China's 150 humanoid firms face reality

China’s humanoid-robot industry is having its first real adulthood test. The hype part came easy — dozens of startups, splashy demos, investor money, and a national manufacturing machine that already knows how to scale hardware fast. But the hard part is commercial use. This week that gap got harder to ignore as reporting around Unitree and AgiBot’s IPO plans collided with a blunt Morgan Stanley survey result: Chinese companies want humanoids, but most buyers are not happy with what they can actually do yet. ### What changed this week? The immediate news is not a new robot breakthrough. It is the market mood. Unitree and AgiBot — two of China’s biggest humanoid names — are preparing listings that could value them at a combined roughly $13 billion, even as the commercial case still looks shaky. That tension is why the story matters now. Capital is racing ahead of customer satisfaction. ### Why is China still ahead? Because China is very good at the parts of this business that look like manufacturing. Chinese firms shipped nearly 90% of the world’s humanoid robots in 2025, and industry tallies put the country at more than 150 humanoid companies. AgiBot alone shipped more than 5,100 units in 2025, giving it a reported 39% global share. That is a real lead — not just a PowerPoint lead. ### So what is broken? Buyer experience, basically. Morgan Stanley’s China survey found only 23% of potential enterprise buyers were satisfied with current humanoid offerings, even though 62% said they were likely to adopt within three years. That is the tell. Interest is high. Product-market fit is not. The robots can demo, walk, sort, and carry — but many still struggle on battery life, reliability, integration, and the boring service work enterprises care about most. ### Why does satisfaction matter more than output? Because robots are not smartphones. A flashy launch does not matter if the machine fails on shift three, needs constant babysitting, or cannot slot cleanly into warehouse software and factory workflows. Enterprises buy labor economics, not vibes. If a humanoid cannot stay up, stay safe, and save money, shipment growth just means more expensive pilots. That is why low satisfaction is more dangerous than low volume. ### Who is most exposed? Probably the long tail. TrendForce expects China’s humanoid output to jump 94% in 2026 and says Unitree and AgiBot could capture nearly 80% of shipments. That sounds bullish, but it also hints at consolidation. When a market scales before the use case is settled, leaders with supply chains, software stacks, and service networks tend to widen the gap fast. Smaller firms get trapped in demo mode. ### What kind of company wins from here? Not necessarily the company with the coolest robot. More likely the one that can deliver a whole system — fleet management, maintenance, uptime guarantees, workflow integration, and maybe financing. The machine is only one layer. Turns out the real product is dependable labor in a box. That favors integrated operators more than pure hardware shops. ### Does this mean the boom is fake? No — just early. China’s lead is real, and the country’s manufacturing depth gives it a serious edge if humanoids do become broadly useful. But the sector is moving from spectacle to spreadsheets. That is where a lot of hot categories cool off. The next phase is harsher. It is about whether those robots can earn their keep. The companies that solve uptime and integration will survive. The rest are about to discover that a robot market is not the same thing as a robot business.