Humanoid robotics: what are the top startups now?

Why is everyone suddenly investing in humanoid robotics?

Humanoid robotics is attracting billions of dollars because robots are finally starting to work in real factories, and investors believe recent advances in AI could make them useful across many more jobs.

Figure AI raised more than $1 billion at a $39 billion valuation in September 2025. Apptronik followed with a $520 million financing extension in February 2026, bringing its Series A to more than $935 million. Germany's NEURA Robotics then announced a financing round of up to $1.4 billion.

These investments are arriving alongside measurable growth. Counterpoint Research reported that more than 22,000 humanoid robots shipped worldwide in the first half of 2026, nearly four times the volume recorded a year earlier. The firm expects annual shipments to exceed 50,000 units. Shipments nearly quadrupled in a year, and we go through what has changed over that year in our humanoid robots report.

Yet most humanoids still work in research, entertainment, training, or controlled industrial environments. Factories are beginning to use them, but there is little evidence that companies can already deploy thousands of autonomous humanoid workers economically.

The race is taking an interesting shape. American startups are attracting enormous amounts of venture capital, Chinese manufacturers are shipping most of the actual robots, and several smaller companies are finding practical uses that may prove more valuable than spectacular demonstrations.

Who's actually winning the humanoid robot race right now?

Figure AI, AgiBot, Unitree, Apptronik, and Agility Robotics are currently the strongest humanoid robotics contenders, although the evidence points to different leaders depending on what we measure.

Figure has a $39 billion financing valuation and is developing its own general-purpose robotic intelligence. AgiBot leads global shipments, while Unitree combines large production volumes with a profitable broader robotics business. Apptronik has strong industrial partnerships, and Agility has demonstrated repeated warehouse operations.

China's Galbot and Germany's NEURA Robotics are also worth watching. Meanwhile, 1X Technologies and Sunday Robotics are targeting households, where the technical and commercial challenges are quite different.

For our ranking, we focus on companies developing humanoid robots as their core business. We include companies such as Unitree and UBTECH that have already gone public, but distinguish them from independent private startups. Tesla's Optimus and Hyundai-owned Boston Dynamics are major competitors, although neither is an independent startup.

There is another distinction worth making. A smaller research robot selling to universities, a wheeled dual-arm machine working in a factory, and a full-size bipedal robot capable of walking around a warehouse can all appear in industry shipment statistics. They are not equally difficult to build, nor do they serve the same market.

Company	Country	Strongest evidence	Current position
Figure AI	United States	Funding, BMW deployment, proprietary AI	Leading private AI contender
AgiBot	China	Global shipment leadership	Largest by robot shipments
Unitree	China	Production volume and profitability	Leading established manufacturer
Apptronik	United States	Funding and industrial partnerships	Major US industrial challenger
Agility Robotics	United States	Commercial warehouse operations	Strongest logistics track record
Galbot	China	CATL industrial deployment	Fast-growing industrial challenger
NEURA Robotics	Germany	Financing and robotics platform	Leading European contender
UBTECH Robotics	China	Disclosed humanoid sales and revenue	Established commercial manufacturer
1X Technologies	US/Norway	NEO consumer reservations	Leading home-focused contender
Sunday Robotics	United States	Household manipulation and AI training	Emerging home robotics contender
Humanoid	United Kingdom	Funding and industrial development	Emerging European challenger
Leju Robotics	China	Research and training deployments	Significant Chinese supplier

We would not give every company an equal claim to leadership. AgiBot and Unitree already have manufacturing scale, Figure has exceptional financing and AI ambitions, and Agility has a particularly useful operating record. Those achievements deserve more weight than announcements of robots that are still years away from commercial production.

Is Figure AI really the world's top humanoid robotics startup?

Figure AI is currently the strongest candidate for leadership in general-purpose humanoid intelligence, but its $39 billion valuation puts it much further ahead financially than commercially.

The California company reached that valuation after announcing more than $1 billion in committed Series C capital, backed by investors including NVIDIA, Brookfield Asset Management, Intel Capital, and Qualcomm Ventures. A Series C above $1 billion shows how big rounds get; we look at who raised the most, and from whom, in our humanoid robots report.

Figure also has unusually concrete manufacturing experience. Its Figure 02 robots worked at BMW's Spartanburg plant, loading more than 90,000 sheet-metal components across approximately 1,250 operating hours. BMW confirmed that the deployment contributed to the production of more than 30,000 X3 vehicles.

Those figures are meaningful because BMW was testing a repetitive production task on an active assembly line. According to BMW, the robots achieved the required precision, although the project also required changes to factory safety systems and infrastructure.

Figure has since shifted toward Figure 03, its newer humanoid designed around the Helix AI system. BMW has been evaluating additional applications for that generation.

One qualification is easy to overlook. Figure retired Figure 02 after the BMW deployment, so its earlier operating record does not automatically establish Figure 03's reliability or manufacturing economics.

Figure is pursuing an ambitious combination of proprietary AI, general-purpose hardware, and robot production through BotQ. Its financial backing gives it considerable room to develop that system, but the company has yet to disclose sales and profit figures that would support its valuation through present commercial performance.

Is AgiBot bigger than Figure AI in the real world?

AgiBot is already far ahead of Figure AI in independently tracked robot shipments, making the Chinese company the clearest current leader by manufacturing volume.

Counterpoint Research estimated that AgiBot shipped approximately 9,700 humanoids in the first half of 2026, representing more than 43% of global shipments.

The growth becomes more striking when we compare periods. Omdia estimated that AgiBot shipped 5,168 humanoids throughout 2025. On those two research firms' estimates, the company shipped almost twice that full-year volume in the following six months.

We should be careful with that comparison because Omdia and Counterpoint do not necessarily count every robot category in exactly the same way. However, several independent trackers, including Smart Analytics Global, agree that AgiBot now holds first place.

AgiBot's production is spread across different markets. Its X-series robots are used in entertainment, education, research, and data collection, while its G-series robots are moving into factories. Counterpoint identified G2 deployments with Longcheer Technology, an electronics manufacturer, and Joyson Electronics, an automotive supplier.

That product mix explains both AgiBot's lead and its limitations. The company has found ways to sell robots in thousands of units, but many buyers are purchasing tools for research, training, or relatively simple commercial activities.

Figure is betting heavily on future general-purpose intelligence, while AgiBot is already building a large market for the hardware that exists. For actual manufacturing volume, the winner is pretty clear.

Is Unitree the most successful humanoid robot company in China?

Unitree has one of the strongest commercial records in humanoid robotics today, with thousands of robots sold, a fast-growing business, and actual profits.

The company's prospectus, reported by the Shanghai Stock Exchange, shows revenue rising from RMB 159.13 million in 2023 to RMB 392.77 million in 2024 and RMB 1.70 billion in 2025.

That represents more than tenfold growth in just two years. Unitree also reported RMB 278.21 million in net profit for 2025.

Its humanoid business has become increasingly important. According to its IPO disclosures, humanoids represented only 1.88% of revenue in 2023. By 2025, their contribution had reached approximately RMB 868 million, or nearly 52% of total revenue.

This change is particularly revealing. Unitree started with a strong position in quadruped robots, then managed to make humanoids its largest product category within two years.

Its G1 model has become widely used in research and education, while the H1 and newer platforms serve more advanced development and commercial applications. Unitree also benefits from its existing experience manufacturing actuators, joints, and integrated robot systems.

The company's profitability should still be interpreted at group level because quadruped robots and other products contribute to its results.

Nevertheless, Unitree has achieved a combination that most humanoid startups are still working toward: meaningful manufacturing volume, a rapidly growing humanoid business, and positive company-wide earnings.

Among the companies selling humanoid hardware today, Unitree has one of the most convincing business cases. There is more to show here than a funding valuation and a collection of impressive videos.

Is Apptronik catching up with Figure AI?

Apptronik has become one of Figure AI's most serious American rivals, especially in industrial robotics, but Figure still commands a much higher private valuation.

Apptronik develops Apollo, a humanoid designed for logistics, manufacturing, material handling, and other repetitive physical tasks. Its February 2026 funding extension valued the company at approximately $5.3 billion, according to TechCrunch.

That was about three times its earlier Series A valuation of roughly $1.75 billion. The increase shows how quickly investors have become more interested in Apptronik, even though its latest reported valuation remains well below Figure's. After Apptronik's valuation roughly tripled in one round, we show what each company is worth, then and now, in our humanoid robots market report.

Its industrial partners help explain the interest. Mercedes-Benz is testing Apollo in manufacturing environments, while GXO and Jabil are connected to its commercial deployment efforts. Google DeepMind is helping develop the robot's intelligence.

Apptronik has also been investing in training infrastructure. In June 2026, the company introduced Robot Park, where fleets of Apollo 2 robots collect real-world operating data alongside deployments at customer sites.

Learning new tasks requires far more than video demonstrations. Apptronik needs repeated manipulation attempts, information about failures, and enough variation in environments to improve Apollo's performance.

The company has convincing industrial relationships and substantial capital. What we still lack is a disclosed fleet operating record comparable to Agility's warehouse results or Figure's BMW production figures.

Has Agility Robotics proved that humanoid robots can actually work?

Agility Robotics has delivered some of the strongest commercial evidence in humanoid robotics, particularly through its Digit robots working inside GXO warehouses.

Digit is designed for material handling and logistics. In a deployment at GXO's Flowery Branch facility in Georgia, Agility reported that its robots had moved more than 100,000 totes.

Completing that many repeated movements in an operating warehouse reveals considerably more than a short demonstration. The robots had to work with existing equipment and perform a narrow physical task often enough for the total to reach six figures.

Agility has since expanded its commercial relationships to include Toyota Motor Manufacturing Canada and automotive supplier Schaeffler.

The company's financial position also changed in June 2026 when it announced plans to go public through a merger with Churchill Capital Corp XI. The transaction assigned Agility a $2.5 billion pre-money equity valuation and included more than $300 million in reported multi-year contracted Digit v5 orders.

Those contracts suggest customers are willing to commit money beyond small experiments, although the full order value has not necessarily been converted into delivered robots or recognized revenue.

Agility's advantage is unusually concrete: we can connect its technology to thousands of repeated warehouse operations. Its unresolved challenge is expanding that success into larger fleets and a broader range of profitable tasks.

Are Chinese startups already dominating humanoid robotics?

Chinese manufacturers currently dominate humanoid robot shipments, accounting for approximately 86% of the global market through the five largest suppliers.

According to Counterpoint's first-half 2026 research, those suppliers were AgiBot, Unitree, Galbot, UBTECH, and Leju Robotics. The two largest alone represented roughly three-quarters of worldwide shipments.

The rankings also reveal a gap between the market leaders and the companies chasing them. Third-place Galbot shipped more than 1,100 humanoids, compared with approximately 9,700 for AgiBot.

Galbot is an interesting challenger because it is developing robots for physically demanding work. Battery manufacturer CATL confirmed in June 2026 that Galbot's S1 was operating on its production lines. The robot offers a combined dual-arm payload capacity of 50 kilograms and a claimed operating duration of up to eight hours.

UBTECH provides another useful comparison. Its half-year financial disclosures show that full-size humanoid robot products and services generated RMB 590.3 million, up from RMB 38.2 million a year earlier. The company sold 921 full-size units during the period.

Galbot is gaining industrial experience at a major battery producer, while UBTECH is reporting rapidly increasing sales of full-size robots. Both are making progress that extends beyond the research market.

Still, China's shipment dominance comes with an important caveat. Counterpoint found that entertainment, performances, research, and data production together accounted for more than 60% of worldwide shipments. Intelligent manufacturing represented 13%, and warehousing and logistics another 5%.

The manufacturers are growing quickly, but the market for robots doing productive industrial work remains much smaller than total shipment figures suggest.

Company	H1 2026 shipments	Approximate global share
AgiBot	9,700	Over 43%
Unitree	Over 7,000	31%
Galbot	Over 1,100	5%
UBTECH	Around 1,000	4.4%
Leju Robotics	650	2.9%
Other manufacturers	Remaining shipments	Around 14%
Total global market	Over 22,000	100%

Figures are based on Counterpoint Research estimates. Categories include different humanoid designs and applications, and percentages are rounded.

China is winning the hardware-volume race by a wide margin. Whether that translates into leadership in general-purpose autonomous labor is still an open question, particularly when so many robots are being sold for activities that require relatively little autonomy.

Can NEURA Robotics become Europe's leading humanoid robot startup?

NEURA Robotics is currently one of Europe's strongest humanoid robotics contenders, with unusually large financial backing and an ambition to build a shared intelligence platform for multiple robots.

The German company announced a Series C financing round of up to $1.4 billion in June 2026. Its named backers included Tether, Qualcomm Technologies, Amazon, NVIDIA, Bosch, and Schaeffler. Tether, Qualcomm, Amazon and NVIDIA all backed one round, and we look at which investors are making the biggest bets in our humanoid robots report.

The stated amount is an upper limit for the financing round, so it should not automatically be interpreted as cash already received.

NEURA is developing its 4NE-1 humanoid alongside cognitive robotics systems and its Neuraverse ecosystem. The idea is to help robots learn and share capabilities across different environments and hardware designs.

It is also building NEURA Gyms, training facilities intended to produce the physical interaction data needed for more advanced robotic AI.

Figure concentrates heavily on its own humanoid platform, while NEURA wants to support a broader robotics ecosystem. That could make the German company attractive to partners developing different types of machines.

The United Kingdom also has a growing contender. Humanoid raised $152 million in July 2026 at a $1.35 billion valuation, bringing its reported financing to $270 million.

European funding is clearly becoming more substantial, but the region has yet to produce an independent humanoid manufacturer with shipment volumes comparable to AgiBot or Unitree. NEURA is among the best-funded candidates to change that.

Is 1X Technologies close to selling humanoid robots for the home?

1X Technologies is closer to a consumer launch than most humanoid robotics startups, although its NEO robot will initially offer a mix of autonomous assistance and human-supported operation.

The company introduced NEO at a planned purchase price of $20,000, alongside a $499 monthly subscription option. Customers could place refundable reservations for $200.

In April 2026, 1X said it had booked its planned first-year production capacity of 10,000 robots within five days of opening reservations.

That points to meaningful early interest, especially for a product priced far above ordinary home appliances. Still, reserving a robot is a much smaller financial commitment than paying $20,000 and using it every day.

NEO's autonomy is another important limitation. Its Expert Mode allows human operators to remotely assist with difficult tasks, helping the company complete work that the robot cannot reliably handle alone.

For early customers, this could make the robot more useful. It also creates questions about privacy, operating costs, and how much of the advertised household assistance is truly autonomous.

1X is targeting one of the largest imaginable markets for humanoids, but its near-term success will depend on actual deliveries and customer experience rather than reservation totals.

Could Sunday Robotics beat 1X in household robots?

Sunday Robotics is one of the most interesting challengers to 1X because its Memo robot concentrates on practical household manipulation, although Sunday is currently behind 1X in consumer sales readiness.

In March 2026, Sunday announced a $165 million Series B at a $1.15 billion valuation. The company said the money would help it move beyond demonstrations and begin placing Memo robots into homes through a beta program.

Its approach focuses heavily on training data. Sunday developed a Skill Capture Glove that records how people manipulate objects, giving its AI examples of the movements needed for household tasks.

The team has also expanded quickly. Sunday reported tripling its engineering organization, quadrupling its research team, and increasing data operations staffing fivefold.

Those investments reflect a specific technical challenge. A robot needs to handle fragile dishes, differently shaped objects, changing kitchen layouts, and countless variations of familiar tasks.

Sunday is betting that collecting large amounts of real human manipulation data can help Memo learn these activities faster.

1X has a clearer consumer offer, with published pricing and reservations. Sunday has yet to demonstrate that its manipulation-focused approach can be manufactured and delivered economically at scale.

Both companies are serious contenders, but the household robot market remains much less commercially proven than industrial robotics. A robot folding laundry in a demonstration is one thing. Having it do useful work in thousands of unpredictable homes every day is quite another.

Are humanoid robotics startups making real money yet?

Several humanoid robotics companies are now generating substantial revenue, although profitability remains much rarer than sales growth.

Unitree reported RMB 1.70 billion in company-wide revenue in 2025, with humanoids accounting for about half. UBTECH reported RMB 590.3 million in full-size humanoid products and services revenue in the first half of 2026 alone.

UBTECH's detailed financial results show how fast the business is changing. Full-size humanoid revenue increased 1,445% year over year, while unit sales climbed from 45 to 921. UBTECH's full-size revenue up 1,445% shows how fast one company can move; we compare who is growing the fastest in our report on humanoid robots.

That is a genuine jump in commercial activity. But UBTECH also reported a group loss of RMB 338.8 million for the period, even after reducing its losses by 23% from the previous year.

We can also calculate what its disclosed figures imply. Dividing RMB 590.3 million in humanoid-related revenue by 921 units gives approximately RMB 641,000 per unit. This is a blended revenue figure, including related products and services, rather than a published average robot selling price.

It nevertheless illustrates the size of the transactions involved. Industrial humanoids remain expensive equipment, and customers need enough useful work from them to justify the investment.

Private companies such as Figure and Apptronik have not published similarly detailed financial results. Their fundraising figures therefore tell us far more about investor expectations than present revenues.

Unitree and UBTECH give us some of the best evidence that a genuine humanoid robotics business is emerging, though neither proves that general-purpose humanoid labor has become broadly profitable.

Are humanoid robotics startups already overvalued?

The highest humanoid robotics valuations currently assume a much larger commercial market than companies have actually demonstrated.

Figure is the clearest example. Its valuation increased from $2.6 billion in 2024 to $39 billion in September 2025, a fifteenfold rise in roughly a year and a half.

Over the same broad period, the robotics industry made major technical advances, but large-scale autonomous humanoid employment remained limited.

Apptronik's approximately $5.3 billion financing valuation and Agility's announced $2.5 billion pre-money valuation are much lower. Those differences are especially interesting because all three companies have industrial relationships, yet the available public data does not allow a clean comparison of their recurring revenue or operating profits.

Investors are evidently assigning enormous value to Figure's possibility of developing widely useful robotic intelligence.

There is a plausible reason for that enthusiasm. A robot able to learn many new tasks could eventually generate revenue from hardware, software, fleet management, and ongoing AI services. The addressable market would extend far beyond traditional robotics.

But the costs are substantial too. Humanoids require expensive electromechanical components, computing hardware, maintenance, safety systems, and large quantities of training data.

Today's valuations therefore depend heavily on improvements that have not yet been demonstrated at commercial scale. Figure could grow into its valuation if it develops a widely adopted robotic platform, but its present valuation is far ahead of publicly verifiable business activity.

At $39 billion, investors are paying for a future robotics market that barely exists today. The potential is enormous, but the commercial evidence is nowhere near as large as the valuations suggest.

Which humanoid robotics startup has the best AI?

No humanoid robotics startup has yet proved that its AI is consistently better across the full range of physical tasks, although Figure, AgiBot, and several robotics AI specialists have developed credible approaches.

Figure's Helix system connects visual perception, language instructions, and robot actions. The company has demonstrated increasingly complex manipulation and coordinated whole-body movement.

AgiBot has been developing its own embodied AI models, supported by data collected across its growing hardware ecosystem. 1X uses Redwood to control NEO, while Sunday's approach relies heavily on household manipulation data captured from people.

Apptronik has chosen to work closely with Google DeepMind, which could give Apollo access to advances in Gemini Robotics without building every AI component internally.

There is also a company that could change the competitive balance without manufacturing a dominant humanoid of its own: Skild AI. In January 2026, it raised $1.4 billion at a valuation above $14 billion to develop general-purpose intelligence for different robotic platforms.

That raises an important question for the whole sector. If powerful robotics models become available to many manufacturers, advantages in hardware cost, reliability, and distribution may become more valuable.

On the other hand, companies operating large fleets can gather proprietary information about physical tasks, robot failures, and successful movements. Those datasets could make their AI progressively harder to match.

We cannot yet rank these systems fairly from company-produced videos. The missing evidence is repeated independent testing across the same unfamiliar tasks, with comparable success rates, operating speeds, and human intervention requirements.

Can humanoid robots compete with normal industrial robots?

Humanoid robots can already perform useful industrial tasks, but conventional industrial robots remain the better choice for many fixed, repetitive operations.

A standard robotic arm can work with exceptional speed and precision when performing the same movement thousands of times. Humanoids bring extra mechanical complexity, particularly when walking, balancing, and coordinating two hands.

Their advantage becomes more interesting in workplaces designed around people or where the required tasks change frequently. Workplaces built around people are where humanoids look most promising, and we look at the biggest opportunities in our humanoid robots market report.

BMW's work with Figure demonstrated accurate sheet-metal handling on an automotive production line. GXO's Digit deployment showed that a humanoid could repeatedly move warehouse containers. CATL's partnership with Galbot offers another example involving demanding material-handling tasks in battery manufacturing.

These cases share an important feature: each involves a specific, measurable job. They provide much stronger evidence of commercial usefulness than demonstrations of robots performing dozens of unrelated actions.

The next question is whether humanoids can do those jobs economically. Customers need to consider purchase prices, maintenance, electricity, software, installation, supervision, and actual productive hours.

Public disclosures still provide limited information about complete fleet-level operating costs and customer payback periods.

For now, factories and warehouses offer the clearest route to commercial growth, especially when robots handle repetitive tasks that are difficult to automate with existing equipment.

Could Boston Dynamics and Tesla beat the humanoid robotics startups?

Boston Dynamics and Tesla could become more powerful humanoid robotics competitors than many venture-backed startups because they have access to large industrial organizations and established manufacturing resources.

Boston Dynamics, owned by Hyundai Motor Group, introduced the production version of its electric Atlas humanoid in January 2026.

Hyundai has outlined plans to deploy 25,000 Atlas robots across its manufacturing operations and create capacity to manufacture 30,000 units annually.

Those targets remain future ambitions, but Hyundai can provide something valuable immediately: factories, engineers, supply-chain relationships, and a substantial potential customer base.

Tesla's Optimus program has similar structural advantages. Tesla can test robots within its own automotive and battery operations while drawing on its existing AI, electronics, and manufacturing teams.

However, neither company has yet published evidence of economically productive humanoid fleets operating at the scale suggested by its longer-term ambitions.

The competitive threat is still serious. A startup may develop an excellent humanoid but struggle with manufacturing, financing, and customer access. Hyundai and Tesla already possess much of that infrastructure.

Independent startups will need advantages strong enough to survive competition from companies that can deploy robots internally before relying on external sales.

So, what are the top humanoid robotics startups now?

Figure AI, AgiBot, Unitree, Apptronik, and Agility Robotics are the leading names in humanoid robotics today, with AgiBot and Unitree ahead in manufacturing scale and Figure standing out in financing and AI ambition.

AgiBot has the strongest claim to current global shipment leadership. Unitree is especially impressive commercially, with humanoids becoming its largest revenue category and its wider robotics business already profitable.

Figure remains one of the most important American startups because of its Helix AI platform, substantial funding, and real industrial experience. Apptronik is developing a serious competing platform for manufacturing and logistics, while Agility has one of the clearest operational records in warehouses.

Galbot deserves attention for its industrial work with CATL. NEURA is among Europe's strongest funded contenders, and 1X and Sunday are exploring a potentially enormous household market that is still at an earlier stage.

The biggest gap in the industry remains the distance between manufacturing success and autonomous labor. China currently sells most humanoid robots, but research, entertainment, and training still account for the majority of industry shipments. Companies developing more general-purpose intelligence may eventually capture larger commercial opportunities, but that outcome has yet to be proved.

As things stand, AgiBot leads in units shipped, Unitree has one of the strongest established business models, and Agility and Figure have produced unusually useful evidence from real industrial operations.

If we had to choose one private startup with the broadest long-term ambition, Figure would be our pick. If we were choosing based on commercial manufacturing evidence today, AgiBot and Unitree would rank ahead.

The eventual industry leader will need to bring those strengths together: advanced AI, reliable hardware, competitive production costs, and customers willing to deploy large fleets. No company has convincingly achieved all four yet.