China drafts $295 billion AI grid with domestic chip mandate

China is drafting a five-year, 2 trillion yuan ($295 billion) plan to connect thousands of data centers into a unified national AI computing grid, with at least 80% of core technology sourced from domestic suppliers. The National Development and Reform Commission plan would put state-owned telecom operators including China Mobile and China Telecom at the center of the buildout, targeted for completion by 2028.

The mandate extends a policy shift that has already restricted foreign accelerators in state-funded projects, effectively reducing the role of Nvidia, AMD, and Intel in China’s largest planned computing procurement. Nvidia reported $19.7 billion in China revenue in its fiscal year ending January 2026, but later disclosed no Data Center Hopper product shipments to China for the quarter ending in April 2026.

Huawei is positioned as the leading domestic beneficiary through its Ascend chips, alongside suppliers including Alibaba, Shanghai Biren Technology, and Moore Threads. Significant bottlenecks remain: SMIC’s advanced production is limited to an N+2 process roughly equivalent to 7-nanometer technology, high-bandwidth memory supply is constrained, and data-center power demand is projected to rise by 300 to 500 billion kilowatt-hours between 2026 and 2030.

Legal and operational risks also shape the grid’s outlook. China’s National Intelligence Law (2017), Cybersecurity Law (2016, amended January 2026), and Data Security Law (2021) create potential government access obligations for operators, while Huawei’s software ecosystem still trails Nvidia’s CUDA platform in developer adoption and production maturity.