Unitree Opens +629%, Closes +460% (19 August 2026): The 10% Float and the 8,288x Lottery That Set the Price

Unitree Robotics sold 10% of itself at 150.80 yuan a share on Wednesday 19 August 2026 and opened at 1,100 — a 629% gain in the first print, valuing a company with 1.70 billion yuan of 2025 revenue at about US$66 billion. It closed at 845 yuan, up 460%. The number everyone quoted is real, and it is not a valuation. It is the output of a rationing mechanism: a fixed offer price, an online tranche oversubscribed roughly 8,288 times, and an exchange rule that removes the daily price limit for the first five trading days.

Once you separate those two things — the price at which shares were sold and the price at which they first traded — the debut stops being a story about robots and becomes a story about market design. That is the more portable lesson, because the same mechanism produces the same headline every time a scarce, popular listing meets an administratively anchored offer price, and it will produce it again.

What actually happened

Unitree Robotics, the Hangzhou-based maker that began as a quadruped "robot dog" company and became a humanoid platform, listed on the Shanghai Stock Exchange's STAR Market on 19 August 2026. The Shanghai exchange had accepted the application earlier in the year with a stated fundraising plan of just over 4.2 billion yuan; strong demand let the company raise 6.1 billion.

The offer was 40.4464 million shares — 10% of the enlarged share capital — at 150.80 yuan. Multiply that price by the resulting share count and the company was sold at roughly 61 billion yuan. The opening print of 1,100 yuan put it at 444.9 billion. The gap between those two numbers, about 384 billion yuan, was created in a single auction and was captured by whoever held an allocation.

Strategic placements went to DeepSeek, Tencent's Qishan Investment and PetroChina's Kunlun Capital, per Forbes. Meituan's early stake, originally about 420 million yuan, was worth roughly 30 billion at the close — around 70 times the original investment. Founder Wang Xingxing holds 68.8% of voting rights.

The price was set by a lottery, not by price discovery

This is the mechanism, and it has three parts that compound.

Part one: the offer price is fixed in advance and is not raised to clear the book. In an offering where the sponsor can lift the price until demand and supply meet, extreme oversubscription is resolved before trading starts. Where the price is anchored — by convention, by regulatory expectation, by a target raise — the excess demand cannot be cleared at pricing. It survives, intact, into the first session.

Part two: the float is small and the retail slice smaller still. Ten per cent of the company was sold. Of that, roughly 9.7 million shares went to the online retail tranche, against about 9.8 million applying accounts — one share's worth of allocation per account, on average. Strategic-placement stock is not part of day-one tradable supply. So the shares genuinely available to trade at the open were a fraction of an already narrow 10%.

Part three: no price limit. The Shanghai Stock Exchange removes the daily price limit for a STAR Market listing's first five trading days, applying a 20% band from day six. The exchange's rationale is efficiency — let the equilibrium price form at once rather than through a queue of limit-up sessions. On a normal listing that is sensible. On a listing where demand was rationed at 0.018%, it means the entire adjustment happens in one print.

Read in that order, +629% is not a surprise. It is what the design produces. The useful test is the one the session itself supplied: the stock closed 255 yuan below its open. Roughly a quarter of the pop did not survive contact with a full day of two-way trading, which is the clearest available evidence that the opening number was a rationing artefact rather than a settled price.

The arithmetic the pop leaves behind

None of the following is a view. It is division, and it defines what would have to be true for each price to be justified by earnings rather than by scarcity.

The underlying business grew fast: 2025 revenue of 1.70 billion yuan against 392.77 million in 2024, with net profit of 278.21 million yuan, gross margins near 60%, more than 5,500 humanoid robots shipped in 2025, and humanoids rising from 27.6% of revenue in 2024 to more than half. Profitability genuinely distinguishes it — the private US developer Figure has carried a US$39 billion valuation without it, and Hong Kong-listed UBTech has been loss-making at comparable scale.

Hold the absolute number in view all the same. 1.70 billion yuan is roughly US$250 million of annual revenue. To pull the closing capitalisation to 40 times trailing earnings at an unchanged share price, net profit would need to rise about thirtyfold. That is the "what has to be true" — not a forecast that it will or will not happen, but the size of the gap the business has been asked to close.

A 460% debut on a day the market fell 7%

The contradiction is the most interesting fact of the session. While Unitree closed up 460%, the STAR Market Composite Index fell 7.2% and the Shanghai Composite 2.4%.

Two forces, running in opposite directions.

The mechanical one is local. A debut turning over 23.2 billion yuan pulls cash out of everything else on the same board in the same hours. Positions get sold to fund allocations and to chase the open, and in a retail-heavy market that rotation is visible at the index level.

The larger one was global and arrived from outside China entirely. Semiconductor shares sold off worldwide that day: the US session saw the PHLX Semiconductor Index fall 628.54 points, or 4.98%, to 11,992.46, with all 30 components lower for the first time since late July, as long-dated government bond yields pressed higher and questions about the pace of AI-related capital spending resurfaced. Chinese chip and robotics names sat inside both stories at once.

That second force is the one with a transmission channel a macro reader can actually use, and we set it out in detail in the long-end selloff piece: when the 30-year yield rises, the discount rate applied to cash flows arriving a decade out rises with it, and the assets repriced hardest are the ones whose value is furthest in the future. A humanoid-robotics company earning 278 million yuan today and valued on what it might earn in the 2030s is the purest possible example of that duration. The rate factor is one of the five the meter scores across the majors, and it does not stop at the currency market's edge. The financing side of the same trade — how the AI build-out is actually being paid for — runs through the Nvidia–OpenAI guarantee structure.

What it does — and does not — tell you about robotics

It confirms one thing solidly: demand to own listed humanoid exposure in China vastly exceeds the supply of it. Nearly 9.8 million retail accounts chasing 9.7 million shares is a measurement, not a mood. Caixin reported 30 to 50 robotics companies preparing Hong Kong listings around the same period, and that is the natural response — scarcity this extreme is an invitation to supply.

It tells you nothing about unit economics at scale, adoption rates outside China, or whether the humanoid form factor wins the tasks it is being pointed at. A first-day price formed under rationing carries almost no information about those questions.

The channel that does reach instruments outside China is the supply chain. A humanoid is an assembly of actuators — precision reducers, servo motors, encoders, force sensors — and that component base is concentrated in Japanese and Chinese suppliers, with the Japanese incumbents holding the high-precision end. If humanoid volume compounds, it lands first as an industrial-export and capital-goods story, which is a yen story long before it is anything else, and it lands there through order books rather than through headlines. The policy dimension is real too: hardware of this kind moves inside the same tariff and export-control architecture we traced in the Section 232 drone piece. And the discount rate that repriced the whole complex on debut day is a dollar-rates story.

Note what is absent from that list: a direct link to the eight major currencies. Unitree lists in Shanghai, trades in renminbi, and does not sit in any index the majors are scored against. Forcing an FX conclusion out of it would be worse than not drawing one. The method behind that discipline is on our about page.

What would change the picture

Three observables, in order of how soon they arrive.

The first is the end of the no-limit window after five sessions, which changes the regime the price trades in rather than the company. The second is supply: further robotics listings, and eventually the expiry of restrictions on stock not sold in the offering, both work against the scarcity that set the debut price. Scarcity premia decay as the scarce thing stops being scarce, and that is a mechanical process, not a sentiment one.

The third is the only one about the business: the next reported revenue and profit line. A company valued at roughly 201 times trailing revenue is being asked a question that only its income statement can answer, and the answer arrives on a schedule.

For everyone else, the transferable lesson is the first-day gap itself. Whenever an offer price is set administratively rather than cleared, the excess demand does not disappear — it is stored, and it is released the moment a price limit is lifted. The size of the pop measures how much was stored. It does not measure what the company is worth.

Educational macro context only — not investment advice.