Nvidia CEO Jensen Huang delivers the keynote tackle throughout the Nvidia GTC 2025 at SAP Heart on March 18, 2025 in San Jose, California.
Justin Sullivan | Getty Photographs
Nvidia mentioned on Tuesday that it’s going to take a quarterly cost of about $5.5 billion tied to exporting H20 graphics processing models to China and different locations. The inventory slid 6% in prolonged buying and selling.
On April 9, the U.S. authorities advised Nvidia it could require a license to export the chips to China and a handful of different international locations, the corporate mentioned in a filing.
The disclosure is the strongest signal thus far that Nvidia’s historic progress might be slowed by growing export restrictions on its chips, which the U.S. authorities says can be utilized to create supercomputers for army makes use of.
Throughout President Biden’s administration, the U.S. restricted AI chip exports in 2022 after which up to date the foundations the next yr to stop the sale of extra superior AI processors. The H20 is an AI chip for China that was designed to adjust to U.S. export restrictions. It generated an estimated $12 billion to $15 billion in 2024.
Nvidia CEO Jensen Huang mentioned on the corporate’s final quarterly earnings name in February that income from China had dropped to half of pre-export management ranges. Huang warned that competitors in China is rising, and for the second straight yr, Nvidia listed Huawei as a competitor in its annual submitting.
China is Nvidia’s fourth-largest area by gross sales, after the U.S., Singapore, and Taiwan, in accordance with Nvidia’s annual report. 53% of its gross sales went to U.S. corporations in its fiscal yr that resulted in January.
Nvidia’s H20 chip is akin to the H100 and H200 AI chips used within the U.S. and different international locations, nevertheless it has slower interconnection speeds and bandwidth. It is based mostly a earlier era of AI structure referred to as Hopper launched in 2022. Nvidia is now specializing in promoting its present era of AI chips, referred to as Blackwell.
DeepSeek, the Chinese language firm whose aggressive AI mannequin R1 unveiled earlier this yr upended markets, used H20 chips in its analysis.
Along with the prevailing Chinese language export controls, Nvidia additionally faces new restrictions on what it could actually export beginning subsequent month, below “AI diffusion rules” first proposed by the Biden administration.
Nvidia has argued that additional controls on its chips would stifle competitors and probably even erode U.S. competitiveness in know-how. The corporate beforehand mentioned it moved a few of its operations, together with testing and distribution, out of China after the 2022 export controls.
On the firm’s convention final month, when requested about Chinese language export controls, Nvidia CEO Jensen Huang mentioned that the corporate works to adjust to the legislation, however he additionally famous that about half of the world’s AI researchers are from China, and plenty of of these work at U.S.-based AI labs.
In Tuesday’s SEC submitting, Nvidia says that the U.S. authorities mentioned on Monday that the license requirement for H20 chips could be in impact “for the indefinite future.”
Nvidia reviews fiscal first-quarter outcomes on Could 28.
WATCH: Nvidia says U.S. requires license to export H20 products to China
