China manufactures 90% of the world’s humanoid robots and the reason is not its industrial policy: it is crossing the street

On Chinese New Year, 16 Unitree humanoid robots danced a folk dance before almost a billion viewers. The West reacted as always: some with panic, others with disdain, others with an undisguised admiration that sometimes tends to concoct theories with more clichés regarding China than real analysis.

None of those answers is entirely true and that blindness has a cost.

The context. China manufactures about 90% of the humanoid robots sold in the world. In 2025, about 13,000 units were shipped, with Chinese companies (AgiBot, Unitree, UBTech…) dominating the ranking by volume, according to Omdia data collected by Bloomberg.

Tesla, with all its brand reputation and all its industrial apparatus, internally deployed around 800 units of the Optimus that same year.

The figure. He Unitree G1 It costs $13,500. He Tesla Optimus will exceed 20,000. That gap is the difference between being able to iterate ten times with the same budget or staying at one.

Between the lines. The story circulating in the West has two versions, equally lazy:

1. The first: all this is the five-year plan, the hand of the State, industrial policy made robot.
2. The second, reserved for the most condescending: it is because they copy.

Neither of them explains what is really happening.

China’s advantage in robotics does not come from the Communist Party. It comes from the Pearl River Delta and the Yangtze Delta: the two densest manufacturing ecosystems on the planet. Motors, actuators, sensors, custom PCBs… everything is available within walking distance. Is what it describes Rui Xuan engineer who has worked in robotics startups in China and Silicon Valley.

• When Unitree wants to test a new joint design, it crosses the street and comes back with the right component.
• A team in San Francisco has to wait weeks to receive the same component from China.

The background. That difference in iteration speed changes everything in hardware engineering. It stops being a problem of talent, because Chinese and American engineers are equally capable, and becomes a problem of infrastructure.

Breaking a robot, learning, replacing it, and trying again: that’s what builds cumulative technical advantage. If breaking a robot costs three weeks of logistics, learning stops and times become longer.

Yes, but. China does have state support, and it is completely legitimate to point this out. The government has injected a lot of money into that sector and has set production targets. But it’s not that Silicon Valley is an impoverished region: it has more capital, investors with more experience and resources, and more decades of experience financing high-risk bets. If this were a war to see who has the fattest checkbook, the United States would win handily. But it is not.

Furthermore, Chinese state money comes with strings attached: it is classified as “state asset” and founders assume personal liability if the company fails. That pushes capital toward politically safe bets, not necessarily toward the most innovative ones.

The question. Can the West make up ground in robotics? Yes, but not like he’s trying.

• Attracting foreign talent helps on the margin, but does not solve the underlying problem.
• The equalization involves building local supply chains capable of delivering a spare part in two days, not two weeks.
• And that is not an immigration or R&D problem. It is an industrial-based problem, and solving it takes many years of work. And of thankless work, from which those who arrive later may reap the fruits.

Until then we are going to see many more viral videos of Chinese robots doing pirouettes with increasing naturalness. And it’s because they’ve built the best environment in the world to break things and try again. In engineering, that explains almost everything.

Featured image | CCTV

In Xataka | Folding clothes or taking apart LEGOs has always been a tedious task. Xiaomi’s new AI for robots has put an end to it