By Stuart Kerr, Technology Correspondent, LiveAIWire
AI robotic hands have become the single most valuable component in China’s humanoid robotics industry, and one Beijing startup is proving it. LinkerBot, founded only three years ago, is now seeking a valuation of six billion dollars for a single product category: robotic hands. The company holds more than 80 percent of the global market for high-dexterity robotic hands and is scaling production from roughly 5,000 units a month toward 10,000, according to Reuters interviews with the company reported by eWeek.
The reason a hand-only company can command that kind of money comes down to a problem robotics has never fully solved. Humanoid robots can now walk, balance and even race half marathons, but almost none of them can reliably crack an egg, thread a needle or button a shirt. China has decided that whoever solves AI robotic hands first controls the most valuable chokepoint in the entire humanoid robot supply chain, and a wave of Chinese startups is racing to get there.
The scale of the bet is unusual even by the standards of China’s current AI investment boom. LinkerBot’s founder, Zhou Yong, has described building a capable hand as harder than building the rest of a humanoid robot combined, a view echoed publicly by Tesla’s Elon Musk, who has called hands the majority of the engineering difficulty in the entire machine. That difficulty is precisely why a company that makes nothing but hands can be worth billions while still selling a single unit for a few hundred dollars.
Why AI Robotic Hands Are the Hardest Problem in Robotics
A human hand can identify an object by touch, adjust its grip in milliseconds to avoid crushing something fragile, and coordinate more than twenty joints without conscious thought. Replicating that in hardware means packing extraordinary mechanical complexity into a space roughly the size and weight of, well, a hand. Engineers building robotic hands are not simply shrinking existing robotic arms. They are solving an entirely different category of problem, one where the ratio of moving parts to available space is far higher than anywhere else on a robot’s body.
The International Federation of Robotics addressed this gap directly in a 2025 position paper, noting that humanoids are well placed to automate complex tasks current robots struggle with using traditional programming, but that mass adoption as universal household helpers is unlikely in the near or medium term. The IFR’s president, Takayuki Ito, has been candid that whether humanoids reach mass adoption at all remains genuinely uncertain. That uncertainty has not slowed Chinese investment, because even if fully autonomous household robots remain years away, dexterous hands have an immediate industrial market of their own, sold as add-ons to existing robotic arms in factories that do not need or want a full humanoid.
What This Means for Anyone Watching the Robotics Race
For manufacturers, investors and policymakers outside China, the practical takeaway is that AI robotic hands, not full humanoid bodies, are where the component-level race is quietly being won or lost. It is easy to focus on flashy demonstrations of robots dancing or running marathons, but the companies capturing durable value are increasingly the ones supplying the specific, hard-to-replicate parts that every humanoid maker needs regardless of who builds the rest of the robot. That dynamic mirrors what LiveAIWire has tracked in how AI and robotics are reshaping the construction industry, where the companies capturing the productivity gains are often the specialist component and systems suppliers rather than the household names doing the marketing.
China’s advantage here is not primarily about algorithms. It is about manufacturing. The country’s electric vehicle supply chain, already capable of producing lithium-ion batteries and miniaturised motors at enormous scale, has given robotics startups a ready-made industrial base that companies in the United States and Europe have struggled to match. One Shenzhen-based founder building competing dexterous hands has said plainly that starting a hardware company in the United States was effectively impossible because of supply chain constraints he did not face at home.
The Software Problem Nobody Has Solved
Building the hardware is only half the challenge, and by most accounts, not even the harder half. Teaching a robotic hand to actually use its fingers the way a human does requires vast amounts of training data that simply does not exist yet in the way text data exists for large language models. Researchers are pursuing this through teleoperation, where a human remotely operates a robotic hand to generate training examples, and through wearable sensor gloves that capture how a human hand moves and what it feels while gripping objects of different textures and weights.
Academics studying the field describe the challenge bluntly: the engineering problem of building a dexterous hand is becoming tractable, but controlling it well enough to match human-level manipulation remains largely unsolved. That gap between hardware capability and software control is the real frontier, and it is one reason China’s push into embodied AI, the effort to give physical robots the kind of general intelligence large language models have brought to text, has become a stated national priority rather than a purely private-sector pursuit.
A National Strategy, Not Just a Startup Story
Beijing has folded humanoid robotics into its broader industrial policy, treating it as one of several sectors expected to open genuinely new markets over the next decade. The government’s interest is not abstract. China faces a shrinking and ageing workforce, and policymakers see capable robots, hands included, as a partial answer to a demographic problem that automation alone cannot fully solve. That framing connects directly to the wider question LiveAIWire has explored in China’s rapid, state-encouraged adoption of generative AI, where the same pattern repeats: a technology moves from research curiosity to national priority once policymakers conclude it addresses a structural economic problem rather than a novelty.
The ambitions of individual founders reflect that same mix of commercial and almost utopian motivation. Zhou has spoken about wanting to build a real-world version of Doraemon, the Japanese cartoon cat whose pocket produces an endless supply of gadgets, framing the ultimate goal as a future where dexterous hands are cheap and common enough that ordinary households, not just factories, can afford them. Whether that vision arrives in three years or thirteen is genuinely unknown. What is measurable right now is the money moving into the sector and the speed at which Chinese manufacturers are converting laboratory demonstrations into units shipped per month, a metric that matters more to industrial buyers than any dancing robot on a stage.
The Autonomy Question That Remains Unanswered
AI robotic hands that can manipulate the physical world with something close to human skill raise the same governance questions that have followed every other advance in autonomous systems, questions LiveAIWire has examined in the context of AI-piloted drones taking on tasks once reserved for human operators. A hand capable of assembling electronics or performing delicate manual tasks without a human in the loop is also a hand capable of displacing exactly the kind of skilled manual labour that has historically been considered safe from automation.
The IFR’s own position paper is careful to frame humanoids as complementing existing industrial robots rather than replacing the human workforce outright, but that framing reflects an industry association’s caution as much as a settled fact about where the technology is headed.
What happens next depends less on any single breakthrough than on whether the software problem, teaching machines to actually use the hands being built for them, closes at anything like the pace the hardware has. China’s manufacturers have shown they can build the parts. Whether they, or anyone else, can teach a machine to use AI robotic hands the way a human uses theirs is still an open question, and it is the one that will determine whether this becomes the decade robots finally got hands, or another chapter in robotics history where the demonstration outran the deployment.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.