The U.S. Department of Commerce imposed a new approval requirement on data-center operators buying advanced AI chips from American manufacturers, extending federal oversight from exports into domestic procurement. The rule inserts Washington directly into capacity planning for GPU clusters, potentially slowing purchases and adding compliance costs for cloud providers, model labs, and chipmakers. It also makes compute allocation an explicit industrial-policy lever, not merely a supply-chain decision. Even GPUs now need a permission slip from the Ministry of Magic 😉