Audit passed, but logic flawed.
That's the signal from Yushu Technology's (Yushu) historic IPO on Shanghai's STAR Market. 73 days from filing to approval. Fastest track record in the exchange's history. A lottery rate of 0.0181% — the lowest ever. Retail investors are frothing. The narrative is crystalline: 'A-share's first humanoid robot stock.'
But here's what the market is missing. The company's core business is still quadruped robots. Humanoid units? Unreported. The AI partnership with DeepSeek? A strategic label, not a product roadmap. The valuation? 609.93 billion RMB on an estimated H1 2026 revenue of 6-18 billion RMB. That's a price-to-sales ratio of 34-100x. For a hardware company. With no disclosed margins.
This isn't a bet on a robot. It's a bet on a narrative. And narratives, in a bear market, are the first to bleed.
Context: The Robot That Came to Market
Yushu Technology is a robotics firm based in Hangzhou, China. It builds both quadruped (dog-like) and humanoid robots. The company claims 90% core component self-sufficiency — motors, reducers, controllers. In H1 2026, it shipped 5,900 units, capturing 31% of the global market. That's a real achievement.
But the IPO pitch is built on the humanoid thesis. The 'humanoid robot first stock' tag is the hook. The strategic investors include DeepSeek (the AI lab), Tencent, Alibaba, Meituan, and state-owned energy giants like CNPC Kunlun Capital and Southern Grid. The message: this is the next Tesla, the next AI platform.
The market is buying it. The IPO raised 60.99 billion RMB. The lottery rate signals extreme demand. Media estimates suggest a first-day gain of 265-398%, making the single-lot profit 200,000-300,000 RMB.
But the data underneath is thin. Very thin.
Core: The Numbers That Don't Add Up
Let's crack the code.
First, the shipment mix. 5,900 units globally. But the analysis — based on public data — strongly suggests the vast majority are quadruped robots. Humanoid robots are still in early prototype stage. Yushu has not disclosed humanoid unit sales or average selling price. The 'humanoid robot first stock' label is a marketing construct, not a commercial reality.
Second, the valuation math. Assuming an average price of 100,000-300,000 RMB per unit (optimistic for quadruped), H1 revenue is 6-18 billion RMB. Annualized that's 12-36 billion RMB. At a market cap of 609.93 billion RMB, the forward P/S ratio is 17-51x. For a hardware company without software subscription revenue, that's extreme. Compare to Tesla's automotive P/S at ~5x.
Third, the strategic investors. DeepSeek's allocation of 933,400 shares is a tiny stake. The partnership is announced, but no milestones are shared. No model integration. No joint technical roadmap. The Yushu-DeepSeek link is a 'strategic tag' — a branding exercise, not an engineering reality.
Fourth, the lottery rate. 0.0181% — the lowest in STAR Market history. This is not a signal of long-term conviction. It's a signal of short-term IPO arbitrage. Retail investors are betting on a first-day pop, not on a five-year hold. The lock-up periods for strategic investors (12-36 months) will create a massive overhang when they expire.
The core insight: Yushu's IPO is a liquidity event for early investors, not a capital-raising for growth. The 2016 seed investor who put in 2 million RMB now holds shares worth 16.85 billion RMB — an 840x return. That's the story. Not the robot.
Contrarian: The Blind Spot Everyone Is Ignoring
The market is obsessing over the hardware. 90% self-sufficiency. 31% market share. But the next competitive frontier in robotics is not hardware. It's the AI stack. The 'brain.'
Tesla Optimus is training its neural networks on real-world data from Tesla's factory fleet. Figure AI is backed by OpenAI and has access to their vision-language models. Yushu? It has a partnership with DeepSeek. But DeepSeek is not exclusive. And there is no evidence of a data flywheel — Yushu's robots are not deployed in large-scale, diverse environments that generate the training data needed for generalist robot intelligence.
The '90% self-sufficiency' claim is also misleading. It counts components by type, not by cost. The high-value items — chips, LiDAR, precision sensors — are likely still imported. In a decoupling environment, that supply chain risk is real.
And the ethical and safety blind spot is deafening. 5,900 units are already in the real world. Human-robot interaction at scale. Yet the article contains zero information about safety certifications (ISO 13482, CE, UL), AI alignment testing, or data privacy protocols. If a single robot causes a serious accident, the liability will cascade. The company has no disclosed framework for responsibility attribution.
The contrarian view: Yushu is a hardware company riding an AI narrative. Its moat is manufacturing, not intelligence. In a market where AI models are commoditizing, hardware margins will compress. The valuation is pricing in a future where Yushu becomes the Android of robotics. But Android succeeded because of open ecosystem and developer buy-in. Yushu's hardware is closed. The contrast is stark.
Takeaway: The Next Watch
Forget the first-day pop. Watch the first quarterly report.
What matters: - Humanoid robot unit sales and ASP. - Gross margin. - Revenue breakdown: hardware vs. software/service. - DeepSeek integration milestones — actual product demos, not press releases. - Lock-up expiry dates for strategic investors.
The market is pricing Yushu as a winner in the humanoid race. But the race hasn't started. The data shows a company still proving its business model. The IPO is a fast-track ticket to a high-risk zone.
A fork is coming. Volatility is imminent. The only question is which direction the price breaks.