Nvidia has approval to ship some H200 AI accelerators to approved customers in China, but shipments under the U.S. licensing program represented less than 1% of data center revenue in the quarter ended July 26. The company said it had shipped only a fraction of the amount allowed, while Chinese authorities have restricted additional H200 deliveries. The H200 is an older-generation product that trails Nvidia’s current Blackwell chips, leaving the company potentially missing billions of dollars in sales to China’s AI startup sector because Blackwell shipments remain unavailable. U.S. inspection requirements also expose licensed chips to a 25% import tariff that Nvidia has been unable to pass to customers, contributing to a $400 million charge for excess inventory and purchase obligations during the first half of its fiscal year. Nvidia’s data center revenue still reached $89 billion, up 117% year over year, and its stock has risen 21% year to date. The company has excluded China data center computing revenue from its outlook because of geopolitical uncertainty. Nvidia CEO Jensen Huang is expected to attend a state dinner with President Trump and Chinese President Xi Jinping, creating a potential setting for discussion of the company’s access to China, although no agreement expanding shipments has been announced. Gary Cohn, IBM vice chair and former economic adviser to President Trump in his first term, said the United States should preserve its advantages in chips and chip manufacturing because China has significant infrastructure and cost advantages.