Chinese companies that spent years building humanoid robots able to run, dance, and perform backflips are now pouring resources into a different challenge: the embodied AI that could make those machines useful across more jobs. The shift is fuelling an investment boom in embodied AI, software that allows robots to understand surroundings, learn unfamiliar tasks, and adapt without being programmed for every job. Chinese embodied AI companies raised 93.5 billion yuan, about $13 billion, across 322 deals in the first half of 2026, five times a year earlier, according to ITjuzi.
Funding is increasingly reaching specialist developers of robot intelligence alongside manufacturers. As competition drives down hardware prices, robot makers have concluded that profit will come from software rather than machines alone. Dermot McGrath, founder of Shanghai based consultancy ZenGen Labs, said the margin will not come from hardware but from the brain.
How embodied AI is measured in tests
American startups are developing software intended to teach robots different skills without extensive retraining. Generalist says its GEN-1.5 model completed around 60 percent of simple tasks after viewing a demonstration lasting between 3 and 12 seconds. China’s X Square has published research showing robots acquiring handling skills from a single human video without retraining the model for each task.
Georg Stieler, a robotics consultant, said the leading American labs are setting a demanding benchmark in how quickly robots can acquire and reuse skills, and he considers China’s strongest startups credible competitors. Chinese companies can work closely with manufacturers and industrial customers to test and improve their systems.
Startups and manufacturers hiring for embodied AI
Specialist Chinese startups X Square and Spirit AI have each reached valuations of roughly $3 billion, according to company disclosures, and Galaxea is reported at a similar level. Spirit AI co-founder Gao Yang said the brain is the weakest link in the complete robotics stack. X Square co-founder Yang Qian said the company wants robots to shift from being driven by code to being driven by common sense.
Established manufacturers are investing in embodied AI too. Hong Kong listed UBTech advertised a chief scientist position overseeing AI model development with compensation of up to 124 million yuan, about $18 million, including equity and benefits. Hangzhou based Unitree, one of the world’s largest humanoid robot makers by shipments, has about 30 researchers developing AI that predicts how objects behave, according to a person briefed by the company.
Where embodied AI training data comes from
Unlike chatbots, which learn from internet text, robots need examples of physical interactions, from gripping objects to handling flexible wires. Chinese companies collect that information by employing people to demonstrate tasks, using computer simulations, or combining both. No company is established as the clear leader in embodied AI.
Spirit AI has hired around 1,000 people to record everyday and industrial activities for embodied AI training, Gao said. Ant Group’s robotics unit Robbyant said it trained a model using 160,000 hours of material, including 50,000 hours of human guided robot movements, across 20 robot configurations from 17 brands.
DexForce uses computer generated environments to teach robots, and chief operating officer Nicole Yang said some industrial tasks can be prepared for deployment within 6 to 8 hours. Simulations cannot perfectly reproduce physical interactions, said Jerry Wang of Manycore Tech.
Chips, funding, and limits on growth
Chinese robot firms also face headwinds: shortages of advanced Nvidia chips due to United States export controls, and a squeeze on initial public offering funding amid regulatory concerns over inflated valuations and overcrowding. Zhu Mingxuan, who leads development of JoyIn Technology’s Aether robot AI, said the chip shortage will become a serious problem for China’s embodied AI industry as robots become more powerful.
Real world deployments in China leave little room for improvisation. Battery maker CATL said Spirit AI robots connect flexible wires during battery pack testing at its Zhongzhou plant, reporting insertion success above 99 percent. Robbyant said its robots handle orders in two Shanghai pharmacies, with staff checking medicines before dispatch. Zhu described an experiment in which a robot repeatedly failed to place headphones in a cupboard beyond its reach. Engineers showed it videos of how people managed the task, and the robot pushed a box towards the cupboard and tried stepping onto it.
Unitree founder Wang Xingxing said last month the company had tested OpenAI’s GPT-6 on robots with encouraging results. He said GPT-7, if released next year, could bring robotics its own “ChatGPT moment”. Wang said he would consider that threshold reached when robots completed 80 percent of tasks in 80 percent of unfamiliar environments from spoken instructions. Stieler said a breakthrough in demonstrations would not necessarily mean robots were ready for broad commercial use, and he would add years, not months, to the timeline for a customer to count on a model.