Nvidia will stop selling its H200 chips in China, acting CEO Colette Kress said Tuesday, ending the company's participation in a contested loophole in the US ban on advanced chips. The announcement came on the company's first earnings call since CEO Jensen Huang stepped down last week, and on the same day Nvidia created a board-level committee to oversee national security policy, export-control compliance and global trade risk.
The end of the H200 arrangement
The H200 deal had been brokered between Huang and President Trump: chips would ship through the United States, letting the government collect a 25 percent tariff. China hawks attacked the pathway from the start, while Huang argued until recently that conceding the Chinese market would be a disservice to American technology, telling podcaster Dwarkesh Patel in April that abandoning it would be a disservice to the country's technology industry. In practice the arrangement was already hollowing out. Chinese buyers had largely refused H200 deliveries even before Tuesday, according to Jacob Feldgoise, a researcher at Georgetown's Center for Security and Emerging Technology, because Beijing prefers to build demand for its domestic chip industry. Chinese frontier labs, meanwhile, have been training on Nvidia Grace Blackwell systems that were never legal to ship to China, apparently at data centers abroad.
The remote-access crackdown closes in
Those offshore routes are narrowing. Commerce's Bureau of Industry and Security on Monday issued an interim final rule barring advanced US chips from foreign data centers that serve Chinese customers remotely, an extension of export rules this publication covered when it was first proposed. Under the new rule, a large training data center that wants Nvidia silicon must certify it serves no Chinese customers, export-control attorney Dan Fisher-Owens told TechCrunch. Nvidia has also warned buyers in recent months that the Remote Access Security Act, which passed the House in January and is still stuck in the Senate, could expose both the company and its customers to legal action, and it told clients it will not sell to data centers that rent more than half of their GPUs to Chinese companies. Jefferies estimated the rules could trim $44 billion to $62 billion from Nvidia's data center revenue next year, but Kress said demand backlog for upcoming systems such as Vera Rubin means the rules will not dent top-line revenue.
What Nvidia stands to lose
Critics of the company's China exposure have claimed that demand linked to the country drives as much as a fifth of revenue; Nvidia's Singapore subsidiary alone booked $104 billion in 2025 before the company stopped breaking out those figures, and analysts now see Singapore, Malaysia, Thailand and Indonesia customers as the next exposure. The political pressure behind the shift is intense: Treasury Secretary Bessent has publicly demanded a dramatic shakeup and deeper scrutiny of Nvidia's compliance culture, and Huang is scheduled to testify before the Senate Banking Committee on December 3. A company that spent two years defending its China business as a strategic asset is now dismantling it one disclosure at a time.