Nvidia acting CEO Colette Kress said the company would no longer sell H200 accelerators in China. That takes Nvidia out of a disputed arrangement that allowed it to supply advanced chips to the country while working around US restrictions. The news came on the company’s first earnings call since co-founder and CEO Jensen Huang left Nvidia. That same day, Nvidia announced a board committee to oversee national security policy, export compliance and international trade risks.
Nvidia ends H200 sales in China after Jensen Huang’s departure
How the supply arrangement worked
Jensen Huang and US President Donald Trump agreed on a plan to sell H200 chips in China. The chips were to pass through the US so the government could collect a 25% tariff.
The arrangement drew criticism from supporters of strict export controls, who wanted to deny Chinese labs computing power and preserve US leadership in AI. Huang, by contrast, argued that it was important for the American technology industry to compete in the Chinese market. He made the case to podcaster Dwarkesh Patel in April.
Huang’s departure, linked in part to his affair with the head of a company under investigation on suspicion of smuggling chips into China, coincided with new export rules. On Monday, the US banned shipments of advanced American chips to overseas data centers that remotely serve Chinese customers.
Impact on revenue and the market
Investment bank Jefferies estimated that the new restrictions could cut Nvidia’s data center revenue by $44 billion to $62 billion next year. Kress said demand for advanced chips, including Vera Rubin, already exceeds available supply, so the new rules would not affect the company’s overall revenue.
Over the long term, Nvidia could lose potential customers in Singapore, Malaysia, Thailand and Indonesia. Critics of the company’s reliance on Chinese demand say it accounts for up to 20% of Nvidia’s revenue. In 2025, the annual revenue of the company’s Singapore subsidiary reached $104 billion; Nvidia stopped disclosing such figures afterward.
Even before Nvidia decided to stop shipping H200s, China had largely refused to accept the chips. Jacob Feldgoise, a researcher at Georgetown University’s Center for Security and Emerging Technology, said the country’s authorities want to develop demand for domestic chips, though major consumers of AI computing power are still more interested in Nvidia’s higher-performance hardware.
Meanwhile, leading Chinese labs are finding ways to obtain Nvidia chips outside the country. All leading Chinese AI models show signs of having been trained on Nvidia Grace Blackwell accelerators, which are banned from export to China. Training may have taken place in data centers outside the country, but the new rules will also ban shipments to those facilities if they serve Chinese customers.
According to export control lawyer Dan Fisher-Owens, large data centers designed to train AI models will have to confirm that they do not work with any Chinese customers. If they cannot do so, they may have to switch to chips from other manufacturers, and Nvidia’s market share could shrink.
Dataist’s take: The new rules restrict not only direct shipments to China but also Chinese customers’ access to overseas computing capacity powered by Nvidia chips. Jefferies’ forecast and Kress’s position differ: the company does not expect an effect on overall revenue, while the bank estimates a possible multibillion-dollar drop in data center revenue.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X