They can sprint, box, and dance for viral video feeds, but humanoid robots are still struggling to land actual day jobs.
Around 7,000 humanoid robots were sold globally in 2025 for industrial and professional service applications, according to the International Federation of Robotics’s first global tally of the category. The figures, reviewed by Reuters, mark the first time the Frankfurt-based organization has tallied humanoid units in its three decades of tracking robotics data.
To qualify under IFR criteria, a robot must possess a human-like appearance and operate autonomously in spaces built for people, though legs are not required. The tally compiles supplier and national industry association data, excluding consumer and military machines while categorizing medical robots separately.
Despite billions in venture backing, the commercial footprint remains modest. Susanne Bieller, secretary general of the IFR, told Reuters that humanoids represent “a fraction” of the world’s automated workforce. For comparison, about 542,000 industrial robots were installed in 2024, while sales of professional service robots reached nearly 200,000 units, according to earlier IFR data.
Data collectors, not line workers
The most telling detail is not how few units shipped, but what they are doing. Bieller noted that many humanoids sold in 2025 were not doing productive work. Instead, universities, research labs, and software developers bought them to harvest physical motion data for training artificial intelligence models.
Reuters reported that nearly three-quarters of Unitree’s humanoid sales went to developers and academic institutions. Even automotive leaders, the sector’s primary commercial testers, remain cautious. Bieller told Reuters that carmakers are operating pilots with “single-digit or sometimes double-digit numbers of robots in their plants.”
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
The hardware-utility disconnect
Some analysts expect shipments to increase rapidly. Bank of America Global Research projects that global humanoid shipments could climb to 90,000 units in 2026 and reach 1.2 million annually by 2030, with China’s manufacturing base and government support expected to drive much of that growth.
Yet an unspoken tension defines the current market: physical capability has outpaced cognitive reliability. While acrobatic demonstrations show significant progress in balance and actuation, reliably completing useful tasks in unpredictable industrial environments remains a bigger challenge.
Buying a humanoid today may introduce integration, maintenance, supervision, and software costs without the predictable output of specialized machinery. Unless a deployment has a clearly defined task and measurable return, it may function more as a testbed than a production investment.
What this means for businesses and workers
The figures suggest that broad humanoid deployment remains at an early stage. Business leaders should assess narrow use cases, integration requirements, safety controls, and measurable returns before making large capital commitments.
Early workplace deployments are more likely to involve limited tasks and continued human supervision, safety monitoring, and physical troubleshooting than fully autonomous replacements for entire roles.
For technology leaders, the 2025 tally is less a signal to rush into humanoid procurement than a benchmark for evaluating future claims. The metrics to watch are productive operating hours, task-completion rates, integration costs, safety incidents, and measurable return on investment.