Unitree IPO proves real robots ship; humanoids remain early

Unitree’s STAR Market filing proves it ships real hardware, but the humanoid story isn’t there yet.

The Shanghai filing signals a rare thing in robotics: a hardware company with scale. Unitree Robotics is seeking roughly $610 million in the public market, a move that rests as much on cash flow and manufacturing leverage as on any hype about future “humanoid” breakthroughs. The company reports about $248 million in revenue for 2025 and a striking gross margin of 59.8%. That margin stands in stark contrast to what you’d expect if the business were sliding on price erosion alone, suggesting a real mix of cost control and production efficiency rather than a mere price war.

Public markets don’t value a robot purely on a cute demo reel, and Unitree’s numbers reinforce that. The average selling price (ASP) fell dramatically—from around $85,000 in 2023 to about $25,000 in the first nine months of 2025—yet gross margins widened. In other words, the company isn’t just discounting its way to higher volume; it’s simultaneously squeezing unit costs and extracting more value from its product line. The filing underscores a hardware business with a repeatable manufacturing engine, not a one-off, high-cost prototype play.

That signal matters because the broader “humanoid” narrative in robotics has long been crowded with flashy demonstrations and fragile business models. The report emphasizes that while Unitree is proving a real hardware story, the case for humanoids as a scalable, profitable product line remains nascent. The G1 humanoid family, as a reference point, is part of the company’s narrative, but the IPO’s strongest takeaway is manufacturing scalability and unit economics, not a proven path to mass-market humanoid labor.

There’s a conspicuous data gap for any buyer assessing true humanoid capability. The filing and public coverage don’t disclose key specification metrics—degree-of-freedom counts, payload capacity, power source details, runtime, or charging requirements—for the G1 or other humanoid variants. Those figures matter a lot when you’re evaluating whether a humanoid can assist in warehouses, office environments, or elder-care settings rather than simply walk in a lab. Without them, the humanoid portion of Unitree’s business remains a promising concept rather than a proven, field-ready product line.

Engineering documentation shows that the company’s core strength today is hardware execution at scale, not cinematic elec­trical gymnastics. The real question ahead is whether the humanoid platform can translate that hardware discipline into reliable, safe, and affordable end-use capabilities. Demonstration footage and glossy tech claims can’t substitute for performance in real workplaces across multiple shifts, with maintenance cycles, spare-parts logistics, and repairability in the mix.

Two practitioner-style takeaways stand out. First, the margin story accompanies scale: price pressure on the ASP alongside rising gross margins hints at a cost-down trajectory and a broader push into higher-volume, lower-margin markets that can still preserve profitability if design-for-manufacture disciplines are maintained. Second, investors should watch for disaggregated data on unit sales by product family, service revenue (if any), and post-sale support costs. Those are the levers that determine whether the hardware business stays healthy as the humanoid narrative evolves.

In the end, Unitree’s IPO data seeds credibility for hardware in robotics—but it doesn’t close the book on humanoids. If the company can sustain margin gains while continuing to grow revenue, it will force a real conversation about mass-market humanoid viability. Until then, expect investors to weigh the undeniable hardware reality against the still-fluid humanoid roadmap.

Sources

Unitree IPO shows a real hardware business, but the humanoid case is still early