China wants to solve the hardest problem in robotics – making hands

A cluster of Chinese startups is racing to solve what engineers call the hardest problem in robotics: building fully dexterous hands. The effort has pushed the country’s dextrous-hand industry past 50 billion yuan ($7.4 billion) in 2025, up from 13 billion yuan a year earlier, according to Chinese media reports [1]. China has registered more than 1 million robotic companies, with registrations in 2025 climbing 40 percent over the previous year [1]. The surge follows a government push to dominate “embodied AI,” a sector the Chinese Communist Party’s theoretical journal Qiushi described in May as “opening up new trillion-yuan markets” [1]. Much of that activity is concentrated in Shenzhen and other manufacturing hubs where supply chains built for the electric-vehicle industry now produce miniaturized motors, lithium-ion batteries and other components at scale [1]. LinkerBot, founded in 2023 by Huazhong University of Science and Technology graduate Zhou Yong, now makes about 5,000 hands a month and plans to double that figure as it pursues a valuation of $6 billion [1]. Zhou calls the hand “one hundred times more difficult” to build than a humanoid, noting that its dexterity is ten times that of other body parts while its volume is only one-tenth [1]. “Human hands are the most important ability of human beings,” Zhou said. “If we focus on this one point, it is easier to realise many human skills” [1]. Tesla CEO Elon Musk has acknowledged the same bottleneck. Hands, he said last year, represent the “majority of the engineering difficulty of the entire robot” [1]. Tesla’s Optimus humanoid program has drawn on deep-learning expertise from researchers such as Andrej Karpathy, who served as director of artificial intelligence and Autopilot Vision at Tesla before co-founding an AI education platform and later joining Anthropic [8]. On the hardware side, China’s advantage is clear. Pan Yunzhe, founder of Shenzhen-based Wuji Technology, returned to China after graduating in the United States in 2018 because, he said, “it was really impossible to do hardware in the United States because the supply chain problem is just so constraining” [1]. Wuji’s flagship product is a sensor-filled glove that captures both movement data and information about pressure and touch — data that is essential for training spatial-intelligence models [1]. “The two most fundamental problems in dextrous manipulation in terms of data collection are capturing how a human moves and what humans are touching or feeling,” Pan said [1]. Nathan Lepora, a professor of robotics and AI at the University of Bristol, said the hardware challenge is being solved but controlling the hands “is a whole different game … nobody knows how to do that” [1]. The software problem echoes broader AI-alignment challenges: researchers must specify complex goals and ensure systems do not find unintended shortcuts, a difficulty that grows with capability [2]. Zhou Yong said LinkerBot aims to bring prosthetic-hand prices down to $1,000 per unit, a fraction of current costs that can reach tens of thousands of dollars [1]. “We are not creating robots to replace labour,” he said. “We are creating robots so that humans can live a better and more prosperous life” [1].