The GPU race: AI, access, and the US–China picture
Training and running AI models takes computing capacity. GPUs can perform many calculations in parallel, making them useful for workloads that benefit from that approach.
The US and China are both important to the AI computing story. Access to advanced chips, manufacturing capacity, and policy can affect how organizations in each country build their systems. A lead in one part of the supply chain does not automatically mean a lead everywhere.
Memory is part of the picture too. AI workloads can place heavy demands on VRAM capacity and bandwidth, while gaming and other workloads have different priorities. Our articles explore those trade-offs and link to further reading.

