Humanoid Robots Sold Just 7,000 Units Worldwide Last Year
Videos of Tesla's Optimus, Unitree and Figure make the humanoid era look like it has already arrived. The real 2025 sales number tells a much smaller story.
The humanoid era has begun, but not yet in the labor market
The International Federation of Robotics (IFR) has started tracking humanoid robot sales as a separate category, giving the industry its first real baseline. IFR defines a humanoid as a robot with a human-like form that can work autonomously without major changes to environments built for people. Two legs are not required. A wheeled robot can still qualify if it has a human-like torso, arms, sensing and autonomy.
The numbers are still small. According to IFR data cited by Reuters, humanoid robot sales came to about 7,000 units in 2025. For comparison, the most recent full-year count of traditional industrial robots, from 2024, showed roughly 542,000 installations, alongside about 199,000 professional service robots. The years don't line up exactly, but the gap is wide enough to make the point: humanoids remain a tiny, early-stage slice of the broader robotics market.
Humanoid shipment outlook
| Year | Units | Note |
|---|---|---|
| 2025 (IFR, actual) | 7,000 | Industrial and professional service sales |
| 2026 (BofA forecast) | 90,000 | About 12.9x the 2025 actual figure |
| 2030 (BofA forecast) | 1.2 million | About 171x the 2025 figure |
Chaining these forecasts together implies a compound annual growth rate of roughly 180% from 2025 to 2030. The 2026 and 2030 figures are Bank of America Global Research forecasts, not confirmed numbers.
The more surprising question is where these robots are actually working
Many of the humanoids sold today are not replacing a single worker on a production line. Reuters reports that research institutions and companies are buying a large share of them for AI development and behavior-data collection, and automaker deployments remain mostly small-scale pilots.
That distinction matters. A robot being sold is not the same thing as a robot delivering economically useful labor. The unit could be collecting pick-and-place data in a lab, running a demo in a showroom, or being tested on a factory floor for a few supervised hours a day.
Take an automaker that buys 100 humanoids. If the stated reason is "AI research," it is still an experiment. But if the numbers start to read "100 units running three shifts of logistics work, with minimal human intervention, paying back the investment in 2.3 years," the character of the industry changes. At that point the robot stops being a tech demo and becomes capital equipment on the factory floor.
From 7,000 units to 1.2 million: what has to be proven
| Metric to track | Why it matters | What a good signal looks like |
|---|---|---|
| Actual sales / shipments | Shows whether the market is growing from real orders, not announcements | Manufacturing and logistics customers growing faster than research buyers |
| Robot-hours | Shows how many hours robots are actually doing productive work | Uptime that looks like paid labor, not demo time |
| Task success rate | The share of tasks completed successfully | High success rates in real deployments, not controlled demos |
| Human intervention rate | How often a person has to take over remotely or step in | Steadily falling intervention rates |
| Cost per hour / payback period | The final gate between technology and economics | Total cost comparable to human labor, with shrinking payback periods |
Watching Tesla's Optimus, the numbers matter more than the dance moves
For Tesla investors, Optimus is a large long-term option. But a big long-term total addressable market is a different question from how much of that eventual success current valuation should already reflect. What matters right now is less how natural a demo video looks and more what it costs to build one unit and what that unit actually earns or saves.
Signals to watch:
- Manufacturing cost — how far actuator, gearbox, hand, sensor, battery and compute costs fall as production scales
- Uptime — how much time is spent on actual work, after subtracting charging, breakdowns and maintenance
- Autonomy — how fast the rate of task completion without human teleoperation climbs
- Payback period — how many years it takes a customer to recoup the purchase price and operating costs
In the end the question is simple: when does a single robot start working more cheaply and more reliably than a single human worker? That date is likely to determine when the humanoid market moves from promise to real commercialization.
Insight Times Editorial Desk
