A wave of Chinese startups says the next bottleneck in humanoids is not legs but fingers, and that manufacturing depth, not Silicon Valley engineering culture, will decide who solves it.
The world's most advanced humanoid robots can now dance in formation, sprint without falling, and pose with swords. What they still cannot do reliably is button a shirt, tie a shoelace, or pick up a coin from a flat surface without crushing it. That gap between polished locomotion and clumsy manipulation is the defining problem in robotics right now, and a wave of Chinese startups believes it has the manufacturing muscle to close it first.
The inflection point, in the telling of Chinese robotics executives, was a 2025 New Year's Eve. Unitree, a Hangzhou-based humanoid maker, dispatched a troupe of its H1 robots to dance alongside human performers at the CCTV Spring Gala, China's most-watched broadcast, in front of hundreds of millions of viewers. The performance turned a niche engineering demo into a national mood: humanoid robots had arrived as a symbol of Chinese industrial ambition, a moment the Guardian's forthcoming long read treats as the cultural starting gun for the country's dexterous-hand push (Guardian).
What has happened since is more telling than the dance. The bottleneck in humanoid robotics has migrated from legs to fingers. Walking and balance, hard a decade ago, are now largely solved at the demo level. What remains stubbornly unsolvable is the hand: a five-fingered, tendon-driven mechanism with enough tactile resolution to fold laundry, hold an egg, or unscrew a jar lid. The robots can move through the world. They cannot yet manipulate it.
A cohort of Chinese startups is betting that the bottleneck has finally become a market. Reuters reported in May 2026 that Linkerbot, a Beijing-based robotic hand specialist, is reportedly targeting a roughly $6 billion valuation in a new funding round, a figure echoed by trade press. Linkerbot sits at the center of a broader pattern: rather than build full-stack humanoids, a generation of Chinese vendors is positioning itself as a dexterous-hand component layer for the rest of the industry (industry context).
The bet has three structural foundations, none of them unique to China on its own. The first is manufacturing depth. Chinese suppliers already make the reducers, harmonic drives, and BLDC motors that go into a humanoid arm. A domestic hand maker can iterate on a design and have a new prototype in physical form within weeks rather than months. The second is supply-chain integration: the same factory cluster that supplies the actuators can supply the tactile sensors, the tendon sheathing, and the housing. The third is state-level coordination. Beijing has named embodied AI a strategic frontier, and Qiushi, the Communist Party's main theoretical journal, has framed the sector as a trillion-yuan opportunity (industry context).
That is the constructive case. The skeptical case is at least as strong. The International Federation of Robotics, the industry's main trade body, has published a position paper that reads in part as a warning shot: humanoid robots remain a vision rather than a market, with general-purpose multipurpose humanoids "far off" and most working deployments still confined to fenced industrial cells (IFR). BBC reporting has documented the same gap from the other direction, noting that international robotics firms broadly still struggle with hand dexterity even after years of effort. A robot that can fold a towel on a flat table under controlled lighting is not a robot that can help an elderly person out of a chair.
Tesla has called the hand the hardest problem in robotics and is engineering Optimus on the assumption that the bottleneck is solvable in software and custom actuators. The Chinese wager is that it is solvable primarily in manufacturing. Both bets may be partly right. Neither has yet produced a hand that works reliably outside a demo, a disconnect the Guardian's reporting makes vivid (Guardian).
China's working-age population is shrinking, and the country is aging faster than its social safety net was built for. If reliable humanoid hands arrive in this decade, the first major deployments will probably be in eldercare, light factory work, and home assistance, not in the household robot butler of consumer marketing. Whichever country's manufacturers can ship dexterous hands at scale, and at a price a service-sector employer can absorb, will set the terms of that transition.
Linkerbot has not commented publicly on the timing of its next round. The IFR's next humanoid market update is due in late 2026, and will be the first real benchmark for whether the Chinese cohort's manufacturing bet is translating into shipped product or only into headlines.