E-Paper
China university buys Nvidia H200 AI chips despite US export restrictions and trade blacklist
A Chinese university on the US blacklist bought Nvidia H200 AI chips despite export restrictions, raising concerns over China's access to advanced technology.
Prefer HTon Google
A Chinese university with strong links to China's military has bought two powerful Nvidia H200 artificial intelligence (AI) chips, despite being on a US trade blacklist. The purchase raises questions about how Chinese institutions with military links are still getting access to advanced American technology, according to procurement records reviewed by Reuters.
The university, Northwestern Polytechnical University, bought the two H200 chips last month for 874,700 yuan, or around $123,000. The purchase was recorded on the university's website on October 21, according to Reuters.
Northwestern Polytechnical University is part of China's Seven Sons of National Defense, a group of universities known for their close links to the country's defence and military research programmes. The university has a history of conducting defence-related research for the Chinese government.
Deep Dive
The two chips were purchased for the university's management school. The university bought them from a Chinese company called Shaanxi Bosuan Network Technology instead of purchasing them directly from Nvidia, according to the procurement record.
The Nvidia H200 is one of the company's most powerful AI chips. It is designed to train and run advanced AI systems and is subject to US export licence requirements. Such licences are normally denied to blacklisted institutions like Northwestern Polytechnical University.
Also read: Nvidia CEO Jensen Huang's China spy probe explained: Background, allegations and key details
The university also obtained access to powerful Nvidia AI chips through rented computing services this year. It signed five contracts with Chinese companies to rent computing capacity using H200 and other Nvidia processors, according to procurement records reviewed by Reuters.
Reuters reported the purchases of the two H200 chips and the university's computing contracts for the first time. Northwestern Polytechnical University, Nvidia and the US Commerce Department's Bureau of Industry and Security did not respond to Reuters' requests for comment. Reuters also could not find contact details for Shaanxi Bosuan Network Technology.
Why is the Chinese university on the US blacklist?
Northwestern Polytechnical University is based in Xi'an, a city in northwestern China. It conducts research related to military aircraft, missiles and satellites, according to Reuters.
The US Commerce Department added the university to its trade blacklist, officially known as the Entity List, in 2001. The US government took this step over the university's alleged involvement in missile-proliferation activities and military research.
Washington tightened the restrictions in 2019. It made the export of controlled American technology to the university subject to a policy of denial in most cases. This means companies generally cannot export restricted US technology to the institution without government approval.
Reuters reviewed 26 Chinese procurement and corporate records from this year. The review found that Chinese universities and research institutions were buying US-restricted Nvidia chips or renting access to computing services powered by these processors through Chinese suppliers. Several of the institutions identified in the records had close links to China's defence industry or military research programmes.
How did the university access restricted Nvidia chips?
US President Donald Trump's administration started allowing some exports of Nvidia H200 chips to China on a case-by-case basis in January. However, this policy did not remove the separate restrictions on organisations listed on the US Entity List, including Northwestern Polytechnical University.
Apart from buying two H200 chips for its management school, Northwestern Polytechnical University's cybersecurity school also signed two contracts in January to rent computing services. Each contract covered access to servers containing eight H200 chips for seven months.
The two contracts were worth 441,700 yuan and 441,000 yuan. The rented computing power was meant to be used for fine-tuning AI models and reinforcement-learning training, according to procurement records.
The contracts were awarded to several relatively unknown Chinese technology companies. However, the procurement records provided very little information about where the Nvidia processors originally came from or where the rented servers were located.
Other Chinese universities also bought restricted Nvidia chips
Shenzhen University purchased three Nvidia H100 processors in September for 769,800 yuan. The H100 is another high-end Nvidia AI chip that the US has banned from export to China.
Also read: US midterms could fuel AI trade as data centre curbs favour Microsoft, Amazon
Wuhan University of Technology purchased a server containing an Nvidia H100 chip for 355,000 yuan in June. It also bought four Nvidia A100 chips in a separate transaction worth 344,000 yuan, according to procurement records.
Beihang University, another member of China's Seven Sons of National Defense group, purchased four Nvidia A800 processors in April for 489,000 yuan.
Two government procurement announcements published in July and September named companies selected to provide computing services using Nvidia H800 and A800 chips to the University of Chinese Academy of Sciences. Both chip models are restricted from export to China.
Restricted Nvidia chips are reaching China through other routes
Last week, Reuters reported that small quantities of Nvidia H200 chips had already reached China through indirect routes. However, Beijing had not approved large-scale imports of the processors.
The latest purchases highlight the challenges facing Washington as it tries to stop advanced American AI chips from reaching Chinese institutions with military links. Despite export restrictions, procurement records show that some Chinese universities are continuing to access these processors through domestic companies and rented computing services.
