While American robotics companies chase the general-purpose humanoid grail, Chinese firms are shipping units. Lots of them. The gap isn't just about labor costs — it's about a manufacturing ecosystem that compresses the entire R&D-to-deployment cycle.
Unitree shipped roughly 36 times more humanoid robots than US rivals Figure and Tesla combined last year. Over 140 Chinese firms released more than 330 different models. And the government has committed £100 billion to strategic technologies including robotics. Something structural is happening here.
The Scale Gap
Global humanoid robot shipments totaled just 13,317 units in 2025. That's a tiny base for an industry projected to reach 2.6 million units by 2035. But within that small number, Chinese dominance is already pronounced.
The top humanoid makers by 2025 shipments were China's Agibot and Unitree, followed by UBTech, Leju Robotics, Engine AI, and Fourier Intelligence. Unitree alone shipped over 5,500 units — more than any company in the world. Their robots start at about $1,600, while comparable American machines cost tens of thousands.
The explanation isn't mysterious. China has two sprawling metropolitan areas — the Yangtze River Delta near Shanghai and the Pearl River Delta in Shenzhen — home to hardware supplier networks so dense that robot-makers can sometimes walk next door for replacement parts. Tweaking a prototype can take less than a day in Shenzhen but weeks in Silicon Valley.
Why Speed-to-Scale Matters
The Chinese approach prioritizes commercial reliability over research ambition. As one Tsinghua University researcher put it: "We commercialize one generation of robots. Then we build the next generation." Many inevitably fail. The survivors iterate rapidly.
This creates a different kind of innovation dynamic. Where American companies like Tesla and Figure pursue the sci-fi vision of a single general-purpose humanoid, Chinese firms often specialize — building robots that do one thing extremely well. The US may eventually produce a single robot that can mow your lawn, walk your dog, and babysit your children. But while you wait, you can buy three Chinese robots that each do one task at a fraction of the price.
The government role is explicit. Municipal officials sit in meetings with robotics startups, offering facilities and incentives. The Leju training facility — over 10,000 square meters — was provided by the Beijing district government as part of a joint venture. Cities compete to host robotics champions: Hangzhou has Unitree, Shanghai has AgiBot, Beijing has Galbot, Shenzhen has UBTech.
The Software Question
Hardware dominance doesn't automatically translate to AI leadership. When it comes to vision-language-action models and world models — the software brains that will determine long-term robot capabilities — the picture is more uncertain.
Most Chinese humanoid startups currently run on NVIDIA Orin chips. Domestic alternatives are in development, but NVIDIA still leads the end-to-end software stack. The data scarcity problem affects everyone: unlike LLMs, robots can't simply scrape the internet for training material. Real-world data collection remains essential, and simulation environments are still imperfect substitutes.
As one analyst noted: "The hardware is currently ahead of the software — the robot body can handle a lot more dexterity today than years ago, but the brain is still nascent."
The Droid Brief Take
The China robotics story is often told as a tale of government subsidy and cheap labor. That's incomplete. The real advantage is ecosystem density — the ability to iterate hardware quickly, manufacture at scale, and deploy in real environments faster than competitors elsewhere.
This doesn't mean Chinese firms will necessarily win the race to general-purpose humanoids. That prize may still go to companies with deeper AI research capabilities and more patient capital. But it does suggest that the robotics market will bifurcate: China supplies the world with affordable, reliable single-purpose robots while the US pursues the generalist grail.
For factory operators and logistics managers, the Chinese option is increasingly compelling. For those betting on the long-term shape of the industry, the question is whether software advantages eventually overcome hardware ecosystem benefits — or whether the two approaches converge in unexpected ways.
What to Watch
• Unitree's IPO plans — The company is reportedly targeting a $7 billion valuation. The offering would test international appetite for Chinese robotics stocks and provide capital for global expansion.
• Domestic chip development — How quickly can Chinese firms reduce dependence on NVIDIA Orin? Progress here would indicate true technological independence.
• US market access — American lawmakers have already suggested restricting Unitree's access to US technologies. Whether Chinese robotics firms can sell into Western markets remains an open question.
• Software convergence — Will Chinese firms develop competitive VLA models, or will they remain dependent on Western AI infrastructure? The answer shapes long-term competitive dynamics.
Sources
TechCrunch — "Why China's humanoid robot industry is winning the early market"
The Guardian — "Inside China's robotics revolution"
Xinhua — "China's first national standard system for humanoid robotics poised to spur industry development"