Compute

Compute

Compute is the supply story from chip to watt. Accelerators and custom silicon, fab and packaging constraints, datacenter buildouts, energy deals, cloud capacity, and the logistics that decide who can train or serve at scale belong here. Hardware and infrastructure are one continuous chain for AI, so this hub does not split them. A new GPU architecture or a nuclear PPA for a training cluster is Compute. A chip startup raising a round is Business: the news is the money. Export controls on accelerators are Policy. This page is for capacity, scarcity, and the physical layer underneath model progress.