After Unitree's IPO, what is Embodied Artificial Intelligence actually worth?

After pouring hundreds of billions of yuan into the primary market, how to price embodied AI companies has become one of the venture capital world's biggest headaches in 2026.

Unitree's IPO provided the humanoid robotics industry with its first genuine valuation benchmark — within five trading days, Unitree's market cap nearly halved from a peak of 440 billion yuan. Before this, the global market lacked a listed company focused primarily on humanoid robot manufacturing with actual profits and order records on the books.

Three points merit attention.

I. The Hundred-Billion-Yuan Divide

Even on Unitree alone, CITIC Securities assigns a fair market cap of 50.6–55.9 billion yuan over the next 6–12 months; CCB International prices it at 269 yuan per share based on 32x 2026 price-to-sales, implying a market cap of 109 billion yuan; Nomura sets a target price of 370 yuan using 25x 2027 price-to-sales, corresponding to roughly 149.6 billion yuan. All fall below Unitree's current valuation. Unitree trades at 134x price-to-sales, roughly 7x UBTECH's 19x and 15x the 9x multiple of China's robot component suppliers.

The premium stems from two factors: first, the market treats Unitree as a technology company rather than a manufacturer, baking in expectations for AI-driven industry transformation; second, "scarcity" — for the past two and a half years, UBTECH was virtually the only investable humanoid robot OEM in the secondary market, commanding a premium despite cumulative losses exceeding 5 billion yuan over six years. But with Unitree now listed and at least 20 related companies queueing for IPO this year, scarcity is depreciating rapidly.

II. The Two-Year Payback Threshold

Compared with distant market size projections, whether humanoid robot unit costs can approach labor costs is the more practical metric. Guotai Haitong Securities calculates: assuming annual labor costs of 80,000 yuan, a robot costing 160,000 yuan or less (including maintenance) pays for itself in two years. Currently, Unitree's G1 base model retails for 85,000–99,000 yuan but cannot be deployed in industrial settings; industrial-grade humanoid robots typically sell for 300,000–500,000 yuan. The bottleneck isn't merely cost — it's capability: one report projects that in 2026, a single humanoid robot will operate at just 50% of human efficiency, not reaching 104% until 2034.

III. Three Business Models

Embodied AI companies cannot be lumped together: first, OEMs competing on hardware manufacturing and cost control, valued on shipment volumes and gross margins — their ceilings are visible but their upside constrained; second, "brain" developers monetizing through licensing, SaaS, or per-call fees, valued on the generality of their models, capable of earning operating-system-like premiums if they become industry standards; third, robot labor platforms that don't sell robots at all but charge by the hour, task, or output, prized for recurring revenue. In reality, the boundaries are blurring: OEMs are developing brains, while brain developers are binding with hardware for end-to-end optimization.

Ultimately, what the market lacks isn't analytical frameworks but methods for pricing uncertainty — a puzzle for both primary and secondary markets, and also an enormous opportunity. As one investor quipped: the companies ranked at the top today may not even crack the top ten in the end.