BEIJING – Chinese military bodies, state-run artificial intelligence research institutes, and universities have found ways to bypass the US export ban on Nvidia chips, as revealed by a recent review of tender documents conducted by Reuters. Despite the restrictions imposed by the US, Chinese entities have managed to purchase small quantities of Nvidia semiconductors, which are essential for advancements in AI and sophisticated computer systems for military applications. These findings highlight the challenges faced by Washington in completely cutting off China’s access to advanced US chips.
While the export of high-end US chips to China is not illegal, the ban has made it difficult for Chinese firms to acquire them. However, the tender documents show that many Chinese entities, including elite universities and organizations subject to US export restrictions, have managed to purchase Nvidia chips since the ban was imposed. These chips include the A100 and H100, which were banned in September 2022, as well as the A800 and H800 chips, which were developed specifically for the Chinese market but were also banned last October.
Nvidia’s graphic processing units are widely regarded as superior to rival products for AI work, as they can efficiently process large amounts of data required for machine-learning tasks. Despite the nascent development of rival products from companies like Huawei, Chinese firms still lack good alternatives to Nvidia chips. Prior to the bans, Nvidia held a dominant 90% share of China’s AI chip market.
Among the purchasers mentioned in the review are elite universities and two entities subject to US export restrictions: the Harbin Institute of Technology and the University of Electronic Science and Technology of China. These entities have been accused of involvement in military matters or having affiliations with military bodies that go against US national interests. While the specific purposes of some purchases are unclear, it is evident that the demand for banned Nvidia chips remains.
The review also found that neither Nvidia nor its approved retailers were among the identified suppliers, raising questions about how these suppliers have managed to procure the chips. It is believed that an underground market for such chips has emerged in China, with vendors seizing excess stock or importing through locally incorporated companies in other countries.
Nvidia has stated that it complies with all applicable export control laws and expects its customers to do the same. The US Department of Commerce declined to comment on the matter. US authorities have pledged to close loopholes in export restrictions and have taken steps to limit access to chips by Chinese company units located outside China.
Experts acknowledge that it is unrealistic to expect US export restrictions to be foolproof, given the small size of chips and the ease with which they can be smuggled. The primary goal of these restrictions is to impede China’s AI development by hindering the construction of large clusters of advanced chips needed for training AI systems.
While the tenders reviewed show that Chinese entities have been able to procure banned Nvidia chips, the quantities purchased are small and insufficient for building sophisticated AI models from scratch. However, these chips can still contribute to running complex machine-learning tasks and enhancing existing AI models.
Overall, the continued demand for and access to banned Nvidia chips in China reflect the limited alternatives available to Chinese firms and the importance of these chips for AI advancements. Despite the US export ban, Chinese entities have managed to find ways to acquire these chips, posing a challenge to Washington’s efforts to control China’s access to advanced US technology.
