53 Humanoid Robot Companies Compared: The 2026 Directory

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.