
At least six Chinese universities placed on the US government’s Entity List have reported purchasing Nvidia processors covered by American export restrictions, raising fresh questions about how effectively Washington can prevent sensitive artificial intelligence hardware from reaching institutions considered a national security risk. A review of nearly 300 Chinese procurement records has identified transactions involving restricted Nvidia chips, including purchases reported by universities with close links to China’s military establishment.
The findings highlight a persistent challenge for US export-control authorities: restricting the sale of advanced processors is one part of the problem, but tracking hardware after it enters a complex international supply chain is considerably harder. The reported purchases were largely routed through lesser-known Chinese suppliers rather than directly through Nvidia or major authorised distributors.
However, the procurement records do not prove that every processor was delivered, establish how suppliers obtained the hardware or demonstrate that the transactions violated US export regulations. They also do not reveal the intended use of the chips. Those distinctions are important because a reported purchase, a completed delivery and an unlawful export are separate matters.
What the Investigation Found About Nvidia Chip Purchases
A CNN review of nearly 300 Chinese procurement records found that at least six universities on the US Entity List had reported purchases of Nvidia processors subject to American export restrictions. The transactions occurred over several years, with some of the most recent records involving institutions closely associated with China’s military research ecosystem.
The review examined procurement notices published by Chinese universities and other public institutions. Under China’s procurement rules, public universities and government-linked organisations are required to publish notices for certain purchases. These records can provide a window into technology acquisitions that might otherwise be difficult to track from outside the country.
Of the 90 buyers examined, nearly 80% were universities, while research institutes accounted for close to 14%. Some government agencies also appeared in the records as buyers of Nvidia hardware.
The transactions involved several generations of Nvidia processors, including the A100, A800, H100, H200 and H800. These chips are used in demanding computing workloads, particularly artificial intelligence model training and inference, scientific research and large-scale data processing.
Most of the reported purchases were made through relatively obscure Chinese suppliers. The records did not identify a complete supply chain showing where the processors originated, how they entered China or whether the final institutions received them.
Military-Linked Chinese Universities Appear in the Records
One of the most significant aspects of the findings is the involvement of universities associated with China’s defence research establishment. Two institutions belonging to the country’s so-called Seven Sons of National Defence featured in the procurement records.
The group consists of universities with close connections to China’s defence and military research sectors. Their academic work spans engineering, aerospace, materials science, electronics and other fields that can have both civilian and military applications.
Harbin Institute of Technology Reportedly Bought an A100
Harbin Institute of Technology reportedly recorded the purchase of an Nvidia A100 processor for approximately $7,000 in September. The A100 is an advanced data-centre GPU that became widely used for artificial intelligence development and high-performance computing.
The university’s appearance in the records is significant because it is part of China’s defence-linked university network. Its research capabilities and institutional connections have made it a subject of US export-control restrictions.
Northwestern Polytechnical University Recorded H200 Purchases
Northwestern Polytechnical University reportedly recorded purchases of four Nvidia H200 processors worth approximately $240,000 in total across September and October. The institution is particularly sensitive from Washington’s perspective because of its links to China’s military research programmes.
The H200 is a high-performance AI processor designed for demanding computing workloads. Its capabilities make it valuable for training and running advanced AI systems, although the procurement records do not establish what the university intended to do with the reported hardware.
Gregory Allen, chief executive of Decision Tree Research, said both Harbin Institute of Technology and Northwestern Polytechnical University had strong connections to China’s military establishment. He also noted that research into Chinese military AI capabilities would likely involve universities of this kind.
Beihang University Also Appears in Earlier Transactions
CNN’s review identified purchases by Beihang University in Beijing, another member of the Seven Sons of National Defence. The records showed more than a dozen transactions between 2022 and 2024, largely involving Nvidia A100 and A800 processors.
Beihang has been included on the US government’s Entity List since 2001. Its appearance in the procurement records illustrates how restricted technology can remain part of purchasing activity involving institutions that Washington has identified as sensitive.
The investigation also found purchases by three other universities on the Entity List. Together, the six institutions raise questions about the effectiveness of restrictions designed to prevent controlled American technology from reaching organisations that may contribute to China’s military or advanced technology capabilities.
Why the US Entity List Matters for Nvidia Chip Sales
The US Department of Commerce’s Entity List identifies foreign organisations and individuals subject to additional export restrictions. Companies generally need government authorisation to supply listed entities with goods, software or technology covered by US export-control rules.
For advanced semiconductor products, the restrictions are particularly important because high-performance GPUs can provide the computing power required to develop sophisticated AI systems. The same technology can support civilian applications, such as scientific research and commercial AI services, while also having potential military uses.
Being placed on the Entity List does not automatically mean that every product associated with an organisation is prohibited under every circumstance. Instead, the applicable restrictions depend on the product, its classification, the transaction and the relevant licensing requirements. For the institutions identified in the investigation, however, purchases involving controlled Nvidia processors raise questions about whether the necessary approvals were obtained.
The procurement notices alone cannot answer that question. They show that institutions reported purchases, but they do not provide enough information to determine whether a transaction involved an authorised exception, previously acquired inventory, an intermediary or an unlawful transfer.
How Restricted Nvidia Processors May Be Reaching Chinese Buyers
The use of lesser-known Chinese suppliers is central to the enforcement challenge. Rather than identifying Nvidia or a major authorised distributor as the direct seller, the procurement records point to smaller domestic companies acting as vendors.
That does not establish that these suppliers smuggled the processors. A supplier could potentially be reselling hardware acquired through an earlier lawful transaction, sourcing products through an intermediary or handling equipment whose original purchaser is not visible in the procurement notice.
Nevertheless, complex intermediary networks can make it harder for regulators and manufacturers to identify the ultimate destination of restricted technology. Once a processor has passed through multiple companies, transactions and jurisdictions, tracing its original source and final recipient becomes more difficult.
Experts cited in the investigation said chip-smuggling networks can rely on shell companies, shipments routed through multiple countries, changes in currency transactions and falsified paperwork. Such methods can obscure the identity of the buyer and make it harder to establish whether a shipment is intended for a prohibited organisation.
Kate Koren, a former US export-control official who is now research director at the Center for Strategic and International Studies, told CNN that once these processors are inside China, preventing their resale becomes extremely difficult.
She also cautioned that some reported A100 purchases may involve processors acquired before US restrictions took effect in 2022. At that time, A100 chips were still widely available in China. This creates an important distinction between newly exported restricted hardware and older stock that may have entered the country before the rules changed.
Why A100, A800, H100 and H200 Processors Matter
The processors identified in the procurement records belong to Nvidia’s data-centre GPU portfolio. Unlike ordinary consumer graphics cards, these products are designed to handle large-scale parallel computing workloads and are widely used in AI infrastructure.
The A100 became an important processor for training large AI models and operating high-performance computing systems. Its successor generations and related products, including the H100 and H200, offer capabilities designed for increasingly demanding AI workloads.
The A800 and H800 were developed for the Chinese market with specifications adjusted in response to earlier US restrictions. Their inclusion in the procurement records demonstrates that export-control enforcement involves more than blocking a single flagship processor. Regulators must monitor several chip generations and configurations, including products designed to comply with earlier restrictions.
The H200 has attracted particular attention because of its advanced computing capabilities. Access to processors of this class can help organisations expand their AI infrastructure, although the actual impact depends on the number of chips available, the supporting systems and the computing resources that can be assembled around them.
For universities, access to powerful GPUs can support research in areas ranging from language models and computer vision to engineering simulations. The same computing resources can also be relevant to military research, which is why US authorities treat certain institutional buyers as sensitive end users.
US Export Rules Have Changed, but Restrictions Remain
Washington has progressively tightened controls on advanced computing hardware destined for China. The restrictions are intended to limit Beijing’s access to processors that could strengthen its artificial intelligence capabilities and support military modernisation.
Sales of Nvidia A100 processors to China remain subject to US export restrictions. The controls introduced in 2022 marked a major shift in the semiconductor trade, requiring companies to pay closer attention to the technical specifications of chips and the identity of their customers.
The policy surrounding the H200 has been more complicated. The US introduced a framework allowing certain H200 sales to Chinese entities on a case-by-case basis, subject to government approval and conditions that include reporting requirements concerning end users.
Such a policy does not amount to unrestricted permission to sell the processors throughout China. Organisations on the US Entity List face separate restrictions, and sales to the blacklisted universities identified in the investigation would be unlikely to receive approval, according to the report.
This distinction matters because the existence of a licensing pathway for some Chinese customers does not remove restrictions on prohibited institutions. Export-control compliance depends on who is receiving the technology, not simply whether a particular chip model is eligible for consideration under a broader policy.
Why Monitoring the Supply Chain Is So Difficult
Export controls are most effective when authorities can identify the original exporter, the immediate buyer and the final user. However, the semiconductor supply chain can involve manufacturers, distributors, resellers, logistics companies, financial institutions and customers operating across different countries.
Every additional intermediary creates another point at which ownership and destination information can become difficult to verify. A company may appear to be purchasing equipment for a permitted commercial purpose, while the final destination or end user may be different from the one originally declared.
US regulations require Nvidia and other chipmakers to assess the end users of products sold overseas and take steps to prevent restricted hardware from reaching prohibited organisations. These obligations are intended to reduce the risk of diversion, but they cannot eliminate every possibility of resale or unauthorised transfer after a product has entered the market.
The investigation also highlights a limitation in the available procurement data. Some Chinese institutions publish notices only through local or internal platforms, while others remove records quickly. The commercial database examined by CNN does not include every procurement notice issued by Chinese organisations.
As a result, the nearly 300 records provide evidence of reported purchasing activity but cannot establish the full scale of Nvidia chip acquisitions by restricted Chinese institutions. The actual number of transactions could be higher or lower than the available records suggest.
China Rejects Washington’s Export Restrictions
The findings emerge amid continuing tensions between Washington and Beijing over semiconductors, artificial intelligence and access to advanced computing infrastructure.
China has repeatedly objected to US restrictions, arguing that Washington is using national security concerns to limit the development of Chinese technology companies and research institutions.
Chinese Foreign Ministry spokesperson Mao Ning said Beijing opposed what she described as the misuse of national security concerns to target Chinese institutions through discriminatory lists.
For China, access to advanced processors is closely connected to its broader ambition to develop domestic AI systems and reduce dependence on foreign technology. US restrictions, meanwhile, reflect Washington’s concern that powerful computing hardware could contribute to military applications and other strategic capabilities.
The dispute is therefore about more than individual chip purchases. It concerns who can access the computing infrastructure needed to compete in advanced AI research and how governments can control technologies that have both commercial and security applications.
What the Findings Mean for Nvidia and the Global AI Industry
For Nvidia, the reported purchases underline the difficulty of managing global demand for advanced processors while complying with increasingly complex export regulations. The company must navigate a market in which the same products can be commercially valuable to legitimate customers and strategically sensitive to governments.
The findings also illustrate why export controls cannot rely exclusively on restrictions at the point of sale. Even when direct shipments to certain organisations are prohibited, older inventory, resales and intermediary networks can complicate enforcement.
For US policymakers, the challenge is to improve visibility across the supply chain without assuming that every purchase recorded by a restricted institution automatically represents a completed illegal transfer. Better end-user verification, stronger documentation and cooperation among exporters, distributors and regulators are likely to remain central to enforcement efforts.
For China, the issue highlights the continuing importance of access to high-performance computing hardware as competition in artificial intelligence intensifies. Universities and research institutes remain important centres for technical development, making their access to advanced processors a significant part of the wider technology dispute.
Do the Procurement Records Prove That US Export Rules Were Violated?
No. The records reviewed by CNN show that at least six US-blacklisted Chinese universities reported purchasing Nvidia processors subject to American export restrictions. They do not independently establish that the chips were delivered, identify their original source or demonstrate that the transactions breached US law.
Some purchases may involve older hardware acquired before restrictions took effect, while other transactions may have passed through intermediaries whose role is not clear from the available notices. Determining whether a violation occurred would require further evidence about the origin of the processors, the parties involved, applicable licences and the final recipients.
Even with those limitations, the reported transactions expose a significant weakness in the practical enforcement of semiconductor controls. Washington can restrict direct exports, blacklist institutions and impose compliance obligations on manufacturers, but monitoring hardware after it enters a complex distribution network remains a much harder task.
The central question is no longer only whether Nvidia can sell restricted processors to Chinese buyers. It is whether US authorities can reliably trace where those processors go after leaving authorised supply channels. The procurement findings suggest that this remains one of the most difficult problems in the global race to control advanced AI technology.
For breaking news and live news updates, like us on Facebook or follow us on Twitter and Instagram. Read more on Latest World on thefoxdaily.com.

COMMENTS 0