H200 sales and geopolitics

- U.S. officials said Nvidia had not yet sold any H200 AI chips to Chinese companies as of this week. - The shortfall reflects permission and policy frictions that make cross-border chip transfers difficult right now. - That underlines compute availability as a geopolitical variable, complicating capacity planning and cloud-cost forecasting for AI projects (reuters.com).

U.S. officials said this week Nvidia still had not sold any H200 artificial-intelligence chips to Chinese companies, months after Washington opened a licensing path. (reuters.com) Commerce Secretary Howard Lutnick made the comment on April 22, 2026. The Commerce Department’s Bureau of Industry and Security said on January 13, 2026 that H200 export applications for China would be reviewed case by case if security conditions were met. (reuters.com) (bis.gov) The H200 is one of Nvidia’s Hopper data-center chips for training and running large artificial-intelligence models. Nvidia says it is the first of its graphics processors to use HBM3e memory and that it carries 141 gigabytes of memory with 4.8 terabytes per second of bandwidth. (nvidia.com 1) (nvidia.com 2) Washington’s chip rules did not start with the H200. The Bureau of Industry and Security tightened advanced-computing controls on China in October 2023, and Nvidia disclosed on April 9, 2025 that the U.S. government had begun requiring a license for H20 exports to China, Hong Kong and Macau. (bis.gov) (sec.gov) That H20 shift hit Nvidia’s China business before any H200 approvals arrived. Nvidia said on May 28, 2025 that the new H20 licensing requirement led to a $4.5 billion charge, while H20 sales had reached $4.6 billion in the quarter before the rule took effect; the company later said there were no H20 sales to China-based customers in the quarter ended July 27, 2025. (nvidia.com) (sec.gov) The policy now leaves Chinese buyers with permission that exists on paper but not yet in shipped hardware. That gap matters because cloud operators and model developers buy computing capacity months ahead, and the H200 sits in the class of chips used to build those plans. (bis.gov) (nvidia.com) Nvidia has argued that export curbs can push customers toward local alternatives, while U.S. officials have said the controls are aimed at military and advanced-computing uses in China. Those two positions have shaped the market since the Biden administration’s 2022 and 2023 rules and the Trump administration’s January 2026 case-by-case revision. (bis.gov 1) (bis.gov 2) (sec.gov) Nvidia’s latest public filings show how large the company has become even without a restored China flow. Nvidia reported $215.9 billion in revenue for fiscal 2026, up 65% from a year earlier, after a year in which China-specific licensing rules had already cut off H20 shipments. (investor.nvidia.com) (sec.gov) For now, the clearest fact is still the opening one: Washington created a route for H200 sales to China on January 13, 2026, and as of April 22, 2026, that route had not produced a sale. (bis.gov) (reuters.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.