NVIDIA's acting CEO, Colette Kress ( Colette Kress ), stated on Tuesday that the company will no longer sell H200 chips to China, ending its involvement in a controversial loophole in the US ban on advanced chip sales to China.
The issue of China obtaining leading AI accelerators was an important topic during NVIDIA's first earnings call since the resignation of Jensen Huang, its CEO and co-founder, last week. On the same day, NVIDIA announced that it would establish a committee at the board of directors level to oversee national security policies, export control compliance, and global trade risks.
The arrangement for NVIDIA to sell H200 chips to China was facilitated by Jensen Huang and U.S. President Donald Trump. This arrangement requires that the chips be shipped through the United States, allowing the government to collect a 25% tariff.
This approach has angered the hardliners against China who wish to deprive Chinese laboratories of their computing power and maintain the United States' leading position in the field of AI; however, Jensen Huang has long argued that it is more important for the United States to compete in the Chinese market. In April, he stated in a podcast hosted by Dwarkesh Patel: "To give up that market is detrimental to our country's technology industry."
Jensen Huang's resignation, which is partly due to an extramarital affair with the CEO of a company that is under investigation for smuggling chips to China, coincides with new regulations introduced by the United States on Monday that prohibit overseas data centers providing remote services to Chinese customers from being supplied with advanced American chips.
Investment bank Jefferies ( Jefferies ) expects that these regulations could cut NVIDIA's data center revenue by $44 billion to $62 billion next year. In response to the new export controls, Cressey stated that the backlog of demand for advanced chips such as Vera Rubin means that these regulations will not affect the company's revenue.
In the long run, NVIDIA may lose potential customers in Singapore, Malaysia, Thailand, and Indonesia. Critics argue that NVIDIA is highly dependent on China, with up to 20% of its revenue coming from demand related to China – NVIDIA's annual revenue from its subsidiary in Singapore reached $104 billion in 2025, after which the company stopped disclosing this data – but this situation is clearly going to change.
Even before NVIDIA decided to stop shipping H200, China largely refused to accept these chips. Jacob Feldgois, a researcher at the Center for Security and Emerging Technologies at Georgetown University, stated that the Chinese government prefers to foster the demand for its own domestic chip industry, even though large AI computing power users still desire more powerful NVIDIA devices.
Meanwhile, China's cutting-edge laboratories are seeking other sources to obtain NVIDIA chips. All of China's leading AI models show signs of having been trained using NVIDIA Grace Blackwell chips, which are prohibited from being exported to China.
These trainings may take place in data centers outside of China, but according to the new regulations, sales to these facilities will be prohibited. Export control lawyer Dan Fisher-Owens ( Dan Fisher - Owens ) told TechCrunch that if large data centers wish to purchase NVIDIA chips for training AI models, they must prove that they are not providing services to any Chinese customers.
If they are unable to provide this proof, the new regulations may force them to turn to other chip options, further eroding NVIDIA's market share.











