For most of the last decade, the public image of humanoid robotics was a single company's YouTube channel. In 2026, the image is a spreadsheet.
Chinese industry data cited by state media in August 2026 puts the number of domestic humanoid robot products at more than 400 models — over half of all humanoid models worldwide — with Chinese humanoid and quadruped robots accounting for about 80% of global unit sales.
That is a different kind of claim than "our robot can do a backflip." It is a claim about industrial capacity, and it is the reason the centre of gravity in robotics has moved.
Key takeaways
- Model count: 400+ domestic humanoid products, more than half of the global total (as reported by Chinese state media, August 2026).
- Market share: roughly 80% of global humanoid and quadruped robot sales by volume.
- Company count: more than 6,600 AI companies in China as of June 2026, about 15% of the world's total, spanning base infrastructure, model frameworks and industry applications.
- Commercial proof: at the 2026 World Artificial Intelligence Conference in Shanghai, 177 international procurement delegations attended, with intended purchase value reported at about 20.36 billion yuan, up roughly 25% year on year.
Why the supply chain decides this race
A humanoid robot is roughly 40 categories of precision components: harmonic reducers, planetary roller screws, frameless torque motors, dexterous hands with tactile sensing, six-axis force sensors, and the batteries to make all of it mobile.
China's advantage is not that it invented these parts. It is that it already manufactures them at consumer-electronics cost curves. The same industrial belt that produces smartphone actuators and electric-vehicle drivetrains can produce a robot knee actuator at a fraction of the cost of a bespoke Western equivalent.
This is why the price of a capable Chinese humanoid has fallen into the range of a mid-sized car, while comparable Western platforms remain priced like laboratory equipment. Price is not a marketing detail in robotics — it determines whether a university can buy twenty units or one, and whether a factory can justify a pilot line at all.
From stages to shop floors
The criticism of humanoid robots — that they are performing rather than working — is fair, and the industry's own deployment data confirms that most units today go to research, education and public exhibition rather than production lines.
But the trajectory matters more than the snapshot. Chinese state broadcaster reporting in 2026 documents industrial cases that were proofs-of-concept two years earlier:
- A "navigator-level" smart factory in Xuzhou, Jiangsu, received simultaneous international orders for nine crane models. Its production system replanned thirty days of operations autonomously, and production lines reconfigured in about ten minutes — a changeover that previously took five to six hours.
- Midea's humanoid robot "Meiluo 2" works quality inspection on a water-heater line: production cycle time cut by 20%, production loss down 50%, daily capacity up 33%.
- Acoustic large models and inspection robots at a Ningxia wind farm are projected to cut inspection labour by about 3,000 hours a year and reduce labour cost by roughly 60%.
None of that requires a robot that can dance. It requires one that can show up for a shift.
The honest limits
Two constraints are real and openly discussed inside China's own policy research:
Bottleneck technology. High-end chips, core algorithms and basic software remain described as "chokepoint" problems in mainstream Chinese analysis. The robot body is not the hard part forever; the compute inside it is.
Deployment depth. Official commentary notes AI in Chinese manufacturing is still moving from local pilots to scaled deployment. Smart factory counts are impressive — 15 "navigator-level", 500+ "excellent-level", 8,200+ "advanced-level" plants per Ministry of Industry and Information Technology data — but penetration is uneven across sectors.
The fair reading: China has won the manufacturing layer of humanoid robotics, is mid-transition on the deployment layer, and has not yet won the intelligence layer. That is exactly the sequence a hardware-first industry follows.
Figures are as reported by Chinese state media, MIIT disclosures and conference organisers in 2026.