.47 m, 80 kg, 32 DoF) from South Korea's KAIST, introduced the concept of transforming locomotion to disaster robotics. The robot's revolutionary design allows it to walk upright on two legs for tasks requiring human-like mobility (stairs, doors) and then drop to its knees to roll on integrated wheels for fast, stable traversal of flat surfaces. This approach directly addressed the single biggest failure mode in the DRC: falling. While competitors tumbled repeatedly, DRC-HUBO completed all 8 disaster tasks without a fall, finishing in 44 minutes and 28 seconds — the fastest of the three perfect-scoring teams. Professor Jun Ho Oh's team at KAIST demonstrated that solving disaster robotics may require abandoning strict human mimicry in favor of practical engineering solutions. All Types of Search and Rescue Robots Humanoid robots in disaster response are one part of a broader ecosystem. Effective robotics in disaster recovery deploys multiple robot types, each optimized for specific aspects of rescue operations: Ground Crawlers and Snake Robots These are the most commonly deployed search and rescue robots because they access spaces no human or conventional machine can reach: CMU Snakebot: Developed by Professor Howie Choset at Carnegie Mellon University's Robotics Institute. Features 16+ articulated joints, head-mounted camera with LED illumination, distance-measuring laser, and the ability to crawl through pipes, gaps in rubble, and along poles. Deployed in the 2017 Mexico City earthquake to search collapsed buildings.MIT SPROUT (2025): A flexible, vine-like robot built by MIT Lincoln Laboratory and Notre Dame researchers. Grows by inflating a tube-like body, navigating through tight spaces in rubble. Equipped with sensors to map the environment and relay 3D data to rescue commanders. Designed specifically for post-earthquake and building-collapse scenarios.Quince (Japan): Developed by Chiba Institute of Technology. Tracked robot with radiation sensors deployed inside the Fukushima Daiichi nuclear plant in 2011 to measure radiation levels in areas where human exposure would be lethal within minutes. Aerial Drones for Disaster Response Drones are now standard equipment for disaster response teams worldwide, providing the fastest situational awareness of any robotic platform: Teledyne FLIR SkyRanger: Launches in under 3 minutes, carries infrared and daylight cameras for detecting human heat signatures in rubble or wilderness. Used by military and emergency response teams globally.DJI Matrice 350 RTK: The most widely deployed disaster response drone platform. Used for post-earthquake damage mapping, flood extent assessment, and wildfire monitoring. DJI's thermal cameras detect humans at distances exceeding 100 meters through smoke and dust.Lockheed Martin K-MAX: Autonomous heavy-lift helicopter with 2,700 kg (6,000 lb) cargo capacity. Delivers supplies, humanitarian aid, and firefighting water to inaccessible locations. Proven in wildfire suppression and military logistics. Aquatic and Amphibious Rescue Robots Hydronalix EMILY: 4-foot, 26-lb remote-controlled flotation robot. Speeds up to 23 mph. Rescued 240+ refugees off the Greek coast during the 2016 European migrant crisis in its first 10 days of deployment. Carries up to 5 people. Kevlar-reinforced hull withstands heavy waves. Co-developed by U.S. Office of Naval Research.Pliant Energy Velox: Amphibious robot using undulating silicone fins for propulsion. Swims through water, skates on ice, and pushes through snow — ideal for cold-water rescues where ice conditions prevent human approach.VideoRay ROV: Remote-operated submersible with multi-beam sonar imaging, GPS, high-power lights, and video. Used by law enforcement and rescue teams for underwater search and recovery in lakes, rivers, and harbors. Firefighting Robots Shark Robotics Colossus: 500 kg tracked firefighting robot. Pumps 660 gallons of water per minute. Fireproof construction with 360° HD thermal camera. Saved Notre-Dame Cathedral in 2019 — Paris Fire Brigade Commander Jean-Claude Gallet credited Colossus with saving his crew's lives when conditions became too dangerous for human firefighters.Howe & Howe Thermite RS1/RS3: 24-HP diesel-powered firefighting robots on industrial tank treads. Climb slopes up to 70 degrees. Blast 2,500 gallons of water and foam per minute. Purpose-built for industrial fires: oil refinery blazes, HAZMAT incidents, and BLEVEs (boiling liquid expanding vapor explosions). Quadruped Robots Boston Dynamics Spot: The most commercially deployed inspection robot (no longer publicly priced; formerly $74,500). 14 kg payload, 90-minute battery, autonomous navigation. Used by utilities, construction firms, and nuclear facilities for hazardous environment inspection. Equipped with gas detectors, thermal cameras, and radiation sensors for disaster assessment.DEEP Robotics Jueying X20: Quadruped with 85 kg (187 lb) working load, autonomous navigation, and ability to climb 35-degree slopes. Carries oxygen tanks and emergency equipment. Full IP-rated for extreme weather operations. Master Comparison: Search and Rescue Robots by Type For more on humanoid robot capabilities across industries, see our guide to the best humanoid robots in 2026 and our applications of humanoid robots overview. Key Technologies Powering Search and Rescue Robots AI and Machine Learning for Autonomous Navigation Modern rescue robots use deep learning for real-time obstacle avoidance, path planning, and terrain classification. Reinforcement learning — the same technique used to train humanoid robots like the Unitree G1 to walk — enables rescue robots to navigate rubble fields they've never encountered before. The AI processes LiDAR point clouds, depth camera feeds, and IMU data to build 3D maps and plan safe routes in real time. The latest foundation models for robotics — including Google DeepMind's RT-2, Figure AI's Helix, and NVIDIA's GR00T — represent a paradigm shift. These vision-language-action models enable robots to understand natural language commands from rescue workers ("go through that doorway and check the room on the left") and execute complex multi-step tasks autonomously. In disaster scenarios where communication with operators is intermittent, this capability is transformative. Thermal and Biological Sensors Finding survivors is the core mission. Modern search and rescue robots deploy multiple sensor modalities simultaneously to maximize detection probability: Thermal cameras (LWIR): Detect body heat through dust, smoke, and total darkness. Effective range exceeds 100 meters for human-sized targets. The temperature differential between a living human and surrounding rubble is detectable even hours after burial.CO₂ sensors: Detect elevated carbon dioxide concentrations indicating breathing humans trapped under debris. More reliable than thermal in deep-burial scenarios where thermal radiation is blocked.Acoustic sensors / microphones: Ultra-sensitive audio sensors detect tapping, voices, breathing, or heartbeat sounds from survivors buried under meters of rubble.LiDAR (3D mapping): Creates precise three-dimensional maps of collapsed structures, identifying voids where survivors may be trapped and planning safe routes for human rescue teams.Gas detectors: Identify hazardous gas leaks (methane, CO, hydrogen sulfide, chlorine) that endanger both survivors and rescuers. Critical in industrial disaster and post-earthquake gas line rupture scenarios.Ground-penetrating radar: Experimental but promising for detecting human bodies and voids through concrete and steel — effective at depths up to several meters. Communication Systems for Denied Environments One of the biggest unsolved challenges in robotics for search and rescue is maintaining communication inside collapsed buildings where GPS, cellular, and radio signals are blocked by concrete and steel. The DARPA Robotics Challenge deliberately degraded communications to force teams to build partially autonomous robots. Current and emerging solutions include: Mesh networking: Multiple robots create ad-hoc wireless networks, relaying data through each other to reach external operators. Each robot serves as a communication node.Fiber-optic tethers: Physical cables providing guaranteed bandwidth for video and control signals — proven in nuclear facilities like Fukushima but limiting robot range.Acoustic communication: Sound-based data transmission through solid structures — experimental but promising for reaching deeply buried robots.Drone relay stations: Aerial robots hovering above collapsed buildings relay ground robot signals to command centers, bridging the communication gap between buried robots and surface operators.Edge computing / onboard autonomy: Rather than relying on communication, next-generation robots process data and make decisions onboard using AI chips, only transmitting results when a link becomes available. Real Disaster Deployments: Case Studies Fukushima Nuclear Disaster (2011) — The Most Extensive Robot Deployment in History The March 11, 2011 earthquake and tsunami triggered the worst nuclear disaster since Chernobyl. Human entry into the Fukushima Daiichi reactor buildings was impossible — radiation levels exceeded 10 sieverts per hour (lethal within minutes of exposure). This became the most extensive deployment of robotics in disaster recovery ever documented. iRobot deployed its PackBot (31 kg tracked robot) and Warrior (136 kg heavy-duty platform) to: Measure radiation levels inside reactor buildings 1, 2, and 3Transmit video footage of structural damage to TEPCO engineers planning decommissioningClear small debris to create pathways for larger decontamination equipment Japan's Quince robot, developed by Chiba Institute of Technology, later entered areas even PackBot couldn't reach, mapping radiation distribution across multiple floors. Over the following decade, more than 30 different robot models were deployed in the ongoing Fukushima cleanup — including swimming robots to inspect flooded reactor containment vessels and remote-controlled demolition machines to remove melted fuel debris. The Fukushima experience proved that robotics isn't just helpful in nuclear disasters — it's the only option. Notre-Dame Cathedral Fire (2019) — Robot Saves an 800-Year-Old Landmark When fire engulfed Notre-Dame Cathedral on April 15, 2019, the Paris Fire Brigade (Brigade de Sapeurs-Pompiers de Paris) faced an impossible situation. The cathedral's medieval wooden roof was fully ablaze, structural collapse was imminent, and entering the building risked both firefighter lives and the loss of an irreplaceable cultural treasure. Shark Robotics' Colossus — a 500 kg tracked firefighting robot — was deployed to: Spray water directly at the fire's hottest points from inside the cathedral naveUse its 360° HD thermal camera to identify hidden fire pockets within the stone and wood structureTransport firefighting hoses and equipment through areas where structural collapse made human entry suicidal Brigade Commander Jean-Claude Gallet stated publicly that Colossus saved his firefighters' lives. The incident was a watershed moment for disaster robotics — not experimental, not theoretical, but a real robot saving real lives and an 800-year-old monument in real time. Turkey-Syria Earthquake (2023) — Robotics at Scale The February 6, 2023 earthquake (magnitude 7.8) killed over 50,000 people and damaged more than 160,000 buildings across southeastern Turkey and northern Syria. Multiple robot teams deployed within hours: Thermal imaging drones identified heat signatures in collapsed buildings, directing rescue crews to buried survivorsMapping drones created 3D photogrammetric models of destroyed urban areas, helping commanders prioritize search zonesGround robots assessed structural stability of partially collapsed buildings before sending human teams insideSocial media AI tools accelerated coordination of international rescue efforts The deployment highlighted both the potential and limitations of rescue robotics at scale. While drones proved effective for rapid aerial assessment, the sheer scale of destruction overwhelmed available robotic resources. The disaster reinforced the need for cheaper, more numerous rescue robots — driving renewed interest in swarm robotics and low-cost drone fleets. WALK-MAN Disaster Scenario Test (2016) — Humanoid Proves Concept In a collaboration with Italy's Protezione Civile (Civil Protection), IIT staged a realistic disaster scenario in its Genoa laboratories: a simulated earthquake-damaged industrial plant with gas leaks, fire, and blocked corridors. WALK-MAN successfully: Opened a heavy industrial door and traversed the entry autonomouslyLocated and closed a gas shut-off valve to stop the simulated leakCleared debris from its path using its 19-DoF manipulator handsIdentified the fire source and activated a CO₂ fire extinguisher While performed in a controlled laboratory setting, this test validated that a full-size humanoid robot could perform the complete chain of tasks required in a real industrial disaster — a significant milestone for humanoid robots in disaster response. Challenges Facing Robotics in Disaster Recovery Battery Life and Endurance Most rescue robots operate for 1–4 hours before requiring recharge or battery swap. In disaster scenarios where operations continue for days or weeks, this remains the single most limiting factor. Current and emerging solutions include: Hot-swappable battery systems (available on robots like the Unitree G1 with 30-second battery swaps)Deployable solar charging stations for forward operating basesHydrogen fuel cells extending operation to 8+ hoursWireless charging pads at staging areas for autonomous recharge cyclesTethered power delivery for stationary or semi-stationary operations Terrain Adaptability and Robustness Disaster environments are inherently unpredictable — rubble fields with unstable footing, flooded basements, extreme heat from fires, toxic atmospheres from chemical spills, and structures threatening secondary collapse. No single robot design handles all conditions. The current best practice deploys mixed robot teams (drones for aerial recon, quadrupeds for terrain traversal, snake robots for confined spaces, humanoids for manipulation) coordinated from a single command post. Communication in Collapsed Structures Inside collapsed buildings, wireless signals attenuate rapidly through concrete and steel rebar. GPS doesn't work underground. Maintaining reliable two-way communication between buried robots and surface operators remains one of the biggest unsolved problems — and the reason why increasing onboard autonomy is critical for next-generation rescue robots. Cost and Accessibility Advanced rescue robots remain expensive. Boston Dynamics' Spot is enterprise quote-only (its old $74,500 public price ended) and Atlas is not commercially available. The Colossus firefighting robot costs well over $100,000. These prices put advanced robotics out of reach for most local fire departments and emergency agencies worldwide. However, more affordable options are emerging — the Unitree G1 starting at $13,500, consumer drones under $5,000, and open-source robot platforms are beginning to democratize basic robotic capabilities. See our humanoid robot cost guide for current pricing across all platforms. Regulatory and Operational Integration Even when robots are available and capable, integrating them into existing emergency response workflows is challenging. First responders need training. Radio frequencies must be coordinated. Liability questions arise when robots make autonomous decisions in life-or-death scenarios. Organizations like CRASAR and NIST's Engineering Laboratory are developing standard protocols for robot-assisted search and rescue to address these operational gaps. The Future of Search and Rescue Robots Swarm Robotics for Large-Scale Disasters Instead of deploying one expensive robot, future disasters may see swarms of dozens or hundreds of small, cheap robots collectively searching large areas. The Zebro project at TU Delft has demonstrated swarming algorithms where robots autonomously coordinate to cover earthquake-damaged areas, sharing map data in real time. Each robot is expendable — if one gets stuck in rubble, the swarm continues without interruption. This approach directly addresses the scale problem exposed by the Turkey-Syria earthquake. AI-Powered Full Autonomy Current rescue robots mostly require human operators via remote control or teleoperation. Next-generation systems will operate with full autonomy — entering a collapsed building, navigating to search zones, identifying survivors using multi-modal sensors, and reporting back without constant human guidance. The AI breakthroughs powering companies like Figure AI, Google DeepMind's robotics division, and NVIDIA's Project GR00T are building the foundation models that will enable this level of autonomous disaster response. Humanoid Robots as First Responders As humanoid robots mature — Boston Dynamics Atlas, Apptronik Apollo, Tesla Optimus, Unitree H1, and DEEP Robotics DR02 — their ability to navigate human-built environments makes them natural candidates for disaster deployment. A humanoid can open doors, climb stairs, turn valves, operate elevators, and use human tools — capabilities that specialized crawlers and drones fundamentally lack. The latest humanoid robots of 2026 are approaching the dexterity, battery life, and robustness needed for real disaster missions. CES 2026 showcased 9 humanoid robots from companies already shipping or deploying units, including models designed for all-weather outdoor operation (DEEP Robotics DR02 with full IP66 protection). The convergence of rugged hardware, AI autonomy, and declining costs suggests humanoid first responders may be operational within this decade. Integration with Smart Building Infrastructure Future disaster response will leverage IoT-connected building sensor networks. Smart buildings could automatically transmit floor plans, structural load sensor data, fire detection locations, and last-known occupant positions directly to rescue robot AI systems — dramatically accelerating search operations. Building Information Modeling (BIM) data, already standard in new construction, provides the 3D maps robots need to navigate efficiently. Soft Robotics and Bio-Inspired Design The next frontier includes soft robots that can squeeze through gaps rigid robots cannot, inspired by organisms like octopi and worms. MIT's SPROUT vine robot is an early example. Researchers are also developing robots with gecko-inspired adhesive feet for climbing walls in damaged buildings, and snake robots with force-sensitive skins that can detect survivors through touch. --- # Humanoid Robot Price in 2026: List Prices, Quotes & Total Cost URL: https://blog.robozaps.com/b/humanoid-robot-price Author: Radica Maneva Published: 2025-02-13T00:00:00.000Z Updated: 2026-08-01T23:05:18.976Z Last fact-checked: 2026-07-20T00:00:00.000Z Category: insights TL;DR: Published prices range from compact education products and developer platforms to full-size humanoids, while many enterprise systems remain quote-only. Compare the exact configuration, seller currency, availability, source date and exclusions—not just the headline number. For a buying decision, compare landed, task-ready three-year total cost with the advertised price. Key takeaways: - The guide separates public list prices, preorders, subscriptions, company targets, estimates and systems without an adequate public price. - A price is only comparable when the configuration, seller currency, market, exclusions and availability are clear. - Dollar budget filters apply to USD-denominated offers; prices in other currencies remain visible without silent conversion. - Task-ready cost includes tooling, freight, tax, integration, safety work, software, training, support, spares and downtime. - Use the same acceptance test and a three-year downside case when comparing a humanoid with established automation. FAQ: Q: How much does a humanoid robot cost in 2026? A: List prices in July 2026: $4,290 to $10,500 for Unitree's R1 line, $13,500 for the Unitree G1, $20,000 (or $499 a month) for the 1X NEO preorder, $29,900 for the Unitree H2 and $100,000 for the H2 Plus. The compact Noetix Bumi launched at ¥9,998 (about $1,400) in China. Enterprise robots such as Figure 03, Atlas and Digit are quote-only, with third-party estimates around $250,000. Q: What costs are excluded from a humanoid robot list price? A: Common exclusions include tax, shipping, customs, hands and tools, extra computing, batteries and chargers, integration, fixtures, site changes, safety engineering, training, software, support, maintenance, insurance, downtime and human supervision. The exact exclusions depend on the seller and configuration. Q: How should you calculate humanoid robot total cost of ownership? A: Model a defined task over a fixed period. Add acquisition or subscription, landed cost, configuration, integration, safety, training, support, maintenance, software, energy, spares, downtime and human intervention. Divide the total by safely completed useful work, not by advertised operating hours. Q: Can you lease or subscribe to a humanoid robot? A: Sometimes. 1X publicly lists NEO at $499 per month as well as $20,000 ownership. Enterprise vendors may offer pilots, leases or robot-as-a-service agreements by quote. Compare deposit, minimum term, included support, usage limits, upgrades, insurance, cancellation and who carries downtime risk. Q: How much does humanoid robot insurance cost? A: There is no defensible universal premium. Cost depends on jurisdiction, use, value, people and property exposed, autonomy, remote access, operating history and policy limits. Ask a qualified broker to price the actual deployment and confirm product, public, employer, cyber and property coverage rather than relying on a generic percentage of robot price. Q: What is the cheapest humanoid robot? A: The cheapest current listing from a major manufacturer is the Unitree R1-D at $4,290, with the R1 AIR at $4,900. The compact Noetix Bumi launched lower at ¥9,998 (about $1,400), but it sold through Chinese retail and Noetix now quotes prices on contact. Q: How much will Tesla Optimus cost? A: Tesla has not set a price. Elon Musk's stated target is $20,000 to $30,000 per unit, repeated through 2026, with public sales targeted for the end of 2027. Q: Where can you buy a humanoid robot in 2026? A: Unitree, EngineAI and 1X sell or take orders directly online, and NEURA takes reservations. Enterprise platforms such as Figure 03, Atlas and Digit are sold through direct sales conversations. Official store prices generally exclude tax and shipping. Q: What is a task-ready humanoid robot price? A: Task-ready price is the complete cost to put the robot into useful operation: the correct hardware configuration, tooling, shipping, tax, integration, safety work, software, training, support and initial spares. It is usually higher than the advertised robot price. Q: How often do humanoid robot prices change? A: Prices can change with configuration, market, launch offers, production stage and vendor terms. Check the source date and exact configuration before relying on a figure, and request a fresh landed quote for any procurement decision. Humanoid Robot Price Ranges in 2026 Humanoid robot price ranges in July 2026 run from about $1,400 to more than $100,000 at published list prices. The cheapest widely cited model, the compact Noetix Bumi, launched at ¥9,998 (about $1,400) in China. Unitree's R1 line starts at $4,290, the Unitree G1 developer platform is $13,500, the 1X NEO home robot is $20,000 (or $499 a month) on US preorder, and Unitree's flagship H2 Plus lists at $100,000. Mid-band machines such as the Kepler K2 (CNY 248,000, about $34,800, company-reported) and EngineAI T800 (from $40,500 international) fill the gap between them. Enterprise machines such as Figure 03, Boston Dynamics Atlas and Agility Digit have no public prices; third-party estimates cluster around $250,000. List price is only the entry ticket, though: the sections below show what a working deployment adds on top. There is no single humanoid robot cost because the market includes compact education products, developer platforms, home pre-orders, full-size hardware and enterprise deployments. The examples below use published manufacturer or primary-source prices; they are not an average for the category. Prices checked July 20, 2026; the Booster K1 and AgiBot X2 rows were added and checked July 29, 2026. Seller currencies are retained rather than converted at a moving exchange rate. Published figures may exclude tax, shipping and configuration; quote-only systems should not be assigned an invented retail price. What Your Budget Buys Advertised hardware prices are useful for shortlisting, but they do not show the cost of a complete working deployment. Keep seller currencies intact when comparing offers, record the exchange rate separately, and ask for a landed, task-ready quote before comparing vendors. If the initial hardware budget is capped, start with the cheapest humanoid robots under $25,000. For commercial work, keep enough budget for tooling, integration, training, spares and downtime. How Much Does Each Major Humanoid Robot Cost? Verified against manufacturer listings and on-record statements on July 20, 2026. Each linked review carries the full teardown. How Much Does Tesla Optimus Cost? Tesla has never published an Optimus price. Elon Musk's on-record target is $20,000 to $30,000 per unit, stated at Davos in January 2026 and repeated on Tesla's 2026 earnings calls, with public sales pointed at the end of 2027. Treat that as a target, not a price you can pay today. Full context in the Tesla Optimus review. How Much Does Figure 03 Cost? Figure has never published a Figure 03 price; units go to enterprise customers under private agreements. CEO Brett Adcock has floated a home model at roughly $600 a month, and the $130,000-plus figures circulating online are unsourced estimates. Details in the Figure 03 review. How Much Does Boston Dynamics Atlas Cost? Boston Dynamics has not announced an Atlas price. A January 2026 KED Global report, citing unnamed sources, put the planned price below roughly $320,000; the company has not confirmed it, and early production is reportedly committed to Hyundai plants. More in the Boston Dynamics Atlas review. How Much Does the Unitree G1 Cost? The Unitree G1 lists at $13,500 direct from Unitree's official store, excluding tax and shipping. The base model does not support secondary development; the programmable G1 EDU is quote-only. Full breakdown in the Unitree G1 review. How Much Do the Unitree H2 and H2 Plus Cost? Unitree lists the full-size H2 at $29,900 and the H2 Plus at $100,000, currently the highest published list price from a major manufacturer. The older H1 still shows about $90,000 on the store but is flagged “contact us for the real price”. See the Unitree H2 review. How Much Does the 1X NEO Cost? The 1X NEO costs $20,000 to own or $499 a month by subscription, on US preorder with a $200 refundable deposit. 1X says deliveries start in 2026. See the 1X NEO review. How Much Will Apptronik Apollo Cost? Apollo has no list price and is sold through enterprise pilots. CEO Jeff Cardenas's on-record goal is under $50,000 per unit at scale, a future target rather than today's cost. See the Apptronik Apollo review. How Much Does Agility Digit Cost? Agility Robotics does not publish a Digit price and sells mainly through Robots-as-a-Service agreements. Third-party estimates put an outright unit around $250,000, unconfirmed by the company. See how it stacks up in the Optimus vs Digit comparison. List Price vs Total Cost A useful planning formula is: acquisition or subscription + landed cost + configuration + integration and safety + training and process change + ongoing operation and support + downtime and human intervention − resale value. Each term should be tied to a specific task, site and contract. Landed Cost Start with the exact configuration, seller currency, tax, shipping, customs, duties, brokerage and destination support. Unitree explicitly labels its R1, G1 and H2 prices as excluding tax and shipping. A price copied from the product page is therefore not a delivered quote. Hands, Tools, Compute and Batteries Low base prices may exclude dexterous hands, higher-compute modules, secondary-development access, extra batteries, chargers, tooling and sensors. For example, Unitree reserves development access on R1, G1 and H2 for quote-only EDU configurations. Compare a task-ready bill of materials, not the cheapest chassis. Integration and Safety Budget for end-effectors, fixtures, workcell changes, networking, software connections, risk assessment, guarding or separation controls, acceptance testing and documentation. Home deployments also need a realistic plan for stairs, children, pets, privacy and remote access. These costs depend on the task; a universal percentage is not credible. Training, Supervision and Process Change Include operator and maintenance training, workflow redesign, teleoperation, exception handling and the time people spend resetting or supervising the system. A robot that needs frequent assistance can shift labour rather than remove it. Require intervention logs during a pilot before modelling savings. Support, Maintenance and Software Ask what the warranty covers, where repairs occur, spare-part lead times, software-update terms, service response, remote access, consumables and post-warranty support. Do not assume an annual maintenance percentage without a vendor quote and a parts-and-labour schedule. Downtime, Charging and Useful Throughput Battery life is not productive time. Model charging, battery swaps, setup, faults, waiting for people or materials, software updates and scheduled maintenance. The economic denominator should be safely completed useful work under the proposed conditions, with human assistance counted. Insurance and Compliance There is no standard humanoid-robot insurance rate. A broker needs the application, location, system value, people exposed, operating controls, autonomy and remote-access model. Confirm who carries product, public, employer, cyber and property risk under the purchase or service contract. Treat tax or regulatory incentives as jurisdiction-specific professional advice, not a guaranteed discount. A Three-Year TCO Model Run a base, downside and upside case. The downside should use slower throughput, more intervention and longer repairs than the vendor demo. Do not approve a fleet from a single success rate; require a pilot long enough to capture variation, failures and maintenance. For a sanity check on the running-cost side, Roland Berger's Humanoid robots 2026 study projects an operating cost near $2 per hour at scale, covering energy, software, network, scheduled maintenance and depreciation — but excluding the purchase or lease itself. That is a forward-looking figure at volume, not what a first deployment costs today, so treat it as the floor your own model should be measured against rather than a number to budget from. Purchase, Subscription, Lease or Robot-as-a-Service Purchase Purchase makes hardware ownership clear but places configuration, obsolescence and resale risk on the buyer. Confirm which software, service and remote features continue after the warranty or support term. Price is moot if the machine cannot lawfully be sold to you, so check that first: humanoid robots legal in the US lists what holds a US authorization and what does not. Subscription or Lease 1X publicly lists NEO at $499 per month as well as $20,000 ownership. A lower upfront number is not automatically cheaper: compare deposit, minimum term, included delivery and support, upgrades, usage limits, cancellation and the total paid over the expected holding period. Pilot or Robot-as-a-Service Enterprise vendors may price a pilot, lease or service agreement privately. Ask whether fees are per robot, hour, shift or completed task; what uptime and response commitments apply; who pays for integration and site changes; and who bears the cost when the robot cannot perform. US Tax Treatment: Section 179 and Bonus Depreciation For US businesses, a purchased humanoid robot is ordinary depreciable equipment. For tax years beginning in 2026, Section 179 allows expensing up to $2,560,000 of qualifying equipment, phasing out dollar for dollar once total purchases exceed $4,090,000 (IRS Rev. Proc. 2025-32). Separately, 100% bonus depreciation applies permanently to qualified equipment acquired and placed in service after January 19, 2025. Eligibility depends on more-than-50% business use, taxable income and how the robot is acquired, and there is no robot-specific IRS rule. Confirm treatment with a tax professional before building it into a purchase decision. What Drives Humanoid Robot Cost? Actuation and mechanics: joint count, torque, precision, durability and thermal management.Manipulation: hands, tactile sensing, wrists, tools and payload requirements.Perception and compute: cameras, depth sensors, lidar, onboard accelerators and development access.Production maturity: low-volume parts, quality control, testing and supplier scale.Safety and reliability: redundant controls, certification work, validation and documentation.Software and data: autonomy, teleoperation, updates, fleet tooling and task integration.Service: warranty length, spare parts, field technicians, response time and geographic coverage. These are the drivers behind the price you are quoted. For what the machine costs its maker to build — the bill-of-materials split across actuators, hands, compute and sensing, and how analysts expect that to fall through 2035 — see the economics of humanoid robot production. Humanoid vs Traditional Automation Cost A humanoid only earns its cost when its form solves a real constraint—such as using tools, stations or spaces built for people—better than a fixed robot, cobot, conveyor, autonomous mobile robot or process redesign. Compare safely achievable throughput, flexibility, integration and service for the exact task. For the return calculation rather than the cost inventory, use the separate humanoid robot ROI guide. This page stops at the inputs required to build that model. Humanoid Robot Quote Checklist A low headline price can become an expensive pilot when the quote excludes the configuration or work needed for the task. Send every shortlisted vendor the same brief and ask them to separate included, optional and recurring costs. Define the task, environment, throughput target and acceptance test before requesting a quote.Name the exact robot configuration, hands or grippers, sensors, compute, batteries and SDK access.Request the landed price, including freight, duties, tax, commissioning and required site work.Price integration, safety assessment, workflow changes, training and any teleoperation or supervision.Set pilot measures for uptime, successful cycles, human interventions, speed and recovery from failure.Confirm warranty, spare-parts lead times, software terms, service territory and response times.Compare the same task with established automation, cobots or process redesign—not only another humanoid.Model base and downside three-year TCO, including downtime, replacement parts and a failed-pilot exit. For a high-stakes deployment, use RoboZaps advisory to structure the shortlist, vendor questions and acceptance criteria. How RoboZaps Verifies Humanoid Robot Prices A price without its configuration, market and date is incomplete. RoboZaps reviews the comparison table and model sections against attributable manufacturer, distributor and reporting sources, then labels targets and estimates instead of presenting them as purchase prices. Manufacturer stores, product pages and primary announcements are preferred over reseller summaries.The seller's currency is retained; displayed prices are not silently converted at a moving exchange rate.Configuration and market are named when the source limits a price to a model, region or launch offer.Tax, shipping and other known exclusions are shown beside the price instead of buried in a footnote.List prices, preorders, subscriptions, company targets and third-party estimates remain separate evidence types.When no adequate public price can be verified, the guide says so instead of filling the gap with false precision.Each tracked price carries a source, confidence label and verification date, with public history when available. Read the full RoboZaps research methodology and follow the linked evidence in the relevant model section. Bottom Line The public market now spans a ¥9,998 (≈$1,400) compact launch price, $4,290 to $13,500 entry and developer platforms, a $20,000 home preorder, $29,900 to $100,000 full-size list prices, and quote-only enterprise systems. The price that matters is the three-year cost of safely completing the target work. Build that number from a task-ready configuration, a real integration scope, pilot intervention data and enforceable support terms. Everything else is a headline. For the ranked budget list, see the cheapest humanoid robots. For what you can actually order today, see humanoid robots for sale. For overall rankings, see the best humanoid robots. For buyer-side help turning these numbers into a decision, see our robotics advisory service. New to the category? Start with what a humanoid robot is. Compare on RoboZaps: Bumi, Unitree H2, Kepler K2, Unitree R1, Booster K1 and Figure 03. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Agility Robotics Digit Review: The Humanoid Robot Already Doing Paid Work (2026) URL: https://blog.robozaps.com/b/agility-robotics-digit-review Author: Radica Maneva Published: 2025-02-13T00:00:00.000Z Updated: 2026-08-01T10:47:47.867Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: reviews TL;DR: Digit is the humanoid already doing paid work: 100,000+ totes at GXO, 65,000+ hours across nine committed facilities, RaaS contracts with GXO, Toyota, Mercado Libre and Schaeffler. No rate card; Agility's investor deck models ~$8,500/month per robot (the quoted $30/hour is the human-labor comparator, not the price). $2.5B SPAC (AGLT) in progress; fence-free Digit v5 targets late 2026. Key takeaways: - Digit has the deepest verified work record of any humanoid: 100,000+ totes at GXO (official, Nov 2025) and 65,000+ real-world hours via deployment commitments across nine customer facilities (SEC filings, as of May 2026). - There is no current list price: Digit rents through RaaS contracts. The June 2026 investor deck models ~$8,500/month per robot (~$100k/year), priced below a ~$30.50/hour fully burdened human labor rate — the widely quoted $30/hour is that human comparator, not Digit's price. The $250,000 was Digit's real 2020 purchase price, not a current sticker. - The current robot carries 35 lb with a 4-hour self-charging battery; the 50 lb payload and fence-free "cooperative safety" belong to Digit v5, announced for late 2026 and not yet shipping. - Digit passed an OSHA-recognized NRTL safety field inspection at a live customer site (Nov 2025), which Agility calls a precedent for the humanoid industry — site-specific field labeling, not a blanket product certification. - Agility is going public via a ~$2.5B SPAC merger (ticker AGLT; draft S-4 confidentially submitted July 13, 2026), aiming to be the first US-listed pure-play humanoid company — $300M+ booked orders, revenue undisclosed, ~$111M preliminary 2025 operating expenses. FAQ: Q: How much does the Agility Digit cost? A: Agility has never published a rate card. Digit is offered through Robots-as-a-Service contracts bundling the robot, software, workcell and maintenance. The June 2026 SPAC investor deck models illustrative pricing of $8,500 per month per Digit (about $100,000 a year), set at a discount to a fully burdened human labor rate of ~$30.50/hour — the widely quoted "$30 an hour" is that human-labor comparator, not Digit's price. The $250,000 seen on directories was Digit's 2020 direct-purchase price, and the $45,000 some sites list is simply wrong. Q: Can you buy an Agility Digit? A: Yes, through a RaaS contract or direct purchase (Schaeffler signed a purchase agreement), arranged with Agility or its integrator channel (Zion Solutions Group, with Ricoh servicing fleets). Digit is the only humanoid today with a genuine commercial path for Western operators. Q: What does the Digit robot actually do? A: Structured tote and case handling: picking totes on and off autonomous mobile robots, feeding conveyors, and stacking at floor locations, integrated with MiR and Zebra AMR platforms through Agility Arc cloud software. It is a logistics specialist by design, not a general-purpose home robot. Q: Where is Digit working today? A: At GXO's Spanx facility in Georgia (paid RaaS work since June 2024, 100,000+ totes moved), Toyota Motor Manufacturing Canada (a year-long three-robot pilot converted into a commercial agreement covering seven Digits in February 2026), Mercado Libre's San Antonio fulfillment site, Schaeffler plants, and Amazon's R&D operations — deployment commitments across nine customer facilities per Agility's 2026 SEC filings (as of May 2026). Q: How many Digit robots are deployed? A: Agility does not disclose a deployed-robot count. Its June 2026 SEC investor materials report deployment commitments across nine customer facilities and more than 65,000 hours of operation (as of May 2026). Booked orders exceed $300M, including a multi-year contract for about 1,000 robots with an undisclosed customer — commitments to be delivered, not the current fleet. Q: Is Agility Robotics on the stock market? A: Not yet, but soon: Agility announced a merger with Churchill Capital Corp XI on June 24, 2026 at a ~$2.5 billion implied valuation, expecting to list as AGLT before the end of 2026. A draft S-4 was confidentially submitted to the SEC on July 13, 2026 (announced July 14); the public filing with audited financials comes later. It would be the first US-listed pure-play humanoid company. Q: Who owns Agility Robotics? A: It is a private venture-backed company founded by Damion Shelton and Jonathan Hurst (now chairman and Chief Robot Officer), led by CEO Peggy Johnson since 2024. Investors include WP Global, SoftBank, Amazon's Industrial Innovation Fund, DCVC, Playground Global and Schaeffler, with roughly $640M raised. Q: What is Digit v5? A: The announced next generation: 50 lb payload (up from 35), and "cooperative safety" designed to let Digit work outside fenced workcells alongside people, built on NVIDIA IGX Thor with the Halos safety platform. It targets late 2026 and is not yet shipping; treat any site listing its specs as current as mistaken. Q: Is the Digit robot safe? A: It carries a genuine industrial safety package (CAT1 stop, Safety PLC at PL d, E-stops, FailSafe over EtherCAT), and in November 2025 passed an OSHA-recognized NRTL safety field inspection at a live customer site — a first Agility describes as a precedent for the humanoid industry, and site-specific rather than a product certification. No public safety incident involving Digit appears in credible reporting through July 2026. Q: Is Digit better than Tesla Optimus? A: They are not comparable as products: Digit does paid customer work today with published metrics, while Tesla's Optimus has no shipped product, no price and no published performance data. Optimus carries bigger ambitions and a bigger factory story; Digit carries the receipts. See our full Optimus vs Digit comparison. How much does the Agility Digit cost, and can you actually get one? Digit has no sticker price, and that is not evasion, it is the business model: Agility Robotics rents Digit through Robots-as-a-Service contracts that bundle the robot, its Agility Arc cloud software, workcell, fleet management and maintenance. The most concrete economics come from the June 2026 SPAC investor deck, which models illustrative RaaS pricing of $8,500 per month per Digit, about $100,000 a year, set at a discount to a fully burdened human labor rate of roughly $30.50 an hour (SEC investor presentation). Read CEO Peggy Johnson's much-quoted benchmark the same way: "under 2 year ROI vs a human at a fully loaded $30 an hour" (The Robot Report, July 2024) — the $30 an hour is what the human costs, not what Digit rents for, a distinction most coverage misses. The "$250,000" you will see on directory sites is real but historical: it was Digit's direct-purchase price when it first went on sale in 2020. The "$45,000" one prominent directory lists is simply wrong. And yes, you can genuinely get one, which makes Digit nearly unique: it is the humanoid with the deepest verified paid-work record for named customers, GXO, Toyota Canada, Mercado Libre, Schaeffler, with an integrator channel and a service partner. If your operation moves totes, Digit is the one humanoid where the conversation starts with a contract instead of a waitlist. One disclosure up front: this is desk research built on the sources linked throughout, primary wherever they exist, not a hands-on test. The misinformation audit: what circulates vs the record What is Digit? Digit is Agility Robotics' bipedal warehouse humanoid, the machine with the backwards-looking "bird legs" that make roboticists smile, and the industry's least hypothetical robot. It descends from Cassie, the walking research platform Agility spun out of Oregon State University, and the current fourth-generation logistics Digit (launched March 2023, substantially upgraded at ProMat in March 2025) is built for one job family done reliably: picking up totes and putting them down somewhere else, between AMRs, conveyors and racking. That narrowness is the strategy. While rivals demo laundry folding and kung fu, Digit clocked sixty-five thousand paid hours, and Agility is about to take that record to the public markets. The proof ledger Very few humanoids have receipts like these; here they are, dated and sourced. Ledger sources: the SEC investor presentation (June 2026) and the June 24, 2026 joint press release. Context matters both ways: those numbers dwarf every Western rival's shipped evidence, and they are still small against the ambition. Agility has not disclosed revenue; its preliminary figures show about $111 million of 2025 operating expenses ($37M SG&A plus $74M R&D) and roughly $100 million of cash burn, and the deck is explicit that the $300M+ order book "is not a measure of current period revenue." Digit proves humanoids can earn; Agility has not yet shown how much, or that they earn profitably. Official specs vs the directory layer The pattern to note: Agility publishes what operators need (payload, uptime mechanics, safety hardware, integrations) and keeps the robot-nerd numbers gated. Directories fill the gap with drift, which is how a 4-hour battery becomes "8 hours" and a v5 target becomes a shipping spec. Verification status per claim is tracked in the Digit record in our robot database. The deployment ledger, dated Around the customers sits a real channel: Zion Solutions Group as the first humanoid systems integrator (May 2024) and Ricoh USA servicing Digit fleets (Sept 2024), both official partnerships. That channel is itself evidence: you build integrator and service networks for products that ship, not products that demo. The price question, in full What a Digit engagement actually is: a RaaS contract bundling hardware, Agility Arc, the workcell, fleet management and maintenance, priced per deployment and never published as a rate card. The most concrete numbers exist in the June 2026 investor deck: an illustrative model of $8,500 per month per Digit, about $100,000 a year and roughly $500,000 over a lifetime engagement, "priced at discount to customer's fully burdened labor rate" of about $30.50 an hour. That is also the correct reading of Johnson's much-quoted line, "under 2 year ROI vs a human at a fully loaded $30 an hour": the $30 is the human's cost, the comparator Digit undercuts, not Digit's rental price. The GXO rate has never been disclosed. Direct purchase exists (Schaeffler signed a purchase agreement), also unpriced. And the $250,000 is history rather than rumor: that was Digit's direct-purchase price when it went on sale in 2020, then-CEO Damion Shelton having put it in the "low-mid six figures"; by 2024, IEEE Spectrum was told the newer generation would cost less than $250,000, and Agility has published no number since. What the filings show instead of revenue: $111 million of preliminary 2025 operating expenses and $300M+ of multi-year bookings, a contracts-ramping engine, exactly what a RaaS model predicts. For the cross-market picture, our humanoid robot cost guide puts these numbers next to everyone else's silence. Digit v5: the next generation Announced alongside the SPAC in June 2026, Digit v5 is the bigger bet: 50 lb payload (up from 35), and "cooperative safety," the ability to work outside a fenced workcell alongside people, built on NVIDIA IGX Thor and the Halos safety platform, with Agility as the first company to integrate NVIDIA's Halos for Robotics into a humanoid safety system. Target: late 2026. Precision matters here: v5 is announced, not shipping; its 50 lb payload keeps getting misprinted as a current spec; and TÜV Rheinland's inspection role covers the NVIDIA components, not Digit itself. If v5 lands with fence-free clearance, it changes Digit's addressable work from tote cells to general floor space, which is the entire growth thesis behind the 1,000-robot order. Safety: the standards story, precisely In November 2025 Digit passed an OSHA-recognized NRTL safety field inspection at a live customer fulfillment site, assessed against ANSI/RIA R15.08 and ISO 13849, a first that Agility describes as "setting an important precedent for the humanoid industry" (Agility, official). Keep the scope exact, as Agility itself does: NRTL field evaluations are site-specific, so this is field labeling at one site, not a blanket product certification. The current robot also carries a genuine industrial safety package (CAT1 stop, PL d Safety PLC, E-stops, FSoE) from the ProMat 2025 refresh. The certification ambition, ISO functional-safety clearance for fence-free operation, rides on v5. No public safety incident involving Digit appears in any credible outlet through July 2026, a cleaner sheet than several rivals. RoboFab and the new Fremont hub Digit is built at RoboFab in Salem, Oregon, a 70,000 sq ft plant Agility calls the world's first humanoid factory, announced September 2023 with a capacity plan of hundreds in year one scaling toward 10,000+ a year and 500+ jobs at full tilt (Agility). Current actual output is unpublished; the honest proxies are the nine committed customer facilities and the ~1,000-robot booked order that SPAC proceeds are meant to scale into. And as of yesterday, July 16, 2026, Agility opened a second front: a 60,000 sq ft "Physical AI" software hub in Fremont, California, around 200 jobs, complementing the Oregon factory (announcement). The company, and the SPAC Agility is run by CEO Peggy Johnson (ex-Microsoft, ex-Magic Leap) since March 2024, with founders Damion Shelton as chairman and Jonathan Hurst as Chief Robot Officer still aboard. The funding ladder: roughly $400M Series C closed in early 2025, led by WP Global with SoftBank, Amazon's Industrial Innovation Fund, DCVC and Playground Global, at about $2.1 billion post-money (the circulating "$1.75B" was pre-money reporting). Then the big move: on June 24, 2026 Agility announced it is going public by merging with Churchill Capital Corp XI at a ~$2.5 billion implied valuation, expecting $600M+ in proceeds, under the ticker AGLT. A draft S-4 was confidentially submitted to the SEC on July 13, 2026 and announced the next day; the public S-4 filing, with audited financials, comes later in the process. If the deal closes as planned before year-end, Agility becomes the first US-listed pure-play humanoid company, which will drag the whole industry's claims into audited daylight, a development this page is built for. For the sector map, see our humanoid robot companies guide. Digit vs the field The two humanoids actually working for customers, Digit and the Walker S2, split the world between them: Digit owns verified per-site metrics and Western logistics; Walker owns audited volume and Chinese industry. Everything else on the list is a program, not a product. Head-to-head: Optimus vs Digit, and the full rankings in our best humanoid robots guide. Who should contract Digit, and who shouldn't Contract-ready: warehouses and 3PLs with tote-based workflows, AMR-to-conveyor handoffs, and enough volume to justify a workcell, the GXO template, plus manufacturers with structured material-handling tasks, the Toyota template. The path runs through Agility directly or the Zion integrator channel, with Ricoh servicing fleets. Not yet the right buyer: operations needing fence-free mobility (wait for v5's certification), tasks beyond structured tote/case handling (that is the whole industry's frontier, not just Digit's), or anyone shopping on sticker price, because there is none, and the ~$100,000-a-year modeled cost only pencils where the robot works long shifts. Households: Johnson herself says home robots are not the near-term plan. What to watch, dated Four threads move this page. The SPAC close, expected before end-2026, which converts management's metrics into audited filings and gives the industry its first public humanoid financials. Digit v5's late-2026 arrival and its fence-free safety clearance, the real capability unlock. The next metrics milestones, tote counts and fleet hours have been updated roughly twice a year. And expansion at the newest customers: Toyota beyond the contracted seven robots, Mercado Libre into Latin America. We re-verify this page at each. The bottom line Digit is the humanoid industry's existence proof: the robot that ended the "will anyone ever pay for this" argument with a tote count and a customer list. It is also deliberately unsexy, a 35-pound-payload specialist with no sticker price, no viral chores, and a spec sheet behind a download gate, and that is precisely why it works. Judge it as the industry's most verifiable robot: real contracts, real hours, real safety paperwork, real financials on their way to public markets, and real limits, thin margins for now, one job family, and a next generation still unshipped. The Digit record in our robot database tracks the verified state as it changes. --- # Humanoid Robots in Retail: The Complete 2026 Deployment Guide URL: https://blog.robozaps.com/b/humanoid-robots-in-retail Author: Radica Maneva Published: 2025-02-12T00:00:00.000Z Updated: 2026-08-01T23:04:40.179Z Category: insights TL;DR: Retail robots are past the pilot stage: Pepper drove a documented 15% sales lift across 1,000+ Nestlé Japan stores, Simbe's Tally scans shelves in every BJ's Wholesale Club, and Agility's Digit moves goods in a Spanx warehouse. With hardware costs down about 40% since 2023, Goldman Sachs sees a $38 billion humanoid market by 2035. Key takeaways: - Goldman Sachs projects the humanoid robot market at $38 billion by 2035, with retail and service among the early adopters. - Proof it sells: Pepper deployed across 1,000+ Nestlé Japan stores drove a documented 15% increase in Nescafé machine sales. - Non-humanoid shelf-scanners are today's workhorses: Simbe's Tally runs in every BJ's Wholesale Club, catching out-of-stocks and pricing errors. - Agility's Digit works a Spanx warehouse in Georgia — 5'9", 141 lb, carrying up to 35 lb alongside human staff. - Hardware costs fell about 40% between 2023 and 2025, putting units at $5,900-$150,000. FAQ: Q: How much does a humanoid robot cost for a retail store? A: Prices range widely in 2026. Entry-level humanoid robots like the Unitree R1 start at $5,900. Mid-range customer service robots like SoftBank Pepper cost $25,000–$30,000 plus monthly software fees. Advanced humanoids with dexterous manipulation (UBTECH Walker S, Agility Digit) range from $50,000 to $150,000. Robot-as-a-Service (RaaS) models start at $999–$2,000/month, eliminating upfront costs. Q: Which retailers are currently using humanoid robots? A: As of January 2026, notable deployments include: BJ's Wholesale (Simbe Tally for inventory), Amazon (Agility Digit for fulfillment), Nestlé Japan (SoftBank Pepper for sales), Denny's and Chili's (Bear Robotics Servi for food delivery), and numerous Chinese retailers piloting UBTECH and Unitree humanoids. Walmart, Kroger, and Target have all conducted robotics pilots of various scales. Q: Will humanoid robots replace retail workers? A: The evidence points to augmentation rather than wholesale replacement. Most successful deployments use robots for repetitive, physically demanding tasks (shelf scanning, material handling, basic customer greeting) while human workers focus on complex customer service, creative merchandising, and problem-solving. McKinsey estimates humanoid robots could address labor shortage gaps rather than displace existing workers, particularly in markets with persistent unfilled retail positions. Q: How long does it take for a retail robot to pay for itself? A: Payback periods range from 12 to 24 months for purchased robots, based on documented deployments. RaaS models can deliver positive ROI from month one if they displace labor costs exceeding the monthly lease fee. Key factors include operating hours (robots can work 16+ hours/day), the local labor market, and whether the robot generates additional revenue through enhanced customer engagement. Q: What are the biggest challenges of deploying humanoid robots in retail? A: The primary challenges are: (1) integration with existing store systems (POS, inventory management, security), which adds 20–40% to deployment costs; (2) battery life limitations requiring 4–8 hour charging cycles; (3) navigation in crowded, dynamic environments; (4) consumer acceptance, which requires thoughtful introduction and transparent communication; and (5) ongoing maintenance costs of 10–15% of purchase price annually. Q: Are humanoid robots safe to operate around customers? A: Yes, modern retail humanoids are designed with extensive safety systems including force-limiting actuators, proximity sensors, emergency stop mechanisms, and compliant (soft) body surfaces. Robots like Agility Digit and SoftBank Pepper are specifically engineered for "cage-free" operation alongside humans. They meet ISO 13482 safety standards for personal care robots. However, retailers must still conduct site-specific risk assessments and maintain appropriate insurance coverage. Q: What's the difference between humanoid robots and other retail robots? A: Humanoid robots are designed with a human-like form factor — bipedal locomotion, arms, hands, and often a head with facial expressions. This makes them more intuitive for customer interactions and capable of operating in environments designed for humans (stairs, doorways, standard shelving heights). Non-humanoid retail robots (like Simbe Tally or Knightscope K5) are typically wheeled, optimized for specific tasks, and generally less expensive. The choice depends on the use case: customer-facing roles benefit from humanoid form; back-of-store operations may not require it. Related: Humanoid Robots in Hospitality: Enhancing Guest Experiences · Applications of Humanoid Robots Ready to buy? Browse humanoid robots for sale on Robozaps. Humanoid robots in retail are no longer a futuristic concept — they're actively reshaping how stores operate, how customers shop, and how retailers compete. From SoftBank's Pepper greeting customers at Nestlé stores in Japan to Agility Robotics' Digit handling logistics at Spanx warehouses, the retail robotics landscape in 2026 looks radically different from just two years ago. This guide is part of our series on the applications of humanoid robots across 12 industries. In this comprehensive guide, we break down every major humanoid robot deployment in retail, compare the leading platforms, analyze real costs and ROI data, and forecast where the industry is headed through 2030. The State of Humanoid Robots in Retail: 2026 Market Overview The global retail robotics market has surged past expectations. According to Goldman Sachs research, the broader humanoid robot market is projected to reach $38 billion by 2035, with retail and service applications representing approximately 22% of that figure. More immediately, the retail automation market — encompassing humanoid and non-humanoid robots — is valued at approximately $58 billion in 2026, up from $19 billion in 2015. Several forces are converging to accelerate adoption: Manufacturing costs dropped 40% year-over-year between 2023 and 2025, bringing humanoid robot prices from $50,000–$250,000 down to $5,900–$150,000 per unitLabor shortages persist: The U.S. retail sector had approximately 531,000 unfilled positions as of late 2025 (per BLS data), making automation economically essentialAI capabilities matured: Large language models and multimodal AI now enable robots to hold natural conversations, understand context, and make real-time decisionsProduction is scaling: Tesla targets 100,000 Optimus units by 2026; Figure AI's BotQ facility in Austin has 12,000-unit initial capacity scaling to 100,000 annually; Agility Robotics' Oregon factory produces 10,000+ Digit robots per year McKinsey's October 2025 analysis described the industry as "crossing the chasm from concept to commercial reality," noting that several hundred humanoid robots were deployed industrially in 2025, with projections scaling to low thousands by 2026–2027. How Humanoid Robots Are Used in Retail Today Humanoid robots in retail serve functions across four primary categories: customer-facing service, inventory management, logistics and fulfillment, and security/loss prevention. Here's how each works in practice. 1. Customer Service and Engagement Customer-facing humanoid robots are the most visible retail application. These robots greet shoppers, answer product questions, provide wayfinding assistance, and even process simple transactions. SoftBank Pepper remains the most widely deployed customer-facing humanoid in retail. Since its launch in 2014, over 27,000 Pepper units have been sold globally. Major retail deployments include: Nestlé Japan: Pepper deployed across 1,000+ stores to recommend Nescafé machines, generating a documented 15% increase in sales at equipped locationsHSBC branches: Pepper serves as a lobby concierge, reducing wait times by directing customers to appropriate service desksCarrefour (France): Pepper assists shoppers with store navigation and promotional information across select locationsLowe's (U.S.): The LoweBot — based on Fellow Robots' NAVii platform — helped customers locate products across 11 San Francisco Bay Area stores UBTECH Walker S is emerging as a next-generation retail service robot. With 41 degrees of freedom, advanced hand dexterity, and integration with large AI models, Walker S can pick up objects, navigate complex store environments, and interact naturally with customers. UBTECH secured a $1 billion strategic financing facility in 2025, underscoring investor confidence in commercial deployment. 2. Inventory Management and Shelf Scanning While not always humanoid in form, inventory robots often operate alongside humanoid customer-facing units. The key players include: Simbe Robotics' Tally: Deployed across all BJ's Wholesale Club locations nationwide. Tally autonomously navigates aisles, scanning shelves to detect out-of-stock items, misplaced products, and pricing errors. Tally operates during store hours with minimal disruptionBadger Technologies' Marty: Operational in over 500 Giant Eagle and Stop & Shop locations. Marty identifies spills, debris, and inventory gaps, alerting store associates in real timeZebra Technologies' SmartSight: An AI-powered shelf-scanning system deployed by major grocery chains to monitor on-shelf availability with 95%+ accuracy 3. Logistics, Fulfillment, and Back-of-Store Operations This is where humanoid robots are making the fastest inroads in 2026. The economics are clearest here — repetitive, physically demanding tasks in controlled environments. Agility Robotics' Digit: A leading humanoid robot deployed commercially, working at a Spanx warehouse in Georgia for material movement. Digit stands 5'9", weighs 141 lbs, can carry up to 35 lbs, and operates alongside human workers. Amazon has also been testing Digit in its fulfillment centers since late 2023Tesla Optimus (Gen 2): Tesla deployed Optimus units in its own Fremont factory for parts sorting and bin transport, with plans to lease units to external retailers and warehouse operators by late 2026. Target cost: under $20,000 per unit at scaleFigure AI Figure 02: Working with BMW at their Spartanburg, SC manufacturing facility. Figure 02 performs sheet metal insertion and parts transport, demonstrating the warehouse-to-retail-backroom pipeline 4. Security and Loss Prevention Robot security patrols are expanding in retail environments. Knightscope's K5 autonomous security robot patrols parking lots and store perimeters at several major U.S. shopping centers. While not humanoid, the trend toward humanoid-form security robots is accelerating, with companies like 1X Technologies (formerly Halodi Robotics) developing bipedal security robots for indoor retail environments. Comparison: Top Humanoid Robots Used in Retail (2026) Real-World Case Studies: Humanoid Robots in Retail Case Study 1: Pepper at Nestlé Japan — 15% Sales Lift SoftBank's Pepper was deployed across more than 1,000 Nestlé Japan stores starting in 2014, making it one of the largest humanoid retail deployments in history. Pepper's role: recommend Nescafé Dolce Gusto machines to browsing customers using natural-language dialogue and emotion recognition. Results: 15% increase in Nescafé machine sales at Pepper-equipped locationsCustomer dwell time increased by an average of 3 minutes per interactionBrand recall improved measurably in post-visit surveysStores reported increased foot traffic driven by curiosity about the robot The Nestlé deployment demonstrated that humanoid robots don't just assist — they draw customers into the store and create memorable brand experiences. Case Study 2: Agility Robotics Digit at Spanx — First Revenue-Earning Humanoid In 2025, Agility Robotics' Digit achieved a major commercial milestone, operating at a Spanx warehouse in Georgia. Digit handles tote movement between workstations, picking up bins and transporting them across the facility floor. Key metrics: Digit operates 16-hour shifts vs. 8-hour human shiftsCarries up to 35 lbs per trip with consistent speedWorks safely alongside human coworkers (no cage required)Estimated 18-month payback period based on displaced labor costs Case Study 3: Simbe Tally at BJ's Wholesale — Chain-Wide Inventory Intelligence BJ's Wholesale rolled out Simbe Robotics' Tally robot across its entire network of 240+ warehouse club locations. Tally autonomously scans shelves three times daily, generating real-time inventory data that feeds into BJ's replenishment and merchandising systems. Results: Out-of-stock detection accuracy: 95%+Planogram compliance improved by 20%Labor hours redirected from manual shelf auditing to higher-value customer service tasksShrinkage (theft/loss) reduced through continuous monitoring Case Study 4: Bear Robotics Servi at Denny's — Autonomous In-Store Delivery Bear Robotics' Servi — a wheeled service robot with a friendly face display — has been deployed at hundreds of Denny's, Chili's, and other restaurant locations. While not fully humanoid, Servi represents the service robot bridge between traditional automation and full humanoid deployment in retail food service. Results: Servers handle 3–4 additional tables when Servi handles food delivery and bussingTips increased an average of 10–15% as human servers focused more on customer interactionTable turnover time reduced by approximately 10 minutes per seating The Economics: How Much Do Retail Robots Actually Cost? Robot pricing in 2026 varies enormously based on capabilities, form factor, and deployment model. Here's the realistic breakdown: Purchase vs. Robot-as-a-Service (RaaS) Outright Purchase: Full humanoid robots range from $5,900 (Unitree R1) to $150,000+ (UBTECH Walker S). Mid-range options like SoftBank Pepper cost approximately $25,000–$30,000 with ongoing software fees of ~$360/month. Robot-as-a-Service (RaaS): Many retailers prefer RaaS models that bundle hardware, software, maintenance, and support into a monthly fee. Simbe Robotics offers Tally at approximately $2,000/month. Bear Robotics leases Servi at $999/month. This model eliminates upfront capital expenditure and reduces deployment risk. ROI Calculation: When Does a Retail Robot Pay for Itself? Consider a mid-market deployment scenario: Robot cost: $50,000 (purchased) or $2,000/month (RaaS)Displaced labor equivalent: 1.5 full-time employees at $15/hour = ~$47,000/yearAdditional revenue from enhanced customer experience: 5–15% sales lift at equipped locations (documented in Nestlé case)Reduced shrinkage: 2–5% reduction in inventory loss through continuous monitoringPayback period: 12–24 months for purchase; immediate positive ROI for well-deployed RaaS Industry data from successful humanoid deployments suggests 18-month average payback periods, with logistics and warehouse applications delivering the fastest returns. Key Technologies Powering Retail Humanoids in 2026 Large Language Models (LLMs) and Conversational AI The integration of models like GPT-4, Gemini, and Claude into humanoid robots has transformed customer interactions. Rather than following rigid decision trees, retail robots can now hold natural, context-aware conversations — answering nuanced product questions, handling complaints, and even upselling based on customer preferences. Computer Vision and Spatial Intelligence Modern retail robots use a combination of LiDAR, depth cameras, and AI-powered computer vision to navigate crowded store aisles, identify products on shelves, read price tags, and detect spills or hazards. Simbe's Tally uses a 30+ camera array to capture 3D shelf imagery with millimeter-level precision. Dexterous Manipulation The ability to pick up, move, and place objects is critical for retail applications like shelf stocking. Agility Robotics' Digit features custom end-effectors optimized for tote handling. UBTECH's Walker S has 41 degrees of freedom across its whole body, enabling fine manipulation of retail products. Tesla's Optimus Gen 2 demonstrated smooth, human-like hand movements capable of handling eggs without breaking them. Edge Computing and 5G Connectivity Retail robots increasingly process data at the edge rather than relying on cloud connectivity, enabling faster response times and operation in areas with limited internet access. 5G connectivity allows robots to upload inventory data, receive software updates, and coordinate with fleet management systems in real time. Challenges and Limitations of Humanoid Robots in Retail Technical Hurdles Battery life: Most humanoid robots operate 4–8 hours per charge. Digit achieves approximately 4 hours of active operation. This requires charging infrastructure and shift schedulingNavigation in dynamic environments: Crowded retail aisles with shopping carts, children, and unpredictable movement patterns remain challenging for autonomous navigation systemsDexterity limitations: While improving rapidly, most robots still cannot match human dexterity for tasks like folding clothing, arranging delicate displays, or handling irregular-shaped products Economic Barriers Upfront costs: Despite price reductions, a $50,000+ humanoid robot is a significant investment for small and mid-size retailersIntegration costs: Connecting robots to existing POS systems, inventory management software, and store networks adds 20–40% to deployment costsMaintenance: Humanoid robots require regular maintenance, software updates, and occasional hardware repairs. Annual maintenance typically runs 10–15% of purchase price Consumer Acceptance Research published in the Journal of Retailing and Consumer Services (2022) found that consumer anxiety toward robots moderates acceptance. Key findings: Customers respond more positively to robots with humanoid features (face, arms, voice) than purely mechanical designsSocial capability (ability to make small talk, show empathy cues) is the strongest predictor of positive consumer attitudesOlder demographics initially show higher resistance but adapt quickly after 2–3 positive interactionsTransparency matters — customers prefer knowing they're interacting with a robot rather than being deceived Workforce Concerns The narrative around robots replacing human workers remains politically sensitive. However, most successful deployments follow an augmentation model — robots handle repetitive, physically demanding, or low-value tasks, freeing human workers for customer engagement, problem-solving, and creative merchandising. Companies emphasizing augmentation over replacement report higher workforce acceptance and smoother implementations. The Future of Humanoid Robots in Retail: 2026–2030 Forecast Near-Term (2026–2027) Expect 50,000–100,000 humanoid robots deployed globally across warehouses, fulfillment centers, and select retail storesRobot-as-a-Service pricing will drop below $1,000/month for basic service robotsMajor grocery chains will begin chain-wide humanoid deployments for shelf stocking and inventory managementTesla's Optimus will become available for commercial leasing outside Tesla's own operations Mid-Term (2027–2029) Consumer-facing humanoid robots become standard in flagship stores of major retailers (think Apple Store, Nike, and luxury brands)Multi-robot coordination enables teams of humanoids to collaborate on complex tasks like overnight store resetsRobot-to-robot communication and fleet management systems mature, enabling centralized control of hundreds of units across a retail chainHumanoid robots begin handling returns processing and basic checkout assistance Long-Term (2029–2030+) Goldman Sachs projects the humanoid robot market reaching $38 billion by 2035Fully autonomous stores staffed primarily by humanoid robots become viable for specific retail formatsHumanoid robots achieve near-human dexterity, enabling roles like visual merchandising and product demonstrationHome delivery by humanoid robots expands beyond pilot programs How Retailers Should Prepare for Humanoid Robot Adoption Based on analysis of successful deployments, here's a practical roadmap for retailers considering humanoid robots: Step 1: Identify High-Impact Use Cases Start with tasks that are repetitive, physically demanding, or data-intensive. Inventory scanning, material handling, and customer greeting are proven starting points with documented ROI. Step 2: Start with a Pilot Program Deploy 1–3 robots in a single location for 3–6 months. Measure specific KPIs: labor hours saved, customer satisfaction scores, sales lift, and operational accuracy improvements. Step 3: Choose the Right Deployment Model RaaS (Robot-as-a-Service) is recommended for first-time deployers. It minimizes upfront risk, includes maintenance and support, and allows scaling up or down based on results. Step 4: Invest in Workforce Transition Train existing employees to work alongside robots. Create new roles like "robot supervisor" or "automation coordinator." Frame the narrative around augmentation, not replacement. Step 5: Scale Based on Data Use pilot program data to build the business case for chain-wide deployment. Successful pilots at companies like BJ's Wholesale and Amazon led to rapid, chain-wide scaling. Compare on RoboZaps: Unitree R1, Bear Robotics Servi, Pepper, Walker S, Tesla Optimus and Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # UBTech Walker S2: Price, Verified Specs & the 3-Minute Battery Swap (2026) URL: https://blog.robozaps.com/b/ubtech-walker-s-review Author: Radica Maneva Published: 2025-02-12T00:00:00.000Z Updated: 2026-08-01T23:05:48.549Z Last fact-checked: 2026-07-17T00:00:00.000Z Category: reviews TL;DR: The UBTech Walker S2 has no public price; the closest real signal is ~US$106,000 of full-size-humanoid revenue per unit sold, derived from UBTech's FY2025 annual report (1,079 robots delivered, ~¥1.4B in 2025 orders, including a ¥264M border-security contract). Its signature is a 3-minute autonomous battery self-swap. Enterprise and government buyers only; no consumer sales. Key takeaways: - UBTech has never published a Walker S2 price; every circulating dollar figure ($80k to $300k) is unsupported. The only real signal: ~¥760k (~US$106k) of products-and-services revenue per unit sold, derived from the FY2025 annual report, a proxy rather than a list price. - By audited delivered volume the Walker S2 is the most commercially proven humanoid we track: 1,079 full-size units delivered in 2025 and roughly ¥1.4 billion (~$200M) in orders, per filings and disclosed tenders. - Its signature 3-minute autonomous battery self-swap is an operating model, not a spec: the robot stays powered, chooses swapping vs charging itself, and makes continuous shifts mechanically plausible. - Official specs are thinner than the internet suggests: UBTech publishes payload (15 kg), waist articulation, workspace and the swap; height and DoF come from launch press, and no weight, battery or speed has ever been published (the "43 kg" claim is a different robot). - A ¥264 million (~US$37M) contract covers Walker S2 customs and patrol work at China-Vietnam border crossings, with deliveries expected from December 2025; mid-2026 reports describe work underway, not yet primarily confirmed. FAQ: Q: How much does the UBTech Walker S2 cost? A: UBTech has never published a price. The only filings-based signal is roughly ¥760,000 (~US$106,000) of full-size-humanoid products-and-services revenue per unit sold in 2025, a portfolio-wide proxy that includes swap stations and solution work, so the bare robot likely costs less. Directory prices like $80,000 or $150,000-$300,000 have no manufacturer source. Q: Can you buy a UBTech Walker S2? A: Only as an enterprise or government buyer, through contracts and tenders, overwhelmingly in China so far. There is no consumer channel, no list price and no online ordering. The closest consumer products are UBTech's separate U1 line, launched June 2026 with published prices. Q: What are the Walker S2's specifications? A: Officially published: 15 kg payload, ±162° waist rotation with 125° pitch, a 0-1.8 m working envelope, RGB-only stereo vision and the autonomous 3-minute battery swap. Launch materials via Chinese tech press add 1.76 m height, 52 degrees of freedom and 4th-generation dexterous hands. UBTech has never published its weight, battery capacity, runtime or speed. Q: How does the Walker S2 battery swap work? A: The robot carries a dual-battery arrangement, walks itself to a swap station when charge runs low, removes the depleted pack from its own back and inserts a fresh one in about three minutes, staying powered throughout. UBTech says the robot autonomously decides whether swapping or charging makes more sense. Q: How is the Walker S2 different from the Walker S1? A: The S1 (October 2024) was the pilot-era workhorse deployed for training at BYD and Zeekr factories. The S2 (July 2025) added the autonomous battery swap, upgraded 4th-generation hands and a reported 52 degrees of freedom, and became the mass-production model, with deliveries from November 2025. Q: Where is the Walker S2 deployed? A: UBTech's named customers and partners include BYD, Dongfeng Liuzhou, Geely, FAW-Volkswagen Qingdao, Audi FAW, BAIC New Energy, Foxconn and SF Express, with earlier Walker pilots at Nio and Zeekr. An Airbus agreement (January 2026) covers early-stage concept testing, and a ¥264M (~US$37M) contract covers customs and patrol work at China-Vietnam border crossings, with deliveries expected from December 2025 and work reported underway in mid-2026. Q: How many Walker S2 robots have been delivered? A: UBTech's FY2025 filings record 1,079 full-size humanoids delivered in 2025, with more than 80% going to industrial settings; the 1,000th Walker S2 came off the Dongfeng Liuzhou line in late December 2025. Capacity guidance is 5,000 units in 2026 scaling toward 10,000 in 2027. Q: Is there a Walker S3? A: No. As of July 17, 2026 UBTech has announced no Walker S3 or Walker S2 Pro; those names circulate only on investor blogs. The newest UBTech launches are the Walker S2 (industrial) and the U1 consumer line. Q: Who makes the Walker S2, and can you invest? A: UBTech Robotics, founded in Shenzhen in 2012 by Zhou Jian, listed in Hong Kong as ticker 9880. FY2025: ¥2.0B revenue (+53%), a narrowed ¥790M net loss, and a humanoid segment that grew 2,204% to become its largest business. The stock is the only direct way to invest; the Walker S2 itself is not a consumer product. Q: Is the Walker S2 better than Tesla Optimus? A: They are opposites: the Walker S2 has audited deliveries (1,079 units in 2025), real orders and a live government deployment but publishes few specs and no independent performance data; Optimus has no shipped product, no price and no published specs at all. By commercial evidence the S2 leads the industry; by independently verified capability, no humanoid yet leads anyone. How much does the UBTech Walker S2 cost, and can you buy one? UBTech has never published a Walker S2 price, and you cannot buy one as a consumer; it sells through enterprise contracts and public tenders, overwhelmingly in China. Every dollar figure you will see attached to this robot, $80,000 flat, $90,000 "early adopter," $150,000 to $300,000, is unsupported: directory and reseller estimates with no manufacturer source behind any of them. What does exist is something better than a guess: UBTech's own audited numbers. Its FY2025 annual report shows ¥820.6 million of full-size-humanoid products-and-services revenue against 1,079 units sold, which works out to roughly ¥760,000, about US$106,000, of revenue per unit sold. Call it what it is: a directional commercial proxy, not a Walker S2 list price or unit cost, since the figure spans the portfolio and bundles swap stations and solution work. It is still the only price-relevant Walker S2 number on the internet that traces to a regulatory filing rather than a blog. The full math, and every circulating fake price, is below. One disclosure up front: this is desk research built on the sources linked throughout, primary wherever they exist, not a hands-on test; where a figure is derived rather than published, we show the arithmetic. The misinformation audit: what circulates vs the record Walker S2 coverage is a provenance minefield, partly because UBTech sells several very different robots whose numbers get blended. Before this page uses anything, the audit. The estimate rows above deserve a concrete example: a widely-cited robot directory displays "$80,000" directly beside its own "Not Verified" label. What is the Walker S2? The Walker S2 is UBTech's third-generation industrial humanoid, unveiled July 17, 2025, and the robot behind the video that defined it: walking to a dock, pulling the depleted battery out of its own back, clicking in a fresh one, and returning to work in about three minutes, unpowered at no point. UBTech markets it for factory and logistics work under its "BrainNet 2.0" AI architecture, and it is the machine carrying China's most aggressive humanoid commercialization push: real tenders, real filings-verified deliveries, and now a border-security deployment, from a Shenzhen company that has been building humanoids since long before the current boom. If Tesla's Optimus is the world's most-watched humanoid promise, the Walker S2 is arguably the most-purchased humanoid reality, and this page holds it to the same sourcing standard we apply to everyone: official claims, labeled inference, and nothing invented. Official specs vs the media layer UBTech's official Walker S2 page, which we fetched directly, publishes far less than the spec sheets circulating online. Here is the honest split. The FY2025 annual report also confirms the 52 DoF, 15 kg load, 0-1.8 m range and three-minute swap in its business review. The pattern still matters: like Tesla, UBTech publishes selectively, and the directory sites fill silence with confident numbers. Where the official page is silent, this page says so. UBTech's records sit alongside every maker we track in our robot database. The 3-minute battery swap, explained The S2's signature is not a spec, it is an operating model. The robot carries a dual-battery arrangement; when charge runs low, it walks itself to a swap station, removes the depleted pack from its own back, seats a fresh one, and carries on, staying powered throughout the exchange and deciding autonomously whether swapping or charging makes more sense at that moment (the decision logic is on UBTech's official page; the launch demonstration was covered worldwide, from CnEVPost to Forbes). Why it is the feature that sells: every humanoid maker promises around-the-clock labor, but a robot that must dock and charge for an hour every few hours delivers at best a fraction of a 24-hour day. Fleet operators solve this with spare robots or spare batteries and human battery-changers; the S2 removes the human from that loop. It is the same insight behind Figure's hot-swap-between-units idea and Apptronik's swappable packs, executed as full self-service. The honest caveats: sustained real-world duty cycles have not been published by any customer, runtime per pack is not officially stated, and a swap station is one more thing a deployment must buy, which is precisely why our derived price anchor (revenue per robot, including stations) likely overstates the bare unit. The price question, in full Since UBTech publishes nothing, here is every price-relevant fact that actually exists, and the junk it displaces. Read together, the honest range for a Walker S2 engagement is "six figures in US dollars, a high-six-to-seven-figure yuan sum per robot, depending on fleet size, stations and integration," and anyone quoting a tighter number than that is guessing. Two useful comparisons frame it: Unitree sells research humanoids from $13,500 because they do no industrial work, and Western industrial rivals publish no prices at all; UBTech's filings-derived ~$106k revenue per unit sold is, ironically, the most transparent industrial-humanoid price signal in the world right now (annual report). For the cross-market picture, see our humanoid robot cost guide. The orders ledger, dated Sources for the ledger: UBTech's official mass-production release, China Daily on the ¥250M contract, Sina Finance on the ¥264M contract and 21jingji on the ~¥1.4B year-end tally, and SCMP on the border deal. One nuance most coverage misses: the two largest disclosed tenders (Zigong's ¥159M and a related ¥126M Guangxi contract) are government-linked data-collection and testing centers, not factory production lines. A meaningful slice of China's humanoid order boom is the state buying training data capacity, which is real revenue and a different thing from factories buying robot workers. Both are in the ledger; this page keeps them distinguishable. Where Walker robots actually work Automotive plants. The Walker S-series' pedigree runs through China's car industry: Walker S testing at Nio (announced February 2024), Walker S1 in "practical training" at a BYD factory, and the well-documented multi-robot deployment at Zeekr's Ningbo 5G smart factory, where dozens of Walker S1 units worked sorting, material handling and precision assembly in what SCMP called the first multi-humanoid collaborative training of its kind (March 2025). UBTech's official customer list for the S2 era adds Dongfeng Liuzhou, Geely, FAW-Volkswagen Qingdao, Audi FAW, BAIC New Energy, Foxconn and SF Express. The framing that survives scrutiny: through most of 2024-2025 these were pilots and training programs; paid mass delivery began November 2025. Airbus. In January 2026 UBTech announced a cooperation agreement that includes Walker S2 purchases for Airbus facilities, the first humanoid entry into a global aviation OEM's production ecosystem; Airbus itself framed it as early-stage concept testing, a caveat worth keeping attached (Aerospace Global News). The border deployment. The most striking development: a ¥264 million (~US$37 million) contract to put Walker S2 units to work at China-Vietnam border crossings around Fangchenggang, Guangxi, guiding passenger queues, patrolling corridors, and supporting cargo-lane logistics checks, with deliveries expected from December 2025 (SCMP). A July 2026 report describes live customs work as begun, which lacks primary confirmation, so we log it as reported rather than verified. It is, by our assessment, the first disclosed government-customs humanoid deployment anywhere, it exists because the battery swap makes continuous shifts plausible, and it will generate the kind of public, observable performance record factory deployments never provide. Production: the 1,000-robot year The S2's production story is the strongest verifiable claim in the entire humanoid industry, because it runs through audited filings rather than keynotes. Mass production and first deliveries were announced November 17, 2025; the 1,000th Walker S2 rolled off the Dongfeng Liuzhou line in late December (Interesting Engineering); and UBTech's FY2025 results recorded 1,079 full-size humanoids delivered in the year, up from essentially none in 2024, with more than 80% going into industrial settings. UBTech calls itself the first company to deliver 1,000+ full-size humanoids in a single year, a claim worth scoping precisely: rival AgiBot shipped more robots overall (Omdia ranked it first in 2025 shipments, and it passed its 15,000th unit in June 2026), but mostly smaller platforms; UBTech's claim is specifically full-size industrial humanoids. Capacity guidance is 5,000 units a year in 2026 scaling toward 10,000 in 2027, per the official November release; monthly ramp figures circulating on investor blogs are unverified. The Walker family tree UBTech, the company behind it UBTech was founded in Shenzhen in 2012 by Zhou Jian, who still runs it, and listed in Hong Kong in December 2023 as ticker 9880, the "first humanoid robot stock." The FY2025 results that anchor this page's price math: revenue ¥2.0 billion, up 53%; net loss ¥790 million, narrowed 32% year over year; gross margin 37.7%; and a humanoid segment that grew 2,204% to ¥820.6 million, now 41% of the company, its largest business. Strategy signals point the same direction: UBTech spent ¥1.67 billion acquiring control of Shenzhen-listed component maker Fenglong (completed April 2026) to pull actuator supply in-house, and launched the consumer U1 line in June. The credibility contrast worth knowing: crosstown rival Unitree is already profitable (¥591M net profit on ¥1.7B revenue in its IPO filing, approved for listing July 2026), while UBTech is buying market leadership with losses. For the sector map, see our humanoid robot companies guide. Walker S2 vs the field Row sources, July 2026: S2 figures per the FY2025 annual report and the orders ledger above; Figure per its production update and BMW; G1 per Unitree's store (backordered); Digit and Apollo per our linked reviews. The honest comparison: by delivered-and-audited volume, nothing in the West is close, and the S2's real competition is domestic (AgiBot on volume, Unitree on price and profitability). What the Western rivals hold is autonomy evidence in unstructured tasks (Figure's demo ledger) and third-party-verifiable customer metrics (Digit's published throughput). The S2's deployments are the largest, and also the least independently measured; the border project may change that, since it happens in public. Head-to-head: Optimus vs Walker S2, and the wider field in our best humanoid robots guide. Who should care Manufacturers and logistics operators in Asia can treat the Walker S2 as an actually procurable industrial humanoid, with the practical caveats that pricing arrives by negotiation, integration includes swap-station infrastructure, and support outside China is unproven; Western operators should watch the Airbus pilot as the test of whether UBTech can serve a non-Chinese production environment. Investors get the industry's most auditable humanoid story, 9880's filings disclose real units and real revenue, plus real losses, with the metrics to watch being 2026 delivery counts against the 5,000-capacity claim, humanoid gross margin, and whether new orders resume beyond the border deal. And anyone tracking the industry should watch Fangchenggang: the first humanoids doing public-facing government work, observable by anyone with a camera. What to watch, dated Four threads move this page. The border deployment's first months, as reports of live work firm up, potentially the most public performance test any humanoid has faced. UBTech's H1 2026 results (typically late August), which will show whether deliveries are tracking the 5,000-unit capacity claim and how the ASP is moving. U1 consumer deliveries from September 16, 2026, the second front in UBTech's expansion. And any new Walker-series order disclosures, notably absent from April to June 2026 before the border deal, a gap itself worth watching. We re-verify this page at each. The bottom line The Walker S2 is the humanoid that ships. It has no public price, no published weight, and less official spec transparency than its own marketing suggests, and this page treats all of that with the same audit we give Tesla. But it also has what no Western humanoid can show: over a thousand full-size units delivered in a year backed by audited filings, a ¥1.4 billion order book, a self-swapping battery that turns uptime from a promise into a mechanism, and a ¥264 million government border contract moving into public view. The fair summary for a buyer or investor: by audited delivered volume, the most commercially proven humanoid we track, with the least independently verified performance record, and 2026 is the year those two facts collide. The UBTech records in our robot database track the verified state as it changes. Compare on RoboZaps: Walker S2, Figure 03, Walker X, Walker S, Apptronik Apollo and Tesla Optimus. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Automakers and Humanoid Robots: Early Factory Pilots, Real Risks, and Too Much Hype URL: https://blog.robozaps.com/b/the-silent-surrender-traditional-automakers-risk-missing-another-tech-revolution Author: Radica Maneva Published: 2025-02-06T00:00:00.000Z Updated: 2026-08-01T23:04:12.128Z Last fact-checked: 2026-07-06T00:00:00.000Z Category: opinion TL;DR: Automakers are not ignoring humanoid robotics. Several are testing pilots or partnerships, but useful factory deployment still depends on safety, reliability, integration cost, and repeatable ROI. Key takeaways: - The old 'silent surrender' framing was too strong. - BMW, Mercedes, Hyundai/Boston Dynamics, and Tesla all show some automaker activity in robotics. - The real risk is not missing a headline. It is failing to learn from controlled pilots while competitors build integration data. - Humanoid robots still need to prove uptime, safety, dexterity, and economics in real factories. FAQ: Q: Are traditional automakers ignoring humanoid robots? A: No. Some have not announced dedicated programs, but others are running pilots, partnerships, or internal robotics work. The better question is which pilots produce durable factory value. Q: Why would automakers test humanoid robots? A: Factories have repetitive movement, ergonomics problems, labor constraints, and layouts built for people. A humanoid can be attractive if it handles useful tasks without rebuilding the whole line. Q: What is still unproven? A: Reliable autonomy, safe human-robot collaboration, uptime, battery life, task flexibility, service support, and total deployment cost are still the big questions. The surrender story is too simple The automaker story is not a clean split between bold robotics companies and asleep-at-the-wheel carmakers. It is messier and more useful than that. Some automakers are quiet. Others are already testing humanoid robots or backing robotics through partnerships. That matters because factories are one of the first plausible homes for humanoids. They have structured tasks, safety procedures, repeatable workflows, and expensive downtime. They also expose the hard part quickly: a robot has to do useful work over and over, not just look impressive in a demo. What is actually happening BMW has tested Figure robots in factory work, and Figure has published its own BMW deployment material.Mercedes has worked with Apptronik around Apollo-style logistics and manufacturing tasks.Hyundai owns Boston Dynamics, whose Atlas program is explicitly aimed at industrial use and customer testing.Tesla continues to position Optimus as part of its AI and robotics work, although external availability and repeatable factory ROI need proof. Why automakers care Cars made the modern factory, and factories are where industrial robots already have the most history. IFR still shows automotive as one of the central robot-using sectors. Humanoids appeal because they might work around existing tools, carts, doors, shelves, and human-scale layouts. The value is not magic labor replacement. The value, if it arrives, is flexibility: moving parts, inspecting, carrying, loading, unloading, and doing ergonomic work where a fixed industrial robot is too rigid or too expensive to integrate. China and EV-linked robotics Chinese EV and robotics companies are moving quickly, and XPeng's IRON, Unitree's lower-cost platforms, UBTECH, AgiBot, and others deserve close attention. But the article should not treat every Chinese automaker as having a proven humanoid strategy unless the company has actually published one. What remains unproven Can the robot complete a useful factory task for a full shift with low supervision?Can it operate safely around people without slowing the process down?Does it reduce cost after support, integration, downtime, insurance, and teleoperation are included?Can vendors service fleets quickly enough for production environments?Does the automaker keep useful data and process knowledge, or only buy a flashy pilot? The factory lesson Automakers have not silently surrendered. They are in the early part of an uneven industrial robotics race. The winners will not be the companies with the loudest demo. They will be the ones that turn narrow pilots into safe, boring, repeatable work. Compare on RoboZaps: Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree G1 vs Boston Dynamics Atlas: Full 2026 Comparison URL: https://blog.robozaps.com/b/unitree-g1-vs-boston-dynamics-atlas-your-guide Author: Radica Maneva Published: 2025-02-06T00:00:00.000Z Last fact-checked: 2026-07-12T00:00:00.000Z Category: comparisons TL;DR: Different leagues: the Unitree G1 ($13,500 base, ~$43,900 EDU) is the accessible research platform shipping in weeks, while the Boston Dynamics Atlas is a premium enterprise industrial robot — roughly 26 G1s for the price of one Atlas. For most buyers the G1 is the practical choice; Atlas is for seven-figure robotics budgets. Key takeaways: - Unitree G1 wins on accessibility: $13,500 base ($43,900 EDU), ships in 4–8 weeks, ROS2 compatible, 23 DoF base (up to 43 on the EDU dex-hand configuration). - Boston Dynamics Atlas wins on raw capability: world's most advanced locomotion, Google DeepMind AI partnership, premium industrial positioning. - Different categories entirely: G1 is a research platform for universities and developers; Atlas is an enterprise industrial robot for Hyundai/Fortune 500 deployments. - Price difference: 26x — you could buy 26 Unitree G1s for the price of one Atlas. - For most buyers: Unitree G1 is the practical choice. Atlas is for organizations with seven-figure robotics budgets. FAQ: Q: How much cheaper is Unitree G1 than Boston Dynamics Atlas? A: Unitree G1 costs approximately $13,500 while Boston Dynamics Atlas costs around ~$150K — making Atlas 26 times more expensive. For the price of one Atlas, you could purchase an entire fleet of 26 G1 robots. This price difference reflects their different target markets: G1 for research and education, Atlas for industrial manufacturing. Q: Can Unitree G1 do backflips like Atlas? A: No. Boston Dynamics Atlas remains the only humanoid robot capable of performing backflips and parkour-level athletic movements. The Unitree G1 is designed for controlled lab environments with walking speeds around 2 m/s (~7.2 km/h, 4.5 mph). For dynamic locomotion research, Atlas is in a class of its own. Q: Which robot is better for university research? A: Unitree G1 is vastly better for university research. Its $13,500 price point fits academic budgets, the open SDK enables custom algorithm development, ROS2 compatibility integrates with standard research tools, and it's actually available for purchase. Atlas's restricted availability and ~$150K price make it inaccessible for most academic institutions. Q: Is Boston Dynamics Atlas available to buy? A: Boston Dynamics Atlas is currently pre-order only, with deliveries expected in 2026. However, availability is restricted — Boston Dynamics prioritizes strategic partners (primarily Hyundai facilities and select manufacturing partners) over general buyers. Even with a ~$150K budget, you may not be able to purchase an Atlas. Q: What can Unitree G1 actually do? A: The Unitree G1 can walk, navigate obstacles, perform manipulation tasks with its 43-DOF arms, and serve as a platform for AI training and research. It has demonstrated tasks like cracking walnuts, handling delicate objects, and basic locomotion. However, it's primarily a research platform — not a deployable worker. Battery life is about 2 hours, payload is 3 kg, and autonomous capabilities require developer implementation. Q: Why is Atlas so expensive? A: Atlas's ~$150K price (post-CES 2026 third-party estimate; Boston Dynamics has not published an official MSRP) reflects decades of R&D — including DARPA funding in the hydraulic era — plus state-of-the-art actuators and sensors, limited production volume, and Boston Dynamics' premium positioning for industrial deployment. Boston Dynamics isn't optimizing for price — they're building the most capable humanoid possible and pricing for enterprise buyers who need maximum performance. The Google DeepMind AI partnership adds further value. Q: Can I use Unitree G1 for commercial applications? A: Technically yes, but the G1 is primarily designed for research and development rather than production deployment. Its 3 kg payload capacity, 2-hour battery life, and limited autonomous capabilities make it impractical for most commercial applications. For commercial humanoid deployments, consider Unitree R1 or enterprise-focused robots. Q: Which robot has better AI? A: Boston Dynamics Atlas has more advanced built-in AI, especially with the Google DeepMind partnership. However, Unitree G1's open architecture makes it superior for developing AI systems. If you want state-of-the-art autonomous capability out of the box, Atlas wins. If you want to build and train your own AI, the G1's accessibility wins. Unitree G1 vs Boston Dynamics Atlas — the G1 and Atlas live at opposite ends of the humanoid market — $13.5K consumer-edu vs ~$150K enterprise. The G1 starts at $13,500 and ships today; Atlas costs around ~$150K and reflects decades of robotics R&D — DARPA-funded in the hydraulic-Atlas era (2013–2024), self-funded by Boston Dynamics/Hyundai for the current electric Atlas. One democratizes humanoid robotics for researchers and developers; the other pushes the absolute boundaries of what's mechanically possible. This comprehensive 2026 comparison breaks down every spec, capability, and trade-off to help you understand which robot serves your needs. Head-to-Head Comparison Unitree G1: Complete Overview The Unitree G1 represents a paradigm shift in humanoid robotics accessibility. Unveiled at ICRA 2024, it brought full humanoid capabilities to researchers and developers at a fraction of what competitors charge. At $13,500 (with configurations from $13,500), the G1 is the most affordable complete humanoid robot on the market — and one of the few you can actually purchase and receive within weeks. Design Philosophy Standing just 132 cm (4 ft 4 in) tall and weighing 35 kg (77 lbs), the G1 was designed for lab environments, not factory floors. Its compact size means it fits through standard doorways, works safely alongside researchers, and requires minimal infrastructure. The robot folds from 1270x450x200mm to just 690x450x300mm for transport — a practical consideration that larger humanoids simply can't match. Unitree built the G1 for iterative research, not production deployment. That philosophy shows in the open SDK, ROS2 compatibility, and imitation learning capabilities. Researchers can train new behaviors, modify control algorithms, and push updates without waiting for manufacturer approval. Technical Capabilities The G1 base ships with 23 DoF; the EDU configuration adds optional dexterous hands for up to 43 DoF total, giving it strong articulation for the price. The hands feature force-torque sensing for manipulation tasks, though the 3 kg (6.6 lb) payload capacity limits heavy-duty applications. Walking speed maxes out at 2 m/s (~7.2 km/h, 4.5 mph) — sufficient for indoor navigation but not dynamic tasks. The sensor suite includes 3D LiDAR (Unitree's L1/L2 module on EDU), an Intel RealSense D435 depth camera (RGB-D, not LiDAR), an IMU, and force-torque sensors at the joints. This enables spatial mapping, obstacle avoidance, and basic manipulation without additional hardware. Limitations The G1 is explicitly a research platform, not a deployable worker. Battery life around 2 hours limits continuous operation. The 3 kg payload means it can't lift meaningful objects. And while the imitation learning framework is powerful for training, out-of-the-box autonomous capabilities are limited compared to enterprise systems. Boston Dynamics Atlas: Complete Overview The Boston Dynamics Atlas is arguably the most famous humanoid robot in history — and for good reason. Decades of robotics R&D (DARPA-funded in the hydraulic era; BD/Hyundai-funded for the current electric Atlas), MIT Leg Lab origins, and viral YouTube videos have made Atlas synonymous with humanoid capability. The 2024 all-electric redesign (replacing the previous hydraulic version) represents the platform's commercial debut under Hyundai ownership. Design Philosophy At 190 cm (6.2 ft) and 89 kg (196 lbs), Atlas is built for power, not portability. The electric redesign prioritizes safety over raw hydraulic force — featuring body padding, minimal pinch points, and human-safe operating modes. But make no mistake: Atlas remains the most dynamically capable humanoid ever built, now packaged for industrial deployment. The Google DeepMind partnership signals Boston Dynamics' AI strategy. Rather than building AI internally, they're leveraging the world's leading AI research lab to develop Atlas's cognitive capabilities. This combination of Boston Dynamics' mechanical expertise and DeepMind's AI prowess is formidable. Technical Capabilities Atlas's capabilities are unmatched in commercial humanoids. Parkour runs, backflips, 180-degree jumps, and recovery from pushes that would topple any competitor — Atlas has demonstrated all of these publicly. The electric redesign maintains this athletic capability while enabling commercial deployment. Precise specifications for the electric Atlas remain under wraps, but payload capacity is 50 kg (110 lbs) instant and 30 kg (66 lbs) sustained, walking speed surpasses 5.5 km/h (3.4 mph), and the platform can navigate unstructured environments that would stump other humanoids. The sensor suite includes LiDAR, RGB cameras, depth sensors, and proprioceptive feedback across all joints. Limitations Price is the obvious constraint. At approximately ~$150K per unit, Atlas is exclusively for organizations with substantial robotics budgets — Hyundai facilities, select manufacturing partners, and Fortune 500 logistics operations. Availability is also limited; Boston Dynamics carefully controls who can purchase Atlas, prioritizing partners who advance the platform's commercial viability. Category-by-Category Comparison 1. Mobility and Locomotion Winner: Boston Dynamics Atlas This category isn't close. Atlas can perform parkour, execute backflips, recover from dynamic pushes, and navigate terrain that would be impossible for the G1. Where the G1 walks at 2 m/s on flat surfaces, Atlas runs, jumps, and maintains balance through aggressive perturbations. The G1's compact form and 35 kg weight do offer advantages in constrained lab environments — it's less likely to cause damage if it falls, and it navigates tight spaces more easily. But for raw locomotive capability, Atlas represents the state of the art. 2. Dexterity and Manipulation Winner: Tie (context-dependent) The G1 EDU's 43 DoF (with the dex-hand option) is competitive with Atlas's hand-articulation count, though Atlas has more body-axis DoF overall — Atlas Electric is 56 DoF total. The G1 is the better hand-research platform for the price. The G1's force-torque sensing in the hands supports delicate manipulation tasks like cracking walnuts or handling soft objects. However, Atlas's greater payload capacity (~25 kg vs 3 kg) means it can manipulate objects the G1 simply cannot lift. For industrial tasks involving heavy parts, Atlas wins decisively. For research requiring fine motor control development, the G1's higher DOF is advantageous. 3. AI and Software Winner: Boston Dynamics Atlas (capability) / Unitree G1 (accessibility) Atlas benefits from the Google DeepMind partnership — representing arguably the world's most advanced AI research applied to humanoid robotics. The Boston Dynamics Orbit platform integrates fleet management, learning, and autonomous operation at enterprise scale. But the G1's open SDK and ROS2 compatibility make it infinitely more accessible for researchers developing their own AI systems. You can modify the G1's control stack, implement custom algorithms, and iterate rapidly. Atlas's software is largely a black box. For cutting-edge autonomous capability, Atlas leads. For AI research and development, the G1's openness is the decisive advantage. 4. Sensors and Perception Winner: Boston Dynamics Atlas Both robots feature capable sensor suites — LiDAR, depth cameras, IMU, force-torque sensing. But Atlas's perception system has been refined through decades of development and real-world deployment. The integration between sensing and dynamic control is tighter, enabling Atlas's remarkable reactive behaviors. The G1's RealSense D435 (depth camera) plus 3D LiDAR are standard research-grade sensors. Capable for R&D, but not built for the harsh industrial environments Atlas targets. 5. Price and Value Winner: Unitree G1 You could purchase 26 Unitree G1 robots for the price of a single Atlas. For universities, research labs, and developers, the G1 offers extraordinary value. It's the only full humanoid robot accessible to small teams and individual researchers. Atlas's price reflects its capabilities — it's genuinely 10x more capable in many dimensions. But for most use cases outside Fortune 500 manufacturing, that premium is impossible to justify. 6. Build Quality and Durability Winner: Boston Dynamics Atlas Atlas was designed for high-impact industrial environments and falls that would destroy a research-grade humanoid. Early demonstrations showed Atlas being pushed, kicked, and knocked down — recovering flawlessly. The electric redesign adds safety padding while maintaining structural integrity for demanding industrial environments. The G1 is built for research labs, not rough handling. It's durable for its intended purpose but would not survive the treatment Atlas routinely absorbs in demonstrations and deployments. 7. Availability and Deployment Readiness Winner: Unitree G1 You can order a Unitree G1 today and receive it within weeks. It ships globally, requires minimal infrastructure, and can be operational in a lab setting immediately after delivery. Unitree has proven supply chain capability through massive robot dog production. Atlas is pre-order only, with 2026 shipping to select partners. Boston Dynamics controls who can purchase, prioritizing strategic manufacturing partners over general buyers. For most organizations, Atlas isn't actually available regardless of budget. Which Should You Choose? Choose Unitree G1 if you: Have a research or development focus — The G1's open SDK and ROS2 compatibility make it ideal for AI research, control algorithm development, and academic projects. You can modify everything.Budget under $50,000 — At $13,500, the G1 is the only realistic option for small teams, startups, and university labs. Nothing else comes close at this price point.Need a robot this year — G1 ships now. You can have a working humanoid in your lab within weeks, not years.Want to train your own AI systems — The imitation learning framework and open architecture let you develop custom behaviors from scratch.Value compact form factor — For tight lab spaces, the G1's 127 cm height and 35 kg weight are significant practical advantages. Choose Boston Dynamics Atlas if you: Have a seven-figure robotics budget — Atlas requires ~~$150K upfront plus ongoing partnership with Boston Dynamics. This is Fortune 500 territory.Need maximum physical capability — For heavy payloads, dynamic environments, and tasks requiring athletic locomotion, nothing else compares.Want enterprise-grade AI integration — The Google DeepMind partnership means Atlas's cognitive capabilities will continue advancing rapidly.Are a strategic partner for Hyundai/Boston Dynamics — Atlas availability is restricted. You need to be the right kind of customer.Require industrial-grade durability — Atlas is built to survive real-world manufacturing environments. What about the Unitree H2 and R1? Unitree released the R1 ($5,900 base) in July 2025 and the H2 ($29,900) in October 2025. If you're shopping the Unitree side of this comparison today, the H2 is a closer head-to-head match to Atlas on size and dexterity; the R1 is a cheaper entry point than the G1 for first-time buyers. See our R1 vs G1 comparison for the trade-offs. Final Verdict For most buyers — researchers, students, developers, dev teams under ~$50K budget — the G1 is the only realistic choice. The Unitree G1 democratizes humanoid robotics in a way no other robot has. At $13,500, it brings full humanoid capabilities to researchers, developers, and educators who could never afford previous-generation platforms. The open SDK, ROS2 compatibility, and immediate availability make it the practical choice for almost everyone seriously interested in humanoid robotics. Boston Dynamics Atlas is genuinely more capable — dramatically so in locomotion, durability, and integrated AI. If you have a seven-figure robotics budget and qualify as a strategic partner for Boston Dynamics, Atlas represents the pinnacle of humanoid engineering. But for most organizations, it's simply not accessible. The real comparison isn't "which is better" — it's "which can you actually buy and use today?" The G1 ships in weeks; Atlas is allocated through 2026 to Hyundai and DeepMind. For researchers, developers, universities, and startups, that answer is Unitree G1. Compare both robots: Unitree G1 on Robozaps | Boston Dynamics Atlas on Robozaps Related comparisons: Tesla Optimus vs Unitree G1 | Unitree R1 vs G1 | Unitree G1 Review | Boston Dynamics Atlas Review Last updated: May 13, 2026. Specifications sourced from official Unitree Robotics and Boston Dynamics documentation. --- # Unitree Go1 Price 2026: Discontinued Status & Go2 Alternative URL: https://blog.robozaps.com/b/unitree-go1-review Author: Radica Maneva Published: 2025-02-06T00:00:00.000Z Updated: 2026-08-01T23:05:11.064Z Last fact-checked: 2026-07-22T00:00:00.000Z Category: reviews TL;DR: The Unitree Go1 is discontinued: official sales quietly ended around mid-2023, though unitree.com still shows its $2,700-$8,500 launch prices. Used units trade thinly (~$1,699 NOS). Its 2025 CloudSail backdoor (CVE-2025-2894) was killed server-side, not patched on robots. With Go1 batteries out of stock and a new Go2 Air at $1,600, the Go1 is now a legacy-research machine. Key takeaways: - You cannot buy a new Go1: the store listing was still purchasable in April 2023 and vanished around the Go2's mid-2023 launch and no discontinuation notice was ever published, which is why stale pages still present it as current. - The CloudSail backdoor (CVE-2025-2894, disclosed March 2025) was ended by shutting the service down, not by fixing robots; the client code and pi/123 default credentials remain onboard. - The Go1's backdoor is a different bug from UniPwn: the Go1 is not on the UniPwn list, and Go2/G1/H1/B2 were not affected by CloudSail. - Used buyers face one deal-breaker risk: official Go1 batteries ($600/$850) were out of stock at our check, and third-party stock is intermittent. - A new Go2 Air costs $1,600, less than the Go1 ever did, with in-stock parts and active firmware; the only rational used-Go1 buyer has legacy research code and an air-gapped network. FAQ: Q: Can you still buy a Unitree Go1? A: Not new from Unitree: the store listing lost its buy button around mid-2023 and shows "this product is unavailable" today. What remains is a thin used and new-old-stock market on eBay and similar channels, with Go1 Pros listing around $1,699. Q: How much did the Unitree Go1 cost? A: At its June 2021 launch: $2,700 for the Go1 Air, $3,500 for the Go1 Pro, and $8,500 for the developer-tier Go1 Edu, plus roughly $1,000 shipping and US import duties. Q: Why was the Go1 discontinued? A: Unitree never published a reason or even an announcement. The practical answer is the Go2, launched July 2023 at $1,600 with better sensors, which made the Go1 redundant at more than its price. Q: Is the Unitree Go1 backdoor still dangerous? A: The remote vector is dead: Unitree shut the CloudSail tunnel service down on March 29, 2025. But no firmware update ever removed the client or the default pi/123 credentials from robots, so a Go1 on an untrusted network still needs manual hardening. Q: How do I secure a Unitree Go1? A: Log into the onboard Raspberry Pi, change the default password, disable the CloudSail client service and move its autostart directory, and keep the robot on an isolated network. The researchers' published mitigation steps cover this in detail. Q: How fast is the Unitree Go1? A: Tier-dependent: about 2.5 m/s for the Air and 3.5-3.7 m/s for the Pro. The famous 4.7 m/s (17 km/h) figure is Unitree's calculated "ultimate test" sprint maximum, not a shipping-tier speed. Q: Can you still get Go1 batteries and parts? A: With difficulty. Unitree still lists Go1 batteries at $600 (6,000 mAh) and $850 (9,000 mAh) but both were out of stock at our July 2026 check; third-party stock is intermittent. The GO-M8010-6 joint motor remains officially listed. Everything else is salvage. Q: Should I buy a used Go1 or a new Go2? A: A new Go2 Air at $1,600 costs about the same as a used Go1 and comes with 4D LiDAR, in-stock batteries and parts, warranty and active firmware. Buy a used Go1 only for legacy research-code compatibility, with a battery plan and an isolated network. Q: Do all Unitree robots have the Go1 backdoor? A: No. CVE-2025-2894 (CloudSail) is Go1-specific and its service is dead. The separate UniPwn Bluetooth vulnerability, disclosed September 2025, affects the Go2, G1, H1 and B2 but not the Go1. Press coverage often merges the two stories; they are different bugs. Q: Where can I find Unitree Go1 documentation? A: The official spec page at unitree.com/go1 is still live, the V1.4 user manual survives as a PDF mirror, Trossen Robotics maintains the best English doc hub, and the official SDK, URDF models and camera SDK remain on Unitree's GitHub. Our documentation table links all of them. Can you still buy a Unitree Go1, and what did it cost? No — not new, not officially, and Unitree never actually announced that. The Go1, the $2,700 robot dog that made quadrupeds affordable in 2021, was quietly discontinued: its store listing was still purchasable as late as April 2023 before quietly losing its buy button around the Go2's mid-2023 launch, and yet unitree.com/go1 is still live today with launch prices and no end-of-life notice, which is exactly why so many people still ask whether it's for sale. What exists in 2026 is a thin used and new-old-stock market, a robot with a documented (and now server-side dead) remote-access backdoor, and a successor that costs less than the Go1 did. We verified every price and link on this page on July 22, 2026. What happened to the Go1 The archival spec sheet Unitree's Go1 documentation is scattering, so here is the per-tier record, drawn from the still-live official page, the official user manual PDF, and IEEE's launch coverage. One honesty note found inside Unitree's own reseller datasheet, verbatim: the marketing material is "Partly exaggerated, but basically true." Their words. Read the speed row accordingly. The famous "17 km/h, faster than Spot" headline traces to that 4.7 m/s sprint figure, which Unitree's own page presents as a calculated maximum; the robots people actually received ran 2.5–3.7 m/s depending on tier. The compute descriptions in official material are marketing shorthand ("128-core 0.5T"); community documentation identifies a Raspberry Pi plus, on higher tiers, Jetson-class modules, and the security research below confirmed the Pi. Where the Go1 documentation lives now A large share of Go1 searches in 2026 are people hunting documentation that Unitree buried. Working links, verified today: The backdoor, told straight In March 2025, researchers Andreas Makris and Kevin Finisterre published CVE-2025-2894: every Go1 shipped with a pre-installed tunnel client for CloudSail, a remote-access service operated by China-based Zhexi Technology, launching automatically at boot. Anyone holding a valid CloudSail API key could list connected robots, open a tunnel, view live camera feeds without authentication, log in over SSH with the default credentials the robots shipped with (username "pi", password "123"), and pivot into the owner's network. The researchers enumerated 1,919 Go1 units that had checked into the service, including robots on networks at MIT, Princeton, Carnegie Mellon and Waterloo. Their full teardown is published as a report; the NVD rates the CVE 6.6, Medium, whatever the headlines said. Three precisions the coverage usually mangles. First, there is no documented evidence anyone exploited it maliciously in the wild; the demonstrated risk was the architecture itself. Second, "it was patched" is the wrong verb: Unitree's response was to change the service's management key on March 24, 2025 and shut CloudSail down entirely on March 29 — a server-side kill, also covered by the South China Morning Post. No firmware update ever removed the client from the robots, and the default credentials remain unless owners change them. Unitree's statement denied leaving a deliberate backdoor and said the service "will no longer affect the use of Go1 series products." Third, this is a different vulnerability from UniPwn, the September 2025 Bluetooth root exploit that affects the Go2, G1, H1 and B2 — the Go1 is not on the UniPwn list, and the Go1's CloudSail issue does not affect those newer robots. Press pieces routinely stitch the two stories into one "unfixed backdoor" narrative; they are separate bugs with separate statuses. If you run a Go1 today, the researchers' mitigation still applies: log into the onboard Pi, change the default password, disable the CloudSail client service and move its autostart directory so it cannot re-establish — and keep the robot off untrusted networks regardless. Buying a used Go1 in 2026: the checklist The used market is thin: eBay carries occasional new-old-stock and lab-surplus units, with Go1 Pros listing around $1,699 — genuinely below launch price. Before paying anything: Battery first. This is the deal-breaker risk. Unitree still lists Go1 batteries ($600 standard, $850 long-range) but both were out of stock at our check, and third-party stock is intermittent. A Go1 without a healthy battery and a plausible replacement path is a paperweight with 12 motors. Parts second. The GO-M8010-6 joint motor was still listed at $369 in Unitree's store feed at our check, and is about the only Go1-era part with official availability; everything else is salvage. Security third. Assume the CloudSail client and default credentials are present; apply the mitigation above before the robot touches a network. Firmware last. Treat the software as frozen — there is no evidence of active official Go1 firmware support, so what the robot does today is what it will do forever. Go1 vs Go2: the honest redirect The only rational used-Go1 buyer in 2026 is someone with existing Go1-era research code (the SDK and URDF still work, and MIT's walk-these-ways stack remains canonical), a battery plan, and an air-gapped network. Everyone else should read our Go2 review — the successor is cheaper than the Go1 ever was, and note it carries its own security homework. Misinformation audit: seven Go1 claims, checked Why the Go1 still matters Unitree's IPO prospectus reports more than 30,000 quadrupeds shipped from 2022 through September 2025, a wave the Go1 started; the Go1-era install base is plausibly in the low thousands, many in university labs. That's why it refuses to disappear: it remains a standard reinforcement-learning platform, its SDK and models are still public, and half the low-cost quadruped research of the past five years was validated on its 12 joints. The Go1 is also, less flatteringly, the robot that taught the industry to check what ships in the firmware — a lesson its successors are still learning. The bottom line The Go1 was the moment robot dogs got cheap, and it aged into a cautionary tale with a healthy afterlife in research labs. In 2026 it is not a product; it's a decision about legacy code, battery logistics and network hygiene. If that sentence doesn't describe your situation, buy the Go2 — it costs less than the Go1 did, the parts exist, and the skeletons in its firmware are at least the currently-documented kind. For the wider field, our Unitree family guide and R1 review cover where the company went after its dog days. Compare on RoboZaps: Spot. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree Go1 vs. Go2: Which Robot Dog is Best for You? URL: https://blog.robozaps.com/b/unitree-go1-vs-go2 Author: Radica Maneva Published: 2025-02-05T00:00:00.000Z Updated: 2026-08-01T23:08:42.739Z Category: comparisons TL;DR: The Go2 supersedes the Go1 outright: 4D LiDAR, GPT-backed interaction, roughly 30% better motor performance, from $1,600 (Air) and $2,800 (Pro). With the Go1 effectively discontinued, the Go2 is the default choice for a Unitree robot dog. Key takeaways: - The Unitree Go2 features significant technological advancements over the Go1, including 4D LIDAR technology and GPT empowerment, enhancing its navigation, obstacle avoidance, and AI capabilities. - The Go2 boasts improved motor performance and efficiency, with a 30% increase in motor performance and better thermal management, ensuring smooth operation even during intensive tasks. - Starting at $1,600.00 (Air) and $2,800.00 (Pro), the Go2 offers substantial value with its advanced features, robust performance, and enhanced user experience, making it suitable for various applications, including household, industrial, and educational use. FAQ: Q: What are the key differences between Unitree Go1 and Go2? A: The Unitree Go2 has significant upgrades, such as 4D ultra-wide LIDAR and big model GPT empowerment, making it more functional and adaptable than the Go1. Q: How does the Go2 navigate and avoid obstacles? A: The Go2 navigates and avoids obstacles by using an AI-powered recognition system and advanced AI modes for real-time navigation, which improves efficiency by 200% compared to the Go1. Q: What is the battery life of the Unitree Go2? A: The Unitree Go2 has an increased battery capacity of 8,000 mAh and can provide up to 4 hours of operation. The optional ultra-long life battery of 15,000 mAh can extend the battery life further. Q: Can the Go2 be used for educational purposes? A: Yes, the Go2 Edu model is specifically designed for educational use, offering extensive SDKs and curriculum resources, making it an ideal choice for research and learning. Q: What is the price of the Unitree Go2? A: The official starting price of the Unitree Go2 is $1,600.00 (Air) / $2,800.00 (Pro), with additional shipping costs and customs duties varying by country. ‍ Related: Unitree Go1 Review: Robot Dog · Unitree Go2 Review: Quadruped Robot In-Depth Analysis Ready to buy? Browse humanoid robots for sale on Robozaps. Are you wondering about “Unitree Go1 vs. Go2” and which one is the right robot dog for you? This article compares these two models, highlighting the key differences, improvements, and best use cases for each. By the end, you’ll have a clear understanding of which robot dog in the Unitree Go1 vs Go2 debate meets your needs. ‍ ‍ Key Differences Between Unitree Go1 and Go2 Unitree Robotics has significantly advanced with the launch of the Unitree Go2. This model boasts substantial technological improvements over its predecessor, the Unitree Go1, augmenting its functionality and adaptability. One of the most notable upgrades is the integration of 4D Ultra-wide LIDAR technology, which provides a 360°x90° hemispherical recognition system. This advancement allows the Go2 to navigate and avoid obstacles with unprecedented accuracy. The Go2 also introduces a revolutionary feature: the big model GPT empowerment, utilizing advanced AI to refine the robot’s decision-making and responsiveness to intricate commands. The integration of GPT not only enhances the robot’s interaction with its environment but also makes it more intuitive and user-friendly compared to the Go1. ‍ Navigation and Obstacle Avoidance A standout feature of the Unitree Go2 is its advanced navigation system, which is markedly different from its predecessor. Utilizing an AI-powered hemispherical recognition system, the Go2 enhances obstacle avoidance efficiency by 200% compared to the Go1. This significant improvement ensures that the robot can navigate complex terrains with ease and precision. The Go2 robot features: Enhanced navigation systemAdvanced AI modes, including large-scale simulation trainingImproved gaits and an optimized obstacle avoidance strategyVersatility and capability in various environmentsIntelligent interaction and programming capabilities provided by the Go2 appPrecise control and AI-driven obstacle avoidance These features enhance the user experience and make the Go2 robot a powerful tool. 4D LIDAR Technology The enhanced navigation capabilities of the Unitree Go2 are primarily due to the integration of 4D LIDAR L1 technology and wireless vector positioning. This high-tech device offers a range of features that make it stand out among other high-tech devices: 360×90 degree hemispherical recognitionsignificantly improves navigation accuracy and obstacle avoidancesuper small blind spotcan detect and avoid obstacles with remarkable precision One of the standout features of the 4D LIDAR L1 in the Go2 is its minimum detection distance of just 0.05 meters. This capability allows the robot to detect even the smallest obstacles, ensuring safe and efficient navigation in various environments. The integration of such advanced technology makes the Go2 a reliable companion for traversing complex terrains. ISS2.0 Intelligent Side-follow System The ISS2.0 Intelligent Side-follow System in the Unitree Go2 represents a significant upgrade over the previous model’s system. This advanced system achieves a 50% improvement in positioning accuracy, allowing the robot to avoid obstacles more effectively. This improvement ensures that the Go2 can navigate complex environments with greater precision and safety. Another notable enhancement in the ISS2.0 system is the extended remote control distance, which now exceeds 30 meters. This increased range provides users with greater flexibility and control over the robot’s movements. Additionally, the system’s optimized obstacle avoidance strategy ensures that the Go2 can handle various tasks and terrains with ease. ‍ Performance and Motor Efficiency The Unitree Go2 excels in two fundamental aspects of any robot dog: performance and motor efficiency. Compared to its predecessor, the Go2 presents a 30% improvement in motor performance, boasting a peak joint torque of 45 Nm. An intra-joint circuit design, which helps to increase motor efficiency, further supports this improvement. To ensure optimal performance, the joint motors in the Go2 feature heat pipe coolers. These coolers enhance thermal management, preventing overheating and ensuring that the robot operates smoothly even during intensive tasks. These advancements make the Go2 a powerful and reliable quadriceps robot. Speed and Stability With exceptional speed and stability, the Unitree Go2 proves to be a versatile and agile robot. Here are some of its key features: Maximum running speed of approximately 5 meters per secondCan move swiftly across various terrainsMaintains a speed of 2.5 meters per second even when carrying a 7-kilogram payload These features demonstrate the Go2’s robust performance. The advanced gaits of the Go2, such as upside-down walking, adaptive roll-over, and obstacle climbing, enhance its flexibility and stability. These features ensure that the robot can traverse complex terrain and perform a wide range of tasks with ease and precision. Battery Capacity and Endurance For any robot dog to maintain its performance, battery capacity and endurance are pivotal. The Unitree Go2 features an increased battery capacity of 8,000 mAh, with an optional ultra-long-life battery available at 15,000 mAh. This significant upgrade ensures that the Go2 can operate for extended periods without requiring frequent recharges. The voltage in the Go2 has been increased to 28.8 volts, which contributes to improved motor efficiency, power, and stability. With these enhancements, the Go2 model offers an optional ultra-long battery life of up to 4 hours, making it ideal for long-term tasks and missions. This extended battery life is a result of the battery capacity increased in the Go2. ‍ AI and Programming Capabilities Advanced AI and programming capabilities distinguish the Unitree Go2. The Go2 offers the following features: Integrating cutting-edge AI functionalitiesEnhancing autonomous navigation and decision-making abilitiesExceptionally adept at detailed inspections and mapping tasks in hazardous environments. The Go2 robot has several key features: It supports over-the-air (OTA) upgrades, allowing it to improve and evolve its programs automatically by connecting to a cloud-based service.Users can simulate their code in the Go2 app before deploying it to the robot, ensuring accuracy and functionality.The Go2 is highly adaptable and user-friendly. These features make the Go2 a versatile and easy-to-use robot. GPT Empowerment The big model GPT empowerment in the Unitree Go2 is a groundbreaking feature that sets it apart from the Go1. This advanced AI capability enhances the robot’s decision-making and responsiveness to complex commands. With GPT empowerment, the Go2 can understand and interact with its environment more effectively, making it a highly intuitive and intelligent robot. By leveraging GPT models, the Go2 improves its natural language processing capabilities, allowing for better user interaction and control. This makes the Go2 not only a powerful machine but also a user-friendly companion. App Graphical Programming The Unitree Go2’s graphical programming feature is designed to make programming more accessible and intuitive for users. Utilizing a block-based interface, users can easily drag and drop elements to create custom commands and programs. This approach simplifies the programming process, making it more user-friendly, especially for those who may not have extensive coding experience. The app’s graphical programming feature allows users to create and simulate their own programs before deploying them to the robot, ensuring that everything works as intended. This capability enhances the overall user experience, making the Go2 a versatile and customizable robot. ‍ Physical Design and Sensors The Unitree Go2 features: A design that is as robust as it is functionalConstructed from aluminum alloy and high-strength engineering plasticBuilt to withstand rigorous useDimensions of 70x31x40 cm when standingA weight of around 15 kg, including the batteryCompact yet powerful The design enhancements in the Go2, including the use of advanced materials, ensure that the robot is both durable and lightweight. These features make the Go2 suitable for various applications, from household tasks to industrial inspections. HD Cameras and Force Sensors The Unitree Go2 is equipped with: HD wide-angle cameras that significantly enhance its visual perception capabilitiesThese cameras provide a clear and comprehensive view of the robot’s surroundingsThis enables it to navigate and interact with its environment more effectively. In addition to the HD cameras, the Go2 features foot-end force sensors that improve its gait and balance control. These sensors enhance the robot’s movement precision and stability, making it more reliable in various tasks and terrains. Additional Sensors and Customization The Unitree Go2 offers extensive customization options, including the addition of high-performance computing modules such as NVIDIA Jetson Orin for advanced processing needs. These modules enhance the robot’s capabilities, making it suitable for more complex tasks and applications. Furthermore, the Go2 can be equipped with additional sensors that provide enhanced functionality and performance. These add-ons make the Go2 a versatile and adaptable robot, capable of meeting various user requirements. ‍ Practical Applications and Use-Cases From household chores to industrial and research uses, the Unitree Go1 and Go2 cater to a broad spectrum of practical applications, making them more high tech devices than just simple robots. In household environments, these robots can assist with activities such as surveillance, delivery of items, and aiding elderly or disabled individuals. In industrial settings, the Go2 can be employed in areas like petrochemical, transportation, and emergency services. Its advanced AI and programming capabilities make it a valuable tool for detailed inspections, hazardous material handling, and first response support. Household and Industrial Uses In both household and industrial settings, the Go1 and Go2 models prove helpful across a variety of tasks. For example, the Go1 can help with cleaning, security, and elder care. The Go2, with its advanced AI and programming capabilities, can perform more complex tasks such as inspecting hazardous areas and carrying tools and materials. In industrial applications, Unitree robots are employed for tasks like inspection and data collection in environments that may be dangerous for humans. This makes them invaluable in sectors such as petrochemical and transportation. R&D and Education The Unitree Go2 Edu model is specially designed for educational and research purposes. It includes extensive software development kits (SDKs) and curriculum resources tailored for educational institutions. These resources make it easier for students and researchers to explore and develop new technologies using the Go2 Edu. The Go2 Edu’s advanced AI capabilities and programmable features make it an excellent tool for computational research and educational projects, even surpassing the needs of basic computing power. With built-in 4G connectivity and eSIM, it can easily integrate into research projects requiring remote monitoring and control. ‍ Price and Value The price tag of $1,600.00 (Air) to $2,800.00 (Pro) for the Unitree Go2 is a testament to its advanced features and capabilities. Shipping costs vary depending on the country, ranging from $399 to $1000. Additionally, customers must consider customs duties, which are not included in the price. Despite the higher price tag, the Go2 offers significant value with its advanced AI, enhanced navigation, and robust performance. For those looking for a versatile and powerful robot dog, the Go2 is a worthwhile investment. ‍ User Experience and Feedback Mixed user feedback on the Unitree Go1 Pro underscores the potential and challenges inherent in this advanced technology. Some users found the Go1 Pro complex difficult to set up due to the lack of startup instructions. However, many appreciated the helpful tutorials available on the Unitree Support YouTube channel, which provided valuable guidance for getting started. The Go2, with its technically upgraded AI and programming capabilities, promises a better user experience. The advanced features and intuitive design make it more accessible and user-friendly, addressing many of the issues highlighted by Go1 Pro users. This upgrade ensures that the Go2 is not just a powerful machine but also an enjoyable and practical tool for a wide range of applications. ‍ Summary In summary, the Unitree Go2 represents a significant advancement over the Go1, offering enhanced navigation, performance, AI capabilities, and user experience. Its integration of 4D ultra-wide LIDAR and big model GPT empowerment sets it apart, making it a highly adaptable and intelligent robot dog suitable for various applications. Whether you’re looking for a robot to assist with household tasks, industrial inspections, or educational research, the Unitree Go2 offers unparalleled value and functionality. Its advanced features and robust design make it a worthwhile investment for anyone seeking cutting-edge robotic technology. ‍ Compare on RoboZaps: Unitree Go1, Unitree Go2 Air, Unitree Go2 Pro and Unitree Go2 X. Full specifications and a quote route for each, alongside every robot dog we carry. --- # Unitree Go2 Price 2026: Air ($1,600), Pro ($2,800) & EDU Cost URL: https://blog.robozaps.com/b/unitree-go2-review Author: Radica Maneva Published: 2025-02-05T00:00:00.000Z Updated: 2026-08-01T23:05:14.703Z Last fact-checked: 2026-07-23T00:00:00.000Z Category: reviews TL;DR: The Unitree Go2 costs $1,600 (Air), $2,800 (Pro), or $4,500 (Go2 X) direct from Unitree, while the developer EDU is quote-only and realistically $11,190 to $15,600 through US resellers after mid-2026 price cuts. The budget Air skips the features the viral videos lean on, and buyers should run the latest firmware (V1.1.13) after a year of real, now largely mitigated, security disclosures. Key takeaways: - The Go2 Air is $1,600, the Pro $2,800, and the Go2 X $4,500 direct from Unitree; the quote-only EDU realistically costs $11,190 to $15,600 in the US after Unitree's mid-2026 price cuts. - The $1,600 Air skips the features the viral videos lean on: no voice, no follow mode, no 4G, capped speed. The Pro is the real entry point to the full Go2 experience. - Run the latest firmware: V1.1.11 suppressed the Go2's DDS attack path on consumer units and V1.1.13 (February 2026) fixed the app-side flaw; EDU units keep one hole open by design and belong on an isolated network. - Spot's famous $74,500 price is just the 2020 base figure; a configured deployment with the arm and autonomy software passes $100,000, which is why labs buy fleets of Go2s instead. - The Go2 is Unitree's mature cash cow: humanoids now drive over half the company's revenue, but the Go2 still gets firmware updates and just got its biggest price cuts ever ahead of the July 2026 IPO. FAQ: Q: How much does the Unitree Go2 cost in 2026? A: Direct from Unitree, the Go2 Air is $1,600, the Go2 Pro is $2,800, and the Go2 X is $4,500 with controller. US-delivered prices through authorized resellers run higher, around $2,590 for the Air and $3,990 for the Pro, once shipping and customs are included. The developer-grade EDU is quote-only. Q: How much is the Unitree Go2 EDU? A: Unitree does not publish an EDU price; it sells by quote. After the mid-2026 price cuts, reported list pricing is $11,190 for the EDU Standard (40 TOPS) and $13,205 for the EDU Plus (100 TOPS), and live US reseller listings run $11,200 to $15,637.50. Older figures like $3,790 reflect stale bare configurations. Q: What is the difference between the Go2 Air, Pro, and X? A: The Air ($1,600) is speed-capped at 2.5 m/s and drops voice control, the 4G/GPS module, follow mode, and developer access. The Pro ($2,800) restores full speed, voice, and follow mode. The X ($4,500) adds a faster top speed and partial secondary development, the closest thing to an EDU without a quote; per Unitree's matrix, neither the Air nor the Pro offers developer access. Q: Can you buy a Unitree Go2 in the US? A: Yes. You can order direct from shop.unitree.com (ships from China, with $399 to $1,000 freight plus customs), buy from authorized US resellers with local warranty support, or find third-party importer listings on Amazon, Walmart, and Newegg. Note the 4G module is currently unavailable in North America. Q: Is the Unitree Go2 worth it? A: For hobbyists and researchers, yes: nothing else matches its capability per dollar, and the mid-2026 price cuts made the developer tiers cheaper than ever. Buy the Pro rather than the Air unless you understand the Air's limits, and skip the Go2 for industrial inspection work, where the A2, B2, or Spot fit better. Q: Does the Go2 have ChatGPT built in? A: Not the way the 2023 launch marketing implied. Current Pro and X units ship an offline voice assistant for commands and basic interaction, and Unitree's current spec page no longer mentions GPT, though older store copy still does. Truly conversational behavior requires wiring your own LLM stack through the SDK, which needs the EDU tier. Q: How long does the Go2 battery last? A: Consumer tiers carry an 8,000 mAh battery good for roughly 1 to 2 hours, closer to 1 hour with heavy stunts. The EDU's 15,000 mAh battery runs about 2 to 4 hours. Batteries are removable, so spares are the practical fix for longer sessions. Q: Has the Unitree Go2 been hacked? A: Researchers demonstrated real vulnerabilities: the UniPwn Bluetooth root exploit (September 2025) and two DDS remote-code-execution flaws (February 2026, CVE-2026-27509 and CVE-2026-27510). Consumer firmware V1.1.11 suppressed the network attack path, and V1.1.13 (February 2026) fixed the app-side flaw. All of the attacks require radio or network proximity, not the open internet. Q: Is the Go2 EDU safe to run on a university network? A: Not on a shared one. Unitree indicated the unauthenticated DDS programming interface (CVE-2026-27509) will likely stay open on EDU firmware because it is the tier's core feature, and researchers confirmed EDU firmware V1.1.11 could be rooted by anyone on the same network. Until a verified fix ships, run EDU units on an isolated network or VLAN. Q: Should I buy a Unitree Go2 or a Boston Dynamics Spot? A: They rarely compete for the same purchase. Spot is quote-only, with configured deployments typically passing $100,000, and offers certified payloads, weatherproofing, and enterprise autonomy support. The Go2 is faster, half the weight, and a lab can field five SDK-equipped EDU units for less than one configured Spot. Q: What can the Go2 do out of the box? A: On Pro and above: walk, run, and climb stairs on rough terrain, perform handstands, flips, and dances, follow you via a wearable beacon while avoiding obstacles, and respond to offline voice commands. All tiers ship with 4D LiDAR obstacle sensing, and the app offers graphical programming on every tier. Custom autonomy beyond the app's features requires the EDU and your own code. Q: What was the 'Orion' robot dog controversy? A: In February 2026, India's Galgotias University presented a stock Unitree Go2 at a New Delhi AI summit as 'Orion,' an in-house creation. Online viewers identified the robot within hours and the university was asked to leave the event, later saying it had never claimed to build the robot. The Unitree Go2 is the robot dog that made Unitree famous, and in July 2026 its consumer pricing is refreshingly simple: $1,600 for the Go2 Air, $2,800 for the Go2 Pro, and $4,500 for the Go2 X, all sold directly from Unitree's online store. The confusion starts with the fourth tier. The developer-grade Go2 EDU is quote-only, and after Unitree cut prices across the line in mid-2026, a realistically configured EDU runs $11,190 to about $15,600 through US resellers. The $3,800 EDU figure that still circulates on price-guide sites comes from stale, bare-configuration quotes and should be ignored. Those sticker prices also come with caveats the product pages don't advertise. Ordering direct means the robot ships from China, which adds $399 to $1,000 in freight plus customs duties, so US buyers typically pay around $2,590 for a delivered Air through authorized resellers. More importantly, the $1,600 tier lacks several of the features the viral videos lean on, and the demo repertoire was largely filmed on Pro and EDU units. Importing is still lawful for this model: the Go2's FCC authorization dates from 27 May 2024 and is grandfathered under the July 2026 rule. The position for every other robot is in humanoid robots legal in the US. This guide breaks down every Go2 variant, explains what the EDU actually costs and why, covers the security history that most Go2 buying guides skip (including one vulnerability Unitree has said it will likely leave open on EDU units), and puts the Go2 honestly against Boston Dynamics Spot and Unitree's own newer robots. Every price and claim was checked against live listings and primary sources on July 23, 2026. One disclosure up front: this is a source-based buying guide, not a hands-on test, and we link the first-hand reporting where it exists. Unitree Go2 price by variant (July 2026) The consumer prices are live on shop.unitree.com. The EDU, ENT, and Go2-W figures come from authorized US resellers and reported post-cut list pricing, because none of those tiers appear on the consumer store. Warranty is 6 months on the Air and 12 months on everything else, and EDU or custom-configured units are generally non-returnable, which is worth knowing before you wire five figures. The timing context matters. Unitree cut the EDU Standard by $1,810, the EDU Plus by $2,795, and the wheeled Go2-W by a full $6,000 in mid-2026, in the run-up to its Shanghai STAR Market IPO, which won final regulatory approval on July 2, 2026. If you have been waiting for a dip on the developer tiers, this is it. The Go2 EDU price, demystified The single most searched question about this robot is what the EDU costs, and the honest answer is that Unitree does not publish a number. EDU sales run through a quote process (sales_global@unitree.cc), and the price depends on compute, LiDAR, and accessory configuration. What we can do is show you the real ladder, from live reseller listings checked in July 2026: So when you see "$3,790" or "$6,800" quoted for the EDU, you are looking at either a bare-bones legacy configuration or a price that predates the current lineup; no transactable listing we checked in July 2026 supports those numbers. The realistic budget for a turnkey EDU delivered in the US is $11,000 to $16,000. The principal reported difference between EDU Standard and EDU Plus is compute: the Standard carries a Jetson Orin Nano at 40 TOPS, the Plus an Orin NX at 100 TOPS, roughly 2.5× the AI throughput for about $2,000 more. What both EDU tiers add over every consumer Go2 is the substance of the price gap: four foot-end force sensors (factory-installed and impossible to retrofit), the full low-level SDK with direct joint-torque control, an Intel RealSense D435i depth camera, a 15,000 mAh battery good for roughly 2 to 4 hours instead of 1 to 2, and a 12 kg maximum payload. Exact inclusions vary by quotation, so confirm the depth camera, controller, and LiDAR model line by line on your configuration sheet. If you need the robot as a research platform, the EDU is not optional. Nothing else gives you low-level control. What $1,600 actually buys, and what it doesn't The Go2 Air is a real quadruped with the same core hardware DNA as its siblings, but Unitree caps it hard. Top speed is 2.5 m/s against 3.5 m/s on the Pro and 3.7 m/s on the X (with a 5 m/s sport-mode burst on the upper tiers). Slope handling drops from 40° to 30°. Payload falls to about 7 kg. And the feature deletions are the ones that hurt: no voice assistant, no 4G/GPS module, no side-follow mode, and no secondary development access. The wearable-beacon companion mode where the robot trots beside you avoiding obstacles, the thing most people picture when they picture this product, needs a Pro or above. The "ChatGPT robot dog" framing deserves its own correction. When the Go2 launched in July 2023, Unitree marketed GPT integration as a headline feature, and the tech press ran with it. Three years later, Unitree's own spec page no longer mentions GPT at all, though older store copy still waves at "GPT empowerment"; the spec sheet describes "offline voice interaction" for commands, intercom, and music. New Atlas, which ran the most substantial hands-on review of the Go2 Pro, found the conversational mode works but amounts to a novelty delivered in a weird little voice. Researchers who want a genuinely conversational robot wire their own LLM pipeline through the SDK, which again means EDU. Buy the robot for what it does out of the box, not for the 2023 marketing. Go2 specifications by tier Every tier ships with Unitree's own 4D LiDAR and an HD wide-angle camera, and every tier weighs about 15 kg. One nuance if you compare spec sheets: Unitree's current official page lists the LiDAR as the newer L2 with a 360°×96° field of view, while older listings and most reseller pages still say L1 (360°×90°). Unitree's current spec sheet says L2; many reseller and EDU listings still say L1, so confirm the sensor on your exact unit. Functionally both give the robot hemispherical obstacle sensing that nothing else near this price has. Living with a Go2: the owner reality The good news first. The tricks are real: handstands, hind-leg standing, flips, and two dance routines, all stable on grass, gravel, and stairs thanks to genuinely excellent gait control. The follow mode works, with the robot pathing around obstacles on its own. Real-world battery life lands at the advertised 1 to 2 hours, closer to 1 if you spend the charge on stunts. And after three years of production and more than 30,000 Unitree quadrupeds shipped cumulatively, spare parts, community knowledge, and third-party accessories are abundant, which is not something you can say about most robots. The friction is almost entirely software and logistics. Owner forums consistently report a buggy companion app, and at least one firmware update temporarily broke app-to-robot connectivity for a batch of owners, with slow official support responses. The 4G module is currently unavailable in North America for regulatory reasons, so US buyers should not plan around remote-over-cellular operation. And the buying channel determines your support experience: order from Unitree directly and you deal with China shipping, customs, and remote support; buy from an authorized reseller and you pay a markup (a delivered Air runs about $2,590, a Pro about $3,990) for local warranty handling. The Go2 also sells on Amazon, Walmart, and Newegg through third-party marketplace importers, where prices and warranty terms vary listing by listing. For a first robot, the authorized-reseller premium is usually worth it. The security record every Go2 buyer should know The Go2 has a genuine, well-documented security history that mainstream buying guides have largely ignored. None of it means you shouldn't buy one. All of it is worth five minutes of your attention, especially if you are buying the EDU. UniPwn (September 2025). Security researchers Andreas Makris and Kevin Finisterre published a working exploit chain against Unitree's Bluetooth Wi-Fi setup flow: a hardcoded encryption key shared across every unit plus a trivial handshake let an attacker in radio range inject commands and gain root access. Worse, an infected robot could in principle infect others, making the flaw wormable. It affects the Go2 along with Unitree's B2, G1, and H1. Unitree launched a formal security response program in November 2025, but there is no public confirmation of a firmware version that fully closes the Bluetooth chain, so the practical stance is to treat Bluetooth provisioning as a trusted-environment-only operation. The DDS vulnerabilities (February 2026). A researcher demonstrated two remote-code-execution paths on the Go2. The serious one, CVE-2026-27509, let anyone on the robot's network join its internal DDS message bus with no authentication and push Python code that executes as root. On consumer firmware the attack surface shrank with V1.1.11, which stopped broadcasting the vulnerable DDS topics in the researchers' tests on an Air unit, and Unitree then shipped firmware V1.1.13 on February 24, 2026 to fix the second flaw, CVE-2026-27510, which required a rooted, paired phone. The researchers did not retest every model and firmware combination after the patches, so the practical rule is unchanged: run the latest firmware and treat the robot's network as sensitive. The EDU exception. Here is the twist that matters for the biggest group of Go2 searchers: per the disclosure, Unitree indicated the DDS hole will likely not be patched on EDU firmware, because that open programming interface is precisely what the EDU is sold for. Researchers confirmed EDU firmware V1.1.11 could be rooted by anyone on the same network, and no verified fix has shipped since. If you run one in a lab or classroom, put it on an isolated network or VLAN, never on the campus Wi-Fi. What the Go2 is not guilty of. Two widely shared stories belong to other robots. The "secret backdoor in Unitree's robot dog" headlines from March 2025 (a pre-installed tunnel service with default credentials) concerned the discontinued Go1, not the Go2. And the reports of robots sending telemetry to China every five minutes were documented on Unitree's G1 and H1 humanoids; that behavior has not been established on the Go2. Conflating these helps nobody. The practical checklist: update to V1.1.13 or later, keep the robot on an isolated network (mandatory for EDU), leave Bluetooth alone except when provisioning in a trusted space, buy through an authorized channel so you get firmware support, and don't run the companion app on a rooted phone. Rifles, summits, and a robot named Orion Two non-technical stories follow the Go2 around, and buyers should know what is and isn't true in them. The military one is real but narrower than the headlines. In May 2024, Chinese state media showed a rifle-armed quadruped firing live rounds during the Golden Dragon joint exercise with Cambodia. Futurism identified the robots as Unitree Go2s, defense reporting named the mounted rifle as a QBZ-95, and CNN covered the exercise. Unitree has publicly denied supplying the military. In June 2026 the US Department of Defense nonetheless added Unitree to its Section 1260H list of "Chinese military companies." That listing changes nothing about the robot sitting in your living room, but it is increasingly relevant to institutional buyers: some US government-funded programs restrict purchases from 1260H-listed vendors, which is a procurement question to settle before a university or agency orders a fleet. The other story is funnier. In February 2026, Galgotias University displayed a robot dog called "Orion" at India's AI Impact Summit in New Delhi, with a professor telling national media it was developed in-house. The internet identified it as a stock Unitree Go2 within hours, and the university was asked to leave the summit. The buyer-relevant lesson is that the Go2's silhouette is now the most recognizable robot-dog shape on earth. Nobody will believe you built it. Go2 vs Boston Dynamics Spot: an honest comparison Every Go2 article eventually reaches for Spot, usually with the claim that Spot costs $74,500. That number dates from Spot's 2020 launch and only covers the bare robot: Spot sales run through a quote process today, and adding the roughly $65,000 arm, inspection payloads, and autonomy software subscriptions at $15,000 to $25,000 a year pushes a working deployment past $100,000. The honest gap between a Go2 and a deployed Spot is well beyond what the sticker prices suggest. What that money buys is also honest: Spot is an industrial tool with IP54 weatherproofing, certified payloads, a robotic arm option, autonomous mission software with self-charging docks, and an enterprise support organization behind it. Nobody is inspecting an offshore gas platform with a Go2. What the Go2 buys is speed (it is more than twice as fast as Spot's 1.6 m/s), half the weight at 15 kg versus 32 kg, and a price at which a lab can put five SDK-equipped EDU units on the floor for less than one configured Spot. That math is why the Go2 shows up in research labs everywhere, and why the two robots almost never actually compete for the same purchase order. We cover Spot's real pricing in detail in our Boston Dynamics robot dog guide. Where the Go2 sits in Unitree's 2026 lineup Below it, nothing anymore. The original Go1, which started at $2,700 in 2021, was delisted around the Go2's launch and is superseded on every axis; the Go2 is now Unitree's entry point, and our Go1 vs Go2 comparison exists mostly to talk people out of secondhand Go1s. Above it, Unitree has been building out industrial quadrupeds, the A2 and B2, with higher payloads, industrial LiDAR, and pricing north of $30,000, aimed at the inspection work the Go2 was never meant for. And alongside all of them are the humanoids, the G1 and the R1, which now dominate Unitree's story. They dominate its revenue too. Unitree's IPO prospectus shows humanoids growing from under 2% of revenue in 2023 to just over half in the first nine months of 2025, with the quadruped line that built the company sliding to second place at roughly 42%. Read that correctly: the quadruped line, led by the Go2, is the mature cash cow, not an abandoned product. It got a meaningful firmware release in February 2026, a reported self-charging patrol dock in April 2026, and line-wide price cuts in the IPO run-up. The spotlight moved to the humanoids; the maintenance didn't stop. Should you buy a Unitree Go2? Who should buy it Hobbyists and robotics-curious buyers should get the Pro, which is the cheapest tier that delivers the actual Go2 experience: full speed, voice, follow mode, and the trick repertoire. Developers who want to experiment beyond the app should look at the X, which offers the deepest developer access outside the EDU for $4,500 without the quote process. Labs, universities, and serious builders need the EDU, budgeted at $11,000 to $16,000 and operated on an isolated network. At each of those prices, little else on the market comes close on capability per dollar. Who should look elsewhere Skip the Air if the viral videos are what sold you; it lacks the voice, follow mode, and speed that make them compelling, and the $1,200 saved will feel expensive. Skip the Go2 entirely if you need industrial inspection, outdoor weatherproofing, or vendor-supported autonomy, which is A2, B2, ENT, or Spot territory. And if your institution takes US federal funding, check your procurement rules against the 1260H listing before committing to a Unitree fleet. The bottom line Three years in, the Go2 remains arguably the best price-to-capability deal in legged robotics: $2,800 buys a fast, stable, LiDAR-equipped quadruped that was science fiction at any price a decade ago, and the mid-2026 cuts made the developer tiers the cheapest they have ever been. Buy the Pro unless you can articulate exactly why you need the X or the EDU. Then do the three things most buying guides skip: confirm the firmware is V1.1.13 or later, keep the robot off shared networks, and buy through an authorized channel. The robot has earned its reputation. The software around it still demands a little adult supervision. Compare on RoboZaps: Spot and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Tesla Optimus vs Sanctuary AI Phoenix: Full 2026 Comparison URL: https://blog.robozaps.com/b/tesla-optimus-vs-sanctuary-ai-phoenix Author: Radica Maneva Published: 2025-02-05T00:00:00.000Z Updated: 2026-08-01T11:19:32.146Z Category: comparisons TL;DR: Tesla Optimus (173 cm, 57 kg) targets $20,000-$30,000 at mass production; Sanctuary's Phoenix is a research platform with industry-leading hydraulic hands and a 25 kg payload — and no price, since Sanctuary now sells its Physical AI software rather than whole robots. Optimus bets on scale, Phoenix on dexterity IP. Key takeaways: - Tesla Optimus stands at 173 cm (5'8") and 57 kg (126 lbs), with a target price of $20,000-$30,000 at mass production — the most aggressive pricing in the industry. - Sanctuary AI Phoenix features 25 kg (55 lbs) payload capacity and the revolutionary Carbon AI cognitive system — the most advanced general-purpose AI architecture in any humanoid robot. - Optimus wins on: price, mobility (8 km/h run speed), and manufacturing scale. Phoenix wins on: payload capacity, manipulation dexterity, and real-world deployment maturity. - Both robots target industrial and general-purpose applications, but through fundamentally different strategies — Tesla through mass production and Phoenix through robotics-as-a-service. FAQ: Q: Is Tesla Optimus better than Sanctuary AI Phoenix? A: Neither is universally "better" — they excel at different things. Optimus wins on price and mobility. Phoenix wins on manipulation capability and current availability. Choose based on whether you need an affordable, mobile robot (Optimus) or an available, highly dexterous robot (Phoenix). Q: How much does Tesla Optimus cost vs Sanctuary AI Phoenix? A: Tesla targets $20,000-$30,000 for Optimus at mass production scale — the lowest price point for any full-size general-purpose humanoid. Sanctuary AI has never published a Phoenix price and does not sell it; unofficial estimates run $50,000 to $100,000, and since June 2026 it sells its Physical AI software for other robots rather than the humanoid. Q: Which robot has better AI? A: They use fundamentally different AI approaches. Optimus uses Tesla's FSD-derived end-to-end neural networks — excellent for vision and learning at scale. Phoenix uses Carbon AI's hybrid of neural networks and symbolic reasoning — better for explainable decisions and rapid task automation. Tesla has more training data; Sanctuary AI claims faster new-task learning (24 hours). Q: Can I buy Tesla Optimus or Sanctuary AI Phoenix now? A: Neither is for sale to the public. Phoenix was never sold — Sanctuary ran limited pilots and, since June 2026, sells its Physical AI software rather than the robot. Optimus is not yet available: mass production targets late 2026, with consumer availability in late 2027, and no Optimus pre-orders are currently open. Q: What is the payload capacity of each robot? A: Sanctuary AI Phoenix can handle up to 25 kg (55 lbs). Tesla Optimus is rated for 20 kg (44 lbs). Phoenix's higher payload capacity makes it better suited for heavier industrial material handling tasks. Q: Which robot is faster? A: Tesla Optimus is significantly faster with a walk speed of 5 km/h (3.1 mph) and run speed of 8 km/h (5 mph). Sanctuary AI has not disclosed Phoenix speed specifications, and the robot appears to prioritize manipulation over mobility. Q: What are the main use cases for each robot? A: Both target manufacturing, logistics, and general-purpose automation. Optimus emphasizes Tesla factory integration and eventually home assistance. Phoenix focuses on industrial tasks requiring fine manipulation — assembly, quality inspection, material handling. Phoenix's Magna International partnership demonstrates automotive manufacturing viability. Tesla Optimus vs Sanctuary AI Phoenix — which humanoid robot leads in 2026? This head-to-head comparison breaks down every spec, capability, price point, and real-world deployment so you can see exactly how these two general-purpose humanoid titans stack up. Tesla Optimus vs Sanctuary AI Phoenix: Complete Specification Comparison Here's how every key specification stacks up between Tesla's Optimus and Sanctuary AI's Phoenix humanoid robots: Tesla Optimus: Everything You Need to Know (2026 Update) Tesla's Optimus represents Elon Musk's ambitious vision to create an affordable, mass-produced general-purpose humanoid robot. Now entering its Gen 3 phase, Optimus is positioned to be the most price-accessible full-size humanoid on the market. Design and Build The Tesla Optimus Gen 2/3 stands at 173 cm (5'8") and weighs 57 kg (126 lbs) — proportions deliberately chosen to mirror the average human worker for seamless workplace integration. The robot features 28 degrees of freedom total, with 11 DOF in each hand enabling human-like dexterity for tasks like picking up small objects, using tools, and manipulating materials. Tesla's engineering team reduced the robot's weight by 10 kg from Gen 1 while improving balance and locomotion. The actuators are custom-designed by Tesla, using the same motor technology developed for their electric vehicles. Core Technologies FSD-Derived AI Stack: Optimus leverages Tesla's Full Self-Driving neural network architecture, repurposed for humanoid navigation and spatial understanding. This gives Tesla a significant head start in vision-based perception.Dojo Supercomputer Training: Tesla uses its Dojo supercomputer to train Optimus on millions of hours of simulation and real-world data, accelerating capability development.End-to-End Learning: Unlike many competitors using rule-based systems, Optimus employs end-to-end neural networks that learn directly from sensor inputs to motor outputs. Price and Availability Tesla has stated a target price of $20,000-$30,000 at mass production scale — by far the most aggressive pricing in the humanoid robot market. For context, this would make Optimus cheaper than many new cars. The timeline: Tesla's Fremont factory has been repurposed for Optimus production (Model S/X discontinued Q2 2026). Mass production is targeted before end of 2026, with consumer availability pushed to late 2027. No pre-orders are currently open. Sanctuary AI Phoenix: Everything You Need to Know (2026 Update) Sanctuary AI's Phoenix takes a fundamentally different approach to general-purpose robotics. Based in Vancouver, Canada, Sanctuary AI has focused on creating the most capable general-purpose AI system first, then building the hardware to match. Design and Build The Phoenix humanoid stands at 170 cm (5'7") and weighs 70 kg (154 lbs) — heavier than Optimus due to its more robust industrial-grade construction. The robot features 20+ degrees of freedom in its hands alone, giving it superior dexterity for manipulation tasks compared to most competitors. Phoenix's upper body design prioritizes manipulation capability over mobility. Its tactile sensing system enables it to handle delicate objects with human-like precision — a critical requirement for real-world deployment in manufacturing and logistics. Core Technologies Carbon AI Cognitive Architecture: The heart of Phoenix is Carbon AI — a proprietary cognitive system that combines modern neural networks with symbolic reasoning (using Cyc technology with tens of millions of logical rules). This hybrid approach enables explainable, auditable decision-making.24-Hour Task Automation: Sanctuary AI claims Phoenix can automate new tasks within 24 hours of demonstration, significantly faster than competitors requiring weeks or months of programming.General-Purpose Intelligence: Unlike single-task robots, Carbon AI is designed for general-purpose task execution — the robot can learn to perform any task a human can demonstrate. Price and Availability Sanctuary AI has never published a Phoenix price, and there is no way to buy one; independent estimates put it around $50,000 to $100,000, though that figure is unofficial. Phoenix was a research and data-capture platform run through pilots, and since June 2026 Sanctuary sells its Physical AI software for other robots rather than the humanoid itself. Phoenix is not for sale. Sanctuary ran limited, mostly teleoperated pilots — including a partnership with Magna International announced in 2024 — but no one has ever bought a Phoenix, and in June 2026 the company pivoted away from selling the whole robot. Head-to-Head: Tesla Optimus vs Sanctuary AI Phoenix Performance Comparison 1. Mobility and Agility Winner: Tesla Optimus Tesla Optimus demonstrates superior mobility with a walking speed of 5 km/h (3.1 mph) and running capability up to 8 km/h (5 mph). Recent demonstrations have shown the robot performing squats, yoga poses, and recovering from pushes — evidence of advanced balance and locomotion control. Sanctuary AI Phoenix does not walk: its current Generation 8 rides a wheeled base, and Sanctuary said bipedal legs were “too frail to support a strong torso.” Phoenix prioritizes upper-body capabilities and manipulation over athleticism. For applications requiring the robot to move quickly between workstations or cover large factory floors, Optimus has the clear advantage. 2. Dexterity and Manipulation Winner: Sanctuary AI Phoenix Phoenix wins decisively on manipulation capability. With 20+ degrees of freedom in its hands alone and advanced tactile sensing in every fingertip, Phoenix can perform delicate operations that require human-like precision — handling eggs without breaking them, manipulating small components, and executing complex assembly tasks. Optimus has 11 DOF per hand and demonstrated improved tactile sensing in Gen 2, but Sanctuary AI's focus on dexterous manipulation gives Phoenix an edge for fine motor tasks. Additionally, Phoenix's hands use HaptX-style haptic feedback for realistic touch perception during teleoperation training. 3. AI and Software Winner: Tie — Different Strengths This category is genuinely close, with each robot excelling in different AI paradigms. Tesla Optimus leverages the FSD neural network stack, giving it excellent vision-based perception and end-to-end learning capabilities. Tesla's massive data advantage from millions of vehicles provides training data that no robotics company can match. Sanctuary AI's Carbon AI combines neural networks with symbolic reasoning, enabling explainable AI decisions — critical for enterprise deployment where auditable reasoning is required. Phoenix's 24-hour task automation claim, if accurate, represents industry-leading adaptability. For pure learning speed and scale, Tesla may have the edge. For explainability and enterprise-grade reasoning, Phoenix leads. 4. Sensors and Perception Winner: Sanctuary AI Phoenix Phoenix features tactile sensors throughout its hands and arms, enabling sophisticated touch-based perception that Optimus currently lacks at the same level. This tactile sensing is essential for manipulation tasks where visual feedback alone is insufficient. Optimus relies primarily on camera-based perception (derived from Tesla's FSD sensor suite) with force-torque feedback at the hands. While effective for many tasks, the absence of comprehensive tactile sensing limits fine manipulation capability. 5. Price and Value Winner: Tesla Optimus Tesla's target price of $20,000-$30,000 is revolutionary for the humanoid robot market. For comparison: Unitree G1: $13,500 (smaller research platform)Unitree H1: $90,000Boston Dynamics Atlas: ~$150K-$320KAgility Digit: ~$250,000 Sanctuary never published a Phoenix price (unofficial estimates run $50,000 to $100,000) and you cannot buy one, which makes a direct price comparison difficult. For price-sensitive buyers or mass deployment scenarios, Optimus offers unmatched value potential. 6. Build Quality and Durability Winner: Sanctuary AI Phoenix Phoenix's heavier 70 kg (154 lbs) construction reflects industrial-grade build quality designed for manufacturing environments. Sanctuary announced a Magna International partnership in 2024, though no plant, unit count or results were ever disclosed. Tesla Optimus, at 57 kg (126 lbs), prioritizes lighter weight for mobility and energy efficiency. While Tesla's manufacturing expertise suggests quality construction, Gen 2/3 has not yet been deployed at scale in industrial settings. Phoenix's track record in demanding factory environments gives it the edge here. 7. Real-World Deployment Winner: Sanctuary AI Phoenix Phoenix was never sold and cannot be bought; Sanctuary ran limited, largely teleoperated pilots with partners like Magna International, and in June 2026 pivoted to selling its Physical AI software for other robots rather than the humanoid. Tesla Optimus remains in pre-production, with mass production targeted for late 2026 and consumer availability in late 2027. While Tesla's internal factory deployments are underway, no external commercial deployments have been announced. For buyers who need a humanoid robot today, Phoenix is available; Optimus is a future promise. Which Should You Choose? Choose Tesla Optimus if you: Need mass deployment at scale: At $20,000-$30,000 per unit, you could deploy 5-10 Optimus units for the price of one competitor — transformative for large-scale automation.Prioritize mobility: If your use case requires robots moving quickly between workstations or covering large areas, Optimus's superior locomotion is essential.Are willing to wait for 2026/2027: Optimus isn't available today, but if your timeline allows waiting, the price-to-capability ratio will be industry-leading.Want Tesla ecosystem integration: For facilities already using Tesla energy products or vehicles, Optimus may offer future integration benefits. Choose Sanctuary AI Phoenix if you: Study best-in-class dexterity: Phoenix's hydraulic hands and tactile sensing are among the most advanced anywhere, even though you cannot buy the robot itself today.Require advanced manipulation: For tasks demanding human-like dexterity (assembly, sorting, handling delicate items), Phoenix's 21-DoF hands and tactile sensing are superior.Track dexterous-manipulation tech: Sanctuary is now licensing its Physical AI and hydraulic hands to run on other companies' robots rather than selling Phoenix.Need explainable AI decisions: For regulated industries or enterprise environments requiring auditable AI reasoning, Carbon AI's symbolic reasoning provides transparency that black-box neural networks cannot. Final Verdict: Tesla Optimus vs Sanctuary AI Phoenix Overall Winner: Depends on your timeline and priorities. For precision manipulation, Sanctuary AI Phoenix has arguably the most advanced hands and cognitive AI in the field, but you cannot buy one today, and Sanctuary has pivoted to selling its software rather than the robot. Treat it as the manipulation benchmark, not a robot you can deploy now. If you can wait until 2026/2027 and prioritize cost-effective mass deployment, Tesla Optimus represents a paradigm shift in humanoid economics. No other company can match Tesla's manufacturing scale, and at $20,000-$30,000 per unit, Optimus could make humanoid robots as accessible as vehicles. The market may ultimately have room for both approaches — Phoenix's technology leading on high-dexterity manipulation, and Optimus enabling mass-market humanoid adoption through aggressive pricing. Both companies are pushing the frontier of what humanoid robots can achieve. Compare both robots side by side: Tesla Optimus on Robozaps | Sanctuary AI Phoenix on Robozaps | Browse all humanoid robots Last updated: July 27, 2026. Specifications sourced from Tesla official announcements, Sanctuary AI documentation, and Robozaps robot database. Robozaps is a humanoid robot marketplace — we maintain comprehensive product databases and may earn referral fees from qualifying purchases. Related Articles: Tesla Optimus Gen 2 Review: Full Specs & PerformanceSanctuary AI Phoenix Review: Price, Specs & AnalysisTesla Optimus vs Apptronik Apollo ComparisonThe Best Humanoid Robots of 2026How Much Does a Humanoid Robot Cost? --- # Navigating the Uncanny Valley: Why Almost-Human Robots Make Us Uneasy URL: https://blog.robozaps.com/b/uncanny-valley Author: Radica Maneva Published: 2025-01-23T00:00:00.000Z Updated: 2026-08-01T23:03:50.731Z Category: opinion TL;DR: The uncanny valley hits when robots reach 80-95% human likeness: close enough to trigger human pattern-matching, imperfect enough to unsettle. Coined by Masahiro Mori in 1970, its biggest triggers are eye movement, facial expressions and motion, which is why designers choose between obviously mechanical and near-perfect realism. Key takeaways: - The uncanny valley occurs when robots reach 80-95% human likeness—close enough to trigger human pattern-matching, but imperfect enough to register as "wrong" - Coined by Masahiro Mori (1970): Japanese roboticist who first documented this acceptance dip - Design implications: Companies must choose between obviously mechanical (safe) or near-perfect human realism (risky but potentially rewarding) - Key triggers: Eye movement, facial expressions, and motion are the biggest uncanny valley triggers—not static appearance - Real-world examples: The Polar Express film, some CGI characters, and ultra-realistic robots like Ameca FAQ: Q: Who coined the term "uncanny valley"? A: Japanese roboticist Masahiro Mori coined the term "uncanny valley" (bukimi no tani) in 1970. He proposed it in an essay describing the relationship between human likeness and emotional response to robots, noting the characteristic dip in acceptance as robots approach but don't quite achieve human appearance. Q: What triggers the uncanny valley response? A: Movement and facial expressions are the primary triggers—not static appearance alone. Subtle errors in eye movement, blink timing, lip synchronization, and facial micro-expressions create the strongest uncanny responses. A perfectly realistic still image may look fine, but animation often reveals the uncanny valley. Q: Why did The Polar Express feel creepy? A: The Polar Express aimed for photorealistic human characters but fell into the uncanny valley. The animation captured overall human appearance but failed to replicate subtle eye movements, facial muscle dynamics, and skin texture responses. Audiences subconsciously detected these errors, triggering unease. Q: Do all people experience the uncanny valley equally? A: No—sensitivity to the uncanny valley varies significantly. Some research suggests people with higher empathy or those who work closely with humans (healthcare workers, therapists) may be more sensitive. Age also plays a role, with some studies showing children are less affected than adults. Q: How do robot companies avoid the uncanny valley? A: Most successful humanoid robot companies use one of three strategies: stay obviously mechanical (Boston Dynamics), use stylized humanoid designs that are clearly artificial (SoftBank Pepper), or invest heavily in pushing through to near-perfect realism (Engineered Arts' Ameca). The middle ground—almost-human—is the danger zone. Q: Is Sophia the robot in the uncanny valley? A: Sophia by Hanson Robotics is a controversial case. She sits near the edge of the uncanny valley—some find her fascinating and engaging, while others experience the classic uncanny discomfort. Her creators intentionally push toward human realism, accepting that some people will find her unsettling. Q: Will the uncanny valley ever be overcome? A: Yes, likely through two paths. First, technology improvements will eventually enable robots to cross the valley by achieving near-perfect human likeness, eliminating the subtle errors that trigger unease. Second, cultural familiarity with humanoid robots may reduce sensitivity to the effect over time. Related: What Is a Humanoid? Definition and Examples · The Evolution of Humanoid Robots from Science Fiction to Reality Ready to buy? Browse humanoid robots for sale on Robozaps. Our relationship with robots is about to get complicated. As humanoid robots edge closer to human likeness, we're confronting an unexpected psychological barrier—one that could shape the future of human-robot interaction. Quick Answer: The uncanny valley is a psychological phenomenon where robots that appear almost—but not quite—human trigger feelings of unease or revulsion. Coined by roboticist Masahiro Mori in 1970, it describes the dip in human comfort that occurs when artificial beings look 80-95% humanlike. Robot designers must either stay clearly mechanical or push through to near-perfect human realism to avoid this unsettling zone. What Is the Uncanny Valley? The uncanny valley is a psychological phenomenon describing how human emotional response shifts as artificial entities become more humanlike. The concept is simple but profound: when robots appear clearly mechanical, we accept them easily. Think of R2-D2 or industrial robot arms—they're obviously machines, and we're comfortable with that. As robots become more humanlike, our comfort with them initially increases. We anthropomorphize them, finding them cute or endearing. But then something strange happens. When robots reach a certain threshold of human similarity—looking almost but not quite human—our comfort level plummets. Instead of increasing acceptance, we experience a visceral unease. Small imperfections in appearance or movement that might go unnoticed in more mechanical robots suddenly become deeply unsettling. Why Do Almost-Human Robots Make Us Uncomfortable? The uncanny valley triggers a fundamental conflict in human perception. Our brains are wired to read faces and detect subtle social cues—it's how we survived as social animals for millions of years. When a robot looks 90% human, our brain initially categorizes it as human, then rapidly detects inconsistencies. These inconsistencies create cognitive dissonance. The eyes might move slightly wrong. The skin texture might be too perfect. The timing of blinks might be off by milliseconds. Each discrepancy triggers a subconscious alarm: something is pretending to be human but isn't. What Famous Examples Demonstrate the Uncanny Valley? Remember The Polar Express? The film's characters were meant to be photorealistic, but audiences found them disturbing. Their almost-human faces triggered the same psychological response that makes ultra-realistic robots uncomfortable. The eyes seemed dead, the movements slightly off—just enough to remind us that something wasn't right. In robotics, examples include early android attempts where the realism was impressive but incomplete. Sophia by Hanson Robotics deliberately pushes toward human realism, landing in contested territory. Some find her fascinating; others find her deeply unsettling. How Are Robot Companies Navigating the Uncanny Valley? This isn't just about aesthetics. The uncanny valley has profound implications for robotics development. Companies investing millions in humanoid robots face a crucial design challenge: how human is too human? Some are choosing to sidestep the valley entirely. Boston Dynamics' robots perform incredible athletic feats while maintaining an obviously mechanical appearance. Others, like Hanson Robotics, push toward human realism despite the risks. Each approach reflects different philosophies about human-robot interaction. Does the Uncanny Valley Matter for Home Robots? As we move toward a future where robots become increasingly integrated into daily life, understanding and addressing the uncanny valley becomes crucial. It's not just about making robots that work well—it's about making robots that we can work with comfortably. For home robots, the design choice is critical. A robot helping with chores needs to be accepted by all family members, including those more sensitive to the uncanny effect. Most consumer robot companies are wisely choosing stylized or clearly mechanical designs. Will the Uncanny Valley Eventually Disappear? Two factors could reduce the uncanny valley effect over time. First, as robotics technology improves, robots may successfully cross the valley by achieving near-perfect human realism—eliminating the subtle "wrongness" that triggers unease. Second, as people become more accustomed to humanoid robots in daily life, the novelty and unfamiliarity that amplifies the uncanny effect may diminish. Younger generations growing up with humanoid robots may have higher tolerance. For now, the valley remains a reminder that human perception is complex and often counterintuitive. As we build machines that increasingly mirror ourselves, we're learning as much about human psychology as we are about robotics. Compare on RoboZaps: Pepper, Sophia, Ameca, Tesla Optimus and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Human-Robot Coexistence: Safety, Privacy and Trust URL: https://blog.robozaps.com/b/humanoid-robots-are-here-but-are-we-ready-to-coexist Author: Radica Maneva Published: 2025-01-10T00:00:00.000Z Updated: 2026-08-01T23:03:54.858Z Last fact-checked: 2026-07-14T00:00:00.000Z Category: opinion TL;DR: Human-robot coexistence should be judged deployment by deployment. Require a task-specific safety case, named accountability, privacy and remote-assistance disclosure, usable human override, incident logs and evidence proportionate to the risk. Key takeaways: - A humanoid robot is not safe in the abstract; the safety case must match its task, environment and the people exposed. - Manufacturers, integrators and operators should allocate safety, maintenance, data and incident responsibilities before deployment. - Sensor collection and remote assistance should be minimized, controlled and disclosed to workers and home users. - Trust should follow operating evidence, known limits, intervention logs, incident reporting and tested human override. FAQ: Q: Are humanoid robots safe around people? A: Not by category alone. Safety depends on the robot, task, environment, people exposed and controls in place. Buyers should require a risk assessment, tested safe state, emergency stop, access rules, maintenance isolation, training and incident response for the real deployment. A successful demonstration is not a general safety approval. Q: Who is responsible if a humanoid robot causes harm? A: Responsibility depends on the facts, contracts and applicable law. Before deployment, the manufacturer, integrator and operator should document who owns risk assessment, configuration, supervision, maintenance, software updates, incident response and insurance. A human should remain clearly accountable for stopping and escalating unsafe operation. Q: Can a humanoid robot record or transmit data from my home or workplace? A: Potentially. Cameras, microphones, logs and remote-assistance systems may collect or transmit sensitive information. Users should be told what is captured, why it is needed, where it goes, how long it is retained, who can access it and when a remote operator can view or control the robot. Collection should be minimized and remote access visibly disclosed. Q: How can workers safely share space with robots? A: Start with a task-specific risk assessment and involve the people who do the work. Define operating zones, speed and force limits, stop and reset procedures, maintenance isolation, training, supervision and near-miss reporting. Expand only after the robot meets written acceptance criteria in the intended environment. Q: What makes a humanoid robot trustworthy? A: Trust should be calibrated to evidence. A trustworthy deployment states what is autonomous and what uses remote assistance, publishes relevant limits and failure modes, records interventions and incidents, protects data, provides human override and names the people accountable for operation. Marketing confidence is not a substitute for operating evidence. Key Takeaways Human-robot coexistence is a safety and governance problem, not a forecast about when robots will arrive.A robot is not safe in the abstract. The safety case must match the exact task, environment and people exposed.Responsibility must be allocated before deployment across the manufacturer, integrator and operator, with a named human able to stop and escalate operation.Cameras, microphones and remote assistance make privacy and transparency operational requirements, especially in homes and shared workplaces.Trust should follow evidence: known limits, intervention logs, incident reporting, data controls and tested human override. The Coexistence Question Is About Control Humanoid robots do not need to be everywhere before their safety, privacy and accountability questions matter. A single robot operating near workers, residents, children or patients can create physical and informational risks. The useful question is therefore not whether society is emotionally ready. It is whether each deployment has controls proportionate to the harm it could cause. For forecasts about capability and adoption, use the separate future of humanoid robots analysis. This article owns the conditions under which people should be expected to share spaces and decisions with robots. Five Tests for Responsible Human-Robot Coexistence 1. Physical Safety Must Match the Real Task The NIOSH Center for Occupational Robotics Research studies hazards created when people and robots work together and develops risk profiles, guidance and training. Its focus is practical: the type of robot, work process and human interaction all affect the risk. ISO 13482 sets safety requirements and guidance for personal-care robots, including inherently safe design, protective measures and information for use. A second edition for service robots is still under development. Neither document makes every humanoid safe; the intended use and foreseeable human contact still have to be assessed for the actual product and setting. At minimum, a deployment should define the robot's safe state, emergency-stop and restart procedure, operating boundaries, speed and force limits, maintenance isolation, supervision, training and response to a fault. These controls should be tested before people are asked to work or live beside the system. 2. Responsibility Must Be Named Before Harm Occurs A robot may combine hardware from one company, software from another, site integration by a third party and daily operation by the customer. That makes vague assurances about the vendor being responsible inadequate. The deployment record should name who owns configuration, risk assessment, access control, maintenance, software updates, supervision, incident response and insurance. Legal liability depends on jurisdiction, contracts and the facts of an incident. Operational accountability should not wait for a court dispute: a named human owner must be able to stop the robot, preserve logs, notify affected people and escalate a safety or privacy failure. 3. Privacy Includes Sensors and Remote Assistance Humanoid robots may carry cameras, microphones, depth sensors and system logs. They may also use remote assistance when autonomy fails. That creates questions ordinary product pages often skip: what is recorded, why it is needed, whether it leaves the site, how long it is retained, who can access it and whether a remote operator can see or control the robot. The voluntary NIST AI Risk Management Framework treats privacy, transparency and accountability as connected parts of trustworthy AI risk management. A sensible deployment minimizes collection, limits access, makes remote assistance visible, gives users meaningful controls and documents deletion and incident procedures. 4. Human Oversight Must Be Usable Under Pressure A nominal override is not enough if workers cannot reach it, do not know when to use it or fear being blamed for stopping production. People exposed to the robot need clear authority to pause operation, an accessible emergency stop, a known escalation route and training that covers abnormal behaviour rather than only the happy path. Logs should distinguish autonomous actions, remote commands, human interventions, faults and software changes. That makes incidents easier to investigate and prevents automation from becoming a way to obscure who made or approved a consequential decision. 5. Trust Must Be Calibrated to Evidence NIOSH notes that emerging robotics can reduce hazardous work but can also create injury risks, distraction, mental stress and lack of trust. A trustworthy deployment does not ask people to treat a confident interface as proof of competence. It states where the system fails, how much teleoperation or resetting it needs and what evidence supports its permitted tasks. Useful evidence includes task success under realistic variation, intervention frequency, near misses, safety stops, uptime, update history and independent testing where appropriate. A demonstration can establish possibility. It cannot establish safe, reliable coexistence across other tasks and environments. What Employers Should Require Before Deployment A written description of the task, operating area, people exposed and foreseeable misuse.A task-specific risk assessment and acceptance criteria agreed before the pilot begins.Named owners for safety, configuration, supervision, maintenance, data governance and incident response.Worker consultation and training, including authority to stop work without penalty.Documented safe states, emergency stops, restart controls, maintenance isolation and degraded-mode behaviour.Intervention, fault, near-miss and incident logs that separate autonomous and remote-assisted operation.Data rules covering sensors, retention, access, remote operators, security and deletion.Pilot stop rules and a requirement to re-assess risk after material hardware, software, task or layout changes. What Home Users Should Demand Homes are not controlled factory cells. Children, pets, stairs, clutter, visitors and private conversations change both the physical and privacy risk. Buyers should ask for the intended-use statement, excluded uses, safe operating limits, emergency controls, update policy, repair support and what happens when connectivity or autonomy fails. The seller should also disclose when sensors record, when data leaves the home and whether remote staff can view or control the robot. A product described as autonomous should not quietly depend on undisclosed people looking through its cameras. Policy Priorities That Matter Now Good policy does not need to predict the year of mass adoption. It can improve current deployments by making evidence and responsibility clearer. Priorities include: Safety requirements and procurement rules tied to the intended task and environment.Consistent reporting of serious incidents and preservation of the logs needed to investigate them.Disclosure of remote assistance, sensor use, material limitations and who is accountable for operation.Auditability for important software changes and human interventions.Worker and affected-user participation in risk assessment, pilots and post-deployment review.Clear routes for complaints, stopping unsafe operation and obtaining redress. These controls are technology-neutral enough to survive changes in vendors and models. They focus on observable risk, evidence and responsibility instead of treating either innovation or prohibition as the default answer. Frequently Asked Questions Are humanoid robots safe around people? Not by category alone. Safety depends on the robot, task, environment, people exposed and controls in place. Buyers should require a risk assessment, tested safe state, emergency stop, access rules, maintenance isolation, training and incident response for the real deployment. A successful demonstration is not a general safety approval. Who is responsible if a humanoid robot causes harm? Responsibility depends on the facts, contracts and applicable law. Before deployment, the manufacturer, integrator and operator should document who owns risk assessment, configuration, supervision, maintenance, software updates, incident response and insurance. A human should remain clearly accountable for stopping and escalating unsafe operation. Can a humanoid robot record or transmit data from my home or workplace? Potentially. Cameras, microphones, logs and remote-assistance systems may collect or transmit sensitive information. Users should be told what is captured, why it is needed, where it goes, how long it is retained, who can access it and when a remote operator can view or control the robot. Collection should be minimized and remote access visibly disclosed. How can workers safely share space with robots? Start with a task-specific risk assessment and involve the people who do the work. Define operating zones, speed and force limits, stop and reset procedures, maintenance isolation, training, supervision and near-miss reporting. Expand only after the robot meets written acceptance criteria in the intended environment. What makes a humanoid robot trustworthy? Trust should be calibrated to evidence. A trustworthy deployment states what is autonomous and what uses remote assistance, publishes relevant limits and failure modes, records interventions and incidents, protects data, provides human override and names the people accountable for operation. Marketing confidence is not a substitute for operating evidence. Coexistence Has to Be Earned Humanoid robots can take on dangerous, repetitive and physically demanding work. Those benefits do not cancel the need for a credible safety case, accountable operation, privacy controls and honest disclosure of human assistance. The burden should sit with the organizations introducing the robot, not with workers or households asked to trust it. The practical standard is simple: deploy only what can be bounded, supervised, explained and stopped. Expand when operating evidence supports expansion. That is slower than a demo reel and faster than waiting for a perfect universal law—and it gives human-robot coexistence a chance to deserve public trust. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # Humanoid Robots in the Home: A Sober 10-Year Adoption Scenario URL: https://blog.robozaps.com/b/in-10-years-owning-a-humanoid-will-be-as-common-as-owning-a-smartphone-today Author: Radica Maneva Published: 2025-01-09T00:00:00.000Z Updated: 2026-08-01T23:03:58.172Z Last fact-checked: 2026-07-06T00:00:00.000Z Category: opinion TL;DR: Humanoids are moving from demos into real pilots and early products, but the safest 10-year view is premium, supervised, and uneven adoption. Industrial and logistics settings should mature before normal homes do. Key takeaways: - Smartphones are a useful awareness analogy, not a good ownership forecast for humanoid robots. - Robots are expensive physical machines with safety, privacy, support, and reliability constraints. - The first credible scale should come from factories, logistics, research, and controlled service environments. - Home adoption depends on autonomy, safety standards, remote-operation transparency, maintenance, and cost. FAQ: Q: Will humanoid robots be as common as smartphones by 2035? A: Probably not. A more defensible scenario is that humanoids become familiar and commercially important by 2035, while household ownership remains premium and uneven. Q: What will home humanoids realistically do first? A: The early tasks will likely be narrow and supervised: simple fetching, tidying, reminders, monitoring, and chores that work in prepared environments. Q: What should buyers check before ordering a home humanoid? A: Check real-home demos, safety claims, remote-operator policies, privacy controls, battery life, warranty, repair model, subscription costs, and whether the robot works without human supervision. A real category, not an inevitable smartphone curve Humanoid robots are becoming real enough to take seriously. That does not mean owning one will be as common as owning a smartphone within 10 years. The comparison is useful for public awareness, but it breaks down when you move from software in a pocket to a moving machine in a home. Phones scale because they are small, relatively safe, standardized, and useful every hour. Humanoid robots must handle stairs, pets, children, clutter, privacy, battery limits, service, repairs, and physical safety. That makes adoption slower and more uneven. What service-robot data actually shows The broader service-robot market is already large, but most consumer units are not humanoids. IFR's service-robot executive summary says consumer service robots reached about 20.1 million units in 2024, with domestic-task robots making up almost all consumer sales. Robots for care at home were still a tiny niche. That is the key lesson for humanoids. Homes create value, but unstructured home tasks require higher reliability and safety than a robot vacuum following a floor plan. What exists today Unitree G1 shows developer-grade humanoid hardware priced from $13.5K, but low cost for a humanoid does not make it household-ready.1X NEO is explicitly framed as a home robot, with subscription and ownership options, basic autonomy, and scheduled expert remote assistance for difficult tasks.Many demos show impressive movement, but demos are not the same as durable unsupervised household labor. The sober 2035 scenario By 2035, humanoids may be familiar to consumers and commercially important. The most credible path is not mass ownership first. It is industrial and logistics adoption, then controlled service settings, then premium homes where buyers accept supervision, remote assistance, and maintenance complexity. For normal homes, the adoption curve depends on costs falling, safety improving, autonomy getting less brittle, and vendors proving that support does not collapse after purchase. Buyer checklist Does the robot complete the task in a normal home, not a staged demo room?What can it do without remote human help?What data leaves the home, and who can view camera or audio feeds?What safety standard or risk process does the vendor claim?What is the total monthly cost after subscription, repair, accessories, and support? Bottom line Humanoid robots are too important to dismiss and too immature to call inevitable household appliances. The best 10-year forecast is visible progress, premium early adopters, and serious commercial deployment before smartphone-style ubiquity. Compare on RoboZaps: Unitree G1. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Nvidia’s Great Leap Forward: Will It Usher in a Robotics Golden Age—Or Spark a Tech Dystopia? URL: https://blog.robozaps.com/b/nvidias-next-frontier Author: Radica Maneva Published: 2025-01-05T00:00:00.000Z Updated: 2026-08-01T23:04:01.568Z Category: opinion TL;DR: Nvidia's robotics play is infrastructure, not robots: Jetson Thor compute plus the GR00T foundation model give every humanoid maker a common AI brain, with Figure AI and Boston Dynamics already building on it. Robotics is a $78 billion market heading toward $165 billion by 2029, and Nvidia aims to tax all of it. Key takeaways: - Jetson Thor launches mid-2025 — Near-supercomputer performance designed specifically for humanoid robot autonomy and real-time decision-making - Platform strategy, not product — Nvidia provides the essential infrastructure layer that all robotics companies need, creating market lock-in - $165 billion market by 2029 — Robotics already at $78 billion and accelerating toward humanoid-driven growth - Key partnerships secured — Figure AI, Boston Dynamics, and others already building on Nvidia's ecosystem - Project GR00T provides AI foundation — Universal humanoid AI brain enabling rapid development and deployment FAQ: Q: What is Nvidia's Jetson Thor platform? A: Jetson Thor is Nvidia's computing platform specifically designed for humanoid robots, launching mid-2025. It combines high-performance GPU cores, advanced sensor fusion, and power-efficient design to enable real-time autonomous decision-making in robotic applications. Q: How is Nvidia different from companies like Tesla in robotics? A: Unlike Tesla, which builds complete robots (Optimus), Nvidia provides the underlying platform—chips and software—that any robotics company can use. This infrastructure approach means Nvidia profits regardless of which robot manufacturer wins. Q: What is Project GR00T? A: Project GR00T, launched in March 2024, is Nvidia's universal AI foundation model for humanoid robots. It combines generative AI with reinforcement learning, allowing robots to learn new tasks without extensive reprogramming. Q: Which companies are using Nvidia's robotics platform? A: Major partners include Figure AI (Figure 02 robot), Boston Dynamics, Agility Robotics, and numerous manufacturing companies. The ecosystem is rapidly expanding as Nvidia's tools become industry standard. Q: How big will the humanoid robotics market become? A: The total robotics market is projected to grow from $78 billion today to $165 billion by 2029. The humanoid segment specifically could reach $38 billion by 2030 and potentially $7 trillion by 2050. Q: Will Nvidia's robots take human jobs? A: The technology enables automation of many roles in warehouses, manufacturing, and service industries. However, new jobs typically emerge around robot maintenance, programming, and supervision. The net employment impact remains debated among economists. Q: When will Jetson Thor be available? A: Jetson Thor became generally available on August 25, 2025. Early partners were already developing humanoid robots on the platform ahead of general availability. Related: The Role of AI in Advancing Humanoid Robot Technology: A Detailed Look · The Future of Humanoid Robots: Innovation and Impact Ready to buy? Browse humanoid robots for sale on Robozaps. Inside the Silicon Behemoth's Quest to Rewrite the Rules of AI, Automation, and Our Future with Machines Nvidia is positioning itself to dominate the humanoid robotics industry through its Jetson Thor platform, Project GR00T AI foundation, and strategic partnerships with leading robot manufacturers. The company's approach—providing the essential hardware and software infrastructure rather than building robots directly—mirrors its successful GPU strategy in AI. With the robotics market projected to hit $165 billion by 2029, Nvidia aims to make its silicon the default foundation for every humanoid robot worldwide. What Is Jetson Thor and Why Does It Matter? Jetson Thor is Nvidia's purpose-built computing platform for humanoid robots, launched into general availability on August 25, 2025. It combines high-performance GPU cores, advanced sensor fusion, and power-efficient design specifically optimized for robotic applications. Unlike conventional GPUs, Jetson Thor handles real-time vision, speech, and decision-making simultaneously. This enables humanoids to move, adapt, and interact with human environments at near-human reaction speeds. High-Performance GPU Cores — Real-time vision, speech, and decision-making, all happening faster than you can blinkSensor Fusion on Steroids — Unifies camera, LiDAR, and ultrasound data with minimal latency for seamless environmental awarenessBattery-Sipping Efficiency — Near-supercomputer performance in a robot that can operate for hours without charging Jetson Thor isn't just incremental improvement—it's Nvidia's bid to become the default standard for any serious humanoid project globally. Why Is Nvidia's Platform Strategy Different From Competitors? Unlike Google, Amazon, or Tesla, Nvidia isn't building the robots itself. Instead, it's positioning itself as the irreplaceable hardware and software layer that every robotics company needs to succeed. This platform approach creates ecosystem lock-in similar to what Apple achieved with iOS. Comprehensive Software & AI Training — Isaac Sim and Project GR00T let developers model, train, and refine robots in simulation before productionCross-Industry Collaborations — Partnerships with Figure AI, Boston Dynamics, and dozens of others expand Nvidia's reachAccelerated Go-to-Market — Easier software stacks and validated hardware help manufacturers ship robots faster Building a humanoid robot without Nvidia's help means betting on an uphill battle. That's how you lock down an entire future market. How Big Is the Humanoid Robotics Market Opportunity? The robotics industry stands at $78 billion today and is projected to reach $165 billion by 2029. The humanoid segment specifically could explode to trillions by 2050 as industries from healthcare to hospitality adopt human-shaped automation. Humanoid robots—once a geeky fantasy—are now serious business. They can fit into human-designed environments without infrastructure overhauls. Nvidia is ensuring these future machines all run on its hardware. What Are the Societal Implications of Nvidia's Robotics Push? Nvidia's unstoppable progress raises significant questions about jobs, ethics, and human-robot interaction. These aren't minor speed bumps—they're the moral crossroads of the 21st century. Jobs, Jobs, Jobs — Millions of roles from warehouse workers to restaurant staff could be replaced by robots that cost less, don't need breaks, and don't unionizeEthical Challenges — When a humanoid damages property or injures someone, who's liable? Privacy and unwanted learned behaviors present ongoing concernsHuman-Robot Interaction — Real-time sensor fusion and generative AI will make robots feel more natural, but society may not be ready for human-like machines Tech giants like Nvidia are effectively writing the rulebook for how these challenges get addressed. What Has Nvidia Already Delivered in Robotics? Nvidia isn't just announcing plans—it's shipping products and forming partnerships that cement its position. Project GR00T (March 2024) — A universal AI "brain" for humanoids combining generative AI with deep reinforcement learning for adaptable robotsFigure AI Collaboration (August 2024) — Figure 02 humanoid uses Nvidia's platform for advanced dexterity and real-time problem-solving These are strategic plays to make Nvidia synonymous with advanced robotics, just like it became the default in AI chips. What's Next for Nvidia and Humanoid Robotics? With Jetson Thor launching mid-2025, Nvidia is declaring that the age of humanoid robots isn't coming—it's here. The platform approach rivals Apple's iOS ecosystem for creating lock-in and developer loyalty. This shift is about more than cool gadgets. It's about reshaping factories, hospitals, stores, and homes. How we work, live, and engage with machines could look radically different by 2030. If Nvidia succeeds, it holds a level of influence that's both exciting and concerning. Controlling the "brains" of a global robot workforce grants enormous economic, technological, and political power. Compare on RoboZaps: Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # China Leads Industrial Robot Deployment. The Humanoid Race Is Still Early. URL: https://blog.robozaps.com/b/chinas-ai-robot-revolution-the-rest-of-the-world-is-playing-catch-up Author: Radica Maneva Published: 2025-01-03T00:00:00.000Z Updated: 2026-08-01T23:04:05.052Z Last fact-checked: 2026-07-06T00:00:00.000Z Category: opinion TL;DR: China is the world's largest industrial robot market by installations and operating stock. That is proven. Humanoid robot leadership is plausible, but still early and should be separated from industrial robot deployment data. Key takeaways: - IFR reports China installed 295,045 industrial robots in 2024, 54% of global installations. - Industrial robot leadership does not automatically prove mass humanoid adoption. - China has policy support, manufacturing scale, and fast-moving humanoid companies. - The U.S. still leads private AI investment, so the race is not one-dimensional. FAQ: Q: Is China ahead in robotics? A: China leads in industrial robot installations and operating stock. In humanoids, Chinese companies are moving quickly, but large-scale commercial performance is still being proven. Q: How many industrial robots did China install in 2024? A: IFR's World Robotics 2025 executive summary reports 295,045 industrial robot installations in China in 2024, representing 54% of global installations. Q: Does this mean the West has already lost? A: No. China has major deployment and manufacturing advantages, while the U.S. has strong AI investment and model leadership. The practical lesson is to prepare and measure, not panic. China's industrial robot lead is real China is the world's largest industrial robot market. That is not hype. IFR's World Robotics 2025 executive summary reports 295,045 industrial robot installations in China in 2024, equal to 54% of global installations. It also reports more than 2 million industrial robots operating in China. That is the strongest evidence on this topic. It shows manufacturing scale, automation demand, supplier depth, and policy follow-through. It does not, by itself, prove that China has already won the humanoid robot race. Industrial robots and humanoids are not the same market Industrial robots are mature tools for factories. Humanoid robots are still trying to prove that a human-shaped machine can do useful work safely and economically in spaces built for people. China has advantages in manufacturing and hardware supply chains, but humanoid deployment is still early. Companies such as Unitree and AgiBot make China important to watch. The right conclusion is not that everyone else is doomed. China's wider industrial ecosystem may help some robotics firms move faster from prototype to production, but deployment evidence still needs to be checked company by company. Policy support matters Made in China 2025 is a broad manufacturing-upgrade strategy covering intelligent manufacturing and high-end equipment. It supports the robotics ecosystem, but it is not proof of mass humanoid deployment. Policy support does not guarantee commercial success, but it can help companies coordinate capital, suppliers, talent, and adoption. The U.S. still has strengths The race is not one-dimensional. Stanford's 2026 AI Index reports far higher U.S. private AI investment than China, while also noting China's progress in model performance. For humanoids, AI capability, data, hardware, controls, safety, and production all matter. What manufacturers should do Benchmark Chinese automation density and supplier speed, but separate industrial robots from humanoids.Track official product releases, deployment evidence, and service terms from Chinese humanoid firms.Run narrow pilots only where the task, safety process, and ROI model are measurable.Build workforce transition plans before automation pressure turns into a staffing problem. The practical lesson China's industrial robotics lead is proven. Its humanoid robotics lead is plausible, but not settled. The practical lesson is to prepare, measure, and learn from the fastest-moving manufacturers without turning every prototype into a geopolitical certainty. Browse the robots behind this analysis: every humanoid robot and robot dog RoboZaps carries, with specifications and a quote route. --- # OpenAI’s Humanoid Robot Gambit: Bold Masterstroke or Silicon Valley Overreach? URL: https://blog.robozaps.com/b/openai-humanoid-robotics-ambitions Author: Radica Maneva Published: 2024-12-30T00:00:00.000Z Updated: 2026-08-01T23:04:08.735Z Category: news TL;DR: OpenAI is hedging its humanoid bet: it has invested in Figure and 1X while reportedly exploring in-house robots, chasing a market that could reach $7 trillion by 2050. Hardware is brutally hard — Google sold Boston Dynamics, Amazon's Astro flopped — but sitting out risks watching Tesla, Figure and 1X embed rival AI in their machines. Key takeaways: - $7 trillion market opportunity — The humanoid robotics market could reach this scale by 2050, making it one of the largest technology opportunities in history - Vertical integration play — Owning both AI software and robot hardware means capturing more value and controlling the entire data feedback loop - Hardware is brutally hard — Google sold Boston Dynamics, Amazon's Astro flopped; hardware manufacturing requires entirely different capabilities than software development - Sitting out could be worse — If OpenAI doesn't build robots, competitors like Tesla, Figure, and 1X will embed AI into their hardware and capture the market - Current investments hedge both ways — OpenAI has invested in Figure and 1X robotics startups while reportedly exploring in-house development FAQ: Q: Is OpenAI really building a humanoid robot? A: According to reporting by The Information, OpenAI has explored developing its own humanoid robot. The project remains unconfirmed, and OpenAI already invests in robotics startups like Figure and 1X as an alternative approach. Q: Why would OpenAI want to build robots instead of just providing AI? A: Building complete robots means capturing more revenue per unit, owning the data feedback loop, and creating stronger competitive barriers. Software-only companies risk becoming commoditized suppliers while hardware makers capture consumer mindshare and margins. Q: What robotics companies has OpenAI invested in? A: OpenAI has invested in Figure (maker of Figure 01 and 02 humanoids) and 1X Technologies (developing the NEO humanoid robot). These investments give OpenAI exposure to robotics without the risks of building hardware directly. Q: How big is the humanoid robot market opportunity? A: The humanoid robotics market could reach $7 trillion by 2050 according to industry projections. Near-term estimates put the market at $38 billion by 2030, with applications spanning manufacturing, healthcare, hospitality, and home assistance. Q: Why is hardware so difficult for software companies? A: Hardware requires supply chain management, manufacturing partnerships, quality control, inventory management, and regulatory compliance—entirely different skills than software development. Google, Amazon, and others have struggled with robotics for these reasons. Q: Who would OpenAI compete against in humanoid robotics? A: Major competitors include Tesla (Optimus), Boston Dynamics (Atlas), Agility Robotics (Digit), Figure, 1X, and Chinese manufacturers like XPeng and Unitree. The market is rapidly consolidating around well-funded leaders. Q: When might OpenAI announce a robot product? A: No timeline has been publicly confirmed. Given the complexity of hardware development, any OpenAI robot would likely take 2-3 years minimum from decision to prototype, with commercial production taking even longer. Related: The Role of AI in Advancing Humanoid Robot Technology: A Detailed Look · The Future of Humanoid Robots: Innovation and Impact Ready to buy? Browse humanoid robots for sale on Robozaps. Why OpenAI's rumored humanoid robot project could either cement its dominance or expose its greatest vulnerabilities. OpenAI is reportedly exploring building its own humanoid robot, a move that would combine its industry-leading AI with physical hardware to capture a potential $7 trillion market by 2050. The rumored project represents a strategic crossroads: OpenAI could either become the dominant force in embodied AI by controlling both software and hardware, or it could stumble into the same hardware graveyard that claimed Google's Boston Dynamics investment and Amazon's Astro robot. What Is OpenAI's Rumored Humanoid Robot Project? According to The Information, OpenAI has explored building its own humanoid robot rather than simply licensing its AI to other manufacturers. This would represent a dramatic expansion from pure software into physical hardware. The logic is straightforward: Why just provide the AI brain when you could own the entire robot? Controlling both hardware and software means capturing more revenue, owning the data feedback loop, and building an unassailable competitive position. OpenAI already invests in robotics startups like Figure and 1X. Building its own robot would shift from being an investor to a direct competitor. Why Would OpenAI Enter Humanoid Robotics? The humanoid robotics market could hit $7 trillion by 2050. Whoever dominates humanoid robots effectively owns the future of physical AI—factories, warehouses, healthcare, and home services. Owning the hardware and the software creates a king's ransom in data, revenue streams, and market power. It's the same reason Tesla is building Optimus—vertical integration means bigger profits and total ecosystem control. What Are the Risks of OpenAI Building Robots? Hardware development has humbled even the biggest tech companies. Manufacturing robots requires supply chain mastery, production expertise, and patient capital that software companies often lack. Google's Boston Dynamics Debacle — Google acquired the robotics leader, then sold it citing lack of near-term profitability and "strategic fit"Amazon's Astro Failure — The retail giant's home robot received tepid reviews and demonstrated how hard it is to meet consumer hardware expectations OpenAI has resources, but scaling hardware isn't like training more AI models. It demands manufacturing know-how and patience that have humbled companies with far more experience. What Happens If OpenAI Doesn't Build Robots? Playing it safe might be even more dangerous than the hardware risk. If OpenAI bows out of humanoid robotics, it hands over a massive chunk of the AI+hardware future to competitors. Humanoid robots could dominate factories, elder care, and even home environments. If OpenAI sticks to pure software, its cutting-edge AI gets embedded into someone else's robot—with someone else capturing the profits and brand recognition. Tesla's Optimus, Figure's robots, and Chinese manufacturers like XPeng are all racing ahead. Second place in this market means watching others define the future of physical AI. Who Are OpenAI's Main Competitors in Humanoid Robotics? The humanoid robotics race includes established players and well-funded startups. OpenAI's entry would intensify an already competitive field. Chinese EV makers like XPeng, NIO, and BYD are quickly moving from cars to humanoids. A wave of new players are all jostling to define what robotic labor means this decade. What Strategic Options Does OpenAI Have? OpenAI faces a fundamental choice about its identity: Does it want to be the puppet master, or own the puppets as well? Be the Brain Only — Provide AI backbone to any humanoid, collecting licensing and partnership fees. Low overhead, high margin, but less total controlBe the Brain and the Body — Control both hardware and software, from motors to code. Logistical nightmare, but winning means becoming the unstoppable giant of AI-robotics OpenAI's current investments in Figure and 1X hedge between these approaches. But rumors suggest leadership may be leaning toward the bolder path. Compare on RoboZaps: Figure 02, Digit and Atlas. Full specifications and a quote route for each, alongside every humanoid robot we carry. --- # Unitree H1 vs Boston Dynamics Atlas: Comprehensive Guide URL: https://blog.robozaps.com/b/unitree-h1-vs-boston-dynamics-atlas Author: Radica Maneva Published: 2024-07-07T00:00:00.000Z Category: comparisons TL;DR: The Unitree H1 (about $90,000, quote-only) is the affordable, agile option with a depth camera and 3D LiDAR, suited to research and education; the Boston Dynamics Atlas is the premium enterprise humanoid with superior manipulation and ruggedness. The H1 wins on price and access, Atlas on raw capability. Key takeaways: - Both Unitree H1 and Boston Dynamics Atlas showcase impressive engineering and versatility, with Unitree H1 focused on affordability and agility through electric motors, while Boston Dynamics Atlas emphasizes dynamic movement and precision with its all-electric system. - Unitree H1’s advanced sensor suite, including an Intel RealSense D435i depth camera and 3D MID360 LiDAR sensor, enhances its spatial awareness and navigation capabilities, whereas Boston Dynamics Atlas leverages sophisticated algorithms and control systems for mobility and balance. - Unitree H1 is priced significantly lower at $90,000, making it more accessible for various applications, including education and healthcare, while Boston Dynamics Atlas remains a high-cost option mainly suited for well-funded enterprises and research institutions due to its advanced capabilities and intricate construction. FAQ: Q: How tall and heavy is the Unitree H1? A: The Unitree H1 is about 180 cm tall and weighs around 47 kg. Q: What kind of movements can the Boston Dynamics Atlas perform? A: The Boston Dynamics Atlas can perform dynamic and agile movements such as parkour, backflips, and navigating diverse terrains. Q: What sensors does the Unitree H1 use for perception? A: The Unitree H1 uses a 3D LiDAR sensor and an Intel RealSense D435i depth camera for its perception capabilities. These sensors enable 360° depth perception and real-time spatial data acquisition. Q: How does the cost of the Unitree H1 compare to the Boston Dynamics Atlas? A: The Unitree H1 is more affordable than the Boston Dynamics Atlas, as it is priced at $90,000 compared to the reported cost of over $1 million for the Atlas. Q: What are some real-world applications of the Unitree H1? A: The Unitree H1 is used in various real-world applications such as industrial maintenance, emergency response, healthcare, entertainment, robotics, and AI education. These applications showcase its versatility and wide range of uses. ‍ Related: Unitree H1 Review 2026: Specs, Price, Performance & Competitor Comparison · Boston Dynamics Atlas Review Ready to buy? Browse humanoid robots for sale on Robozaps. Trying to decide between the Unitree H1 and Boston Dynamics Atlas? In this article, we will discuss the “unitree h1 vs boston dynamics atlas” comparison in terms of design, mobility, sensors, and task performance. Learn which robot excels in different scenarios and make an informed decision. ‍ ‍ Overview of Unitree H1 and Boston Dynamics Atlas Unitree Robotics and Boston Dynamics have both made significant contributions to the field of robotics, each with a distinct approach. Unitree Robotics, known for its budget-friendly and versatile robots, introduced the H1, the world’s first mass-produced full-size electric-driven humanoid robot. Standing at approximately 180 cm and weighing about 47 kg, the Unitree H1 is designed for agility and versatility, making it suitable for a wide range of tasks and environments. On the other hand, Boston Dynamics has been a pioneer in showcasing the potential of robotics through its Atlas model. The Boston Dynamics Atlas is renowned for its dynamic capabilities, including obstacle navigation and dexterous manipulation. These demonstrations have not only captivated audiences worldwide but have also highlighted the robot’s potential in various challenging environments. ‍ Design and Build Quality Functionality and versatility of humanoid robots hinge significantly on their design and build quality. Both the Unitree H1 and Boston Dynamics Atlas showcase impressive engineering and material selection, which we will explore in detail through their structural design, mobility, and the choice between hydraulic systems and fully electric motors. Structural Design The Unitree H1 features: Advanced carbon fiber and aluminum alloy materials for a lightweight yet robust structureEnhanced durability, agility, and speedSleek, modern design with streamlined limbs and a white finishFuturistic appearanceEasy component swaps due to its modular design In contrast, the Boston Dynamics Atlas features: A frame made from lightweight, high-strength materials such as aluminum and titaniumConstruction designed to withstand significant physical stress and impactRobust and resilient design for challenging environmentsAdvanced structural designs that enhance durability and agilityAbility to perform dynamic tasks with precision Mobility and Balance Showcasing exceptional mobility and balance, the Unitree H1 can perform backflips, achieve speeds of 3.3 meters per second, and handle dynamic activities like speed-walking, stair climbing, standing jumps, and dancing. Holding the world record for the fastest full-size humanoid robot, the H1 can walk at speeds up to 3.3 m/s (~11.9 km/h) and maintain dynamic balance even on uneven surfaces. Boston Dynamics Atlas, known for its dynamic and agile movement capabilities, showcases impressive feats such as: ParkourBackflipsNavigating diverse terrainsPerforming dynamic motions like jumping and running An advanced control system that ensures balance and accuracy facilitates its ability to carry out these actions. The Atlas leverages its whole body to move with grace and dexterity, making it a highly versatile platform for various tasks. Both robots utilize an array of IMUs to maintain balance and orientation during complex movements. The Atlas, in particular, demonstrates highly dynamic and agile locomotion abilities, making it an impressive contender in the field of humanoid robotics. Hydraulic Systems vs Fully Electric Motors The choice between hydraulic systems and fully electric motors has a significant impact on a robot’s performance and capabilities. Boston Dynamics Atlas employs a hydraulic system, which enables powerful and precise movements. This system is particularly advantageous for tasks that require significant force and control, such as lifting heavy objects or performing complex maneuvers. In contrast, the Unitree H1 uses fully electric motors, specifically the M107 joint motors, which provide high agility and speed. These fully electric motors are designed to optimize the robot’s performance in dynamic activities, offering more efficient and responsive movement compared to traditional hydraulic systems. ‍ Sensors and Perception Humanoid robots rely on sensors to perceive and interact with their environment. Both Unitree H1 and Boston Dynamics Atlas are equipped with advanced sensor technologies that enhance their navigation and perception capabilities. Depth Cameras The Unitree H1 is equipped with an Intel RealSense D435i depth camera, which significantly enhances its perception capabilities. This camera has a depth field of view of 86° x 57° and can capture depth images at a resolution of up to 1280 x 720. With a frame rate of up to 90 frames per second, the Intel RealSense D435i provides real-time, high-precision spatial data, allowing the H1 to navigate and interact with its surroundings effectively. This depth camera technology is crucial for tasks that require detailed environmental mapping and object recognition, making the Unitree H1 a formidable player in scenarios that demand high levels of spatial awareness and precision. Other Sensors In addition to its depth camera, the Unitree H1 utilizes a 3D MID360 LiDAR sensor to enhance its spatial awareness. This sensor provides 360° depth perception and real-time acquisition of spatial data, which is essential for navigating complex environments. The H1 also integrates a Bosch IMU with 6 degrees of freedom for motion sensing, further improving its ability to maintain balance and orientation during dynamic tasks. These advanced sensors collectively enhance the H1’s ability to analyze and interact with its surroundings, making it a highly capable robot in terms of perception and navigation. The combination of LiDAR and depth-sensing cameras allows the H1 to perform complex maneuvers with precision and agility. ‍ Computing Power and Control Systems A humanoid robot’s performance and functionality are deeply reliant on its computing power and control systems. The Unitree H1 is equipped with an Intel Core i5 for platform functions and an Intel Core i7 for user development, with optional configurations including Intel Core i7 or Jetson Orin NX. This robust computing setup ensures that the H1 can handle complex tasks and applications with ease. Boston Dynamics leverages decades of research to develop the Atlas robot’s computing power, focusing on three key areas: Mobility: The Atlas uses an advanced control system and state-of-the-art hardware that enables it to achieve power, balance, and agility, making it a highly versatile platform for various tasks.Perception: The Atlas is equipped with advanced sensors and cameras that allow it to perceive its environment and make informed decisions.Intelligent: The Atlas has sophisticated algorithms and machine learning capabilities that allow it to learn and adapt to various situations. This combination of mobility, perception, and intelligence makes the Atlas capable of performing dynamic and complex maneuvers with precision. ‍ Performance in Various Tasks Capabilities and versatility of humanoid robots are reflected in their performance across various tasks. Both the Unitree H1 and Boston Dynamics Atlas excel in different applications, showcasing their potential to transform industries and improve efficiency. Specific Tasks In industrial settings, the Unitree H1 conducts routine maintenance and inspection tasks, handling complex maneuvers in confined spaces. Its agility and versatility make it ideal for navigating through tight environments and performing detailed inspections. Additionally, the H1 excels in emergency response scenarios, where it can navigate through debris and hazardous environments to perform search and rescue operations. The Boston Dynamics Atlas, on the other hand, is known for its ability to perform tasks such as lifting and carrying objects, demonstrating its utility in manual labor scenarios. Its dynamic movement capabilities and advanced control system make it suitable for a wide range of applications, from industrial tasks to household chores. Real-World Applications In real-world applications, the Unitree H1 is used as an educational tool to teach robotics and artificial intelligence. Its advanced capabilities and affordable price make it an attractive option for academic institutions and research facilities. The H1 is also employed in healthcare applications, where it supports patient care and rehabilitation, demonstrating its versatility and potential impact on society. Boston Dynamics Atlas, the new atlas in the world of robotics, is designed for real-world applications with a control system that allows it to move with athleticism and dexterity. Its ability to navigate complex environments and perform dynamic tasks makes it a valuable asset in various industries, from manufacturing to entertainment. ‍ Innovation and Development Advancements in humanoid robotics are rooted in innovation and development. Both Unitree Robotics and Boston Dynamics have demonstrated significant recent innovations, pushing the boundaries of what is possible with robotics technology. Recent Innovations Unitree Robotics has introduced the G1 humanoid at ICRA 2024. This model stands out for its: Smaller sizeLower cost compared to the H1Fast walking speed of 2,000 mm/sAbility to be equipped with 3D LiDAR and Intel RealSense D435 cameras These advancements reflect Unitree’s commitment to making advanced robotics technology more accessible and versatile. Additionally, Unitree Robotics is rapidly iterating AI-driven programs for the H1, introducing several major version upgrades within six months. These continuous improvements in AI algorithms for motor control and agility highlight Unitree’s dedication to pushing the envelope in robotics development. Future Developments Looking ahead, Unitree plans to continue supporting the H1 robot while developing new features for the G1 model. This includes integrating advanced AI features and improved agility to enhance user interaction and adaptability. The focus on continuous development ensures that Unitree remains at the forefront of robotics innovation. Future developments at Boston Dynamics are expected to include more sophisticated mobile manipulation capabilities and enhanced autonomous functions. These advancements will enable the Atlas to perform even more complex tasks with greater efficiency, further solidifying its position as a leader in the field of humanoid robotics. ‍ Community and User Engagement The success and advancement of robotics companies hinge heavily on community and user engagement. Both Unitree and Boston Dynamics understand the importance of engaging with the public and fostering a sense of excitement and anticipation within the robotics community. Social Media Presence Boston Dynamics utilizes platforms like YouTube to demonstrate the advanced capabilities of Atlas, garnering significant media attention and public interest. These high-profile demonstrations not only showcase the robot’s agility and balance but also attract both fascination and skepticism from the public. By frequently releasing content featuring Atlas performing complex maneuvers, Boston Dynamics effectively boosts its online presence and public engagement. Unitree Robotics promotes its products through social media by: Highlighting the affordability and accessibility of their robotsEngaging with a wide audience on platforms like Twitter, YouTube, Facebook, and BilibiliBuilding a strong community of robot enthusiasts and professionals This strategic use of social media helps Unitree, one of the innovative Chinese companies, build a strong presence and reach within the Chinese market. Developer Support A comprehensive software development kit (SDK) that supports Boston Dynamics’ Atlas robot enables programmers to control and program the robot’s advanced functionalities. This includes a range of developer tools and resources, such as SDKs and APIs, to support the integration and customization of their robots for various applications. This level of support ensures that developers can fully leverage the capabilities of Atlas in their projects. Similarly, Unitree Robotics provides multiple SDK packages, such as unitree_legged_sdk and unitree_sdk2, to facilitate the development of their robots in real environments. Unitree also offers extensive documentation and software support, making it easier for developers to integrate their robots into different projects. Additionally, a Python interface is available for the unitree_sdk2, allowing developers to use Python for robot development, further enhancing the versatility of Unitree’s platform. ‍ Cost and Accessibility The cost and accessibility of humanoid robots have a big impact on their adoption in various industries. The Unitree H1 is positioned as a more affordable humanoid robot compared to many of its counterparts. Priced at $90,000, it offers advanced capabilities at a fraction of the cost of other high-end robots, making it accessible to a broader range of consumers and industries. In contrast, Boston Dynamics’ Atlas is known to be one of the most expensive humanoid robots, with costs reportedly over $1 million per unit. This high price reflects the advanced control systems and state-of-the-art hardware used in its construction. While the Atlas provides exceptional performance and capabilities, its high cost may limit its accessibility to only well-funded enterprises and research institutions. ‍ Environmental Impact When developing new technologies, the environmental impact of advanced robotics warrants significant consideration. Some of the environmental impacts of robotic automation include: Faster resource depletionIncreased waste from planned obsolescenceContribution to environmental hazards, particularly in countries with high levels of robotic implementation like the US and Japan. Green robotics initiatives are essential to mitigate some of the negative environmental impacts of AI and robotics. By focusing on sustainable practices and reducing resource consumption, companies can develop technologies that benefit both humans and the environment. This approach is crucial for ensuring that the advancements in robotics do not come at the expense of our planet’s health. ‍ Summary In summary, both the Unitree H1 and Boston Dynamics Atlas represent significant advancements in the field of humanoid robotics. The Unitree H1 stands out for its affordability, agility, and versatility, making it suitable for a wide range of applications, from industrial maintenance to healthcare. On the other hand, the Boston Dynamics Atlas showcases exceptional dynamic capabilities and a robust design, making it ideal for complex and demanding tasks. While the Unitree H1 offers a cost-effective solution with advanced features, the Atlas provides unparalleled performance at a higher cost. Each robot has its own strengths and is suited to different applications and environments, reflecting the diverse needs of the robotics industry. As we look to the future, continued innovation and development in humanoid robotics will undoubtedly lead to even more remarkable advancements. The potential for these robots to transform industries and improve our daily lives is immense, and it will be exciting to see how they continue to evolve and integrate into society. ‍ --- # Figure 03 vs Tesla Optimus: Who Is Actually Ahead in 2026? URL: https://blog.robozaps.com/b/figure-01-vs-tesla-optimus-gen-2 Author: Radica Maneva Published: 2024-06-23T00:00:00.000Z Updated: 2026-08-01T23:05:36.478Z Last fact-checked: 2026-07-16T00:00:00.000Z Category: comparisons TL;DR: Figure 03 leads Tesla Optimus Gen 3 on evidence available as of 16 July 2026: its hardware, production claims, autonomy tests and BMW demonstration are public. Tesla has a Fremont line and a much larger scale target, but no full Gen 3 specification or disclosed output. Key takeaways: - RoboZaps lists Figure 03 as verified and piloting; Optimus Gen 3 remains announced and partially verified. - Figure leads on disclosed hardware, a company-reported 350-plus fleet, autonomy evidence and a BMW demonstration. - Tesla has shown its Fremont production line and targets initial output, but it has not disclosed current Gen 3 output or a complete specification. - Neither robot has a public MSRP, normal ordering route or independent head-to-head benchmark. FAQ: Q: Is Figure AI ahead of Tesla Optimus? A: Yes, as of 16 July 2026. Figure has disclosed a current specification, a fleet count, a BMW demonstration, a whole-body autonomy test and a multi-robot endurance stream. Tesla has disclosed Gen 3 production-line progress but not a complete current specification or output figure. Q: Is Figure 03 better than Tesla Optimus Gen 3? A: Figure 03 is more verifiable today, but a full technical better verdict is premature. Tesla has not published enough Gen 3 information for a fair comparison of hardware, performance, reliability or cost. Q: Has Tesla revealed Optimus Gen 3 specifications? A: Tesla has described Gen 3 as its first Optimus design intended for mass production, but it had not published a complete official Gen 3 specification sheet as of 16 July 2026. Specifications from earlier Optimus versions should not be relabelled as Gen 3 data. Q: Can you buy Figure 03 or Tesla Optimus? A: No normal public ordering route is open for either robot. The companies may work with selected partners, but neither offers a standard public purchase with a published price and delivery date. Q: How much do Figure 03 and Tesla Optimus cost? A: Neither robot has an official public MSRP. Price targets and third-party estimates should not be presented as sale prices. Figure AI is ahead of Tesla Optimus on the evidence available today. Figure 03 is a revealed, specified robot; Figure reports a fleet of more than 350 third-generation units, has streamed a rotating fleet sorting packages for more than 24 hours, and has demonstrated the platform at BMW. Tesla Optimus Gen 3 is the more ambitious manufacturing bet, but Tesla has not published a complete Gen 3 specification sheet or disclosed current production output. That makes this an uneven comparison. Figure has a current machine we can assess. Tesla has an Optimus programme, a Fremont production line being commissioned, and a very large eventual capacity target. Treating earlier Optimus specifications or executive price targets as confirmed Gen 3 facts would make the table look more complete, but less accurate. This comparison uses the live RoboZaps /robots database and checks material claims against manufacturer and regulatory sources. Official specifications, manufacturer demonstrations, company-reported production, and future targets are labelled separately. Figure 03 vs Tesla Optimus at a glance Figure wins the present-tense comparison. Tesla's case rests on what it plans to build at scale, not on a Gen 3 product that outsiders can fully evaluate today. In RoboZaps data, verified means every audited public specification in the record passed its source check. Partially verified means at least one checked field could not be fully verified or the necessary public data was unavailable. Those labels apply to database fields; they do not independently validate a manufacturer's demonstration or success-rate claim. The verdict: Figure leads today; Tesla is the scale bet Figure 03 leads in the categories that can currently be checked: disclosed hardware, recent autonomy evidence, reported production, and work demonstrated at a partner site. Tesla's manufacturing plan could reverse that lead, but it has not done so yet. Figure has not proved that its robots can perform varied work reliably for months without frequent human help. Its strongest autonomy and endurance evidence remains company-controlled. Its BMW update describes a Figure 03 demonstration, not a large commercial fleet running an entire production process. There is no sensible “which one should you buy?” verdict either. As of 16 July 2026, neither company publishes an MSRP or offers a normal public ordering route for the robot compared here. The evergreen Tesla Optimus review maintains that programme's general capability and availability record. What the RoboZaps /robots database can verify RoboZaps lists Figure 03 as a verified robot in the piloting stage. Tesla Optimus Gen 3 is listed as announced and partially verified. The difference matters: a programme can be real while its current specifications remain too incomplete to verify. Figure 03 has a disclosed specification The Figure figures in RoboZaps match the official Figure 03 product specification. They are headline specifications, not measures of uptime, task success, safety, or how much work the robot completes per charge in a particular environment. Why the Tesla column says “not disclosed” Tesla has published specifications for earlier Optimus designs and shown multiple development robots. Those numbers are useful when tracing the programme, but they are not automatically Gen 3 specifications. Tesla's Q4 2025 investor update described Gen 3 as its first design intended for mass production and said a reveal was planned for the first quarter of 2026. As of this article's fact-check date, Tesla has not published a complete official Gen 3 specification sheet. The RoboZaps Gen 3 record therefore leaves height, weight, payload, speed and runtime blank rather than copying values from a previous version. That is not a judgment that Gen 3 lacks those capabilities. It is a judgment that the information is not public enough to compare responsibly. Hardware and design: Figure has the product we can inspect Figure introduced Figure 03 in October 2025 as a ground-up redesign for its Helix AI system, domestic use, and higher-volume manufacturing. The disclosed changes go beyond a new outer shell. The robot adds palm cameras to help maintain visual coverage when its main cameras are blocked, tactile sensors in the fingertips and palms, an improved audio system, softer external materials, wireless charging, and a redesigned battery system. Figure also says the robot is nine percent lighter than Figure 02 and uses production processes intended to make higher-volume manufacturing more practical. These features give Figure a clear win in disclosed current hardware. They do not prove that Figure 03 is ready to work unsupervised in a normal home. Soft materials and better sensors can support safer, more capable interaction, but safety certification, failure behaviour, service intervals and real-world reliability are separate questions. Tesla has said Gen 3 is designed for mass production, which may ultimately be the more consequential hardware achievement. Yet Tesla has not disclosed enough about the current design to compare its hands, sensing, battery, payload or maintainability with Figure 03. In this category, the honest score is Figure first and Tesla not yet scorable—not Figure first and Tesla last. Autonomy and endurance: Figure has two different kinds of evidence Figure's public evidence now covers two separate questions: can the robot complete a complex whole-body task, and can a fleet repeat a narrow task for a long shift? In January 2026, Figure demonstrated a four-minute dishwasher task involving 61 loco-manipulation actions. The robot walked to the dishwasher, handled racks and dishes, and put items away without human intervention during the published run. Figure says Helix 02 draws on more than 1,000 hours of human motion data and training in more than 200,000 simulated environments. Its palm cameras and tactile sensors are intended to keep the system informed when an object is close to the body or hidden from its head cameras. The dishwasher run is the stronger evidence of task complexity. It shows one current Figure platform completing a continuous, multi-stage task that requires locomotion and manipulation to work together. Figure added endurance evidence in May. In the original eight-hour stream, three Figure 03 robots took turns sorting small packages on a looped conveyor while the other units charged. The task was to find each barcode and place the package face down. Figure described the run as autonomous and unsupervised, then continued it beyond 24 hours. Business Insider reported that the fleet had processed more than 30,000 packages by the 24-hour mark. That is fleet-endurance evidence, not proof that one robot worked continuously for 24 hours. It was also a repetitive task at Figure's headquarters. The stream showed automatic resets and robot handoffs, while an independent roboticist noted wrong barcode orientations and a package being knocked off the belt. In a later 10-hour contest, a Figure intern beat one robot by 192 packages: 12,924 versus 12,732, or 2.79 seconds per package versus 2.83. Together, the two tests say more than either alone. The dishwasher run shows more complex loco-manipulation; the package stream shows sustained repetition close to the measured human pace on one tightly defined task. It is still manufacturer-controlled evidence. A successful video does not tell us the task success rate over 100 attempts, the frequency of remote intervention outside the selected run, performance in unfamiliar kitchens, or the safety envelope around children and pets. No independent benchmark currently lets us compare Figure 03 and Optimus Gen 3 on the same tasks under the same conditions. Tesla has not released a comparable public Gen 3 benchmark. Figure therefore leads the public autonomy evidence, while the broader claim of general household autonomy remains unproved for both. Figure AI vs Tesla Optimus: how the control strategies differ Figure publishes much more about the current robot's control stack. It describes Helix 02 as a Vision-Language-Action system with three layers: System 2 interprets scenes, language and task sequences; System 1 turns sensor input into full-body joint targets at 200 Hz; and System 0 handles balance, contact and coordination at 1 kHz. Inputs include head and palm cameras, fingertip touch sensors and whole-body joint state. Tesla's AI and Robotics description is broader. Tesla says its approach combines AI for vision and planning with efficient inference hardware, and that Optimus needs software for balance, navigation, perception and physical interaction. Tesla groups vehicles and robots under the same autonomy programme, but it has not published a Gen 3 control-stack description or benchmark detailed enough to match Figure's current disclosure. Figure therefore leads on technical transparency and demonstrated current behaviour. That is not proof that Helix is intrinsically more capable than Tesla's unrevealed system; it is the limit of what outsiders can support with evidence. Production and deployment: evidence today versus ambition tomorrow Production claims are easy to blur. A fast cycle time is not the same as sustained factory output, and producing robots is not the same as deploying them successfully. Figure's current production evidence In April 2026, Figure said BotQ had delivered more than 350 third-generation robots into a fleet used for internal research, data collection, housework development and commercial use-case development. It also said its demonstrated production cycle had improved from one robot per day to one robot per hour, with more than 80 functional tests performed on each unit. A demonstrated one-hour cycle suggests that Figure has made serious design-for-manufacture progress. It does not prove that the factory continuously ships one finished robot every hour, 24 hours a day. The disclosed unit count is the better measure of achieved volume. Figure also has evidence outside its own facilities. On 30 June 2026, the company reported that Figure 03 had demonstrated a logistics sequencing workflow at BMW's Spartanburg plant. The robot used Helix 02 to work through a sequencing task at the partner site. This was a demonstration, not proof of a scaled Figure 03 production deployment. It does, however, extend a real industrial relationship. Figure says the previous Figure 02 programme contributed to the production of more than 30,000 BMW vehicles before those robots were retired. That is useful evidence about the programme's factory lineage, but it would be wrong to credit those 30,000 vehicles to Figure 03. Tesla's current production evidence Tesla's Q4 2025 investor update described Gen 3 as the first Optimus design intended for mass production. It said production was planned before the end of 2026 and discussed eventual installed capacity of one million robots per year. In its Q1 2026 update, Tesla said its first Optimus production lines were being installed. On the official Q1 earnings webcast, Musk targeted initial production for late July or August and warned that the early rate would be slow and difficult to predict. There is newer physical evidence of the conversion. On 1 July, Musk posted from the Fremont Optimus production line with the production team. He then rejected speculation that Tesla was secretly ahead of schedule, saying the ramp would be extremely slow because the product and manufacturing system are new. The line exists; disclosed Gen 3 output still does not. So the production verdict has two parts: Figure leads current disclosed output and partner-site evidence.Tesla leads on stated eventual manufacturing scale. There is no contradiction: one describes evidence already reported, while the other describes the size of a company target. Price and availability: neither robot is for sale Neither Figure nor Tesla publishes an official MSRP for the robot in this comparison. Neither offers a public checkout, a standard delivery date, or a normal commercial ordering process. Executive targets and media estimates are not retail prices. The RoboZaps humanoid robot cost guide explains why a buyer comparison needs to include software, support, maintenance, integration, safety work and expected utilisation—not just an aspirational hardware number. For now, the correct price row for both robots is not disclosed. The correct buyer advice is to evaluate deployable alternatives for a defined workflow rather than wait on an unpriced product with no public delivery commitment. Category scorecard This scorecard measures disclosed evidence. It does not score unseen internal prototypes, assume that every company demonstration generalises to daily work, or predict which company will sell more humanoids five years from now. What would change the verdict? Tesla could close the evidence gap quickly. Four developments would justify a substantial reassessment: A full Gen 3 reveal. Tesla publishes the current design, official specifications and the features that distinguish it from previous Optimus versions.Measured production or deployment. Tesla reports actual output, deliveries, or named partner-site work—not only installed capacity and timing targets.Real commercial terms. Either company opens verified orders and publishes pricing, support and delivery details.Comparable independent testing. A third party measures task success, intervention rate, uptime and safety on the same jobs for both robots. Figure's lead could also weaken if production stalls, partner demonstrations do not become repeatable work, or the autonomy shown in controlled runs fails to transfer to less structured environments. Until one of those changes occurs, Figure remains ahead on disclosed evidence. Tesla remains the scale threat. Compare on RoboZaps: Figure 03, Tesla Optimus Gen 3, Tesla Optimus and Figure 02. Full specifications and a quote route for each, alongside every humanoid robot we carry.