Could China win the AI race through robotics?

The next phase of artificial intelligence development is exposing a growing divergence in the technological priorities of the United States and China. While American companies have concentrated enormous resources on developing increasingly powerful AI models, China is building the industrial infrastructure needed to put artificial intelligence to work in the physical economy.

At the center of Beijing's ambitions is “embodied AI,” which integrates artificial intelligence into physical machines such as robots, allowing the devices to perceive their surroundings, interpret information, and perform tasks with greater autonomy.

China's starting position is formidable. Chinese factories installed 354,000 industrial robots in 2025, accounting for 59 percent of new installations worldwide, according to the International Federation of Robotics’ World Robotics 2026 report. The United States installed fewer than 38,500.

These figures demonstrate China’s industrial capacity, not necessarily its lead in autonomous robots. Most industrial robots remain fixed machines performing narrow, repetitive tasks such as welding automobile frames, assembling electronic components, or transferring materials between production lines. Yet this infrastructure gives China an advantage as AI advances from generating information to performing physical tasks.

From language to action

Large language models learn patterns from text, images and other digital information. Embodied AI connects this intelligence to machines that interact with physical environments and learn from their actions.

A December 2025 report from Georgetown University's Center for Security and Emerging Technology (CSET), China's Embodied AI, finds that some Chinese researchers view robotics as a potential route toward artificial general intelligence, rather than merely a commercial application of existing AI models.

Consider an automotive assembly line. A conventional robot repeatedly welds vehicle frames according to predetermined instructions. An embodied system could distinguish between a sedan and SUV, adjust its movements for different components and correct misplaced parts without manual reprogramming. More advanced systems could move between workstations, coordinate with other robots and improve through experience.

Beijing has put government policy behind the idea. Embodied intelligence entered China's national government work report in 2025, while municipal governments have established research centers, industrial parks and data-collection facilities. Beijing's 2025 to 2027 action plan called for roughly 10,000 embodied intelligent robots and an industry cluster worth 100 billion yuan, or about $13.8 billion at the conversion used by CSET.

China's manufacturing advantage

China's advantage extends beyond AI benchmarks. Its electric vehicle and electronics industries have established supplier networks for batteries, motors and other robotic components. The April 2026 report Embodied AI, published by the Mercator Institute for China Studies (MERICS), estimates that China controls 63 percent of key companies supplying components for humanoid robots, machines designed to resemble the human body. Chinese companies also supplied 57 percent of China's industrial robot market in 2024, overtaking foreign manufacturers for the first time.

Scale could affect AI development itself. Intelligent robots need enormous quantities of information about movement, objects and physical environments. In Shanghai, Chinese robotics company AgiBot has established a large-scale data-collection facility where more than 100 human-operated robots perform physical tasks to generate training data. Its AgiBot World dataset captures how robots grasp objects, coordinate movements and respond to physical conditions.

The crossover is already visible. Chinese sensor manufacturer Hesai launched Kosmo in October, a device combining cameras and laser sensors to help AI-powered robots map and interpret their surroundings.

China's industrial scale therefore offers a potential advantage at several stages of development. Its manufacturing networks can lower production costs, its factories offer testing environments, and deployment generates data to improve robotic capabilities.

The intelligence gap

Manufacturing more robots, however, is different from producing robots capable of reliably navigating the physical world.

China produced about 12,800 humanoid robots in 2025, roughly 90 percent of global production, according to the April 2026 MERICS report. Yet this was tiny beside its production of 556,000 conventional industrial robots that year. Humanoids remain concentrated in research, training and limited industrial trials, and MERICS found that Chinese models still struggle with precision and dexterity.

Important dependencies also remain. Chinese robotics companies continue to rely heavily on Nvidia's chips and software ecosystem, while European and Japanese manufacturers retain commanding positions in some specialized mechanical parts. German and Japanese companies supply 90 percent of the ball screws used to control precise robotic movements, according to figures cited by MERICS.

The technological challenge is equally formidable. Robots that perform reliably in demonstrations may fail when lighting changes, unfamiliar objects appear or humans behave unpredictably. Transferring simulated capabilities into real environments remains difficult. CSET cautions that China's momentum guarantees neither successful deployment nor progress toward general intelligence.

A different technological contest

The stakes reach well outside Washington and Beijing. Japan and Germany remain critical suppliers of specialized robotic components, South Korea is a major robotics and semiconductor producer, and European manufacturers face the possibility that China's experience in electric vehicles could be repeated in another advanced industry.

The World Robotics 2026 report counted a record 5 million industrial robots operating in factories globally in 2025, more than twice the number seven years earlier. More than 600,000 were installed during the year alone. As artificial intelligence becomes integrated into those machines, the economic consequences could spread from factories and warehouses into logistics, agriculture, services, and eventually defense.

The competition extends beyond American software and Chinese hardware, testing different approaches to technological development. While American companies prioritize advanced AI models, China is seeking to integrate them into its industrial ecosystem. The competition will test whether advances in artificial intelligence can be converted into industrial capabilities, and whether China's manufacturing strength can compensate for its remaining technological dependencies.