The U.S. is closing a loophole that let Chinese firms buy Nvidia AI chips abroad - qz.com

The U.S. Department of Commerce's Bureau of Industry and Security (BIS) recently clarified that export licenses are required for advanced computing chips sold to entities headquartered in China, regar... Read original →

The U.S. Commerce Department’s rule shift redefines export control around ultimate beneficial ownership, not shipment destination—effectively blocking Chinese AI firms from sourcing NVIDIA’s Blackwell chips via offshore subsidiaries. Technically, this forces AI training workloads onto less efficient, lower-tier compute clusters, slowing adoption of 3nm EUV-based accelerators. Compliance burdens surge: HPE must overhaul global KYC protocols, while TSMC faces heightened scrutiny over advanced-node wafers potentially destined for Chinese shell entities. AMD may exploit this gap with marginally less powerful but license-friendly architectures. NVIDIA, meanwhile, will likely accelerate regional deployments of integrated solutions like Grace-Hopper. Within 18 months, a bifurcated AI hardware ecosystem will solidify—one high-performance but geopolitically constrained, the other ‘de-Americanized’ yet computationally capped. The real threat isn’t lost revenue; it’s the entrenchment of a widening technology gap.