Nvidia China Hold
- U.S. officials say Nvidia has not sold its H200 AI chips to China amid export approval delays. - Commerce secretary Howard Lutnick told lawmakers there were zero H200 shipments to China "as of today." - Limited H200 access shifts where frontier AI can run, favouring efficient systems over GPU-scaled experiments. (reuters.com)
Nvidia has not shipped any H200 artificial intelligence chips to China, even after Washington cleared some sales in January, U.S. Commerce Secretary Howard Lutnick said on April 22. (usnews.com) Lutnick told lawmakers there had been zero sales “as of today” and said Chinese companies were struggling to get approval from Beijing to buy the chips. Reuters reported the comments from Washington on April 22. (yahoo.com) The Trump administration gave formal approval in January for China-bound H200 sales with conditions, but Reuters said shipments then stalled over disagreements on sale terms in both the United States and China. (devdiscourse.com) The H200 is a data-center chip for training and running large artificial intelligence models. Nvidia says it is the first of its graphics processors to use HBM3e memory, with 141 gigabytes of memory and 4.8 terabytes per second of bandwidth. (nvidia.com) That matters because frontier artificial intelligence systems are usually built in clusters of these chips inside cloud data centers, where more memory and faster links let companies train larger models and serve more users at once. Nvidia says the H200 is designed for generative artificial intelligence and high-performance computing workloads. (nvidia.com) The China hold sits inside a wider export-control fight over how much U.S. computing power Chinese firms can access. Reuters said China hawks in Washington have warned that advanced Nvidia chips could be used to strengthen Beijing’s military capabilities. (usnews.com) Lutnick also said on April 22 that he supported reconsidering the “affiliates rule,” a U.S. restriction that would have widened limits on advanced technology shipments to thousands of Chinese companies. Reuters reported that the rule was delayed for one year in November as part of a trade negotiation with China. (devdiscourse.com) Trade lawyers said the Bureau of Industry and Security suspended that affiliates rule on November 10, 2025, through November 9, 2026. The pause did not repeal the rule; it delayed enforcement while the broader U.S.-China trade deal remained in effect. (skadden.com) For Nvidia, the result is a market it can legally approach but still cannot fully serve. For Chinese cloud companies, it means the most advanced U.S. chips remain out of reach on April 23, 2026, despite months of negotiations. (techinasia.com)
