Unitree's Slump Leaves 144x Sales on 5,500 Robots • The Weighted Average

Unitree, China’s best-known humanoid robot maker, has given back about 45% of its value since peaking at 1,100 yuan intraday days after its 19 August Shanghai debut, closing Monday at 603.08 yuan against the 845 yuan it reached at the end of its first session, and erasing more than 200 billion yuan of market capitalisation in under a week. The coverage of the round trip records the arc: opened 629% above its 150.80 yuan issue price, closed day one up 460% at a valuation near 445 billion yuan, and sat around 244 billion yuan by Monday’s close. Even after the fall, the stock trades at roughly four times its IPO price — which is the part operators should be pricing, not the crash.

Unitree gave back 45% of its debut peak in four sessions

Unitree share price in yuan, from IPO issue price to the 24 August 2026 close

Shanghai Stock Exchange prospectus coverage; CoinCentral · Aug 2026

What 244 billion yuan buys you

Divide the survivors’ valuation by the operating record and the enthusiasm is still visible. Unitree’s prospectus, summarised by the Shanghai Stock Exchange, reports 1.699 billion yuan of 2025 revenue, 590 million yuan of net profit excluding non-recurring items, a 60.13% core gross margin, and more than 5,500 humanoid robots shipped — first globally by its own count. Humanoids reached 868 million yuan, or 51.78% of revenue, up from 1.88% in 2023.

Two derived figures follow from those disclosures. At Monday’s roughly 244 billion yuan, the company trades at about 144 times 2025 revenue; at the 445 billion yuan peak it was near 262 times. And dividing 244 billion yuan by 5,500 humanoid units gives roughly 44 million yuan of market value per robot shipped last year — down from about 81 million yuan at the peak. Those are not software multiples applied to a software business. They are software multiples applied to a hardware manufacturer with a 60% gross margin and a physical assembly line.

The mix shift under those numbers is the genuinely impressive part, and it is easy to lose in the price action. Quadrupeds were 75.78% of revenue in 2023; humanoids were 1.88%. Two years later humanoids are the largest segment. A hardware company that re-based half its revenue on a new product category inside 24 months has demonstrated something real about manufacturing execution, whatever the multiple says. The IPO proceeds are pointed the same way: 2.022 billion yuan earmarked for intelligent-robot model R&D, nearly half the raise, against 1.11 billion yuan for robot bodies and 624 million yuan for a manufacturing base. Unitree is spending its listing money on the brain, not the chassis — an implicit admission that the bottleneck is software.

The near-term direction is the wrong one. Reporting on the selloff notes adjusted net profit fell 53% to about 40 million yuan in the first quarter of 2026 as costs rose, even as revenue grew. Founder Wang Xingxing said at the World Robot Conference that humanoids are not yet ready for broad factory deployment, remaining less efficient than humans at simple tasks and poor at adapting across jobs. HSBC analysts had warned before the listing that the recent rise in shipments could be hard to sustain without major AI improvements. The company guided to 1.052–1.128 billion yuan of first-half 2026 revenue, 35.6% to 45.4% growth — strong, and nowhere near 144×.

The buyer’s read, not the trader’s

For anyone evaluating humanoids as capital equipment rather than as a stock, three things follow.

First, the pricing signal is a market signal, not a capability signal. EconoTimes’ account of the slump notes Unitree’s 460% first-day gain against an average 226% first-day pop for newly listed Chinese stocks over the past three years, with nearly 9.8 million retail accounts chasing about 9.7 million shares. A tightly controlled IPO pricing regime, a 10% float and restricted short selling manufacture scarcity; memory maker CXMT surged 466% on debut last month under the same mechanics. None of that tells you whether a robot can do your work.

Second, the founder’s own guidance is the most useful datum in the file. When the vendor with the largest shipment base says the machines are not ready for general factory deployment, the correct procurement posture is a bounded pilot with a measured task, not a fleet order. Our earlier look at Tau’s humanoid cleaning pilot made the same case from the deployment side: humanoids earn budget where the task is narrow, repetitive and already instrumented.

Third, watch the unit economics, not the market cap. 44 million yuan of market value per unit shipped implies investors expect shipments to rise by orders of magnitude; a Q1 profit down 53% on rising costs implies each additional unit is currently getting more expensive to build, not less. Those two expectations cannot both hold for long. The evidence that would resolve it is boring and public: first-half results against that 35–45% guidance, and whether humanoid gross margin holds near 60% as volume scales.

The strongest bull case is duration. Robotics R&D precedes commercial demand by years, and judging a company on the quarter in which it scaled its manufacturing spend is a category error; Nomura analysts argue Unitree’s fast product cycle gives it a first-mover advantage in a sector that barely exists yet. That case is coherent. It also implies the stock should be priced on 2030 shipments, which is precisely what a 144× multiple does — and precisely what leaves no margin for the AI capability improvements the founder says are still missing.

The capital-markets pattern is familiar from today’s lead on the $13 billion contest for Hugging Face and from Lambda’s talks at eight times revenue: buyers are paying forward multiples for positions in AI supply chains, and the discipline arrives afterward. Unitree simply got the “afterward” in six trading days.