Abstract:
A newly disclosed procurement record shows that a Chinese university that has close ties with the People's Liberation Army and is included in the U.S. export control blacklist recently obtained Nvidia's advanced H200 artificial intelligence chip. Relevant findings have once again triggered discussions in U.S. political circles on the flow of high-end AI chips to China and the effectiveness of current export control measures.
The report stated that the university involved was Northwestern Polytechnical University. As a member of China’s “Seven National Defense Programs”, the school has long been involved in research in the fields of defense technology and aerospace, and has been included in the entity list by the U.S. Department of Commerce. Under current regulations, it is often difficult for U.S. companies to obtain approval to export relevant advanced technologies to these restricted agencies.
Public procurement documents show that Northwestern Polytechnical University purchased two Nvidia H200 artificial intelligence accelerators in October this year for a total amount of approximately 874,700 yuan, equivalent to approximately US$123,000. Records show that these chips were not sold directly through Nvidia, but purchased through a company called Shaanxi Boyuan Network Technology.
H200 is one of NVIDIA's previous generation flagship artificial intelligence chips. It has extremely strong computing power in the fields of large model training, inference computing and high-performance artificial intelligence applications. Although NVIDIA has now launched an updated Blackwell series of products, the H200 is still considered one of the most competitive AI computing platforms in the world, and it is also an advanced semiconductor product that is heavily regulated in the United States.
The report pointed out that according to U.S. export control policies, sales of such high-performance AI chips to institutions included in the entity list require special licenses, and such applications are usually rejected. Therefore, how these products eventually enter the hands of relevant universities has become the focus of attention from the outside world.
In addition to directly purchasing hardware, procurement records also show that Northwestern Polytechnical University also purchased computing services based on Nvidia's advanced AI chips this year. In recent years, as restrictions on chip exports have continued to tighten, leasing computing power, remotely accessing servers, or obtaining computing resources through third-party channels has become a common way for some organizations to bypass hardware acquisition restrictions.
The findings come as Nvidia and its chief executive, Jensen Huang, face growing political pressure in Washington. Recently, many members of the Democratic and Republican parties in the U.S. Congress have called for strengthening the supervision of the export of advanced AI chips and demanding the establishment of a more stringent tracking and supervision mechanism to prevent related technologies from being diverted to military uses.
Some U.S. lawmakers believe that advanced artificial intelligence chips can not only be used for commercial AI development, but may also be used for weapons design, intelligence analysis, military simulations, and autonomous combat system development, so exports to China should continue to be strictly restricted. Huang Renxun has publicly stated on many occasions that allowing U.S. companies to sell products to China will help maintain U.S. technology leadership and allow Chinese customers to continue to rely on the U.S. technology ecosystem.
In fact, this is not the first time that Chinese military-related institutions have been found to have obtained restricted AI hardware. In the past few years, multiple investigative reports have revealed that some Chinese universities and scientific research institutions that have been included in the U.S. restricted list have access to advanced AI computing resources through server procurement, third-party resale, or computing power leasing.
The procurement records exposed this time once again highlight the real challenges faced by the U.S. chip control system in China. Although Washington continues to expand the scope of export restrictions, as global supply chains complicate and intermediary networks develop, high-end AI hardware and related computing power may still flow to restricted end users through multiple channels.
Analysts point out that as artificial intelligence gradually becomes the core area of technological competition between China and the United States, the issue of the flow of advanced computing chips is likely to continue to become an important focus of controversy between the US government, Congress and technology companies in the future. The Northwestern Polytechnical University's purchase of H200 chips will further prompt the outside world to re-examine the implementation effects and regulatory loopholes of the existing export control system.
Comments