Features | Economy | East Asia

How China’s AI-Powered Robots Could Reshape the Global Order

China is moving from AI in the cloud to “embodied intelligence” with physical, real-world applications.

When Chinese humanoid robots flipped, sparred, and performed martial arts alongside human performers – including children – on national television during the Spring Festival Gala, it was easy to dismiss the “dancing robots” as mere theater. Days later, German Chancellor Friedrich Merz’s visit to Hangzhou-based Unitree Robotics reinforced a far more consequential interpretation: this was not a viral stunt, but an industrial signal. China is trying to shift the global AI race from model performance in the cloud to “embodied intelligence” in the physical economy.

The global AI debate still tilts heavily toward “AI in the cloud”: frontier models, compute concentration, software platforms, and winner-take-most dynamics. The United States holds structural advantages in this layer – deep capital markets, hyperscalers, and the gravitational pull of top research labs. But China’s strategic emphasis is increasingly moving toward “AI in steel”: AI systems embedded in machines that sense, move, and adapt in the real world.

Just as China’s early electric vehicles were underestimated until scale, supply chains, and iteration ecosystems reshaped global competition, embodied intelligence and humanoid robotics is emerging as the next arena in which AI-driven great-power competition will be decided. And China is industrializing these technologies at speed and scale.

In China, the “AI in steel” strategy frames embodied intelligence and robotics as a general-purpose capability. Humanoids are not primarily luxury gadgets; they represent a strategic bet on labor augmentation and operational resilience in settings where labor is scarce, tasks are repetitive or dangerous, and large-scale deployment generates the data needed to improve the next generation. 

As Premier Li Qiang underscored in the Government Work Report delivered at the opening of the 2026 Two Sessions, embodied intelligence – alongside quantum technology, brain-computer interfaces, and 6G – is now positioned among China’s priority future industries. Notably, this is the first time embodied intelligence has appeared as a distinct category in such a high-level policy document, marking its elevation from niche technology to strategic industrial priority. More importantly, embodied intelligence and robotics are being folded into a longer-term blueprint for AI integration across manufacturing and strategic sectors under China’s 15th Five-Year Plan (2026–2030).

Four features stand out.

First, China is building scenario data at scale. Embodied intelligence improves by doing: the more tasks robots perform in messy, real-world environments, the better the next model becomes. Chinese policy and local government programs have begun funding training and testing sites designed to generate scenario-specific data and accelerate deployment, as Reuters reported. This represents a critical difference from “cloud AI,” where training data is largely scraped, purchased, or simulated.

Second, China is industrializing the component stack. The durable moats in humanoid robotics will lie in supply chains – and the manufacturing tooling that turns lab designs into scaled production capacity. China’s advantage is not mysterious: dense industrial clusters, deep supplier networks, and manufacturing scale that enable fast iteration and cost reduction. Spillovers from EV supply chains – motors, power electronics, precision manufacturing, and quality control – are already lowering the cost of building next-generation robots.

Third, China is writing the rulebook. On March 3, Chinese authorities released the country’s first national standard system for humanoid robots and embodied intelligence, spanning foundational standards, computing, limbs and components, full-system integration, applications, safety and ethics. Standards do more than reassure consumers: they unify interfaces, testing, evaluation, and modularity across a fragmented industry, reducing coordination costs and accelerating diffusion. In a sector where safety, liability, and certification are central, standards become an industrial accelerator.

Fourth, China is compressing learning curves through policy, funding, resources mobilization and closed iteration loops. The Gala performance did not prove humanoids are “ready for prime time.” It functioned as a high-visibility stress test and a national narrative signal. The industrial system then does the incremental, unglamorous work of making the next version cheaper, safer, and more reliable – turning a demonstration into a pipeline of continuous improvement.

Why This Matters for Geoeconomic Competition

Unlike purely software-based artificial intelligence, embodied intelligence and physical robots possess actuation capabilities and physical agency. Deploying these technologies requires distinct regulations as well as safety and liability regimes. In a sense, when robots operate in the real world, safety requirements, certification pathways, and data rules should not be afterthoughts – they need to be market gatekeepers. Thus, the contest in this sector will........

© The Diplomat