Humanoid robotics is attracting billions in investment and increasingly ambitious production targets. But new industry data shows just how early the market still is: roughly 7,000 humanoids were sold in 2025, while factories installed more than 603,000 conventional industrial robots during the same year.
Humanoid robotics has become one of the most visible areas of the technology industry.
Companies are building new factories, raising large funding rounds, launching increasingly capable robots and talking about production in the tens of thousands—or eventually millions—of units.
The latest numbers from the International Federation of Robotics put those ambitions into a useful perspective.
Around 7,000 humanoid robots were sold globally in 2025 for industrial and professional service applications. Just days later, the IFR reported that factories installed approximately 603,000 conventional industrial robots during 2025, a record year for industrial automation. Reuters
That means traditional industrial robots outnumbered new humanoid sales by roughly 86 to one.
The comparison does not mean humanoids are failing.
It shows something more important: the humanoid market is still at the beginning of commercialization, despite already being valued and discussed like a much larger industry.
7,000 Units Is Both Small and Significant
Seven thousand robots is not insignificant for a category that was largely experimental only a few years ago.
But compared with established automation, it remains tiny.
The IFR says more than 5 million industrial robots were operating in factories worldwide by the end of 2025, after global installations climbed 11% to a record 603,000 units during the year. IFR International Federation of Robotics
Professional service robotics is also considerably larger.
The IFR recorded approximately 250,000 professional service robots in its 2025 sample, up 24% from the previous year. Mobile transportation robots represented nearly half of those installations. IFR International Federation of Robotics
Put beside those markets, humanoids are not yet a major automation category.
They are an emerging one.
And that distinction matters when evaluating everything from company valuations to manufacturing announcements.
A Humanoid Sale Does Not Necessarily Mean a Robot Worker
The 7,000-unit number also needs another qualification.
IFR Secretary General Susanne Bieller told Reuters that many humanoids sold in 2025 went to universities, research organizations and companies developing AI systems rather than directly replacing or supplementing workers in production environments. Automotive manufacturers remain among the earliest industrial users, but many current deployments consist of only a handful of robots per facility. Reuters
That means shipments and productive deployments should not yet be treated as the same metric.
A company purchasing 50 humanoids for AI training creates 50 robot sales.
But those robots may be generating demonstrations and training data rather than generating measurable labor productivity.
Both activities are valuable.
They simply represent different stages of commercialization.
This distinction is likely to become increasingly important as humanoid companies announce larger production numbers.
The market ultimately needs to know not only how many robots were manufactured, but also what those robots are actually doing after delivery.
The 2026 Forecast Requires an Enormous Jump
The contrast becomes even sharper when current sales are compared with industry forecasts.
Bank of America Global Research projected in March that humanoid shipments could reach approximately 90,000 units in 2026, followed by 290,000 in 2027, 500,000 in 2028 and 1.2 million annually by 2030. Bank of America Institute
Those projections should be treated as forecasts rather than expected outcomes.
The IFR’s newly reported 7,000 units for 2025 makes the required growth particularly visible.
Moving from 7,000 units to 90,000 would mean shipments increasing by nearly thirteen times in a single year.
Then the market would need to expand by another roughly thirteen times between 2026 and 2030 to reach the 1.2 million-unit forecast.
That is possible in an emerging technology market.
But it requires much more than companies announcing manufacturing capacity.
It requires customers.
Production Capacity Is Becoming a Misleading Metric
This may become one of the most important distinctions in humanoid robotics over the next several years.
Manufacturers increasingly announce the number of robots their factories can produce.
But production capacity is not the same as market demand.
A factory capable of manufacturing 10,000 humanoids annually demonstrates industrial readiness.
It does not demonstrate that 10,000 customers exist.
The automotive industry learned this distinction long ago. Manufacturing capacity only becomes economically valuable when factories can maintain utilization because customers continuously purchase the output.
Humanoid robotics will face the same test.
The stronger signal will therefore not be:
“We can manufacture 10,000 robots.”
It will be:
“Customers deployed 10,000 robots and are ordering more.”
That second number would show that the technology has progressed beyond manufacturing capability into sustainable commercial demand.
Conventional Robots Have One Advantage Humanoids Still Need to Earn
Traditional industrial robots are limited compared with humanoids.
An industrial arm normally performs a narrower range of movements and works inside an environment specifically engineered around it.
That appears to be a weakness.
Economically, however, specialization can be an advantage.
A manufacturer purchasing a welding robot already knows what it is buying the machine to do.
Its workload can be calculated.
Cycle times can be measured.
Maintenance requirements are relatively understood.
The robot’s productivity can be compared against alternatives.
Humanoids promise something fundamentally different: flexibility.
A humanoid could theoretically perform material handling today, parts sequencing tomorrow and another physical task after a software update.
That versatility is potentially extremely valuable.
But until those capabilities work reliably enough in production, customers are being asked to pay for future flexibility rather than established productivity.
That is why reliability matters more than spectacular demonstrations.
Humanoids Do Not Need to Beat Industrial Robots Everywhere
The comparison with 603,000 industrial robots also highlights a mistake that can easily enter discussions about humanoid automation.
Humanoids do not need to replace conventional robots.
In many applications, they probably should not.
A fixed industrial robot performing the same high-speed motion millions of times may remain faster, cheaper and more reliable than a humanoid.
The commercial opportunity for humanoids exists elsewhere.
Factories, warehouses and service environments still contain enormous numbers of tasks designed around human bodies because conventional automation has historically been too expensive or inflexible to handle them.
Boxes change position.
Doors need opening.
Carts need moving.
Different tools need handling.
Production lines change.
Facilities contain stairs, shelves, workstations and controls built for humans.
Humanoids become economically interesting when redesigning the environment around conventional automation costs more than giving the robot a body capable of operating in the environment that already exists.
That is the real business case the industry is attempting to prove.
The Most Important Number May Eventually Be Utilization
As more humanoids enter commercial environments, shipment totals may actually become less interesting.
The next generation of useful market metrics should include:
Deployment size: How many robots are operating at customer locations?
Utilization: How many productive hours does each robot perform per day?
Autonomy: What percentage of tasks can be completed without human intervention?
Reliability: How frequently does the robot fail or require assistance?
Expansion: Do customers order additional robots after completing a pilot?
Economics: How quickly does the deployment recover its purchase, integration and operating costs?
Those numbers reveal something shipment totals cannot:
whether customers actually want more robots after experiencing them.
In software, retention often separates novelty from a sustainable product.
Humanoid robotics may eventually develop its own equivalent metric: fleet expansion after deployment.
The 7,000 Figure Is Actually Useful Because It Establishes a Baseline
There is another reason the IFR data matters.
This is the first year the organization has separately tracked humanoid robot sales within its long-running robotics statistics. The IFR definition includes autonomous robots with a human-like appearance capable of operating in environments designed for people; legs are not required. Consumer, military and medical robots are excluded from the 7,000-unit figure. Reuters
That methodology is important because humanoid market estimates have historically been difficult to compare.
Different research groups count different types of machines.
Some include research platforms.
Others count shipments rather than deployed units.
Some define humanoids primarily by their body design, while others emphasize functionality.
As standardized reporting improves, the industry should become easier to measure.
The 7,000-unit number may therefore matter less because of its absolute size than because it creates something humanoid robotics badly needs:
a baseline.
The Market Is Early — Not Necessarily Overvalued or Undervalued
The numbers alone cannot tell us whether current humanoid company valuations are justified.
That depends on future growth, margins, software revenue, service models and how broadly humanoids eventually penetrate the labor market.
But the data does tell us where commercialization currently stands.
In 2025:
~7,000 humanoids were sold. Reuters
~250,000 professional service robots were recorded in the IFR sample. IFR International Federation of Robotics
~603,000 industrial robots were installed. IFR International Federation of Robotics
More than 5 million industrial robots were already operating worldwide. IFR International Federation of Robotics
Those figures place humanoids much closer to the beginning of their adoption curve than the end.
And that may make the next two years more important than all the demonstrations that came before them.
The Next Phase Is About Converting Expectations Into Orders
Humanoid robotics has already demonstrated that it can attract capital.
It has demonstrated extraordinary improvements in locomotion, manipulation and AI.
Several companies are now demonstrating that they can manufacture robots in larger quantities.
The remaining question is the one that matters most to a market:
Will customers keep buying them?
If humanoid shipments move from thousands to tens of thousands while customers expand successful pilots into permanent fleets, today’s ambitious market forecasts will begin to look more credible.
If production grows faster than real deployments, the industry may discover that manufacturing robots was easier than creating demand for them.
That is why the 7,000-unit figure should not simply be interpreted as either disappointing or impressive.
It should be treated as the starting line.
The humanoid industry has spent years proving that the technology can exist.
The next phase is proving that a market of comparable scale can exist around it.
