China Now Installs Nine of Every Ten Humanoid Robots, Barclays Finds

The uncomfortable fact underneath the humanoid-robot boom is that, on the ground, China is winning. Barclays estimates China accounted for roughly 85% of humanoid-robot installations last year and builds its machines at about half the Western cost — on the order of $50,000 each. The gap widens in industrial robots more broadly: China installs about 300,000 a year against roughly 34,000 in the United States, and its robot density has grown some 600% since 2016.

Part of the lead is manufacturing muscle, and part is data. Training a robot to move usefully in the physical world takes enormous amounts of real-world motion data, and Chinese firms are gathering it at industrial scale. JD.com is reported to be mobilizing as many as 500,000 workers to collect movement data for training robots — a data advantage that Western rivals have not matched. The pattern echoes how China turned smartphone supply chains into a national capability a decade ago, only pointed now at actuators, hands, and legs.

The clearest sign of the shift may be that even Nvidia, the American company whose chips power most advanced AI, reached for a Chinese machine when it needed one. Nvidia selected Unitree's H2 humanoid as the basis for its first research-grade humanoid platform — a choice that says as much about who is shipping capable, affordable hardware today as any market-share chart.

The American counter-argument is real but narrower than it used to be. The United States still leads on the AI models that give robots their judgment, on frontier chips, and on the first genuinely paid commercial deployments — Figure's units at BMW being the marquee example. Volume, though, is a scoreboard that compounds: every robot installed generates data, drives down unit cost, and funds the next generation. Whoever installs the most machines this year is better positioned to install even more next year.

For readers trying to invest around this, the takeaway is not to chase a single Chinese or American name but to recognize that the supply chain — chips, motors, sensors, and reduction gears — gets built regardless of which flag flies over the winning robot. That is where the public-market exposure sits, and it is why so much of the money is flowing to suppliers rather than to any one robot maker.