Beijing has set a target of nearly 10,000 eflops of AI computing capacity by 2030 in its next five-year plan for the information economy.
China’s Ministry of Industry and Information Technology posted the 15th Five-Year Plan for the information and communications sector on September 7. The document, dated August 12, aims for 9,800 eflops of intelligent computing by 2030 against a 2025 baseline of 1,590 eflops, and pairs that with 1,700 exabytes of storage, up from 540 exabytes.
The plan calls for a national computing network linking hubs, regional facilities and edge sites, with AI clusters deployed at scales of 10,000 cards and beyond 100,000 accelerators. It pushes infrastructure toward domestically produced chips, sets a power usage effectiveness target below 1.2, and funds exploratory research into space-based computing.
Cumulative information infrastructure investment over the period is projected at 3.8T yuan, and ministry figures cited in the plan put China’s intelligent compute at 2,185 eflops by mid-2026, up 177 percent year over year. An eflop equals one quintillion floating-point operations per second.
The plan lands as Chinese cloud providers start renting AI compute by the token and Washington keeps tightening export rules on the advanced chips those networks depend on.
