Apptronik Apollo Review 2026: Price, Apollo 2 & Real-World Deployments

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.