"U.S. Weighs Limiting NVIDIA H200 Chip Exports to 75,000 Units per Chinese Company"

"Less Than Half of Chinese Companies' Desired Purchase Volume"
"Upcoming U.S.-China Summit Will Be a Key Turning Point"

EPA Yonhap News

The U.S. government is reportedly considering a plan to limit exports of Nvidia’s H200 artificial intelligence (AI) chip to China to 75,000 units per individual company. The upcoming U.S.-China summit, scheduled for the end of this month, is expected to serve as a key turning point.

According to Bloomberg News on March 2 (local time), citing sources, “the U.S. government’s plan to restrict H200 exports to China to 75,000 units per company is less than half of the unofficial purchase requests made by Alibaba and ByteDance to Nvidia.” The report also noted that AMD’s MI325 chip could be included in the export restrictions.

This measure comes as an additional restriction following the previous total export quota to China of 1 million units. A quota of 1 million units would be enough to build one of the world’s largest supercomputers. However, 75,000 units are only enough for a single company to build a medium-sized data center of about 100 megawatts (MW).

Previously, China refused to import the lower-performance H20 chip. In response, President Trump considered restricting exports of the next-generation “Blackwell” chip. However, this idea was put on hold due to opposition from senior advisers. As a result, the H200 has emerged as a sort of compromise solution.

President Trump stated in December of last year that Chinese President Xi Jinping responded positively to the H200 proposal and mentioned that Chinese regulatory authorities had instructed companies to prepare orders. The upcoming summit between the two countries is likely to mark a crucial turning point. Bloomberg reported that President Trump is hoping to reach an agreement to allow exports of the H200 to Chinese companies for non-military uses.

However, hardliners both inside and outside the administration have expressed concerns that exporting the H200 would only accelerate China’s development and commercialization of AI models while bringing no benefit to the United States. The H200 was considered the industry-standard chip for training and running AI software such as ChatGPT until the release of the Blackwell last year.

Nevertheless, Jensen Huang, CEO of Nvidia, managed to persuade President Trump. In a podcast last January, U.S. Secretary of Commerce Howard Lutnick explained that CEO Jensen Huang argued the export of AI chips would “contribute to positive economic relations with China” and is “overall positive for us.” The idea is that keeping Chinese AI companies reliant on the U.S. would prevent Huawei from establishing a global competitive foundation.

He added, “Many people do not agree with this argument,” but “that is Jensen Huang’s logic and it is the decision the President has made.”

If such restrictions are implemented, uncertainty over Nvidia’s business in China is expected to increase further. Nvidia currently has no data center-related sales in China, and even if the U.S. approves exports, there is no guarantee that the Chinese government will grant approval. Bloomberg noted that China is now faced with the challenge of balancing its domestic semiconductor industry policy with the demand from its AI companies for high-performance chips from the U.S.

This content was produced with the assistance of AI translation services.