Unitree's 53% drawdown stops at 3.4 times its IPO price

The floor of this drawdown is instructive because it sits above anything a private buyer ever paid. Unitree priced its Shanghai offering at 150.80 yuan a share, a valuation near 61 billion yuan, itself about 45% above the roughly 42 billion yuan the company had been targeting before the book opened ^[7][2], and Wednesday's close is 3.4 times that offer price ^[1]. The first day's 460% gain took the stock to about 7.3 times the offer ^[5][12], and what has come out of it since is that auction, or rather the part of the auction priced by people who could not get an allocation.

At roughly $30 billion the market pays about 125 times 2025 revenue and more than 350 times adjusted 2025 earnings ^[11], which backs out to adjusted profit near $86 million, about a third of sales ^[3]. That is a real margin on a small base. Guidance for the first half of 2026 is 1.052 billion to 1.128 billion yuan, growth of 36% to 45% ^[10]; double the midpoint, crudely, and the year is 2.18 billion yuan, about $306 million at the 7.13 yuan to the dollar implied by Wednesday's close, or 98 times ^[7][8]. Three more years of 40% compounding takes revenue to 4.66 billion yuan and the same market value to 46 times ^[9].

The composition of that revenue is the harder question. The Wall Street Journal puts industrial applications at less than 10% of 2025 revenue, so under 170 million yuan ^[14][10], and the Financial Times counts more than 90 humanoid training centres in China, many co-funded by local governments alongside robot manufacturers, buying humanoids to generate training data by teleoperation ^[15]. Average revenue per humanoid shipped works out at no more than about 158,000 yuan ^[4], which is a research budget line rather than a labour-replacement payback, and research budgets get re-approved annually.

Set against Agility Robotics, which opened its books through the Churchill Capital Corp XI SPAC, Unitree's 2025 revenue is roughly 142 times larger, and Agility lost about 78 dollars for every dollar it sold ^[5][6]. That gap is why the securities regulator's reported preference for recurring revenue and loss reduction, per The Information, echoed separately by the Wall Street Journal ^[12][13], works as much in Unitree's favor as against it: it builds a moat around the company at least as much as it disciplines the sector.

The branch points are visible. If the training centres are a one-time restock, second-half growth decelerates below the 36% floor of the first-half guide and 125 times trailing revenue does not survive it. Should they instead seed repeatable industrial orders, the more than 40% of 2025 revenue booked overseas ^[14] argues the demand is not purely a domestic subsidy artefact, and 46 times in 2028 is arguable for a maker that, unlike many Western competitors, is not pre-revenue ^[18]. A listing bar that holds keeps comparables scarce either way. The view here is that the repricing removed the debut premium and left the composition question untouched, so the number deciding the next 50% in either direction is the industrial share of revenue. An industrial share climbing well past 10% in the 2026 accounts, with growth intact, would break that read.