China Targets 9,800 Eflops of AI Capacity by 2030, Backing 100,000-Card Clusters
Updated
Updated · South China Morning Post · Sep 8
China Targets 9,800 Eflops of AI Capacity by 2030, Backing 100,000-Card Clusters
2 articles · Updated · South China Morning Post · Sep 8
Summary
9,800 eflops is China’s 2030 target for intelligent computing capacity under a new five-year MIIT plan, alongside 3.8 trillion yuan in cumulative information-infrastructure investment from 2026 to 2030.
100,000 accelerator-card clusters are part of the buildout, with the plan also calling for orderly deployment of 10,000-card facilities, application-specific inference capacity and better support for domestic chips.
2,185 eflops of capacity were already in place at end-June—up 177% from a year earlier—and China had built 52 intelligent computing facilities with more than 10,000 accelerator cards.
2,450 eflops by end-July shows the expansion is already accelerating, but China still needs to more than quadruple June capacity to hit the 2030 goal.
The push extends the 2022 East Data, West Computing project, which routes power-hungry workloads to western regions through eight national hubs, 10 data-center clusters and three power-computing coordination zones.