The 37% Wipeout

I. Two Days

On August 19, 2026, Unitree Technology (688836.SH) listed on Shanghai's STAR Market.

Offer price: RMB 150.80. The opening auction printed RMB 1,100 — up 629% — and the company touched RMB 444.9 billion in market cap. One lot (500 shares) showed a paper profit north of RMB 470,000.

Then it dumped. Intraday low: RMB 800.08. Range: 27%. It closed at RMB 845, up 460%, with a market cap of RMB 341.8 billion. Turnover: RMB 23.16 billion. Turnover rate: 85.28%.

August 20, day two. Opened down 6.5% at RMB 790. Sunk to RMB 685.01 — down 18.93% at one point. Closed at RMB 687, down 18.70%. Turnover: RMB 9.07 billion. Turnover rate: 41.98%. Market cap: RMB 277.9 billion. Below RMB 300 billion.

Two days. From RMB 1,100 at the open on day one to RMB 687 at the close on day two: a 37.55% retrace. Peak cap to close: RMB 167 billion vaporized.

II. Three Layers

Layer 1: The Free Float

Unitree's total share capital post-IPO: 404.46 million shares. Free float at listing: 30.09 million — 7.44% of the total. More than 92% locked.

Huang Lichong, President of Huisheng International Capital, told International Finance News that a 7.44% free float is low. Add the STAR Market's no-price-limit rule for the first five days, and short-term prices are set by marginal money. Small floats amplify price elasticity, not company value. The deeper problem: too few shares are available for trading. Institutions trying to build or exit large positions can move the price significantly. The total market cap looks enormous, but only a sliver of shares are actually discovering the price. Valuation detaches from fundamentals easily — and snaps back just as fast.

Layer 2: The Chip Shuffle

85.28% turnover on day one. Almost every allocated share changed hands. 41.98% on day two. The float kept rotating. Cumulative turnover over two days: over 127%. The free float was flushed clean.

PaiPaiWang data: 95 mutual funds with 5,117 products participated in offline book-building. Day-one paper profit: RMB 8.38 billion. 134 private funds with 2,833 products: RMB 1.77 billion. Combined: RMB 10.15 billion. E Fund alone: RMB 1.19 billion. Among private funds, Century Frontier, Jiukun, Ningbo Huanfang, and Yanfu each cleared over RMB 100 million.

Ninety percent of offline allocations were freely tradable from listing day — no lock-up. Those firms had paper gains. They cashed out on day two.

August 20: net main-force outflow hit RMB 1.10 billion — the largest on the STAR Market that day. The dragon-tiger board: all five top sell-side seats were institutional. Four of the top five buy-side seats were Lhasa retail desks. Institutions distributing; retail absorbing.

Day one: net main-force inflow of RMB 2.23 billion, with mega-lot orders contributing RMB 5.86 billion. Two days. Complete reversal.

Margin financing on day one: net buy of only RMB 1.56 billion — 6.7% of turnover. Most of the high-price buying came from retail chasing. Some placed orders right in the opening auction. Anyone who bought at RMB 1,100 on day one was sitting on a 37.5% loss by the close of day two.

On social boards: "When do the 1,100 buyers break even?" "Asking for a friend — bought at 1,000, can he get out?"

Layer 3: Valuation vs. Earnings

Offer price: RMB 150.80. Based on 2025 recurring net profit of RMB 591 million, the post-offer P/E was 219 times. The industry's average static P/E for the latest month: 38.6 times.

At the RMB 1,100 opening price, the dynamic P/E hit about 811 times. As of the day-two close, trailing P/E still sat at 475 times.

Revenue history: 2023, RMB 159 million; 2024, RMB 393 million; 2025, RMB 1.70 billion. But 2026 has shown a slowdown: Q1 revenue, RMB 423 million — up 68% year-over-year. Recurring net profit: RMB 40 million — down 52%. The company expects first-half recurring net profit of RMB 236–283 million, a year-over-year decline of 6% to 22%.

Nomura projects revenue of RMB 2.69 billion, RMB 5.40 billion, and RMB 13.18 billion for 2026–2028. That implies 58%, 101%, and 144% year-over-year growth for the next three years.

Ge Shoujing, analyst at Guoxin Securities: a 219-times P/E assumes the company grows far faster than the industry average over the coming years to justify the price.

III. Who Exits, Who Takes

Exiting: IPO flippers and short-term money.

Ninety-five mutual funds. One hundred thirty-four private funds. Ninety percent of their allocations unlocked on day one. Rules allowed it. RMB 10.15 billion in paper profits taken down on day one. They realized on day two. Institutional seats recorded a net RMB 129 million sell on August 20. Some still had chips; they kept selling.

Taking: retail and hot money.

Retail who bought the peak on day one sat on 18% paper losses by the next close. Four of the top five buy-side seats on the day-two dragon-tiger board were Lhasa retail desks. On August 20, CITIC Securities' Shanghai Luyang Road branch bought RMB 90 million net — but hot-money scale couldn't absorb the selling. Shareholder count had grown to 27,666, with an average holding of just 1,087 shares. Chips atomized.

Locked old shareholders: can't exit.

Meituan, Sequoia, Tencent — Pre-IPO rounds. All have 12- or 36-month lock-ups. Not yet unlocked. Strategic placement investors include DeepSeek, the National Social Security Fund, Tencent affiliates, and PetroChina Kunlun — 36-month locks. Wang Xingxing, founder, holds about 31.29% directly and indirectly. His stake dipped below RMB 100 billion.

IV. The Split

The market is split.

Bullish. Humanoid robots are a long-runway sector. Yang Delong, Chief Economist at First Seafront Fund, calls them the fourth major industrial sector after appliances, smartphones, and EVs. Nomura initiated with a "Buy" and a RMB 370 target on day one. Nomura's core case: Unitree's high in-house hardware ratio (80–90% BOM coverage) drives cost down and volume up — a "low price → scale → R&D" flywheel. In 2025, humanoid revenue hit RMB 868 million — 51.8% of total revenue, surpassing quadruped robots for the first time. Full-year shipments: about 5,716 units.

Bearish. A VC partner focused on embodied intelligence: it's far from certain that embodied intelligence will generate meaningful economic returns in the foreseeable future. Application scenarios are unclear. One investor's take: "The industry hasn't even trained a proper brain, and they've inflated a several-hundred-billion-yuan bubble and pushed it to market."

In Unitree's humanoid revenue for the first nine months of 2025, scientific research and education accounted for 73.6%. Commercial consumption: 17.4%. Industrial applications: 9.0%. The share actually used in manufacturing or inspection: 29.3% of the industrial slice. Shen Meng, Executive Director of Chanson Capital: the key challenge post-listing is penetration into manufacturing and services and productivity improvement. The embodied intelligence space has a bubble. The business闭环 isn't fully formed.

V. At the Same Hour

While the stock was tanking, Unitree founder Wang Xingxing was on the main stage at the 2026 World Robot Conference. He didn't mention the stock or market cap. He talked bottlenecks and timelines.

His words: current robot factory efficiency and generalization still lag behind humans. The company hasn't rolled out real-world operations at scale. The biggest bottleneck: embodied intelligence generalization is insufficient. The "last few centimeters" of deviation can crater success rates. The "ChatGPT moment" for embodied intelligence: optimistic case, 2–3 years; pessimistic, 5–10 years. His metric: in 80% of unfamiliar scenarios, with voice or text commands, a robot completes about 80% of tasks.

Same moment: stock crashing. On one side, the secondary market selling off. On the other, the founder talking about "another 2 to 10 years." During his speech, Unitree launched a bionic 7-axis dexterous arm — RMB 9,900, 5.5 kg, for research and education. Two days earlier, the "Superman" prototype showed a 2-meter vertical jump and 12.66 m/s top speed. The company said it won't go into production.

VI. What Comes Next

Short term: sentiment digestion and valuation correction. Jiao Bing, researcher at Gefu Fund: the day-one surge came from extreme free-float scarcity plus sentiment premium. The day-two pullback was inevitable — valuation returning from far-future dreams toward fundamental reality.

Three variables matter.

Chip clearing. In the first three trading days (August 19–21), the SSE recorded zero block trades. Hot money mostly cleared out on days 2 and 3. But the bulk of the loss zone between RMB 845 and RMB 1,100 was institutional chips from day one. How that overhang clears is still uncertain.

Target gap. Nomura's RMB 370 target maps to about RMB 150 billion in market cap. Current price: RMB 687. Still far above the target.

Commercialization pace. The "ChatGPT moment" is 2 to 10 years away. Whether the capital market grants a valuation premium during that window depends on the share of industrial-scenario revenue in quarterly filings.

VII

The two-day swing came from a tiny free float (7.44%) meeting an extreme turnover rate (85%) — a liquidity stampede. IPO institutions cashed out tens of billions on day one. Retail bought the peak. The buying pool dried up on day two. Collapse.

As Huang Lichong put it: small floats amplify price elasticity, not corporate value. Unitree is the world's largest humanoid-robot shipper by volume and one of the few profitable players. That counts. But from a RMB 61 billion offer-day cap to an intraday peak of RMB 445 billion to a day-two close of RMB 278 billion — three orders of magnitude in three days. Fundamentals don't explain that range.

When 7.44% of a company's shares are trading, the price reflects marginal-money games, not a consensus view of enterprise value. The "ChatGPT moment" for humanoids is 2 to 10 years out. The capital market has never been that patient.