What is the real market size of humanoid robotics?
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
The real humanoid robotics market today is roughly $2 billion to $5 billion globally, with a productive factory-and-logistics core that probably remains below $1 billion.
The market is already commercial, but shipment volume overstates economic maturity. Tens of thousands of humanoids are now shipping, yet most still go into research, entertainment, data collection, education and service applications rather than paid industrial labor.
China has built the clearest manufacturing lead. Chinese vendors account for most current unit volume, and companies such as Unitree, UBTECH and AgiBot are already generating hundreds of millions of dollars from humanoid or embodied-robot businesses. Chinese vendors ship most of the units; we look at which segments and regions are ahead in our humanoid robots report.
The sharper divide is between shipped robots and productive robots, not between China and the West. In the first half of 2026, manufacturing and warehousing together represented only about 17.7% of humanoid shipments, or roughly 3,900 machines on a 22,000-plus shipment base.
Hardware prices are splitting into very different markets. Entry platforms can now cost under $15,000, while integrated industrial systems can imply values many times higher once software, support and deployment services are included.
Factory evidence is finally getting harder to dismiss. BMW, Agility Robotics, Hyundai and others now have measurable operating hours, parts moved, totes handled and real production tasks rather than just choreographed demonstrations.
The biggest unresolved issue is repeat demand. Production capacity is rising faster than evidence that customers are returning to buy second and third fleets, which makes capacity utilization and repeat orders more informative than headline factory capacity.
Humanoids do not need to beat conventional automation everywhere. They only need to become cheaper in the awkward middle ground where work is repetitive enough to automate but too variable to justify rebuilding a site around fixed machinery.
The home market is still mostly future upside. Controlled industrial environments are far easier than homes, where clutter, children, stairs, fragile objects and changing tasks make reliability and safety much harder to solve.
The next real inflection point will come when customers expand fleets after proving savings in production. A $100 billion hardware market becomes plausible at several million annual units, but the multitrillion-dollar scenarios belong to a different category: they are labor-economy scenarios built around hundreds of millions of robots.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us
Your time is limited. In our report covering the humanoid robotics market, we put numbers on each part of the market, so you can see where it makes sense to focus.
How big is the humanoid robotics market right now?
The real humanoid robotics market today is probably worth around $2 billion to $5 billion globally, while the part made up of robots doing productive factory or logistics work is still much smaller.
That range is more useful than pretending there is one universally accepted market number. TrendForce expects China's humanoid robot market alone to reach about RMB15 billion, roughly $2.2 billion, while broader global estimates from firms such as MarketsandMarkets land around $5 billion. The difference comes largely from what each estimate includes: hardware, services, research robots, entertainment machines, industrial systems or some combination of them. Because each estimate counts different things, we spell out how big the market really is today in our report on humanoid robots.
The revenue disclosed by major manufacturers fits that order of magnitude. Unitree generated RMB868 million from humanoids in 2025. UBTECH produced about RMB820 million from full-size humanoid products and services. AgiBot reported RMB1.05 billion across its broader embodied-intelligence business. Those are three of the biggest commercial players currently shipping at scale, and their businesses are still measured in hundreds of millions of dollars.
The smaller industrial-worker market is especially important. Most humanoids sold today still go into research, data collection, demonstrations, entertainment and service applications. Only a minority are working in factories or warehouses.
So the current commercial market has clearly moved beyond laboratories and prototypes, but the genuinely productive humanoid labor market remains surprisingly small.
Why do estimates of the humanoid robot market vary so much?
Humanoid robot market estimates vary wildly because researchers often measure completely different things and give them the same “market size” label.
A shipment forecast measures robots leaving suppliers. Revenue measures what companies actually recognize from customers. Production capacity tells us how many robots factories could build. Orders may cover several future years. Long-range forecasts sometimes calculate the economic value of human labor that robots might eventually perform.
Those figures can differ by several orders of magnitude.
Hyundai, for example, plans to deploy more than 25,000 Boston Dynamics Atlas robots across Hyundai and Kia facilities and aims for annual Atlas manufacturing capacity of 30,000 units by 2028. That tells us a lot about where Hyundai thinks the technology is going, but those robots do not belong in today's shipment or revenue totals.
Tesla creates the same problem on an even larger scale. Optimus manufacturing plans have been discussed in extremely large future volumes, while today's external Optimus revenue remains negligible.
Orders require the same caution. Agility Robotics now has more than $300 million of multiyear Digit orders, subject to contractual milestones. That is serious customer commitment, although only delivered and accepted robots turn that backlog into actual revenue.
For the rest of this analysis, we therefore keep four ideas separate: robots produced, robots shipped, revenue earned and economically productive robots actually working.
In our research about humanoid robots, we always try to gather several sources and make estimates that make sense, so you can work with solid numbers.
How many humanoid robots are actually shipping today?
Global humanoid robot shipments are now measured in tens of thousands per year, which makes this a real hardware market, although it remains tiny beside conventional industrial robotics.
IDC counted more than 18,000 humanoid shipments worldwide in 2025. Omdia used a narrower definition and came closer to 13,000. Estimates for the first half of 2026 already range from roughly 19,000 to more than 22,000 units.
The direction matters more than the disagreement between trackers. TrendForce expects more than 50,000 global shipments for the full year, and Chinese production could exceed 100,000 units as manufacturers build ahead of demand.
Compare that with conventional automation. The International Federation of Robotics recorded 542,000 new industrial robots installed worldwide in 2024, including about 295,000 in China alone. The operating industrial robot population already exceeds 4.6 million machines.
Even if humanoid shipments pass 50,000 this year, the category would still represent less than one-tenth of one year's conventional industrial robot installations.
What has changed lately is the credibility of the manufacturing ramp. AgiBot had produced 15,000 robots cumulatively by June. Unitree has committed to annual humanoid capacity of 75,000 units. Several years ago, volumes like these existed mainly inside forecasts. Companies are now building the factories capable of reaching them.
Are today's humanoid robots actually doing useful work?
Most humanoid robots shipped today still spend their lives outside economically productive factory and warehouse jobs.
IDC estimated that more than 85% of 2025 deployments went into performances, education, data collection, guided tours and similar applications.
More recent application data show some improvement. Counterpoint estimated that during the first half of 2026, roughly 12.8% of shipments went into intelligent manufacturing and another 4.9% into warehousing and logistics. Entertainment and performance still accounted for about one-third, while data production and research took another 27%.
Applied to a shipment base above 22,000 robots, manufacturing and warehousing together represented only about 3,900 machines.
That distinction changes how we interpret the boom. A robot sold to a university or used to collect training data still creates real revenue, and some of those deployments help improve future models. But neither tells us much about whether a factory can replace a repetitive human task economically.
For now, tens of thousands of humanoids exist as commercial machines. Only a few thousand appear to be entering the kinds of jobs that could eventually turn humanoids into a large labor market.
| Where humanoid robots are going | Approximate H1 2026 volume | Share |
|---|---|---|
| Total shipments | 22,000+ | 100% |
| Entertainment and performance | ~7,400 | 33.6% |
| Data production and research | ~5,900 | 27.0% |
| Service and guidance | ~4,200 | 19.0% |
| Intelligent manufacturing | ~2,800 | 12.8% |
| Warehousing and logistics | ~1,100 | 4.9% |
| Other uses | ~600 | 2.7% |
| Manufacturing + warehousing | ~3,900 | 17.7% |
In our market report about humanoid robots, we don't want to repeat what boring consulting reports say. Instead, we show you the real and latest market signals that matter.
How much money are the biggest humanoid robot companies making?
The biggest humanoid robot manufacturers currently generate hundreds of millions of dollars in annual sales, and nobody has yet built a multibillion-dollar pure humanoid business.
Unitree gives us one of the cleanest numbers. Its 2025 revenue reached RMB1.699 billion, with humanoids contributing RMB868 million, slightly more than half of the company total. More than 5,500 humanoids were shipped during the year.
UBTECH operates further up the price ladder. The company generated around RMB820 million from full-size humanoid products and services in 2025 on 1,079 units. During the first half of 2026, that business produced another roughly RMB590 million from 921 units. A humanoid line already worth hundreds of millions of yuan is no pilot; we cover which companies lead today in our report covering humanoid robots.
AgiBot reported total company revenue of about RMB1.05 billion in 2025 and shipped more than 5,000 robots. Its product mix is broader, so treating the full amount as humanoid hardware revenue would overstate the comparison.
The latest American disclosures show a different stage of commercialization. Agility Robotics recorded only $1.8 million in net sales in 2025, despite building one of the industry's deepest real-world deployment records. The much bigger number is future business: more than $300 million of multiyear Digit 5 orders across a pipeline that includes manufacturing and logistics customers.
That gap is telling. Chinese companies currently dominate recognized hardware revenue and unit volume. Several American companies have strong industrial customers and substantial future commitments but much smaller disclosed revenue bases.
| Company | Best commercial figure | What it measures |
|---|---|---|
| Unitree | RMB868M | 2025 humanoid revenue |
| Unitree | 5,500+ units | 2025 humanoid shipments |
| UBTECH | RMB820M | 2025 full-size humanoid revenue |
| UBTECH | RMB590M | H1 2026 full-size humanoid revenue |
| AgiBot | RMB1.05B | 2025 total company revenue |
| Agility Robotics | $300M+ | Multiyear orders, subject to milestones |
| Figure | Undisclosed | Commercial deployments, including BMW |
| Tesla Optimus | Undisclosed | Mostly internal development today |
Why can one humanoid robot cost $5,000 and another cost more than $100,000?
Humanoid prices span more than twentyfold because a small development robot and an integrated factory worker barely belong to the same commercial category.
Unitree has pushed entry prices down dramatically. The R1 starts around $4,900 and the G1 around $13,500, putting humanoid hardware within reach of universities, research teams and developers.
Larger systems become expensive quickly. AgiBot has offered models around the $20,000-to-$40,000 range, while higher-end Unitree machines can reach roughly $90,000 to $100,000.
Industrial deployments add another layer. UBTECH's 2025 full-size humanoid revenue divided by units sold works out to roughly RMB760,000 per unit. That should not be read as a retail price because the revenue category includes products and services, but it shows how much more valuable a full industrial deployment can become once integration and support are included.
Unitree sits much lower. Dividing its RMB868 million of humanoid revenue by more than 5,500 units gives a blended figure around RMB158,000 or less per robot.
This is why shipment counts alone can mislead us. Selling 10,000 research robots at $10,000 produces a $100 million hardware business. Selling 10,000 integrated industrial systems at $100,000 produces $1 billion before recurring software or maintenance enters the picture.
| Humanoid | Approximate public price or implied value | Typical position |
|---|---|---|
| Unitree R1 | From ~$4,900 | Entry platform |
| Unitree G1 | From ~$13,500 | Research and development |
| AgiBot X2 | Around $24,000 | Mid-range platform |
| Unitree H2 | Around $30,000 | Larger humanoid |
| AgiBot A2 Lite | Around $45,000 | Commercial/service robot |
| Unitree H1 | Around $90,000 | Full-size humanoid |
| Unitree H2 Plus | Around $100,000 | Higher-end full-size platform |
| UBTECH blended 2025 revenue/unit | ~RMB760,000 | Industrial products + services |
In our document detailing everything you need to know about humanoid robotics, we like to show you the big picture. We know your time is precious, so we make every page quick to read and packed with as much learning as possible per minute.
Is China already winning the humanoid robot race?
China is currently winning the humanoid robot volume race by a huge margin.
IDC estimated that Chinese vendors accounted for roughly 95% of global humanoid shipments in 2025. Omdia put AgiBot near 5,200 units and Unitree around 4,200, giving those two companies alone roughly 70% of the market under its methodology.
The concentration has continued this year. Different trackers produce different numbers, but AgiBot and Unitree remain comfortably ahead of the rest of the industry on shipments.
China also has something the Western market still lacks: a dense price ladder. Buyers can choose relatively cheap developer platforms, mid-range robots and six-figure industrial systems from domestic suppliers. Behind those brands sits a supply chain already good at producing motors, batteries, electronics, sensors, actuators and precision components at scale.
Recent Chinese industry events have also become noticeably more commercial. TrendForce described the latest World Robot Conference as moving away from acrobatic demonstrations toward logistics, assembly, inspection and retail tasks. The conference introduced a procurement-focused program for buyers and suppliers, another sign that companies are now under pressure to show actual customer demand.
The harder question is whether China can turn unit leadership into durable economics. Unitree's revenue grew sharply, but recurring profit is currently under pressure. TrendForce also says the industry remains some distance from standardized repeat purchasing.
China has clearly built the first large humanoid manufacturing ecosystem. The repeat-order test will tell us how valuable that lead becomes.
Are China's huge humanoid production numbers backed by real demand?
China's humanoid boom has genuine customers behind it, although production capacity is still moving faster than proven repeat demand.
AgiBot's ramp is one of the strongest examples. The company moved from its first thousand robots to 5,000 and then 10,000 unusually quickly, reaching 15,000 cumulative units by midyear. Unitree is preparing capacity for tens of thousands more.
Orders are becoming more concrete as well. Galbot has secured a RMB236 million embodied-AI equipment procurement order. UBTECH has deployed robots with manufacturers and has accumulated thousands of orders or preorders across different products. Industrial customers now include companies in automotive, electronics, aerospace and energy.
The weakness appears one step later in the process.
TrendForce says large-scale standardized repeat orders remain scarce. Customers are simultaneously at the stages of formal procurement, preorders, expressions of interest and pilot partnerships. Those categories look similar in an industry presentation but mean very different things financially.
The most useful number from here will be the percentage of early customers that come back and buy larger fleets.
That is also where profitability starts to matter. Unitree's latest revenue continues to grow, while recurring net profit has been falling. If manufacturers need aggressive pricing and heavy support to keep volumes moving, higher shipments will tell us less about the long-term quality of demand.
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.
Have humanoid robots moved beyond factory demos yet?
Humanoid robots are now doing measurable work inside real factories, although deployments remain small compared with normal industrial automation.
BMW provides the cleanest example. Figure 02 spent roughly ten months working at BMW's Spartanburg plant, running ten-hour weekday shifts. According to BMW, the robot moved more than 90,000 sheet-metal components, logged about 1,250 operating hours and contributed to production involving more than 30,000 X3 vehicles.
BMW has now expanded its humanoid experiments to a second factory. At Plant Leipzig, the company recently introduced Hexagon Robotics' AEON into production testing. That is useful evidence because BMW is no longer learning from one robot supplier in one plant. BMW testing humanoids in a second plant shows how fast pilots spread; we follow the past year in our report on humanoid robots.
Agility Robotics has accumulated even more operating time. Digit 4 has now logged more than 65,000 hours across customer environments. At GXO's Flowery Branch facility, Digit passed 100,000 totes handled while maintaining around 98% on-task accuracy.
Boston Dynamics is moving closer to the same stage. The company has just opened a dedicated Atlas testing operation inside Hyundai's Georgia manufacturing footprint as Hyundai prepares for future factory deployments.
These examples are far stronger than choreographed demo videos because we can measure operating hours, parts moved and repeated tasks. What we still lack is fleet scale. A handful of humanoids inside factories containing hundreds or thousands of conventional robots remains an early commercial foothold.
Where are humanoid robots most likely to make money first?
Factories and warehouses are currently the best places for humanoid robots to prove their economics.
The attraction is straightforward. These environments already contain repetitive physical jobs, expensive labor, standardized containers and workstations built around human reach and movement. Companies also know exactly what each task costs them today.
Automotive manufacturing has become the busiest test ground. Figure has worked at BMW. UBTECH has entered multiple vehicle plants. Apptronik has worked with Mercedes-Benz. Boston Dynamics is preparing Atlas with Hyundai. Tesla has the unusual advantage of being able to test Optimus inside its own factories.
Warehouses offer a similar opportunity. Agility designed Digit around moving totes and materials, while its customer list now includes GXO, Amazon, Schaeffler and Toyota Motor Manufacturing Canada.
The strongest jobs today are fairly narrow: moving totes, loading parts, feeding machines, retrieving materials and handling repetitive objects between existing workstations.
Specialized automation remains a tough competitor. Conveyors are excellent on fixed routes. Industrial arms are faster at highly repetitive motions. Autonomous mobile robots already move goods around warehouses cheaply.
Humanoids become interesting when customers need one machine to work across several human-designed tasks without rebuilding the whole environment. That flexibility is the economic argument the industry now has to prove.
Our research about humanoid robots gets updated several times a month, so you always get the latest view of where things stand.
Can humanoid robots actually beat conventional automation on cost?
Humanoid robots can beat conventional automation in flexible environments, but today's economics still depend heavily on the task.
A fixed industrial arm usually wins when the same object follows the same path thousands of times per day. It has fewer moving parts, lower power requirements and decades of reliability data.
The humanoid advantage appears when the surrounding workplace is already designed around people and changes frequently. Rebuilding several work cells with dedicated machines can cost much more than putting a flexible mobile robot into the existing environment.
Hardware prices are also falling quickly. Bank of America has estimated that humanoid bills of materials could eventually reach roughly $13,000 to $17,000. Chinese entry-level models have already pushed retail prices far below what would have seemed realistic a few years ago.
However, the purchase price captures only part of the economics. Customers also pay for integration, maintenance, software, charging, downtime and sometimes remote human assistance.
The metric we care about most is cost per productive autonomous hour.
A $20,000 robot that needs constant intervention can be expensive. A $100,000 robot that works reliably across several shifts and multiple tasks can become cheap.
What is still stopping humanoid robots from working all day?
Battery life, reliability, safety and recovery from unexpected situations still limit how many productive hours humanoid robots can deliver.
TrendForce estimates that most current humanoids run for only around two to four hours on a charge. That is a poor fit for factories running eight-, ten- or twelve-hour shifts.
Battery swapping helps, and Agility's new Digit 5 shows how quickly this is changing. The company says the new generation can operate for more than 20 hours per day with rapid charging. That is a much more commercially relevant specification than simply making a robot walk faster.
Safety is moving forward too. Digit 5 was designed to work near people without the traditional physical barriers required by many industrial robots. It can detect someone approaching and slow or stop accordingly.
The difficult problems then move into software and reliability. Real workplaces contain misplaced objects, blocked routes, changing parts, imperfect lighting and people doing unpredictable things. Every time a robot gets confused and needs remote help, some of the promised labor savings disappear.
BMW's own factory tests make this visible. Its humanoid work produced useful results, while the deployment also required improvements to safety concepts, barriers and 5G coverage.
These practical details will decide the size of the market much more than impressive walking or dancing videos.
In our market report about humanoid robots, you will always find fresh funding data, taken straight from our own verified funding database.
How small is humanoid robotics compared with normal industrial automation?
Humanoid robotics is still tiny beside conventional industrial automation, especially when we compare productive deployments.
The International Federation of Robotics recorded 542,000 new industrial robot installations worldwide in 2024. China alone installed about 295,000. More than 4.6 million industrial robots are already operating globally. Against 542,000 industrial robots installed in 2024, humanoids are tiny; we look at how fast they grow, and why, in our humanoid robots market report.
Humanoid shipments were only around 13,000 to 18,000 in 2025, depending on the tracker.
The gap gets even wider when we focus on industrial use. Counterpoint's latest application mix suggests that only around 3,900 humanoids entered manufacturing or warehousing during the first half of 2026.
That six-month figure equals little more than 1% of the conventional industrial robots China installed in a single year.
Humanoids are growing much faster, of course. A market moving from a few thousand units toward more than 50,000 in a short period deserves attention. But the comparison tells us how early the category remains.
| Robotics market | Approximate volume |
|---|---|
| Global industrial robots installed in 2024 | 542,000 |
| China industrial robots installed in 2024 | 295,000 |
| Global industrial robot installed base | 4.66M |
| Global humanoid shipments in 2025 | ~13,000–18,000 |
| Global humanoid shipments in H1 2026 | ~19,000–22,000+ |
| Humanoids sent to manufacturing in H1 2026 | ~2,800 |
| Humanoids sent to warehousing in H1 2026 | ~1,100 |
| Manufacturing + warehousing humanoids | ~3,900 |
Are humanoid companies building robots faster than customers can use them?
Humanoid manufacturers are currently expanding production faster than customers have proven they can absorb robots economically.
China shows the gap most clearly. TrendForce expects Chinese humanoid output to pass 100,000 units this year, while global shipment forecasts from other researchers remain much lower.
Unitree is building toward annual humanoid capacity of 75,000 units. AgiBot has increased production at extraordinary speed. Hyundai plans capacity for 30,000 Atlas robots annually by 2028. Tesla's ambitions go much further if Optimus reaches mass production.
This aggressive buildout makes sense because higher manufacturing volumes should lower component costs. Cheaper actuators, sensors, hands and batteries can then make more customer use cases viable.
The risk appears if customer economics improve more slowly than manufacturing.
A factory capable of producing 50,000 robots needs a very different market from one supplying a few hundred pilot units. Manufacturers need recurring orders, stable utilization and customers expanding fleets after seeing actual savings.
As pointed out above, TrendForce says standardized repeat purchases are still limited. Unitree's recent profit pressure adds another warning that rapid revenue growth does not automatically translate into attractive economics.
Capacity utilization may end up being one of the most revealing humanoid metrics over the next few years.
In our document detailing everything you need to know about humanoid robotics, we look at what the winners do differently, so you can learn from the best instead of guessing.
Is the home humanoid robot market real yet?
The general-purpose home humanoid market remains tiny today and contributes very little to the current size of the industry.
Homes are much harder than controlled factories. A household robot has to cope with clutter, stairs, pets, children, fragile objects and constantly changing tasks while remaining safe around people who have no robotics training.
Price also becomes more unforgiving. Consumers may pay thousands for a useful appliance or tens of thousands for a car. Spending the same amount on a robot that only handles a few unreliable tasks is a harder sell.
Companies are still pushing toward that market. 1X has made the home a core part of its strategy, and cheaper Chinese hardware is pulling the entry cost down. Better vision-language-action models should also make robots more capable in messy environments.
Yet the commercial evidence remains weak compared with factories and research applications. We do not see millions of households buying autonomous humanoid workers, nor do we have strong data showing that current consumer robots can perform enough useful work to justify their cost.
Home robots could eventually make the humanoid market vastly larger. For today's market-size calculation, they remain mostly future upside.
Why do humanoid robot forecasts disagree by billions of dollars?
Humanoid forecasts disagree by billions because tiny differences in robot price and adoption rates become enormous once researchers project them ten or twenty years forward.
Omdia has projected around 2.6 million annual humanoid shipments by 2035. Bank of America has put the same year closer to 10 million units. Two forecasts nearly four times apart for the same year need explaining; we cover why forecasts disagree so much in our humanoid robots market report.
Goldman Sachs has revised its own expectations sharply as the technology improved. One earlier scenario pointed to roughly 1.4 million annual units and a $38 billion market in 2035. A more recent scenario has moved toward approximately 6.5 million units and about $138 billion.
The long-range estimates become far bigger again. Morgan Stanley has modeled a world with more than one billion humanoids and around $5 trillion of associated economic activity by 2050. Citi has explored a scenario approaching 650 million robots and roughly $7 trillion.
These forecasts answer increasingly different questions as the horizon gets longer.
A five-year shipment forecast mainly asks how quickly factories can build robots and customers can deploy them. A ten-year forecast makes much larger assumptions about falling prices, autonomy and new applications. A 2050 forecast is really asking how much physical human labor could eventually be automated.
That explains why the numbers spread so quickly.
| Forecast | Horizon | Scale envisioned |
|---|---|---|
| TrendForce | 2026 | 50,000+ global shipments |
| Omdia | 2035 | ~2.6M annual units |
| Goldman Sachs, newer scenario | 2035 | ~6.5M units / ~$138B |
| Bank of America | 2035 | ~10M units |
| Morgan Stanley | 2050 | 1B+ robots / ~$5T |
| Citi | 2050 | ~650M robots / ~$7T |
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.
Could humanoid robotics really become a $100 billion market?
A $100 billion humanoid robotics market is plausible within the next decade if annual shipments reach several million units.
The arithmetic is easy. Five million robots at an average $20,000 of revenue already create a $100 billion hardware market. Higher-end systems could reach the same revenue with fewer units. Software, maintenance, fleet management and integration would add more.
Reaching five million annual units is the hard part.
Even a 50,000-unit market has to grow one hundredfold to get there. Customers would need to find millions of jobs where humanoids make financial sense. Reliability would need to improve dramatically. Robots would have to work for long periods without supervision, and deployment would need to spread beyond automotive plants into electronics, logistics, retail, healthcare, hospitality and eventually homes.
There is now a believable path toward that scale. Manufacturing volume is rising, prices are falling, industrial deployments are accumulating real operating hours and recent robot generations are addressing practical problems such as battery life and working safely around people.
Still, the repeat-order evidence is several years behind the production ambition.
A $100 billion market therefore looks credible as an upside case. Calling it today's inevitable trajectory would run well ahead of what customers have demonstrated.
Does a multitrillion-dollar humanoid robot market make sense?
Multitrillion-dollar humanoid forecasts make sense only when we treat them as future labor-economy scenarios rather than estimates of robot hardware sales.
Morgan Stanley's roughly $5 trillion scenario and Citi's roughly $7 trillion scenario assume hundreds of millions of humanoids working across large parts of the economy.
At that scale, robot purchase prices become only one piece of the calculation. A company could pay for robots through subscriptions, robotics-as-a-service contracts, software fees or some version of hourly labor pricing. The relevant value starts looking closer to the economic output of workers than to the price of machines.
The mathematics can therefore reach trillions fairly quickly.
The uncertainty lies in the number of jobs humanoids can actually capture. Specialized robots will remain better at many repetitive tasks. Some physical jobs may be redesigned rather than replicated by a human-shaped machine. Countries with cheap labor will also adopt at different speeds from countries where workers are expensive or scarce.
Software could eventually add a large recurring layer as well. Millions of deployed robots would create demand for fleet management, model updates, teleoperation backup, maintenance and task software. Those businesses are still small because the installed humanoid base itself is small.
So the trillion-dollar scenarios describe what happens if humanoids become global labor infrastructure. They tell us a lot about the ceiling and very little about the current floor.
When you build or invest in the humanoid robotics market, you need clarity, not confusion. In our research about humanoid robots, we always try to put numbers on and quantify whatever is usually vague and confusing.
What is the real market size of humanoid robotics today?
The real humanoid robotics market today is roughly a $2 billion to $5 billion commercial industry, with a productive factory-and-logistics core that probably remains below $1 billion.
The strongest evidence now lines up surprisingly well.
Global shipments have moved from roughly 13,000–18,000 robots in 2025 toward more than 20,000 in the first half of this year. China already has manufacturers generating hundreds of millions of dollars from humanoids. AgiBot, Unitree and UBTECH have created genuine hardware businesses, while American companies such as Figure and Agility are accumulating serious operating experience with major industrial customers.
At the same time, most robots still go into research, entertainment, data collection, education and service applications. Counterpoint's latest mix suggests that fewer than one in five shipments are entering manufacturing or warehouses.
The newest industrial evidence is becoming much harder to dismiss. BMW has now tested humanoids in more than one plant. Figure accumulated 1,250 hours of production-line operation there. Agility has passed 65,000 operating hours across deployments and 100,000 totes at GXO. Boston Dynamics has moved Atlas testing directly into Hyundai's manufacturing environment.
Those numbers show a market that is becoming commercially serious while remaining far smaller than the forecasts surrounding it.
The next jump in market size will come from customers buying second, third and fourth fleets after the first deployments have paid for themselves. If that starts happening across automotive, electronics and logistics, today's few-billion-dollar market can grow very quickly.
For now, humanoid robotics is a small but unusually fast-growing hardware market, dominated by Chinese shipment volume, with a much smaller industrial-worker business sitting inside it. The huge forecasts become believable only after repeat purchases and productive operating hours start scaling as quickly as robot factories already are.
Our methodology
This analysis treats humanoid robotics as an evidence-aggregation exercise rather than a forecast-picking exercise. The market is still early, and the numbers most often used to describe it frequently measure different things: robots produced, robots shipped, revenue recognized, orders announced, production capacity, or the future economic value of labor that robots might eventually replace.
We broke the market into the commercial dimensions that give the clearest picture of its real size: realized revenue, actual shipments, application mix, productive deployments, pricing and unit economics, customer commitment, manufacturing capacity, and longer-term adoption scenarios. We gave the most weight to the freshest 2025 and 2026 evidence.
Our source hierarchy favors stock-exchange filings, audited or official company disclosures, customer deployment data, manufacturer pricing and production records. We then use established robotics research firms for market-wide shipment and application estimates, and major institutional research for long-range scenarios. Where a figure can be calculated directly from disclosed inputs, such as revenue per robot, we treat it as a derived indicator rather than a separately reported fact.
We also weight evidence according to how close it is to actual economic activity. Recognized revenue and delivered robots carry more weight than announced production capacity. Measured operating hours, repeated tasks and physical output at named customer facilities carry more weight than demonstrations. Contracted orders carry more weight than preorders, expressions of interest or partnership announcements.
Where credible market trackers reach different numbers, we preserve that disagreement instead of manufacturing false precision. We do not mechanically average forecasts or shipment estimates whose definitions differ. For the narrower productive-labor market, manufacturing and warehousing/logistics are used as the clearest observable proxy for humanoids already performing economically useful physical work rather than research, entertainment, demonstrations or data collection.
Today's market is kept separate from the much larger 2035–2050 scenarios. Those forecasts are useful for understanding what falling costs, higher autonomy and mass adoption could eventually produce, but they are not used as evidence for the current market size.
Key market-wide sources include IDC on 2025 commercialization and shipment concentration, Omdia's 2026 market radar, Counterpoint Research on H1 2026 shipments and application mix, TrendForce on China's 2026 market and commercial validation, TrendForce's global shipment outlook, and the International Federation of Robotics' World Robotics 2025 benchmark.
Company and deployment evidence comes from Shanghai Stock Exchange disclosures on Unitree, Unitree's official store for current pricing, UBTECH's 2025 annual results, UBTECH's H1 2026 interim results, AgiBot's 15,000th robot production milestone, Agility Robotics' Digit 5 commercial disclosures, BMW's Figure and AEON deployment data, Boston Dynamics' Atlas manufacturing-center update, Hyundai Motor Group's Atlas commercialization roadmap, and Mercedes-Benz on Apollo factory testing.
For longer-range scenarios, we use Bank of America Institute on humanoid costs and 2035 scale, Goldman Sachs Research on the 2035 market outlook, Morgan Stanley's 2050 scenario, Citi's long-range humanoid scenario, and MarketsandMarkets as an independent current market-value reference.
A good product starts with a real problem. In our market report about humanoid robots, we show you the ones customers still live with.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us