China's National Development and Reform Commission used its August regular press briefing to sound a note of caution on the robotics industry. Li Chao, deputy director of the NDRC's Policy Research Office and the commission's spokesperson, said robotics touches multiple frontier technologies — artificial intelligence, advanced manufacturing, new materials — and that regions need to build on their own resource endowments and industrial strengths rather than blindly chasing the trend or piling in indiscriminately.
Set against this year's industry data, that remark carries more weight.
Selling little, producing a lot
According to the Ministry of Industry and Information Technology, revenue at China's above-scale robotics enterprises topped 300 billion yuan in 2025, with average annual growth exceeding 20% over the past five years; in the first half of 2026, revenue reached 165.5 billion yuan, up 24.5% year on year. Another figure the NDRC has cited more often: of every 10 humanoid and quadruped intelligent robots sold worldwide, 8 are made in China.
But zoom in on the humanoid robot segment specifically, and the picture narrows considerably. Global humanoid robot shipments in the first half of 2026 totaled roughly 19,100 units, with China accounting for more than 97% of that.
Put the two numbers side by side and the gap is jarring. The vast majority of that 165.5 billion yuan in revenue comes from industrial robotic arms, AGVs, and collaborative robots — older categories that have already proven their commercial viability. Humanoid robots' share, worked out roughly from the 19,100-unit figure, comes to somewhere around 3.8 billion yuan for the half-year even at an optimistic average price of 200,000 yuan per unit — just over 2% of the industry's total revenue.
Behind that 2% sit more than 150 humanoid robot companies — a figure the NDRC disclosed late last year, noting that more than half are startups or firms that crossed over from other industries. Split 19,100 units across 150 companies and the average comes to just over 100 units each — and that's the generous version of the math, one that credits Chinese manufacturers with the entire global shipment total.
Who "piling in" is really aimed at
The NDRC's language has escalated. Last November, the wording was about "preventing a rush of highly similar humanoid robot products to market, which compresses R&D space" — aimed at product homogeneity. This time, the phrasing shifted to "tailoring approaches to local conditions" and "preventing blind imitation," with the target moving to local governments — the building of industrial parks, dedicated funds, and courtship of assembly manufacturers that has been all but copy-pasted across regions over the past two years.
The other side of the industry's heat is more visible at trade shows and in capital markets. More than 300 companies exhibited at the 2026 World Robot Conference, up 69% from last year, with more than 3,000 items on display, over 300 of them launched for the first time. In the Hong Kong and STAR Market listings, Unitree Robotics has cleared its listing review and plans to raise 4.2 billion yuan, racing to become the first embodied-AI company to go public, with more peers lining up behind it.
The technical progress is real. Xu Xiaolan, chair of the China Institute of Electronics, said domestically made dexterous hands have reached precision of 0.1 millimeters and a single-unit lifespan of over 1 million uses. Since embodied AI was first written into the government work report in March 2025, the industry has also seen an unprecedented density of policy support.
The problem is that between improved component precision and an assembled robot that actually sells, there are still three hurdles: application scenarios, cost, and reliability. One earlier estimate found that China builds 90% of the world's humanoid robots, but only 23% of buyers say they're satisfied — production capacity has outrun demand validation.
Who the cooling shot is aimed at
The NDRC introduced no restrictive measures this time, only a spokesperson's remarks, so the binding force is limited. But for local investment-promotion officials, this kind of language usually functions as a weather vane: the lessons from the previous rounds of overcapacity in solar and lithium battery are still fresh, and when the central government signals early, it typically means project approvals, dedicated funding, and industrial park planning will tighten going forward.
For companies, the pressure transmits more directly. Once financing shifts from rewarding a good story to demanding actual orders, that denominator of 19,100 units becomes a question every company has to answer. Most of the 150 companies won't capture a meaningful share.
The industry's leader-takes-most dynamic should become clear next year. Whoever drives down per-unit cost and gets one concrete use case working will walk away with most of the shipment volume in this category; the rest will likely be acquired, or quietly retreat into the component supply chain.
Sources: National Development and Reform Commission press briefing, Ministry of Industry and Information Technology industry data, Xinhua, CocoLoop, China Institute of Electronics; revenue and shipment figures were verified against official public disclosures, while per-unit price and segment size are rough estimates, not official data.