China has mandated that publicly owned data centres source more than fifty per cent of their computing chips from domestic producers, reflecting Beijing’s intensified efforts to reduce reliance on foreign technology amid tighter US export controls.

The policy draws from guidelines introduced in Shanghai in March last year, which required at least fifty per cent domestic adoption at the city’s intelligent computing centres by 2025—supporting the semiconductor sector across the country.

The measure has now been extended nationwide and is intended to bolster local chipmakers, including those manufacturing AI accelerators such as Huawei’s Ascend series.

Since two thousand twenty-three, China has announced more than five hundred new data centre projects in regions including Inner Mongolia and Guangdong, as it pushes forward with infrastructure expansion to meet growing AI demands.

The shift comes amid challenges tied to integrating mixed chip architectures: data centres combining domestic and foreign chips require distinct software ecosystems, complicating deployment when models trained on one platform must be adapted to run on another.

China’s AI Chip Mandate Could Reshape the Market

China’s requirement that publicly owned data centres use at least half domestic chips is more than just a defensive response to U.S. export controls — it is a strategic bet on technological sovereignty. The policy, which grew from Shanghai’s municipal guidelines, has now become national law, forcing the country’s massive AI infrastructure push — over five hundred new data centres since twenty-twenty-three — to run increasingly on homegrown processors.

The immediate challenge is technical. Most AI models in China are trained on Nvidia’s ecosystem, and shifting to Huawei’s Ascend chips or other local alternatives demands complex software adaptation. For smaller data centres without elite engineering teams, the process is slow and costly. Performance gaps remain, particularly in training advanced AI models where Nvidia’s H100 series still dominates.

Yet the long-term dynamics favour China’s strategy. By pushing through short-term inefficiencies, Beijing is laying the foundation for scale, funding, and real-world deployment of local chips. Companies like Huawei and startups such as SiliconFlow are already proving that Ascend-based architectures can run large models more efficiently than Nvidia’s sanctioned hardware. If these efforts continue, Nvidia’s current grip on the Chinese market could loosen.

For Chinese startups, this mandate creates a new landscape: more affordable domestic chips, preferential government contracts, and less vulnerability to U.S. policy shifts. While the transition may slow near-term AI progress, it accelerates the process of building an independent ecosystem. Over time, this could democratize AI development inside China, cutting costs and reducing reliance on foreign suppliers.