Article — "China Ships 97% of Humanoids as Unitree's ~$9B IPO Validates a Manufacturing-First Playbook": Article — "China Ships 97% of Humanoids as Unitree's ~$9B IPO Validates a Manufacturing-First Playbook": Nine point eight million retail orders landed on Unitree's IPO, oversubscribing the retail tranche 5,526 times—a company best known for making robots dance at a Spring Festival gala, treated like a lottery ticket by a crowd that couldn't all be filled.
Then the number the frenzy skated past: first-quarter 2026 profit, stripped of one-off items, fell 52.6 percent.
That single juxtaposition is the whole story. A record wave of retail money bidding a $9 billion valuation into a company whose margins are moving the wrong way as it spends on research and marketing.
A record wave of retail money bidding a $9 billion valuation into a company whose margins are moving the wrong way as it spends on research and marketing.
Last week we mapped the unit-economics wall using SPAC pipeline slides and shipment forecasts—the theoretical ceiling on where humanoid margins go once volume arrives. This week the abstraction has a ticker. Unitree priced at 150.8 yuan on Shanghai's STAR Market, roughly 61 billion yuan, and became mainland China's first publicly listed humanoid maker. The wall is now something you can buy shares in.
So look closely at what 9.8 million orders are actually buying.
Per its prospectus, Unitree's revenue more than quadrupled to 1.7 billion yuan in 2025, and humanoid sales hit 867.8 million yuan—enough to overtake the robot dogs that made the company famous. Humanoid revenue is now the largest line. That's the good news the crowd fixated on.
The bad news sits in the same document: rising research and marketing spend drove that 52.6 percent profit drop. This is not a company harvesting a moat. It's a company burning to defend a position in a field where more than 140 Chinese firms now ship more than 330 distinct models.
Which raises the question the valuation quietly answers. What is the scarce asset here—the body or the brain?
Unitree's IPO prices the body. The $9 billion rests on the physical machine: actuators, batteries, motors, sensors, and the ability to build them at scale for less than anyone outside China. That's the bet.
Contrast the numbers on the other side of the ocean. Bank of America reportedly values Tesla's Optimus business at about $30 billion; Morgan Stanley has reportedly floated $180 billion. Optimus has shipped essentially nothing to outside customers—Tesla reportedly started Gen-3 production in Q2 2026 for internal testing. Figure's robots run at BMW plants in the low hundreds.
So the West's premium is entirely a brain premium. It assumes intelligence—generalizable, autonomous, reliable manipulation—is the scarce thing, and that whoever owns it captures the value regardless of who welds the chassis.
The shipment counts tell you which bet has volume behind it, and it isn't close.
Global humanoid shipments hit roughly 19,100 units in the first half of 2026, up 272 percent year over year, and Chinese makers took more than 97 percent. AgiBot alone reportedly moved about 8,400—reportedly 44 percent of the entire world. Unitree followed with around 5,900. Together they reportedly shipped three-quarters of every humanoid on Earth.
Figure and Tesla, per multiple trackers, managed a few hundred apiece. Morgan Stanley's arithmetic last year reportedly had one Chinese firm shipping roughly 36 times more units than its American competitors.
That gap is too large to explain with one clever company. It's structural, and the structure has a name that has nothing to do with AI labs.
Agility Robotics co-founder Damion Shelton put it plainly: one reason humanoids "work now" is "the e-bike boom." Cheaper motors, better batteries—the same commodity components that scaled for electric bikes and toys are exactly what a humanoid needs in its joints and its chest cavity.
That's the real engine under the 97 percent. More than half of the world's top 100 robotics component suppliers are Chinese. Local costs run about 20 percent lower than foreign equivalents. A Chinese manufacturer sources most of a humanoid's bill of materials domestically, on a short supply chain, at prices a Western demo-maker importing the same parts simply cannot match.
This is why EngineAI can sell a basic model at $26,600, and why some limited runs of Chinese humanoids go under $6,000. The body has been commoditized by an adjacent consumer-electronics boom, and China owns the adjacency.
Which is precisely what makes the Unitree valuation a trap worth naming.
The market is pricing Unitree as if the body is the moat. But a commoditized supply chain is the opposite of a moat—it's a race. When 140 firms draw from the same 20-percent-cheaper parts bin, hardware margins compress toward the cost of assembly. The 52.6 percent profit drop is what that compression looks like early.
Wall Street loves a manufacturing-scale story right up until it reads the manufacturing-margin story underneath it. Commodity-hardware businesses at scale earn commodity-hardware returns. The crowd bidding 5,526x oversubscription may be buying volume leadership into a business whose economics reward volume with thinner and thinner slices.
The strategic investor list hints that Unitree's own backers understand this. DeepSeek reportedly took a 2.31 percent stake with a three-year lockup, and the partnership exists to build what Unitree calls embodied-intelligence models—robots that can think and act on their own. Tencent's vehicle and the investment arms of China National Petroleum, China Southern Power Grid, and China Telecom joined too.
Read that as a company that shipped its way to a manufacturing lead now spending—hence the profit hit—to acquire the very brain premium the West is being valued on. The body got them public. The brain is what they're buying next.
And here's the uncomfortable symmetry the two camps share, which neither has resolved.
China has won the body, but at manufacturer margins Wall Street may not actually want to own once the growth narrative cools. The West is pricing the brain at $30-to-$180 billion on machines that haven't shipped, betting intelligence is the scarce asset—an unproven claim about which input the physics of unit economics ultimately rewards.
Nobody has demonstrated which bet wins. If intelligence is the bottleneck, the AgiBots and Unitrees shipping thousands of commodity bodies are building the low-margin scaffolding for someone else's software layer. If manufacturing scale and the data it generates is the bottleneck, the $180 billion Optimus number is a fantasy priced on a factory that doesn't yet exist.
What we can say from the priced IPO is narrower and firmer than either camp's pitch.
The era of the SPAC slide is over; there are now hard shipment counts and a real market cap. Nineteen thousand units in six months. Ninety-seven percent from China. A $9 billion valuation. A 52.6 percent profit decline in the same breath as record retail demand.
Those facts don't settle the body-versus-brain question. They sharpen it. The crowd has priced the body, the body is cheap because e-bikes made it cheap, and cheap-because-commoditized is a strange foundation for a $9 billion moat.
The robots can dance. Whether the margins can is the trade nobody in that 9.8-million-order stampede has actually underwritten.