On a trading floor in Shanghai, a retail investor faced odds of about 0.018 percent of actually getting allocated shares. The object of the frenzy was not a chip firm or a battery maker, but a company whose best-known product is a dog that walks on four legs — and a robot that dances on two.
That company is Yushu Technology, known globally as Unitree. Its ascent from a Hangzhou workshop to a STAR Market listing is now the clearest read on what public markets will pay for a Chinese robotics company that actually ships.
From workshop to STAR Market
Unitree was founded in 2016 by Wang Xingxing (王兴兴), who had built an early low-cost quadruped as a graduate student. The company's IPO application was accepted by the Shanghai Stock Exchange on March 20, 2026, and its listing committee approved the deal on June 1 — a 73-day stretch the company says set a record pace for the board.
When subscriptions opened, the retail portion was covered more than 8,000 times over. The shares priced at 150.80 yuan (≈ US$21 / HK$164), implying a valuation above 60 billion yuan (≈ US$8.5 billion / HK$66 billion). Gross proceeds came to about 6.10 billion yuan (≈ US$860 million / HK$6.6 billion). These figures are drawn from the company's regulatory filings and Reuters reporting on the offering.
The numbers behind the hype
The financials are the reason the book was hot. Revenue in 2025 reached 1.708 billion yuan (≈ US$240 million / HK$1.9 billion), up 335 percent from 392.77 million yuan in 2024. On a reported basis the company posted a 2025 net profit of about 278 million yuan, but its net profit excluding non-recurring items — the cleaner operating measure, inflated downward in reported terms by large share-payment expenses — was roughly 600 million yuan (≈ US$85 million / HK$650 million).
Chinese state media described Unitree as the first embodied-AI (具身智能) company to reach scaled profitability, a claim tied to those operating numbers rather than to market cap. Either way, the trajectory is unusual: a hardware robotics firm that turned cash-positive and stayed there.
Scale tells the rest. Unitree says it shipped more than 5,500 humanoid units in 2025, and that its quadruped (四足机器人) sales since 2022 total over 30,000 units. For most of the company's life, the dog was the business. In 2025 that flipped: humanoid revenue first overtook quadruped, with the two segments posting 5.95 billion and 4.88 billion yuan respectively in the first nine months — humanoids at 51.5 percent of revenue.
Why the market paid up
Two things separate Unitree from the crowd of humanoid startups. The first is vertical integration: it builds its own motors, reducers, dexterous hands, lidar and sensors, which is how it reportedly holds gross margins in the 60 percent range. The second is an open-source ecosystem — the prospectus cites 43 open-source projects — that keeps a global developer community building on its hardware.
International demand is real, not decorative. Overseas sales accounted for 43.65 percent of core revenue in 2025, about 731.66 million yuan (≈ US$103 million / HK$800 million). The branding work done by two consecutive Spring Festival TV gala appearances did not hurt domestic recognition either.
The risks the prospectus admits
The filing is candid about exposure. The United States is named as a key market, and the company flags potential American trade restrictions as a meaningful geopolitical risk. It notes the U.S. FCC added foreign-made humanoid and quadruped robots to its Covered List in late July, and that China's commerce ministry responded with countermeasure threats. Reports cited in the filing put China at roughly 85 percent of the global humanoid market — a position that cuts both ways when trade policy turns.
There is also the question of whether a 60-billion-yuan valuation prices in years of flawless execution. Humanoids are still overwhelmingly sold into research, education and demonstration today; the company's own mix shows commercial and industrial use rising but not yet dominant.
Why global investors should care
For readers outside China, Unitree's listing is the cleanest public benchmark yet for what markets will pay for a robotics company that actually ships volume rather than promises it. The 8,000-times retail oversubscription and 60-billion-yuan valuation matter less as a verdict than as a price signal: a profitable, vertically integrated hardware maker with its own motors, reducers and sensors, and 43 open-source projects drawing a global developer base.
The 43.65% overseas share of core revenue shows this is not a domestic-only story; demand from the United States, the market the prospectus itself flags as a geopolitical risk, is real. A useful comparison is with the broader humanoid field, where most peers sell into research and demos while Unitree already reports humanoids at 51.5% of revenue. The concrete risk to watch is the same one the company discloses: the U.S. FCC's late-July addition of humanoid and quadruped robots to its Covered List. That single policy move could reshape the 85% global-market position far faster than any product cycle.
Honest limitations
This article relies on Unitree's prospectus and on Reuters/Yahoo reporting of the offering; the "first profitable embodied-AI company" framing comes from Chinese state media interpretation of the operating numbers. The 8,000-times retail oversubscription reflects heat in the domestic retail book, not institutional conviction, and tells you little about long-term demand. Exchange-rate conversions use approximate rates and are labels, not precise valuations. Humanoid adoption outside labs remains early, and the trade-restriction risk is disclosed by the company itself rather than resolved here.
What readers can do now
- Read the prospectus's risk section directly before treating the valuation as a verdict on the whole sector — the trade-restriction disclosures are the most useful part for outsiders.
- Track the humanoid-to-quadruped revenue split each quarter; it is the cleanest public signal of whether bipeds are moving from demos into paid work.
- Compare Unitree's ~60 percent gross margin against peers' disclosed figures before assuming humanoid hardware is inherently high-margin at scale.