Is the Humanoid Robotics Market growing now?

Why is the humanoid robotics market so hard to read right now?

Humanoid robotics is expanding fast, but the growth is still messy to measure.

The positive side is now hard to ignore.

In 2025, three independent reads put global humanoid volume between about 13,000 and 18,000 units: IDC around 18,000 shipments, Counterpoint around 16,000 installations, Omdia around 13,000 shipments. In the first half of 2026 alone, Counterpoint estimated more than 22,000 shipments. The category has left the “mostly lab prototypes” stage.

The negative side is that most of this volume does not prove normal enterprise demand.

Counterpoint’s first-half mix puts more than 60% of units in entertainment, performance, research and data production. A robot bought by a university or a show venue and a robot moving totes for eight hours a day count as one unit each, but they are not the same business. AP’s June 2026 reporting made the same point from the supply side: China can build humanoids at scale, but the hard part is finding enough buyers.

Vendors also count differently. Shipments, installations, pilot units and paid robots get mixed together, and almost no company publishes utilization or margin per deployed robot. The question is whether customers are pulling robots into useful work, or whether vendors, governments and investors are pushing capacity ahead of demand.

What are analysts saying about humanoid robotics growth now?

Analysts agree humanoid robotics is in a growth phase, but their numbers do not measure the same thing.

The near-term trackers are the most concrete. Counterpoint forecasts more than 50,000 shipments in 2026. IDC’s reported first-half estimate of nearly 25,000 units, up about 432%, is a separate estimate with a different scope, not extra units to add. Both are forecasts or reconstructions, not an audited census.

Barclays starts from today’s market size. Its January 2026 research put the current humanoid robotics market at roughly $2 billion to $3 billion, then modeled upside toward $200 billion by 2035 in optimistic scenarios.

Goldman Sachs is more conservative, forecasting a $38 billion humanoid robot market by 2035, with roughly 1.4 million units. Goldman's $38 billion by 2035 is one view among many; we look at why the forecasts disagree so much in our report on humanoid robots.

Morgan Stanley is much more expansive, pointing to a possible $5 trillion market by 2050, but that includes the wider ecosystem: hardware, supply chains, repair, maintenance and support.

The useful insight is the spread. Goldman is closer to a shipped-product market. Morgan Stanley is closer to a long-term labor-platform thesis. Barclays sits in between. Counterpoint and IDC tell us what is shipping this year.

Forecasts support one conclusion: analysts expect the market to keep growing. They do not prove that customers are adopting humanoid robots at scale today. For that, we need shipments, revenue, deployments, pricing, capacity, regulation and failure signals.

Are humanoid robot shipments actually taking off now?

Yes, humanoid robot shipments are taking off now, but from a very low base.

The 2025 base was about 13,000 to 18,000 units depending on the tracker. Counterpoint then estimated more than 22,000 shipments in the first half of 2026 alone, up nearly 300% year on year, with AgiBot at about 9,700 units (more than 43% share) and Unitree at more than 7,000 (about 31%).

The trackers do not match exactly. IDC’s reported first-half figure is close to 25,000. The direction is the same, and that agreement matters more than the exact number.

What the shipments are used for matters as much as how many there are. By Counterpoint’s estimate, 13% of first-half units went to manufacturing and 5% to warehouses and logistics. The rest is mostly research, data production, entertainment and performance.

The scale is still tiny versus mature robotics. Industrial robot installations are measured in hundreds of thousands per year. Humanoids are in the early volume phase: the market is growing now, but it is not yet a mainstream automation market.

Are AgiBot, Unitree and UBTECH selling robots, or just showing demos?

Yes, they are selling real robots, and their public accounts now show it.

China is the center of current volume. Two Chinese companies, AgiBot and Unitree, lead Counterpoint’s first-half 2026 ranking, with about three-quarters of global shipments between them. AP reported in June 2026 that China produced about 85% of the world’s humanoids in 2025.

The financial evidence is stronger than it was. UBTECH reported first-half 2026 revenue from full-size humanoid products and services of RMB590.3 million, up from RMB38.2 million. That line was close to half of group revenue of RMB1,269.1 million. Unitree reported Q3 2026 revenue of RMB665 million, up about 70%; that is company-wide revenue, not humanoids alone.

The quality of demand is less clear. AP found companies such as Matrix Robotics and EngineAI claiming thousands of orders from government and private customers across hospitality, logistics, entertainment and research. Neither UBTECH nor Unitree shows how much comes from repeat industrial buyers rather than schools, labs, data centers or promotional users.

China is not just staging robot videos; it is building the first real humanoid supply market. The risk is that supply matures faster than durable customer demand.

Are buyers spending real money on humanoid robots now?

Yes, humanoid robot buyers are spending real money, but in a few pockets.

UBTECH is still the clearest example. Its full-size humanoid business reportedly generated about RMB820 million in 2025, then RMB590.3 million in the first half of 2026 alone.

Orders have grown past one-off pilots. UBTECH said Walker S2 orders exceeded RMB800 million after a 159 million-yuan data-center contract. In the US, GXO agreed in October 2026 to deploy Agility’s Digit at two more warehouses, in North Carolina and Tennessee. A repeat order from an existing customer is a stronger signal than a first pilot.

Headline order values still need care. Agility cites more than $300 million in conditional, multi-year Digit 5 orders and more than 30 prospective customers, but its SEC registration statement shows net sales of $0.174 million for the first half of 2026. Conditional orders are not revenue, and prospects are not customers.

The limitation is concentration. Hard spending evidence still comes mostly from UBTECH, Unitree and China-linked deployments, plus a handful of Western logistics and automotive buyers. Customer-level payback data is almost never published.

Are factories using humanoid robots for real work now?

Factories are using humanoid robots for real work, mainly in narrow and controlled tasks.

Figure’s BMW Spartanburg deployment is the best Western proof point. Figure said its F.02 robots loaded more than 90,000 parts, ran more than 1,250 hours, supported production of more than 30,000 BMW X3 vehicles and worked 10-hour weekday shifts. That is a real factory task with measurable throughput, not a keynote demo.

Automotive pilots are spreading. BMW is moving toward humanoid work in Leipzig with Hexagon’s AEON. Mercedes-Benz has a commercial agreement with Apptronik to pilot Apollo. Tesla said in October 2026 that it had started early production-quality Optimus builds at Fremont, with no outside sales disclosed. Several big names are moving toward factory use, and we cover the companies that lead the market today in our humanoid robots report.

In China, Counterpoint cites AgiBot production-line deployments. In the West, Agility lists factory engagements including Toyota and Schaeffler.

The interpretation matters: manufacturing was only 13% of first-half shipments in Counterpoint’s estimate. Robots are being tested on constrained, repetitive tasks. That is task-level growth, not workforce-level replacement.

Are warehouses using humanoid robots beyond pilots now?

Yes, warehouse humanoid robots have moved beyond pure pilots at one customer, but deployment depth is still thin.

Agility’s Digit at GXO is the clearest case. GXO says Digit has moved more than 100,000 totes at its Flowery Branch, Georgia, site since the 2024 commercial deployment. In October 2026, the two companies agreed to expand to Statesville, North Carolina, and Lebanon, Tennessee. Those deployments are expected to start in early 2027. They are contracted, not running yet.

GXO is a useful buyer signal because it has tested several humanoid vendors, including Agility, Apptronik and Reflex Robotics. Large logistics operators are comparing platforms against real warehouse pain points, not running a one-vendor science project.

Scale is still missing. Warehouses took about 5% of first-half 2026 shipments by Counterpoint’s estimate, and public evidence points to a handful of high-duty sites, not hundreds of rollouts. Warehouse humanoids are operational in specific workflows without being a default automation purchase.

Are humanoid robots getting cheaper and easier to buy now?

Yes, humanoid robots are becoming cheaper and easier to buy, especially in China.

Price compression is one of the strongest growth signals. AP reported Chinese humanoid prices ranging from about $6,000 to more than $99,000. Unitree publishes list prices for entry models such as its G1 and R1 families, which pushed full-body humanoids into price bands that were unrealistic a few years ago. Chery’s AiMoga listed an adult-sized humanoid on JD.com at about 285,800 yuan, roughly $42,000.

This matters because adoption is payback-sensitive. A $200,000 humanoid needs a much stronger labor-replacement case than a $15,000 to $40,000 robot used for research, guided service or light commercial work. Lower prices widen the buyer pool before the technology is mature.

The catch is capability. A list price does not include integration, safety validation, supervision or guaranteed useful work, and a comparable cost per productive hour is not published. Cheap robots also help explain why research and entertainment dominate shipments: price compression grows volume faster than it proves labor economics.

Are investors still throwing money at humanoid robotics now?

Yes, investors are still putting a lot of money into humanoid robotics, and public markets have joined in.

NEURA Robotics reportedly secured up to $1.4 billion in June 2026 at around a $7 billion valuation, with investors including Nvidia, Amazon, Qualcomm, Bosch, Schaeffler and the European Investment Bank. NEURA is positioning around industrial AI and humanoid production, not just lab research. Nvidia, Amazon and Bosch backing a round shows the funding window is open; we look at who is funded, and by whom, in our humanoid robots report.

The pace held after June. In July 2026, UK startup Humanoid announced a $152 million Series A at a $1.35 billion post-money valuation. In August, XPENG said its robotics business had signed share-purchase agreements for more than $900 million at a post-money valuation above $6.3 billion; full closing was not publicly confirmed. Unitree listed on Shanghai’s STAR Market in August 2026, and Agility has filed a registration statement with the US SEC.

Earlier rounds set the bar. Apptronik’s Series A passed $935 million in February 2026, with backers including Google, Mercedes-Benz and John Deere. Figure raised more than $1 billion at a $39 billion valuation in September 2025. In September 2026, Figure also signed an initial $3.5 billion compute commitment with Nscale, which is a spending deal, not money raised.

Funding still should not be confused with demand. But the breadth of strategic investors matters: automakers, chip companies, cloud players, industrial suppliers and sovereign capital all want exposure.

Are big companies launching humanoid robot products now?

Big companies are moving humanoids from R&D into product and production, although most launches still come with caveats.

Tesla remains the largest wildcard. In its Q1 2026 update, Tesla said the first Optimus line at Fremont is designed for 1 million robots per year, and a future Texas line for 10 million. In October 2026, its Q3 letter said it had started early production-quality Optimus builds at Fremont. Elon Musk forecast 20,000 robots per week by the end of 2027. That is a target, not output, and no sales to outside customers have been disclosed.

Tesla is also the cautionary example. TrendForce warned in 2026 that Optimus commercialization remains uncertain, citing high-precision applications, cost structure and supply-chain dependency, after reporting in 2025 on battery-life and integration problems.

Elsewhere, Agility unveiled Digit 5 in September 2026, with company-stated specs of 50 lb payloads, a 90-minute runtime and nine-minute charging; commercial availability is planned for 2027. XPENG is developing its IRON humanoid. Chery’s AiMoga is on JD.com, and Unitree sells to developers, schools and small organizations. Productization is real; the gap is between “available” and “economically useful.”

Are humanoid robot companies hiring and building capacity now?

Humanoid robot companies are building capacity, but the capacity plans are clearer than the demand plans.

Tesla’s Fremont line is the biggest signal, now producing early units. Agility opened a new Fremont engineering facility in July 2026 and says its RoboFab plant is designed for up to 10,000 robots a year, which is a design ceiling, not current output.

Hiring and research show up in costs. Unitree’s Q3 2026 research and development expense rose about 250% to RMB127 million, close to a fifth of revenue, driven by AI investment.

The tension keeps returning: capacity can scale faster than usage. AP’s June 2026 reporting found that China has production capability but buyer demand is the constraint. In this market, production ambition is both a growth signal and a bubble signal.

Are humanoid robots still too limited for customers?

Humanoid robots are still too limited for many customers, and this is the main brake on growth.

The bottlenecks are concrete. TrendForce has pointed to battery life and hardware-software integration for Optimus. The Wall Street Journal has highlighted the “hands problem”: fine manipulation is still hard, even when robots can walk, lift or run scripted tasks. Agility’s Digit 5 targets longer runtime and faster charging, but those are 2027 promises, not features in the field today.

The deployment evidence tells the same story. The best real-world examples are narrow: loading parts, moving totes, guided service, data collection, reception and demonstrations. Figure showed its Helix 2.5 model working in 30 unseen homes in September 2026, with a company-reported success rate rising from 9% to 56% in its test, but that is research, not a paid home service. The market is growing because a few use cases now work, and we lay out what is behind it in our report on humanoid robots.

This is why the market can grow and still disappoint. Investors are pricing a general-purpose labor platform. Customers are mostly buying specific, supervised capabilities.

Is Unitree's falling profit a sign of a humanoid robot bubble?

Not on its own, but Unitree’s first quarter as a listed company show the stress points of this market clearly.

Unitree listed on Shanghai’s STAR Market in August 2026. Its Q3 revenue rose about 70% to RMB665 million, but net profit fell 17.4% to RMB112.7 million as research spending rose about 250%. Its shares also gave back much of their debut gains. Growth is real; so are volatile valuations and rising costs.

Agility shows a different gap. More than $300 million in conditional orders sits next to $0.174 million of first-half 2026 sales in its SEC filing. That does not mean the orders are fake, but it means headline numbers need to be read as potential, not revenue.

The older warnings still stand. China’s top economic planning agency warned in late 2025 about a humanoid bubble, pointing to more than 150 companies and similar products. AP’s June 2026 reporting found many Chinese humanoids still expensive, limited and dependent on government or research buyers. When many companies race to launch similar robots, prices fall and weak players survive longer than they should. The market is expanding, but parts of it look overheated.

Are humanoid robots getting pulled by labor shortages now?

Labor shortages are helping the humanoid robotics story, but they are not yet creating mass adoption.

The buyer logic is clear in factories and warehouses. Automotive plants and logistics sites have repetitive tasks, high turnover, safety issues and workflows built around human-scale infrastructure. That is where humanoids are being tested first: BMW with Figure, GXO with Agility, Mercedes-Benz with Apptronik, and Tesla internally with Optimus. GXO’s decision to add two sites suggests the workforce case holds up at least for tote handling.

Labor pressure creates urgency, but it does not close the business case by itself. A robot still needs uptime, safety certification, predictable maintenance, task reliability, integration support and a price that beats human or fixed-automation alternatives. No public study yet measures how much humanoid buying is driven by labor shortages, which is why the market shows many pilots and few large rollouts.

Are humanoid robot use cases becoming more practical now?

Yes, humanoid robot use cases are becoming more practical, mainly because the market is narrowing.

The strongest use cases are tote moving, parts loading, factory logistics, reception, inspection, education, guided service, teleoperation and robot-data collection. That is less exciting than the “general-purpose robot” story, but healthier. A narrow workflow can be measured, priced, trained and repeated, and repeat orders like GXO’s new sites only happen when it works.

The most underappreciated signal is data collection. Counterpoint groups research and data production among the largest application segments in its first-half 2026 mix, and UBTECH’s data-center contract shows robots being bought to generate the training data future autonomy needs. That is not labor replacement, but it is a paying market today. Robots bought to generate training data are a monetizable early market; we compare which business models are winning in our humanoid robots report.

The closer vendors get to boring repeatable tasks, the more real the market becomes.

Are regulations helping humanoid robotics grow now?

Regulation is helping serious buyers by turning chaos into standards, but slowly.

China released its first national standard system for humanoid robots and embodied AI in early 2026, covering the industrial chain and lifecycle of these systems. Enterprise buyers need safety, testing and interoperability rules before deploying robots near workers or the public.

International safety work is moving too. ISO 10218-1:2025 updated safety requirements for industrial robots. A dedicated draft for dynamically stable mobile robots such as legged humanoids, ISO 25785-1, closed its committee-draft comment period in July 2026, but it is still under development, not a published standard or a blanket approval to work next to people.

The EU AI Act adds a different kind of friction, with risk-based compliance that can slow workplace systems. Regulation is not a pure growth driver. It is a filter that favors better-capitalized, more mature vendors.

Are chip and sensor suppliers cashing in on humanoid robots now?

Yes, the humanoid supply chain is benefiting before the finished-robot market is fully proven.

This is one of the cleaner growth signals. Humanoids need actuators, reducers, sensors, cameras, batteries, chips, hands, simulation tools, safety systems and robot-learning infrastructure. Suppliers earn from prototypes, training fleets and production ramps even when end-customer adoption is still immature. Reducers, sensors and batteries form a segment of their own; we show which segments and regions lead in our humanoid robots market report.

Big suppliers are now buying in. In September 2026, Qualcomm agreed to buy PickNik, a robot-manipulation software company, and Cognex agreed to buy RealSense, the depth-camera maker. Both deals were announced, not confirmed closed, so they show buying interest rather than new revenue. Nvidia’s robotics stack and Figure’s compute deal with Nscale point the same way: compute and perception are absorbing money now.

That is why the supply chain can look stronger than customer deployments. Finished humanoid demand is still narrow; the infrastructure to build, train and scale humanoids is already being paid for.

Are humanoid robots becoming a normal procurement category now?

No, humanoid robots are not a normal procurement category yet, but a few large buyers now treat them like one.

The proof is in how buyers talk about them.

Most large-company activity is still framed as pilots, commercial agreements, trials, internal deployments, data collection or strategic investment. Mature automation categories are bought through repeatable tenders. Humanoids are still mostly evaluated through innovation programs and executive-level bets.

Procurement is getting closer, though.

GXO’s multi-site contract with Agility looks like ordinary vendor expansion. Robots-as-a-service reduces upfront risk. Rental models in China lower barriers for live performances and public venues. Public JD.com listings and Unitree price lists make the category easier to buy than two years ago. What is still missing is published fleet-level ROI and comparable tenders across many buyers.

The category is moving toward procurement, but it has not arrived.

So, is the humanoid robotics market growing right now?

Yes, the humanoid robotics market is growing right now, and the evidence is firmer than it was earlier in 2026.

The positive evidence is too broad to dismiss.

First-half 2026 shipments passed 22,000 by Counterpoint’s estimate. UBTECH and Unitree report fast-rising revenue in public accounts. Humanoid and NEURA raised large sums, XPENG signed for more, and Unitree went public. BMW and GXO have measurable deployments, and GXO is adding sites. Tesla has started early Optimus production. Prices are falling, standards are forming and suppliers are attracting buyers.

But the negative evidence still matters.

Most units go to research, entertainment, data production and demos. Profits are falling or absent at the leading listed vendors. Agility’s orders dwarf its recognized sales. Tesla has no disclosed outside customers. Western deployments remain few, and hands, battery life, safety, uptime and ROI still limit scale.

So the answer is yes, with limits. AgiBot and Unitree lead on volume, UBTECH on reported humanoid revenue, and Agility and Figure on measured Western work. Humanoid robotics is growing as an early commercial market. It is not yet growing as a broad labor-replacement market.

Here is a summary table of everything we have analyzed on this page.

Question	Verdict	Comment
Are humanoid robot shipments taking off now?	Yes	Counterpoint estimates more than 22,000 units in H1 2026, up nearly 300%, and forecasts over 50,000 for the year.
Are Chinese humanoid makers selling robots?	Yes	UBTECH and Unitree report fast-growing revenue, but the share from repeat industrial buyers is not disclosed.
Are buyers spending real money now?	Mixed	UBTECH’s humanoid revenue and GXO’s repeat order are real; Agility’s orders are conditional and far above its sales.
Are factories using humanoids now?	Mixed	BMW’s Figure work is measured and Mercedes and Tesla are moving in, but manufacturing was only 13% of H1 shipments.
Are warehouses moving past pilots?	Mixed	GXO’s Georgia site is live and two more sites are planned for early 2027, but depth is still limited.
Are humanoids cheaper now?	Yes	Chinese pricing has moved into lower bands, widening research and light-commercial demand.
Are investors still funding humanoids now?	Yes	NEURA and Humanoid raised large sums, XPENG signed funding agreements, and Unitree listed in Shanghai.
Are big companies launching products now?	Mixed	Tesla started early Optimus production and Agility unveiled Digit 5, but outside sales and 2027 availability are still ahead.
Are companies building capacity now?	Mixed	Production targets are large, but they are plans and design ceilings ahead of confirmed demand.
Are robots still too limited?	Yes	Battery, dexterity, autonomy, safety and integration remain major scale blockers.
Are bubble signals appearing recently?	Yes	Unitree’s profit fell and its shares swung sharply; China warned about overheating in 2025.
Are labor shortages creating demand?	Mixed	Labor pressure explains pilots, but it has not yet created mass fleet purchasing.
Are use cases becoming practical?	Yes	The market is narrowing toward tote moving, parts loading, data collection, guided service and factory logistics.
Are regulations helping growth?	Mixed	Standards support serious procurement, but ISO’s humanoid-relevant draft is not yet published.
Are suppliers benefiting now?	Yes	Qualcomm and Cognex announced robotics acquisitions; compute and perception benefit before finished robots scale.
Is procurement becoming normal?	Mixed	GXO’s multi-site deal, RaaS and listings help, but buying remains pilot-led for most companies.