Nvidia's reported Vera CPU push in China shows how AI infrastructure demand is shifting around export controls.

Nvidia's China strategy is moving into a different part of the server rack.

Mobile World Live, citing Reuters, reported that Nvidia has told Chinese customers its new Vera CPU for AI data centers could be available in August. The pitch comes after shipments of the company's H200 AI chip were stalled by U.S. export controls.

The shift is subtle but important. Nvidia is best known for GPUs that power large-scale training and inference, but Vera is a standalone CPU designed for agentic AI workloads and built into the broader Vera Rubin architecture.

If advanced GPUs remain difficult to sell into China, CPUs may offer a narrower route back into a market where local cloud providers still need performance, software compatibility and scale. Nvidia has said Vera can deliver meaningfully higher performance than rival processors, though real adoption will depend on workload fit and ecosystem support.

For China, the decision is not only technical. Beijing is pushing domestic alternatives while cloud companies weigh whether foreign components can still be trusted in long-term AI infrastructure plans.

For Nvidia, the stakes are financial and strategic. China was once a major growth market for AI accelerators, but restrictions and domestic substitution have changed the addressable opportunity.

The Vera push suggests the AI chip race is no longer a single contest over the fastest accelerator. It is becoming a broader competition over CPUs, networking, software stacks, supply chains and the ability to design around policy constraints.

Image source: i.ibb.co