Humanoid robots: which company is ahead?
Whether you're building or investing in the humanoid robotics market, our report covers everything you need to know: the players, the money, and what's coming next.
Summary
Unitree is now narrowly ahead in the humanoid robot race. It combines the cleanest shipment record, a profitable business checked through a public listing process and a market value of about $31 billion.
UBTECH ranks second on audited revenue and deliveries. Agility Robotics ranks third. It still has the best evidence of robots doing repeated paid work inside customer facilities, but its own merger filings show only $1.8 million of net sales in 2025 and $174,000 in the first half of 2026. Real paid work with just $1.8 million in sales is mixed evidence; we sort real customers from promises in our humanoid robots report.
There is still no uncontested leader across the whole market. Unitree leads clean humanoid volume and profitable hardware, AgiBot leads broad embodied-robot shipments in 2026, UBTECH leads audited humanoid revenue, Agility leads proven industrial workflows and Figure AI leads general-purpose autonomy.
China's manufacturing lead has widened. Smart Analytics Global counted about 19,100 humanoid shipments worldwide in the first half of 2026, with Chinese vendors holding more than 97% of them.
Shipment rankings depend on definitions. AgiBot shipped about 8,400 robots in the first half of 2026 across a broad range that includes compact and wheeled machines. Unitree shipped about 5,900, mostly pure humanoids.
Unitree went public on Shanghai's STAR Market in August 2026, raising about $900 million. Its 2025 revenue reached about RMB1.7 billion, with a reported net profit of RMB278 million. Most of its humanoid demand still comes from research, education and display use rather than factory work.
UBTECH has turned humanoids into real revenue faster than anyone else. Full-size humanoids brought in RMB590 million in the first half of 2026 alone, close to its full 2025 total, although the company still lost RMB339 million over the period.
Figure AI may still hold the technology most likely to change the ranking. Figure 03 now works in logistics at BMW's Spartanburg plant, but Figure has disclosed no revenue, contract values or paid fleet size behind its $39 billion valuation.
Capital is moving to public markets. Agility plans to list on Nasdaq through a $2.5 billion SPAC merger with a shareholder vote set for November 2026, and AgiBot is preparing a Hong Kong listing.
Home humanoids remain much earlier than industrial systems. 1X has the clearest consumer offer and a factory in Hayward, but it has not publicly confirmed meaningful NEO deliveries to homes.
Unitree's lead is narrow. Agility could climb if its Digit V5 orders turn into a large paying fleet, and Figure could jump ahead once its production ramp becomes disclosed customer revenue.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us
In our report covering the humanoid robotics market, we turn what is vague and complex into simple, useful visuals. Market maps like this one show you at a glance where every startup stands.
Which humanoid robot companies should we actually compare?
The credible humanoid robot race still comes down to eight companies with enough funding, working hardware, production or customer evidence for a serious comparison.
We include Agility Robotics, Unitree, UBTECH, AgiBot, Figure AI, Apptronik, 1X and Sanctuary AI. Each is building a human-shaped or closely related general-purpose robot meant to move between tasks rather than repeat one fixed operation.
This is an evidence-led shortlist, not a directory. NEURA Robotics, Fourier, EngineAI, RobotEra and several younger Chinese manufacturers have credible machines, but they disclose too little about shipments, paid deployments or revenue to rank with confidence. Smart Analytics Global's first-half 2026 data shows how fast these smaller Chinese players are growing: vendors outside AgiBot and Unitree took roughly a quarter of global shipments.
Tesla Optimus, Boston Dynamics Atlas and XPeng IRON stay outside the formal ranking. They are programs inside much larger companies, so their funding and economics cannot be separated from the parent. Tesla said in April 2026 that early Optimus V3 production at Fremont could start that summer, with a broader ramp in 2027, but it has not published sustained external customer work.
Funding figures for private companies remain approximate. Databases treat grants, debt, strategic investments and secondary sales differently. Agility's merger filings, for example, cite more than $390 million of equity raised, well below the roughly $641 million that funding databases report including other financing. UBTECH's figure includes money raised after its listing, and Unitree's now includes its 2026 IPO.
| Company | Main humanoid strategy | Approximate cumulative capital raised | Why it belongs |
|---|---|---|---|
| Figure AI | General-purpose humanoids for factories, logistics and homes | About $1.85 billion | Largest private financing and a leading full-body autonomy program |
| UBTECH | Full-size Walker humanoids for industrial work | About $1.34 billion, including post-IPO financing | Largest audited humanoid revenue among the companies assessed |
| Unitree | Affordable humanoids for developers, research and industry | About $900 million from its August 2026 IPO, plus earlier private rounds | Largest clean humanoid shipment record, profitable and now listed in Shanghai |
| Apptronik | Apollo humanoids for manufacturing and logistics | Nearly $1 billion | Deep industrial partnerships and Google DeepMind backing |
| Agility Robotics | Digit humanoids for logistics and manufacturing | About $641 million reported by databases; more than $390 million of equity per its filings | Strongest evidence of repeated paid industrial operation; Nasdaq listing pending |
| Sanctuary AI | Dexterous Physical AI for humanoid and industrial robots | More than $140 million | Strong manipulation technology and industrial validation |
| 1X | Soft humanoids designed primarily for homes | About $126 million | Clearest consumer proposition and published purchase pricing |
| AgiBot | Full-size, compact and wheeled embodied robots | Not fully disclosed; more than a dozen rounds, with a new round of RMB1 billion to RMB1.5 billion reported as close | Largest reported embodied-robot shipments in the first half of 2026 |
Is there a clear humanoid robot leader today?
Unitree is the narrow overall leader today, while Agility, UBTECH, AgiBot and Figure each control a different part of the race.
Unitree's lead rests on evidence that is hard to argue with. It shipped more than 5,500 pure humanoids in 2025, about 5,900 more in the first half of 2026 alone, and went through a full STAR Market review before listing in August 2026. Its prospectus shows a profitable business at a scale no Western rival has reached.
Agility still has the deepest operating record inside customer facilities: more than 65,000 hours across nine sites. Its merger filings, made public in 2026, also show what that record is worth commercially so far: $1.8 million of net sales in 2025, about $1.1 million of it from related parties. The work is real. The business is still tiny.
UBTECH has converted full-size humanoids into revenue faster than its peers. AgiBot now ships more embodied robots than anyone. Figure appears strongest in integrated learned autonomy, especially where one robot must walk, see and handle parts within the same workflow.
China's manufacturing lead keeps growing. Omdia estimated roughly 13,000 global humanoid shipments in 2025, and Smart Analytics Global counted about 19,100 in the first half of 2026 alone, up 272% year on year. IDC counted closer to 25,000.
We place Unitree slightly ahead because it combines volume, profit and public-market scrutiny. UBTECH follows on audited revenue and deliveries, then Agility, whose paid work remains small in money terms.
Our research about humanoid robots gets updated several times a month, so you always get the latest view of where things stand.
Has AgiBot overtaken Unitree in humanoid robot shipments?
Yes, on 2026 volume under a broad definition. Unitree still has the strongest claim to the pure humanoid shipment lead.
Smart Analytics Global estimates that AgiBot shipped about 8,400 robots in the first half of 2026, a 44% share of the global market and a 562% jump year on year. Unitree shipped about 5,900, a 31% share. IDC puts AgiBot slightly higher, at more than 8,600. AgiBot's range includes full-size bipeds, compact humanoids and wheeled embodied robots, and the firm itself attributes AgiBot's lead to that broader product mix. Shipment counts shift depending on who is measuring, so we follow who is ahead and who is falling behind in our report on humanoid robots.
The cleaner number still belongs to Unitree. It says it sold and delivered more than 5,500 complete humanoids to end customers in 2025, excluding wheeled dual-arm systems, and its prospectus shows 5,632 humanoids sold from 2023 to 2025. AgiBot says it had rolled out about 15,000 embodied robots in total by June 2026, a company figure that mixes product types.
UBTECH sold 921 full-size humanoids in the first half of 2026, close to its 1,079 for all of 2025. It also sold more than 15,000 smaller humanoid units in the period, which are not comparable with industrial machines. Figure had delivered more than 350 Figure 03 units from its BotQ facility by late April 2026, mainly into its own fleet.
Agility does not disclose a cumulative unit count. It reports working hours, customer sites and completed tasks instead.
| Company | Most recent disclosed volume | What the figure covers | Our interpretation |
|---|---|---|---|
| AgiBot | About 8,400 (H1 2026, estimate) | Full-size, compact and wheeled embodied robots, per Smart Analytics Global | Largest 2026 volume across a broader portfolio |
| Unitree | About 5,900 (H1 2026, estimate); more than 5,500 in 2025 | Mostly pure humanoids; 2025 figure excludes wheeled systems | Strongest narrowly defined shipment record |
| UBTECH | 921 (H1 2026); 1,079 in 2025 | Full-size humanoid products sold | Smaller volume with audited revenue |
| Figure AI | More than 350 | Figure 03 units delivered from BotQ into its fleet by late April 2026 | Manufacturing progress rather than confirmed external sales |
| Agility Robotics | Undisclosed | Company reports utilization and customer sites instead | Commercial depth cannot be inferred from unit count |
Does Agility still have the strongest commercial traction?
Agility still has the strongest record of repeated customer work, but not the strongest commercial traction in money terms. That title now belongs to UBTECH, with Unitree close behind.
Agility's customer list is still convincing. GXO tested Digit and then signed a multi-year robots-as-a-service agreement, and Digit had moved more than 100,000 totes in a live facility by November 2025. Toyota Motor Manufacturing Canada turned a pilot into a robots-as-a-service agreement, Mercado Libre agreed to bring Digit into a Texas operation, and Schaeffler is both investor and customer.
The filings put a number on all of this. Agility reported $1.8 million of net sales for 2025 and $174,000 in the first half of 2026, when its operating loss reached $363 million, mostly stock compensation. Its headline backlog, more than $300 million of multi-year Digit V5 orders, comes from a single customer contract for 1,000 robots over three years, tied to milestones.
UBTECH leads on recognized humanoid revenue. Full-size humanoid products and services brought in RMB820.6 million in 2025, according to its audited results, and RMB590 million in the first half of 2026. Humanoids made up 46.5% of company revenue in that half.
Figure's BMW relationship remains the strongest quantified automotive case. Figure 02 handled more than 90,000 parts over 1,250 operating hours and supported production of over 30,000 vehicles. In June 2026, BMW said Figure 03 was moving into logistics sequencing at Plant Spartanburg, but neither company disclosed robot numbers or contract terms.
Apptronik has an exceptional group of partners, including Mercedes-Benz, GXO, Jabil and Google DeepMind. Apollo remains mainly a customer-trial and data-collection platform, so the quality of the names still exceeds the disclosed operating scale.
Unitree has far more paying buyers, but exchange materials showed that 73.6% of its humanoid revenue in the first nine months of 2025 came from research and education, and only 9.01% from industrial uses. AgiBot's chairman says 2025 revenue passed RMB1 billion, but its customer mix and recurring usage remain opaque.
Agility leads repeated customer operation, UBTECH leads humanoid revenue, and Unitree leads paying hardware buyers. Agility's lead is about proof of work, not about sales.
In our market report about humanoid robots, we look at what the winners do differently, so you can learn from the best instead of guessing.
Which humanoid robot company is growing fastest now?
UBTECH has posted the fastest measurable commercial growth, AgiBot the fastest shipment growth, and Figure the sharpest disclosed manufacturing ramp.
UBTECH's full-size humanoid revenue rose 1,445% year on year in the first half of 2026, to RMB590 million. That half alone came close to its full 2025 humanoid revenue of RMB820.6 million, which had itself grown more than twentyfold from 2024. Total company revenue doubled to RMB1.27 billion, and the net loss narrowed by 23%.
AgiBot's shipments grew 562% year on year in the first half of 2026, according to Smart Analytics Global. Its chairman has set a goal of several-fold revenue growth in 2026, a target rather than a result.
Unitree is growing more slowly from a larger, profitable base. Its 2025 revenue rose more than fourfold to about RMB1.7 billion. In the first half of 2026, it reported revenue of RMB1.15 billion, up 49%, and net profit of RMB274 million, while profit excluding one-off items fell about 19%.
Figure moved from about one Figure 03 per day to one per hour in fewer than 120 days, a 24-fold rise in stated throughput. Demand from paying customers is still the test of whether that ramp produces commercial leadership.
Agility's growth is still ahead of it. The company projects about 800 Digit units deployed in 2027 and aims to grow from nine to as many as 25 committed customer facilities by the end of 2027. Both are forecasts. Digit V5 early access starts in the first half of 2027.
Which humanoid robots are actually ready for customers?
Digit is still the most mature humanoid for a specific industrial workflow.
Agility has focused on moving material rather than attempting every human task at once. Digit unloads autonomous mobile robots, moves totes, loads conveyors and returns to its charger, and it can work up to 20 hours a day. The machine has logged more than 65,000 operating hours across nine customer facilities. Digit V5, which lifts up to 50 pounds and adds swappable hands, is the model behind its new orders.
UBTECH's Walker S2 has reached mass production and customer delivery. It handles materials, sorting, inspection and autonomous battery swapping. UBTECH management has nevertheless put current productivity at only 30% to 50% of a human worker in selected tasks, with a target of about 80% by 2027, and turning pilots into repeatable line productivity remains its main challenge. Getting from pilots to repeatable line productivity is a hard step, and we cover the biggest unsolved problems in our humanoid robots report.
Unitree's R1, G1, H1 and H2 can be bought as working hardware platforms. Buyers mostly receive a machine for research, software development, education or custom integration, not an autonomous worker ready for an established workflow.
Figure 03 has entered fleet production and moved into logistics sequencing at BMW Spartanburg. Apptronik describes Apollo as a customer-trial and data-collection platform. 1X takes NEO orders and says first-year capacity sold out within five days, but early units come with basic autonomy and scheduled remote expert assistance.
In our document detailing everything you need to know about humanoid robotics, you will always find fresh funding data, taken straight from our own verified funding database.
Who has the best humanoid robot performance and autonomy?
Figure still has the strongest general-purpose autonomy demonstration, while Agility leads dependable autonomy inside a narrower commercial workflow.
Figure's Helix system controls vision, reasoning, walking, balance and manipulation across the whole robot. In September 2026, Figure released Helix 2.5 and said it could work zero-shot in 30 homes it had never seen. That is a company-stated result, without independent testing.
The earlier Figure 02 fleet ran 10-hour weekday shifts at BMW, logged more than 1,250 operating hours, loaded over 90,000 parts and contributed to more than 30,000 BMW X3 vehicles. These are company-reported metrics, but they give a stronger baseline than short edited demonstrations. Figure 03's newer BMW task asks it to pick parts from unsorted containers and place them in sequencing trolleys.
Figure is also buying compute at a scale no rival has announced. In September 2026, it signed a partnership with Nscale for up to 100,000 NVIDIA GPUs, a ceiling rather than a delivered fleet.
Agility's autonomy covers a smaller task range and has lasted longer in ordinary customer sites. Digit navigates around workers and equipment, reacts to conveyor availability, switches between loading and unloading and docks for charging. We give Agility the reliability lead and Figure the generality lead.
Unitree remains exceptional at dynamic walking. Its robots have become common research platforms because developers can buy the hardware, access interfaces and test their own control systems. Walking alone rarely decides industrial purchases, where manipulation accuracy, intervention rates and uptime matter more.
Hands remain hard to compare. In July 2026, 1X introduced a 25-degree-of-freedom tendon-driven hand for NEO. Sanctuary AI has focused on contact-rich dexterity: its Physical AI reached a company-reported success rate above 99.5% at a 2.54-second cycle time while plugging flexible wires for a Tier 1 automotive supplier.
No independent benchmark tests these robots under the same conditions, so Figure's breadth is clearer than the exact gaps between manipulation systems.
Which humanoid robot offers the best economics?
Unitree offers the strongest hardware economics today, while Agility now offers the clearest published price for a working industrial humanoid.
Unitree lists the R1 from roughly $4,900 and has offered the G1 in the low five figures, depending on configuration. These prices put a complete humanoid within reach of universities, developers and small businesses. Its prospectus shows a gross margin of about 60% on its core businesses, so low prices have not meant thin margins. Selling cheap without thin margins is one way to make this work; we compare which business models are winning in our report covering humanoid robots.
The buyer must still supply software, integration, safety work, support and task training. Unitree lowers the cost of experimenting more clearly than the full cost of industrial automation.
Agility's filings give the first detailed price list for Digit V5. Robots-as-a-service costs $8,500 per month plus a deployment fee of about $25,000. A direct purchase runs about $200,000 upfront plus roughly $36,000 a year for software and maintenance. At roughly $100,000 a year per robot, Digit has to replace a meaningful share of a shift's work to pay off, and Agility has not published customer payback data.
1X has the clearest consumer offer: $20,000 to buy or $499 per month by subscription. At almost $6,000 per year, NEO will need to do much more household work before most families see it as good value.
UBTECH's revenue now comes with healthier margins. Its full-size humanoid gross margin reached 66.8% in the first half of 2026, although the group as a whole still lost money.
Learning from others' mistakes is the cheapest lesson there is. In our report covering the humanoid robotics market, we look at what broke startups before you, so you don't repeat it.
Who can manufacture humanoid robots at scale?
Chinese manufacturers own humanoid production. Unitree has the strongest proven pure-humanoid system, and AgiBot now has the largest output overall.
Unitree produced more than 6,500 pure humanoids in 2025 and shipped about 5,900 robots in the first half of 2026 alone. Output at that level proves more than theoretical capacity: suppliers, assembly lines, testing and end-customer delivery all work at multi-thousand-unit scale. Part of its IPO money is earmarked for a new robot manufacturing base.
AgiBot's estimated 8,400 shipments in six months make it the volume leader across its broader portfolio. UBTECH sold 921 full-size machines in the first half of 2026 and said its annual full-size production capacity had passed 6,000 units by the end of 2025. Its achieved volume is still well below that capacity.
Figure has built the fastest visible American ramp. BotQ delivered more than 350 Figure 03 robots by late April 2026 and reached a stated rate of one per hour. Sustained output in the thousands has yet to appear in public data.
Agility's RoboFab has a stated peak capacity of 10,000 robots a year, and Foxconn leads the private investment in its planned listing. 1X says its Hayward factory can build up to 10,000 NEOs a year. Both give a path to scale, while completed output remains far below the Chinese leaders.
Western rivals have credible capacity, but Chinese vendors shipped more than 97% of humanoids in the first half of 2026.
What makes the humanoid robot leaders hard to copy?
The hardest humanoid companies to copy combine hardware, data, production and operating experience rather than one celebrated technical feature.
Unitree's advantage begins with cost. Its in-house component skills, accumulated volume and broad developer base reinforce one another. More robots shipped mean more buying power and more researchers, whose software makes the hardware more useful. Its listing now adds a large cash buffer and a public currency. A rival can copy the outline of a G1 faster than it can rebuild that cost and distribution system.
Agility's defensibility comes from deployment experience. Every customer site exposes Digit to floor layouts, safety stops, damaged containers, workflow changes and integration failures that lab testing misses. Arc fleet software and the support team carry that knowledge from one deployment to the next.
Figure is building the strongest vertically integrated American stack. It controls robot design, manufacturing and Helix autonomy, then collects data from its own fleet, and its planned GPU capacity points to heavy spending on training. Its weakness is the cost of running that whole stack while cheaper manufacturers improve their software.
AgiBot has built the largest open robot-manipulation data resource among the companies assessed. AgiBot World contains more than one million trajectories across 217 tasks. In its published research, policies trained on the dataset improved average performance by 30% over Open X-Embodiment-trained models and passed 60% success on selected complex tasks.
UBTECH holds almost 3,000 authorized patents, takes part in national humanoid standards and gathers industrial data from factory deployments. Sanctuary makes its dexterity software available on existing industrial equipment, which could grow its training base faster than a Phoenix-only strategy.
No moat looks permanent. The strongest protection will come from deployed robots, task data, reliable production and repeat customers.
Your time is limited. In our research about humanoid robots, we put numbers on each part of the market, so you can see where it makes sense to focus.
Who turns capital into results: Figure, Unitree or Agility?
Unitree turns capital into hardware, revenue and profit more efficiently than any rival, although public markets have already cooled on its price.
Unitree listed on Shanghai's STAR Market on 19 August 2026 at 150.8 yuan per share and raised about RMB6.1 billion, or roughly $900 million. The shares opened at 1,100 yuan, a value near $66 billion, and closed up 460%, near $51 billion. By October 2026 they had fallen to about 520 yuan, roughly $31 billion, below Figure's last private valuation. The company reported 2025 net profit of RMB278 million and about RMB591 million excluding one-off items; Fortune cited a profit near RMB600 million, which matches the adjusted figure rather than the reported one.
Figure has raised about $1.85 billion and was valued at $39 billion after its 2025 round. That money has funded Helix, BotQ, several robot generations and a fast production ramp. External revenue and paid fleet size remain undisclosed, so the valuation still rests on future platform dominance. Figure's $39 billion rests on future platform dominance; we examine what humanoid companies are valued at in our humanoid robots market report.
Agility has chosen the public route too. In June 2026, it agreed to merge with Churchill Capital Corp XI at a $2.5 billion equity value, with more than $620 million of expected gross proceeds: about $420 million from the SPAC trust, before redemptions, and about $200 million from a Foxconn-led private placement. The SEC declared the registration effective in October 2026, and shareholders vote on 2 November. The valuation is about 1,400 times Agility's 2025 revenue, a bet on Digit V5 rather than a reward for current sales.
Apptronik raised another $520 million in February 2026, taking its Series A above $935 million and its total close to $1 billion. That money funds Apollo development, manufacturing and Robot Park, where robots gather task data, often under human guidance.
AgiBot is reportedly close to a new round of RMB1 billion to RMB1.5 billion at a valuation near RMB50 billion, according to investors cited in October 2026. That round could push its Hong Kong listing past the end of 2026.
The money produces very different outcomes. Unitree leads capital efficiency, Figure leads technical ambition, Apptronik leads partner-backed preparation, and Agility leads deployed work that has yet to become revenue.
Who leads each part of the humanoid robot market?
The humanoid robot market still has specialized leaders, with no company controlling hardware, intelligence, customers and economics at once.
Our segment judgments give the most weight to achieved results. Completed production counts more than factory capacity. A recurring customer agreement counts more than a pilot. Runtime in an ordinary workplace counts more than an edited demonstration. Recognized revenue counts more than an order backlog.
This breakdown explains why the overall ranking differs from a shipment chart. Unitree's cheap hardware is the strongest platform for developers, while Agility's managed system is the safer choice for a company seeking a proven industrial workflow. Figure leads the race toward general learned behavior, and 1X has the clearest proposition for homes.
| Market dimension | Current leader | Why it leads |
|---|---|---|
| Overall commercial deployment | Agility Robotics | Most convincing record of repeated customer work, operating history and converted agreements, though revenue was only $1.8 million in 2025 and $174,000 in the first half of 2026 |
| Pure humanoid shipment volume | Unitree | More than 5,500 clearly defined end-customer deliveries in 2025 and about 5,900 shipments in the first half of 2026 |
| Audited humanoid revenue | UBTECH | RMB590 million from full-size humanoids in the first half of 2026, after RMB820.6 million in 2025 |
| Broad embodied-robot volume | AgiBot | About 8,400 estimated shipments in the first half of 2026, a 44% global share |
| General-purpose autonomy | Figure AI | Strongest full-body learned-control demonstrations and quantified BMW experience |
| Affordable developer hardware | Unitree | Complete humanoids at prices far below most Western systems, with gross margins near 60% |
| Open manipulation data | AgiBot | More than one million published robot trajectories |
| Home humanoids | 1X | Consumer pricing, direct ordering, soft hardware and home-specific design |
| Industrial dexterity | Sanctuary AI | Strong contact-rich manipulation results against a live production benchmark |
| Strategic partner network | Apptronik | Google DeepMind, Mercedes-Benz, GXO, Jabil and other major backers |
In our market report about humanoid robots, we always try to gather several sources and make estimates that make sense, so you can work with solid numbers.
How reliable is the humanoid robot evidence?
The evidence is now good enough to identify leaders with reasonable confidence, but still too uneven for a precise universal score.
UBTECH provides strong financial evidence. Its revenue, losses and unit sales appear in audited filings, including its August 2026 interim report. Those filings still mix humanoid products with services and say little about uptime, customer concentration or revenue per workflow.
Unitree's prospectus, reviewed by the Shanghai exchange, gives unusually detailed data on revenue, profit, shipments and customer mix. We can be highly confident that Unitree runs a large, efficient hardware business. Confidence falls when judging the productive value of its installed base, since most disclosed humanoid revenue comes from research, education and display use.
Agility's merger filings changed what we know about it. They confirm customer names, hours and facility count, and for the first time disclose revenue and losses. They do not disclose fleet size per site, intervention rates or customer payback, and its order total depends on future milestones. The figures we cite come from its definitive proxy statement/prospectus of October 2026.
Figure releases detailed technical and manufacturing metrics, nearly all from Figure itself, with little independent checking. The BMW work raises their credibility because it happens inside a named plant, though BMW publishes fewer numbers than Figure.
Shipment rankings for 2026 come from Smart Analytics Global, with IDC slightly higher, while 2025 rankings came from Omdia. The two firms use different methods, so year-to-year comparisons are approximate. 1X openly explains that early NEOs rely on remote expert help, which helps transparency while confirming the consumer product is early.
We weight evidence in this order: audited financials and clearly defined deliveries first; named commercial deployments and repeated operating data second; customer pilots, order backlogs and internal fleet output third; demonstrations, reservations and factory capacity last. The ranking uses the strongest evidence available without pretending that every company reports the same variables.
Which humanoid robot company is actually ahead?
Unitree is the company ahead overall, with UBTECH second and Agility Robotics third.
We rank Unitree first because it now leads on the evidence we trust most. It shipped more than 5,500 clearly defined humanoids in 2025, about 5,900 more in the first half of 2026, earned a reported profit of RMB278 million and passed a full exchange review before listing in August 2026. Only about 9% of its disclosed humanoid revenue came from industrial uses, which is why its lead is narrow rather than comfortable.
UBTECH ranks second. It leads audited humanoid revenue, and it delivered 921 full-size humanoids in the first half of 2026, when its humanoid revenue nearly matched its full 2025 total. Task productivity remains well below human output, and the group is still loss-making.
Agility ranks third. It still has the best proof that a humanoid can do useful work often enough for customers to keep it running: recurring operation, converted pilots, several credible industrial buyers and a fleet-management layer. Its filings show that this proof has produced only $1.8 million of 2025 revenue and $174,000 in the first half of 2026, while a single customer accounts for its $300 million order backlog. The SPAC merger and Digit V5 give it momentum, not yet traction.
AgiBot takes fourth. It now ships more embodied robots than anyone and reports revenue above RMB1 billion for 2025. More transparent financials, likely in its Hong Kong listing documents, could move it into the top three.
Figure ranks fifth despite leading frontier autonomy. Its technology and production progress are exceptional, but a $39 billion valuation demands evidence beyond BMW and an internal fleet. Wider paid deployments with low human intervention would quickly change its position.
Apptronik, 1X and Sanctuary follow. Apptronik has the capital and partners to climb fast, although Apollo remains mainly a trial and data machine. 1X has the clearest home strategy but must prove that buyers get enough autonomous help for the price. Sanctuary's manipulation technology is impressive, while its push onto conventional industrial robots makes Phoenix less central.
Several moves could change the order. Agility can move up if the Digit V5 contract turns into hundreds of paying robots in 2027. Figure can challenge once BotQ output becomes disclosed customer revenue. UBTECH can rise if Walker productivity moves closer to human output, and AgiBot if its filings show durable customer economics.
| Rank | Company | Why it holds this position |
|---|---|---|
| 1 | Unitree | Cleanest humanoid shipment record, profitable and exchange-reviewed, with industrial adoption still a small share of demand |
| 2 | UBTECH | Largest audited humanoid revenue, real deliveries and fast growth, offset by weak current productivity and continuing losses |
| 3 | Agility Robotics | Best evidence of sustained paid work and converted pilots, but only $1.8 million of 2025 revenue, $174,000 in the first half of 2026 and a backlog from one customer |
| 4 | AgiBot | Largest 2026 embodied-robot shipments and the strongest open manipulation dataset, with limited financial transparency |
| 5 | Figure AI | Most convincing general-purpose autonomy and fastest American production ramp, but no disclosed revenue behind its valuation |
| 6 | Apptronik | Exceptional financing and strategic partners, though Apollo is still preparing the future commercial fleet |
| 7 | 1X | Strongest home positioning and direct consumer offer, with early utility still dependent on limited autonomy and remote support |
| 8 | Sanctuary AI | High-quality dexterity and industrial AI results, but a smaller disclosed humanoid deployment and manufacturing footprint |
If you build or invest in the humanoid robotics market, you need to know which business models pay off now and which ones are fading. That's why our report covering it has a whole section on it.
Our methodology
The answer to this question is less obvious than it first appears. The company shipping the most humanoids is not necessarily producing the strongest autonomy, creating the most commercial value or operating most reliably inside customer facilities.
We therefore broke the question into the dimensions that currently separate technical promise from competitive leadership: real-world deployment, shipment volume, audited revenue, customer traction, autonomy, economics, manufacturing scale, defensibility, capital efficiency and recent momentum.
For each dimension, we reviewed the freshest available company filings, IPO and merger documents, production disclosures, customer deployments, operating results, technical publications and independent market estimates. We assessed each data point according to what it actually measured, how directly it answered the question and whether it represented an achieved result or a future ambition.
We gave the greatest weight to evidence that was specific, recent and difficult to misinterpret. Audited financial results and clearly defined customer deliveries carried more weight than broad shipment claims. Repeated work inside named customer facilities carried more weight than demonstrations, recognized revenue carried more weight than order backlogs, and completed production carried more weight than stated factory capacity.
The companies do not report the same metrics, so we did not force unlike figures into a mechanical scoring model. We identified the strongest evidence within each dimension, preserved cases where different companies led for different reasons, and then considered which kinds of leadership matter most at the market’s current stage.
Shipment comparisons for 2025 rely on Omdia figures reported by Bloomberg, Omdia figures reported by the South China Morning Post, Unitree’s clarification of its 2025 sales data, and AgiBot’s production disclosures.
Earlier commercial evidence includes UBTECH’s audited public filings and Agility Robotics’ agreements and operational updates with GXO, Digit’s 100,000-tote milestone, Toyota Motor Manufacturing Canada, and Mercado Libre.
For autonomy, technical performance and manufacturing, we used Figure’s published results from its Figure 02 deployment at BMW, its Figure 03 logistics workflow, its Helix autonomy disclosures and its reported production ramp. We also reviewed AgiBot World’s published manipulation research, Sanctuary AI’s industrial manipulation results, 1X’s published pricing and early-support model and Apptronik’s Robot Park announcement.
Key sources include the Churchill Capital Corp XI and Agility Robotics proxy statement/prospectus, The Robot Report on Agility’s 2025 revenue and Digit V5 pricing, the Shanghai Stock Exchange’s summary of Unitree’s IPO and financials, Fortune on Unitree’s market debut, UBTECH’s first-half 2026 results, Smart Analytics Global’s first-half 2026 shipment data, TechNode on AgiBot’s funding and listing plans, and Apptronik’s $520 million extension.
The final ranking is the result of this structured aggregation. It reflects which companies currently combine the strongest recent evidence across technology, manufacturing, customer adoption and commercial execution, rather than rewarding the loudest announcement or the largest isolated number.
In our report covering the humanoid robotics market, we don't want to repeat what boring consulting reports say. Instead, we show you the real and latest market signals that matter.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us