The key facts
- Nvidia is halting H200 sales in China — announced by interim CEO Colette Kress
- New US rules prohibit selling advanced chips to foreign data centers serving Chinese clients
- Jefferies estimates potential data center revenue losses of $44–62 billion next year
Nvidia Exits the Chinese H200 Market
Nvidia will no longer sell its H200 accelerators in China. Interim chief executive Colette Kress announced this on Tuesday, putting an end to the company's participation in a controversial arrangement that allowed advanced chips to be sold to China despite US restrictions. The statement came during the company's first earnings call after chief executive and co-founder Jensen Huang left his post last week. The same day, Nvidia announced the creation of a board-level committee that will oversee national security policy, export control compliance, and global trade risks.
The Arrangement That Will Be No More
The agreement to sell H200 chips to China was reached with the involvement of Huang and President Donald Trump: the chips were to be shipped through the US, allowing the government to collect a 25% tariff. This arrangement irritated China hawks who advocated denying Chinese labs computing power, but Huang consistently defended it. In April, he told podcaster Dwarkesh Patel that 'ceding this market to the US technology industry would be a disservice to our country.' Huang's departure came, in part, against the backdrop of a romance with the head of a company under investigation for chip smuggling into China.
New Rules and Loss Estimates
Huang's resignation coincided with new rules introduced on Monday: they prohibit supplying advanced American chips to foreign data centers that remotely serve Chinese clients. Investment bank Jefferies forecasts that these rules could cut Nvidia's data center revenue by $44–62 billion next year. Kress herself stated that due to a huge order backlog for advanced chips such as Vera Rubin, the new export restrictions will not affect the company's revenue. In the long term, Nvidia could lose potential customers in Singapore, Malaysia, Thailand, and Indonesia. Critics of the company's dependence on China argue that up to 20% of its revenue is tied to demand from that country: Nvidia's Singapore division's revenue reached $104 billion in 2025, after which the company stopped disclosing those figures.
China Was Already Rejecting the H200
Even before Nvidia's decision to halt shipments, China was largely refusing to accept the H200: the authorities prefer to stimulate demand for the products of their own chip industry, even if major AI computing users want more powerful Nvidia hardware, according to Jacob Feldgoise, a researcher at Georgetown's Center for Security and Emerging Technology. Meanwhile, leading Chinese AI labs are finding ways to access Nvidia chips abroad: all leading Chinese AI models show signs of training on Nvidia Grace Blackwell chips, which are banned from export to China. Such training could have taken place in data centers outside China, but the new rules close that channel as well. As export control lawyer Dan Fisher-Owens explained to TechCrunch, large data centers wishing to buy Nvidia chips will have to certify that they do not serve Chinese clients; without such certification, they may have to turn to alternative suppliers, which would further reduce Nvidia's market share.



