Scott Singer is a fellow and Pavlo Zvenyhorodskyi is a research analyst at the Carnegie Endowment for International Peace.
A year ago, the Western world was stunned when DeepSeek, a Chinese artificial intelligence lab, released a large language model that achieved performance comparable to other publicly available frontier models. Many policymakers and analysts had assumed Chinese AI labs were far behind their American counterparts. DeepSeek demonstrated that China lagged only months behind the leading edge.
Yet it should not have been such a shock. Throughout 2024, Chinese labs were systematically gaining ground on key benchmarks, turbocharged by long-standing whole-of-country policy efforts to build a robust AI industry. Western observers simply were not paying sufficient attention to what China was openly saying and doing.
Now, the same pattern is repeating itself. China is making an equally clear push into embodied AI — hardware systems such as robots and drones that use AI to understand their environment, make autonomous decisions and learn from physical interactions rather than simply following preprogrammed instructions. While not yet commercially viable or widely deployed, China’s current efforts in embodied AI could generate an early lead that ultimately translates into substantial economic and geopolitical advantage.
AI-powered robots are among the Chinese leadership’s top tech policy priorities, featuring in several major policy documents from the last year. Most recently, the Chinese Communist Party Central Committee wrote embodied AI into its upcoming 15th five-year plan, framing it as one of the new engines of economic growth for China through the early 2030s.
China’s real estate sector has been in prolonged decline and weak domestic consumption has weighed heavily on economic growth. Against this stagnant backdrop, Beijing hopes that integrating embodied AI into various core industries could help boost productivity and ease looming labor shortages. One emerging example is the deployment of teams of humanoid robots on production lines, where they can operate around-the-clock at complex tasks that up until now could not be easily automated.
The longer-term payoff could be even more substantial. Embodied AI could help Beijing upgrade its autonomous warfare capabilities. For example, drones that operate with minimal human oversight and coordinate their actions in large swarms would offer the Chinese military substantial advantages, including greater resilience to electronic warfare.
Furthermore, China could leverage its existing manufacturing strengths to become the leading global production hub for embodied AI systems. Unlike 5G networks or solar panels, embodied AI could become foundational to directly automating physical tasks across many sectors worldwide. That could position China as a major global supplier of intelligent robots and autonomous platforms, giving it formidable geoeconomic leverage.
The stakes could be even higher. Some influential government-affiliated researchers argue that embodied AI is critical to China’s longer-term pursuit of artificial general intelligence, enabling AI systems to learn through sustained interaction with the physical environment, beyond ingesting the mounds of written data that current AI training relies on.
Directed by Beijing, local governments are allocating resources to support local embodied AI companies, opening data facilities where companies can train their robots and launching investment funds to provide subsidies for promising local embodied AI developers. The private sector is also pushing the embodied AI frontier, with Chinese companies like Agibot, UBTech and Unitree already global front-runners.
To be sure, China faces real challenges. It has limited access to advanced AI chips for training models for embodied AI, relies on Western suppliers for high-end sensors and generates substantial waste because of duplicative research and development efforts across provinces and companies. But underestimating China’s ability to overcome early obstacles is exactly what led to the DeepSeek surprise.
The United States, in contrast, has strategically neglected embodied AI. American attention remains concentrated within a few private companies with limited involvement across the broader AI policy and research ecosystem. At a minimum, the U.S. needs to recognize both China’s AI policy direction and its material progress in achieving its ambitions. That requires building deeper China AI analysis capabilities and diffusing that knowledge throughout the U.S. policy ecosystem.
Once that urgency sets in, execution will have to follow. Silicon Valley can harness its talent and capital to move embodied AI from lab to market at scale. But software alone won’t suffice. The U.S. needs manufacturing infrastructure to match. To bring back that capacity, the U.S. government will have to play a key role.
It can make targeted investments in manufacturing infrastructure, build R & D partnerships between tech hubs and industrial centers, and implement procurement policies that create early demand for American-made embodied AI systems. Working with close partners in Europe and East Asia with deep strengths in hardware components could accelerate America’s competitiveness in what may well be one of the most consequential technologies of the coming decade.
From batteries to solar manufacturing to shipbuilding, the U.S. has repeatedly missed Chinese industrial surges until too late. DeepSeek was a warning shot. If the U.S. fails to learn the right lessons, the next surprise won’t be a chatbot. It will be Chinese robots reshaping the global economy.
The post China is winning the race for embodied AI appeared first on Washington Post.
