Unitree Halved in Three Weeks — and China Just Wrote Its Lesson Into Humanoid-Robot IPO Rules

Unitree Robotics looked like the perfect China-tech story when it began trading on Shanghai's STAR Market on August 19. The maker of the humanoid that does backflips and kung-fu kicks priced its IPO at 150.80 yuan a share, then opened at 1,100 yuan — a 629% jump on day one. It closed that first session 460% higher, raising about 6.1 billion yuan (roughly $900 million) in the process, and China had its first mainland-listed humanoid-robot company.

Three weeks later the stock traded at 513.51 yuan, more than half below that opening peak. The collapse was not a market wobble; it was the price of the story being re-examined against what the company actually sells. And in early September, Beijing responded in a way that tells you what the whole episode was really about.

What the market was paying for

The share count is the place to start. Unitree sold about 40.45 million shares in the IPO, or 10% of its enlarged capital, which puts the total company at roughly 404 million shares. That makes the arithmetic simple at any price: at the 1,100-yuan opening high the company was worth around $60 billion; today, at about 513 yuan, it is worth roughly $30 billion.

Against those numbers, the slide looks less like a panic than like a mismatch being corrected. For 2025, Unitree reported revenue of 1.708 billion yuan — about $240 million, up 335% from a year earlier — and a net profit after one-off items of 600 million yuan. Even after halving, the company trades at more than 100 times trailing sales and hundreds of times trailing profit. That is a price investors pay for a company on the verge of dominating an industry, not one that sells tens of thousands of units.

Here is the number that carries the story, buried in Unitree's own IPO review materials. For the first nine months of 2025, fully 73.6% of its humanoid-robot revenue came from research and education — universities and labs buying the products to study and teach with. Only about 9% came from industry applications like factories and power grids. The company sells the cheapest serious humanoid on the market, the $13,500 G1, and shipped more than 5,500 units in 2025, but roughly 70% of those went to universities and research institutions for algorithm research and teaching.

In other words, the market briefly valued a maker of research tools for robotics labs as if it were already the industrial-scale supplier. That gap — between the backflip-spectacle narrative and a revenue mix that is overwhelmingly academic — is precisely where the halving came from.

The fundamentals were already cooling

Unitree is genuinely profitable, which sets it apart from most peers, and that profit is what let regulators approve its listing. But the strength was already thinning before the debut. In the first quarter of 2026, the company's adjusted profit fell by roughly half from a year earlier even as revenue kept climbing — falling prices and a crowded market compressing margins. The 2025 profit figure itself was eye-watering growth off a small base: revenue up 335%, profit up 674%, numbers that cannot be extrapolated forward.

That pattern is why this is not only a Unitree story. China's humanoid sector has more than 140 manufacturers and over 330 announced products, and the speculation drew a wave of copycats built on near-identical hardware and business models. The company's own success on the tape — surging then halving — turned the sector's hype into evidence of a problem.

Beijing wrote the lesson into the rules

The regulatory response, reported in mid-September, is informal "window guidance" — the China Securities Regulatory Commission signaling to banks and institutional investors rather than publishing a formal rule. The instruction to humanoid-robot IPO applicants is essentially the three things investors failed to price properly in Unitree's case: applicants must now show sustainable, recurring revenue; a clear path to profitability, or at least to narrowing losses; and genuine technological innovation rather than undifferentiated, low-price copies.

That is the significance of the whole episode, and it travels beyond China. Money has been chasing humanoid robotics on spectacle — backflips, dance routines, the promise of a Tesla-Optimus rival — and Unitree showed what happens when the market stops looking at the trick and looks at the revenue column. Less than 10% of the company's income comes from industrial deployment; the rest is labs. The regulatory crackdown exists to make sure the next listing cannot hide that same gap behind a hot debut.

For an investor watching any robot or embodied-AI name, the question Unitree's charts answer is the right one to ask everywhere: separate the demonstration from the customer. A company can be fun to watch and red-hot on its first day and still be, underneath, a research-tools vendor priced as an industrial champion. Unitree collapsed because the two things were attached to the wrong valuation. Beijing just made proving the difference a condition of going public at all.