
The U.S. considers "purchase limits" on Chinese companies buying H200 chips; experts say the U.S. strategy to curb China in AI chips has failed

The U.S. government is considering setting a quantity limit on the export of NVIDIA H200 chips to Chinese companies, aiming to restrict China's development in the AI chip sector. Reports indicate that the purchase limit for each Chinese company may be 75,000 units, and similar MI325 chips will also count towards this limit. Although the total export limit to China is set at 1 million units, it is still lower than NVIDIA's initial proposal. Experts point out that the U.S. restrictions may lead Chinese companies to abandon the purchase of H200 chips, further complicating NVIDIA's return to the Chinese market
The U.S. government is repeatedly simulating the export policy of NVIDIA's H200 chip to China, attempting to achieve a balance where advanced AI chips are not used for military purposes in China while making it difficult for China to reduce its dependence on the U.S. and preventing the faster development of China's independent AI chips. The latest news indicates that U.S. government officials are considering setting a single quantity limit on the H200 chips exported to Chinese companies.
According to Bloomberg, citing informed sources, the U.S. government has discussed setting a purchase limit of 75,000 units for each Chinese company buying NVIDIA's H200 chips. They also stated that the MI325 chip, which has similar functions to the H200, would also be included in this limit. The report mentioned that the current demand for H200 chips is mainly concentrated among a few Chinese tech giants, and under a single company purchase limit, these companies could collectively obtain at most a few hundred thousand chips.
Last December, U.S. President Trump announced that NVIDIA would be allowed to export H200 chips to China. Reports indicate that NVIDIA has become increasingly dissatisfied with a series of restrictions imposed by the U.S. government, believing that the export license conditions are too stringent and may even lead Chinese companies to abandon purchasing H200 chips. These restrictions include that exporting chips to China must not affect domestic demand in the U.S. and must ensure they cannot be used for military purposes in China. However, some White House officials believe that quantity limits are the key means to constrain China. Although the total limit of 1 million H200 chips for China is higher than some officials expected, it is still far below the scale initially proposed by NVIDIA. Nevertheless, if this batch of chips is deployed centrally, it could still support China in building one of the world's top supercomputers. Therefore, this has become one of the reasons for U.S. government officials to discuss setting single import limits for companies.
U.S. media believe that when and to what extent NVIDIA can return to the Chinese market is already unclear, and setting a single quantity limit for companies will complicate the situation further. NVIDIA stated last week that it still has no revenue from Chinese data centers, and even if the U.S. allows it, it is uncertain whether Beijing will permit imports.
Chinese telecommunications and internet expert Ma Jihua told the Global Times on the 3rd that hardliners in the U.S. believe that as long as they continue to contain China, its AI development will stagnate, while moderates believe that it is better to occupy the market with products and maintain generational differences to suppress China's independent development. Both factions ultimately have the same goal, but in the repeated tug-of-war, the ultimate victim may be companies like NVIDIA, which currently cannot sell even if they want to.
For several domestic companies engaged in artificial intelligence, the computing power they possess is already very strong. Since the second half of last year, many companies have become very cautious in their investments in computing power. The U.S. strategy of trying to contain China in AI chips over the past few years has already failed, a reality they must acknowledge.
Wei Shaojun, chairman of the Integrated Circuit Design Branch of the China Semiconductor Industry Association, recently told the Global Times that from the perspective of industry development, moderately introducing advanced computing power products can help accelerate the landing and transformation of AI technology in China, forcing Chinese companies to achieve rapid breakthroughs. However, regarding the export of high-end chips to China, U.S. policies are unpredictable, making it difficult for Chinese users to confirm their true intentions Wei Shaojun believes that only by unwaveringly pursuing independent innovation and continuously enhancing core technological competitiveness can one stand firm and take the initiative in this smoke-free technological game.
Global Times
Risk Warning and Disclaimer
The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk
