China freezes humanoid-robot IPOs — a valuation reckoning hits the makers who ship the most.

On September 21, Reuters reported that Chinese regulators are quietly slowing a rush of humanoid-robot companies toward the stock market, using informal 'window guidance' to hold back listings rather than imposing a formal ban. At least half a dozen firms are preparing to go public, including Deep Robotics, X Square Robot and AgiBot; one person familiar with the matter said humanoid IPOs had effectively been frozen for now. The China Securities Regulatory Commission, which The Information first reported was raising the bar for approving these deals, did not comment. This matters far beyond China's capital markets: Chinese makers ship the overwhelming majority of the world's humanoids, so the financing that funds that volume — and the credibility of the numbers behind it — are now under the same microscope.

The core doubt is whether the revenue is real. Regulators are focused on how much of these companies' sales reflects genuine commercial demand versus state support: local governments can fund 80% to 90% of the robot data-collection centers and joint ventures that book much of the revenue, and one person close to humanoid-robot investors estimated that valuations at some companies could fall 60% to 70% if that project-linked revenue were stripped out. The context sharpens the point. According to the International Federation of Robotics, only about 7,000 humanoid robots were sold worldwide for industrial and professional use in 2025 — against roughly 542,000 conventional industrial robots installed in 2024, nearly 80 times as many. The deployed base is tiny relative to the valuations the sector has been commanding.

Unitree is the cautionary tale. It went public in Shanghai in August at roughly a $9 billion valuation, retail investors ordered more than 8,000 times the shares available, and the stock briefly pushed the company's value to around $67 billion on its first trading day before sliding about 55% from that peak. Its underlying business is growing fast — revenue more than quadrupled to 1.7 billion yuan (about $252 million) in 2025 — but the IPO still valued it at roughly 219 times last year's earnings. Venture capitalist Leo Wang of Qianchuang Capital called the boom 'campaign-style innovation,' and some private robotics projects have already taken valuation cuts of 30% to 50%. None of this means Beijing is retreating from humanoids as a national priority — it has stood up dedicated funds, including a 100 billion yuan Beijing vehicle with a 15-year horizon for AI and robotics — but it is trying to separate industrial policy from market demand.

For a buyer, the lesson is to hold a vendor's shipment and valuation headlines at arm's length from its commercial durability. If a large share of a maker's revenue comes from local-government data centers rather than independent customers, its ability to support, update and stand behind a fleet you deploy is less certain than the top line implies. The practical move is to ask any Chinese vendor for named commercial customers and an order book that is not government-backed, and to remember that these platforms remain shut out of new US sale under the July FCC ban regardless of how the IPO question resolves. The companies shipping the most humanoids are real and their machines work; whether their businesses are as large as their valuations is exactly what Beijing is now forcing them to prove.