DEV Community

武乐丹
武乐丹

Posted on

Unitree Priced Its IPO at ¥150.80/Share — a ¥61B Valuation Anchor for China's 'First Humanoid-Robot Stock'

Unitree Priced Its IPO at ¥150.80/Share — a ¥61B Valuation Anchor for China's 'First Humanoid-Robot Stock'

Subtitle: The pricing sheet reads like a who's who of strategic capital: China Social Security Fund, DeepSeek, Tencent, and CNPC all took placement shares. Meanwhile Zhiyuan revealed nine Huawei-heavy IPO partners, Zoox quietly became the first paid robotaxi service in Las Vegas, and the FCC ban entered its consumer-fallout phase. The embodied-AI sector just moved from hype to pricing — and the two don't mix gently.


On the evening of Aug 6, Unitree set its STAR Market IPO price at ¥150.80 per share — a 219.23x P/E (against an industry average of 38.56x) and a market capitalization of roughly ¥60.99 billion (~US$9 billion). This is the number that every future embodied-intelligence listing in China will be measured against, and it arrived with a strategic placement roster that tells you who the market's big money thinks wins this decade.

Here's what the pricing actually means, who's in the room, and why the timing — with the first batch of embodied-AI concept stocks already 'crashing' — is the most interesting part.

The pricing anchor has landed

A 219x P/E looks absurd until you remember what's being priced: not this year's revenue, but the option on a sector that Chinese policy and capital are both treating as strategic. The comparable anchor is the way CATL and BYD were priced in their early days — except those were electrifying an existing market, while humanoid robotics is being valued before a mass market exists at all.

The placement roster is the signal that matters more than the multiple:

  • China Social Security Fund — the national pension fund taking strategic placement shares is a statement about long-horizon conviction, not quarterly returns.
  • DeepSeek — an AI lab taking a humanoid-robot placement is the market's clearest signal yet that the two fields are converging: embodied models need robot hardware to collect real-world data, and robot companies need frontier-model brains.
  • Tencent and CNPC — consumer-ecosystem capital plus state energy capital in the same book. When those two sit side by side, the deal stops being a tech IPO and becomes a national industrial strategy with a ticker.

Online subscription opens Aug 10. The book was already oversubscribed at the ¥55B end during bookbuilding; at the final ¥61B it's effectively a referendum on whether the sector can hold a triple-digit multiple through the first earnings cycles.

Zhiyuan's nine partners and the Huawei question

The same day, Zhiyuan Robot disclosed its full nine-partner team for the first time on the eve of its own IPO — and six of the nine are ex-Huawei. When a humanoid startup's pre-IPO partner list is dominated by people who built Huawei's supply chain and hardware systems, it's not a coincidence; it's a statement about where the manufacturing discipline in this sector is coming from.

The detail that got the Chinese tech press talking: chief scientist Luo Jianlan has disappeared from the official website, fueling speculation ahead of the listing. Whether that's a standard pre-IPO governance shuffle or something more is unconfirmed — but in a market where the first listing's every footstep is being parsed, it's the kind of loose thread the short-sellers will pull on.

Zoox went paid — the first commercial robotaxi city

Across the Pacific, Amazon's Zoox announced it will begin charging for robotaxi rides in Las Vegas on Aug 10 — its first commercial market. No safety driver, purpose-built vehicle, paid fares. That's a genuine milestone: the autonomous ride-hailing category now has two paid models in the US (Waymo's expansion and Zoox's launch), which changes the competitive frame for everyone — including the Chinese players who have been running pilot fleets in Beijing, Shanghai, and Shenzhen.

For the embodied-AI story, the throughline is data: every paid robotaxi mile is labeled real-world driving data, and every factory robot hour is labeled manipulation data. The companies that monetize first collect the most training signal. Unitree's factory deployments and Zoox's paid miles are two sides of the same flywheel.

FCC ban: the consumer-fallout phase

The FCC's robot-vacuum ban has moved past the policy announcement into its messy consumer phase. Unitree's prospectus confirms existing models are unaffected, but also reveals US revenue peaked at around 20% of the mix — meaning the ban's ceiling on new US sales is now a known quantity for investors. For Shark, iRobot, and Eufy the collective response was immediate; for Chinese makers like Roborock, Dreame, and Ecovacs, "localized production" has shifted from a cost optimization to a compliance necessity, and the financial statements will show it within two quarters.

The throughline: pricing time and cooling-off are arriving together

The most honest framing of this week: the valuation anchor landed at ¥61B while the first batch of embodied-AI concept stocks 'crashed.' Pricing time and cooling-off are arriving together. That's not a contradiction — it's the sector maturing from narrative to numbers.

When a sector stops being valued on story and starts being valued on price, the dispersion between winners and everyone else widens fast. Unitree's placement roster and Zhiyuan's Huawei-heavy partner list are the market picking its horses. The retail investors who pile in on Aug 10 are buying at the anchor price; the strategic funds bought below it and are holding for a decade.

As a developer or investor, the question worth sitting with isn't whether humanoids are overhyped — it's which layer of the stack compounds: the foundation models (DeepSeek's bet), the hardware/supply-chain layer (Zhiyuan's bet), or the full-stack robot company (Unitree's bet). This week's pricing says the market believes all three can win — just not at the same multiple.

What's your read — is a 219x P/E on the first humanoid listing a signal that the sector is priced for perfection, or the entry price for a decade-long industrial cycle? And would you rather own the model, the hardware, or the full stack?


Based on SinoBot Daily Pulse #58 — Aug 7, 2026. Sources: ITHome, TechNode, CNBC, Caixin Global, 36Kr, Sina Finance, Reuters, The Verge.

Top comments (0)