China wants to spend $295 billion on AI data centers, and the plan explicitly locks out Nvidia.
Beijing is drafting a five-year investment blueprint worth roughly 2 trillion yuan to build AI computing infrastructure at national scale. The central design constraint is domestic hardware: American chips, including the Nvidia GPUs that dominate AI training globally, are specifically excluded. The plan fits a trajectory that began in earnest when the US restricted sales of Nvidia's most advanced processors to Chinese buyers in 2022 — a squeeze Beijing has been preparing to work around ever since.
The significance here isn't just the dollar figure. The plan reflects a strategic bet that export controls are permanent rather than negotiable, and that China needs to build its AI industrial base on whatever chips its own manufacturers can supply. That's a harder path than it sounds: current domestic alternatives trail Nvidia's top-end hardware on AI training workloads, and raw capital spending doesn't automatically close a silicon performance gap.
The blueprint is still in draft, which is worth noting. China has announced ambitious technology targets before, and the distance between a government plan and deployed infrastructure can be considerable.