The U.S. is preparing to allow Nvidia to export its H200 AI accelerators to China, marking a shift toward controlled access rather than a full ban. The H200 is intentionally constrained for compliance and trails Nvidia’s most advanced chips by roughly 18 months, but it still offers meaningful AI training and inference capabilities.
Officials appear to be carving out a middle path: allowing limited semiconductor sales to maintain commercial ties while preserving the technological edge of cutting-edge U.S. chips. For Nvidia, the move could restore part of its Chinese revenue stream without violating national security constraints.
Analysts note that China will gain access to capable, though not state-of-the-art, hardware, potentially easing demand pressures in its AI sector. At the same time, the U.S. maintains strict controls on frontier chips, signaling that strategic management of semiconductor exports remains firmly in place.
