China's Humanoid Robot Makers Are Racing to IPO. The Buyers Haven't Shown Up Yet.

Behind Unitree, the pipeline is crowded. EngineAI, a Shenzhen maker founded only in 2023, filed confidentially in Hong Kong after a $200 million round valued it at $1.5 billion. Leju, Deep Robotics, and AgiBot are all queued. Hong Kong alone has at least 46 robotics-related companies in its IPO pipeline, more than a tenth of all applicants. Globally, robotics firms have raised $55.8 billion in 2026 according to Dealroom, nearly double the previous annual record.

This is the same physical-AI thesis that Jeff Bezos just raised $12 billion on, arriving from the opposite direction. Bezos is betting on the software brain. China is shipping the bodies.

The number nobody on the roadshow leads with

Here is what the prospectuses bury: who is buying these robots, and at what margin.

China already accounts for about 85 percent of the world's humanoids produced in 2025 and half of all industrial robots. The factories exist. The shipments are real. But "units shipped" and "units sold into durable demand" are different metrics, and the second one is hard to find. A lot of early humanoid volume goes to research labs, demos, government-linked buyers, and other robotics companies, not to customers deploying them at scale for work that pays for itself.

CNBC put it bluntly in a recent newsletter: humanoid robots are great, but they need buyers too. The bull case assumes factories and warehouses absorb millions of these machines this decade. The current reality is a handful of pilots and a lot of viral videos of robots doing kung fu.

The market is already nervous

The public-market signals are flashing caution even as the IPOs accelerate. Humanoid robot stocks fell roughly 12 percent in 2026 after a 47 percent surge in 2025. The sector trades around 40 times forward earnings against 14 for the CSI 300. The main robotics ETF has seen net outflows for most of the year.

Shen Meng of Chanson & Co. said the quiet part out loud: "Investors trading at such elevated valuations are typically not driven by long-term fundamentals, but rather by the pursuit of short-term price gains." China's own securities press has warned that pre-IPO valuations are outrunning fundamentals, with many firms expected to burn cash for years.

When the companies raising the money and the regulators watching them both signal that valuations have detached from fundamentals, that is worth more attention than any shipment record.

How I'd read the next six months

I am not bearish on humanoids as a technology. The manufacturing base is real, the hardware is improving fast, and the NVIDIA partnerships give these machines a brain they did not have two years ago. The question is timing and price, not whether robots matter.

Three things will tell you which way this breaks. Watch whether Unitree's listing holds its valuation after the first month or fades like the broader sector. Watch for a single named enterprise customer deploying humanoids at scale for paid work, because that proof point still barely exists. And watch the burn rates in these prospectuses, because a company shipping 20,000 units at a loss is a very different investment from one shipping 5,000 at a profit.

The robots are coming off the line. Whether anyone can sell them profitably is the bet every one of these IPOs is asking you to make, usually without saying so.