Procurement records show that Northwestern Polytechnical University — a Chinese institution with close ties to the People’s Liberation Army and listed on the U.S. Commerce Department’s Entity List since 2001 — purchased two Nvidia H200 AI chips last month for 874,700 yuan (about $123,000). The procurement document is dated October 21.
According to the files, the university obtained the chips not directly from Nvidia but via a domestic supplier, Shaanxi Bosuan Network Technology. The H200 is one of Nvidia’s high-end processors designed for training and running advanced artificial intelligence systems; its export is subject to U.S. licensing, and licences for entities on the Entity List are typically denied.
What the purchases reveal
The transactions highlight how institutions with military connections in China can still access restricted U.S. AI hardware — either by buying chips through local intermediaries or by renting compute capacity on servers that contain those chips.
The documents indicate that Northwestern Polytechnical this year also gained access to H200 chips and other high-performance Nvidia AI processors through five contracts with Chinese companies that provided rented computing capacity.
Broader pattern: other universities’ purchases
The procurement records show similar acquisitions at other Chinese universities:
- Shenzhen University in September bought three H100 processors for 769,800 yuan.
- Wuhan University of Technology in June acquired a server containing an H100 for 355,000 yuan, and in a separate 344,000 yuan transaction purchased four A100 chips.
- Beihang University (Peking University of Aeronautics and Astronautics), a member of the so-called “Seven Sons of National Defense,” bought four A800 processors in April for 489,000 yuan.
These institutions are involved in defense-related research or have close ties to China’s aviation and defense industries; some are subject to U.S. restrictions.
Political context and oversight demands
The disclosures come amid a debate in Washington over how tightly to restrict sales of advanced AI chips to China. Nvidia CEO Jensen Huang has lobbied to relax limits, arguing that continued sales would preserve U.S. technological leadership and keep Chinese buyers dependent on American solutions. Some U.S. lawmakers counter that access to the most powerful Nvidia chips could accelerate China’s development of AI systems for weapons design, intelligence analysis and military modernization.
Members of Congress from both parties have urged stricter controls, including chip-tracking requirements and greater congressional oversight of exports. Three Republican committee chairmen have requested information from Jensen Huang by November 18 about alleged ties to a purported Chinese intelligence operative; reports claim Huang had an intimate relationship with the alleged operative.
Conclusion
The procurement records do not indicate any legal action or indictment tied to the cited institutions or suppliers, but they underscore potential gaps in U.S. export controls and the role of domestic Chinese suppliers and compute-rental arrangements in making restricted American AI hardware available to institutions with military links. The episode feeds into a wider policy debate over technology controls, national security and the regulation of global supply chains.
