Blacklisted Chinese University Bought Nvidia H200 AI Chips Despite US Export Ban, Records Show

November 11, 2026 Faizan Altaf

A Chinese university with close ties to the People’s Liberation Army (PLA) bought two of Nvidia’s most powerful artificial intelligence (AI) chips last month, even though it has been on a United States export blacklist for more than two decades. The purchase, revealed in the university’s own procurement records, shows how hard it is for Washington to keep advanced American technology out of the hands of blacklisted Chinese institutions.

Northwestern Polytechnical University, based in Xi’an in north-western China, bought two Nvidia H200 processors for 874,700 yuan (about $123,000), according to a procurement record dated 21 October on the university’s website. The chips were bought through a Chinese supplier, Shaanxi Bosuan Network Technology, rather than directly from Nvidia.

The H200 is one of Nvidia’s most capable chips for training and running advanced AI systems. It carries 141 gigabytes of high-bandwidth memory and can move data at 4.8 terabytes per second, making it one of the most sought-after processors in the world for building AI models. Exports of the chip to China need approval from the US government, and licence applications from blacklisted institutions are normally refused.

The university is no ordinary buyer. It is a member of China’s “Seven Sons of National Defense”, a group of elite universities with deep links to the country’s defence industry, and it has been on the US Commerce Department’s Entity List — America’s trade blacklist — since 2001, over alleged involvement in missile proliferation and military research. In 2019, Washington tightened the rules further, making licences for exports of controlled US technology to the university subject to a policy of denial in most cases.

Reuters, which first reported the purchase on 11 November after reviewing the procurement records, said neither the university, Nvidia nor the US Commerce Department’s Bureau of Industry and Security, which oversees export controls, responded to requests for comment.

More Than Just Two Chips

The two H200 chips were only part of the picture. The procurement records show that the university also bought access to H200 chips and other powerful Nvidia processors through five contracts with Chinese companies this year, renting computing capacity rather than buying the hardware outright.

In January, the university’s cybersecurity school awarded two seven-month contracts for access to servers containing eight H200 chips each. The records also show it rented access to servers containing eight Nvidia A100 processors and eight A800 processors. Both the A100 and the A800 are banned from export to all Chinese entities, and the H200 is the most advanced processor named in the university’s records.

A Reuters review of 26 Chinese procurement and corporate records from this year found other universities and research institutions, including some with links to China’s defence industry, buying restricted Nvidia chips or renting computing services powered by them through domestic suppliers.

It is important to be clear about what the records do and do not prove. They show that the university bought and rented the chips and the computing power. They give limited information about where the chips originally came from or how they reached the Chinese supplier. They do not show that Nvidia itself sold the chips to the university, and they do not say what the chips were used for. What they do show is that the export controls did not stop a blacklisted, military-linked university from getting hold of some of the world’s most advanced AI hardware.

Why the H200 Matters So Much

To understand why this story matters, it helps to understand what the H200 is and why governments treat it like a strategic asset.

The H200 is part of Nvidia’s Hopper family of data-centre graphics processing units (GPUs). A GPU was originally designed to draw images on a screen, but the same design turns out to be extremely good at the repeated calculations needed to train AI models. Modern AI systems, from chatbots to image generators to scientific research tools, are trained on thousands of these chips working together in giant data centres.

The H200 is the memory-upgraded version of the H100, Nvidia’s earlier flagship AI chip. It carries 141 gigabytes of HBM3e memory — nearly double the H100’s 80 gigabytes — with memory bandwidth of 4.8 terabytes per second. In simple terms, it can hold more information and move it around faster, which makes training large AI models quicker and cheaper. That is why it is in such high demand, and why it sits at the centre of the technology rivalry between the United States and China.

The United States has spent years tightening controls on which chips can be sold to China. The aim, as American officials have described it, is to slow China’s ability to build advanced AI systems that could be used for military purposes, from weapons design to surveillance and cyber operations. The controls cover not only the most powerful chips but also the software and equipment needed to make them.

China, for its part, has been working to build its own chip industry and to find ways around the restrictions, including buying through intermediaries and renting computing power from domestic cloud providers. The Northwestern Polytechnical University case shows both of those routes in action: a direct purchase through a local supplier, and rental contracts for server time.

A Loophole That Was Left Open

There is an added twist to this story. In January 2026, the administration of US President Donald Trump began allowing some H200 exports to China on a case-by-case basis. The January licensing rules also require safeguards to prevent blacklisted entities and military end-users from remotely accessing chips exported under the policy.

In other words, the H200 is not banned from China outright. Some sales are allowed, with conditions. But those conditions are supposed to keep the chips away from blacklisted institutions and military users. Northwestern Polytechnical University is exactly the kind of institution the rules are meant to exclude — blacklisted since 2001, defence-linked, and subject to a policy of denial. The fact that it still obtained the chips, and rented access to many more, raises questions about whether the safeguards are working.

Export-control experts have long warned that the system has gaps. Chips are small, valuable and easy to move. Once a chip has been legally sold to an approved buyer, it can be resold, sometimes several times, before it reaches its final user. Cloud computing adds another gap: a blacklisted institution does not need to own a chip at all if it can rent time on a server that contains one. The January rules tried to close the remote-access gap, but enforcement depends on knowing who is really using the computing power — and that is hard to track.

Nvidia Under Pressure in Washington

The disclosure comes at an awkward moment for Nvidia and its chief executive, Jensen Huang, who are already under heavy scrutiny in Washington.

In recent days, Nvidia’s shares have fallen after Huang admitted to an affair with a woman US authorities believe may have been a Chinese intelligence asset, and the FBI opened a probe into the matter. Nvidia has launched an independent investigation into its chief executive, and three Republican congressional committee chairmen have requested a briefing with Huang on 18 November over his reported contacts with the alleged intelligence operative and his lobbying for looser restrictions on sales of AI chips to China.

The new report adds to that pressure, because it touches on the same underlying question: how much control should the US government have over where Nvidia’s most advanced chips end up, and how well are those controls working?

Huang has argued publicly that restricting chip sales to China hurts American companies and pushes Chinese buyers towards domestic alternatives. His critics in Congress argue the opposite: that advanced chips are a national security asset, and that even small leaks can help a rival military. The Northwestern Polytechnical University case will give the critics fresh evidence, even though there is no suggestion that Nvidia itself broke the rules. The chips were bought from a Chinese supplier, not from Nvidia, and Reuters could not determine how they reached that supplier.

For more on the scrutiny facing Nvidia’s chief executive, read our earlier report: Nvidia Shares Fall as Jensen Huang Admits Affair and FBI Opens Probe — What We Know. Nvidia has also launched its own investigation: Nvidia Launches Independent Investigation Into CEO Jensen Huang After Spy Claims.

What This Means for the US–China Chip War

The technology rivalry between the United States and China is often described as a chip war, and the H200 sits near the front line. The US controls the world’s most advanced chip designs — Nvidia is an American company — and the most advanced chip-making equipment, much of it made by American and allied companies. China has the world’s largest market for chips and a fast-growing domestic industry, but it still depends on foreign technology for the most advanced processors.

Washington’s strategy has been to use that dependence as leverage: sell China the chips it needs for ordinary commercial use, but block the most advanced ones, and block blacklisted and military-linked buyers altogether. The approach has slowed China’s progress in some areas, but it has also pushed China to invest heavily in its own chip industry. Chinese companies have made real progress in recent years, though they still lag behind Nvidia at the cutting edge.

The Northwestern Polytechnical University case suggests the controls are leaking. Two chips may not sound like much — a single AI training run can use tens of thousands of them — but the rental contracts for servers with eight H200 chips each, plus A100 and A800 access, add up to serious computing power. And the wider review of 26 records found that this was not an isolated case: other universities and research institutions are using the same routes.

The case also shows how procurement records, which are public by law in China, can reveal what official trade data hides. Universities and government bodies must publish their purchases, and those records leave a paper trail. Investigative reporting based on that trail is now one of the main ways the outside world learns how export controls are working in practice.

What Happens Next

Several things are likely to follow from this report.

First, expect questions in Washington. The congressional briefing with Huang on 18 November was already scheduled before this report, but the H200 purchases will almost certainly come up. Lawmakers who want tighter controls will ask what Nvidia knew, what the Commerce Department knew, and what is being done to close the gaps.

Second, expect the Commerce Department to look closely at the suppliers involved. Shaanxi Bosuan Network Technology, the company that sold the two H200 chips to the university, may face questions about where it got them. Other Chinese suppliers named in the rental contracts may also come under scrutiny, and some could themselves be added to the Entity List.

Third, expect a debate about cloud computing. Renting access to chips is harder to control than selling the chips themselves, and the rules on remote access by blacklisted entities are new and largely untested. If the rules cannot be enforced, Washington may have to choose between tightening them further or accepting that some leakage is unavoidable.

Finally, expect China to keep pushing its own chip industry. Every report of blocked or leaked chips is used in Beijing as evidence that China cannot rely on foreign suppliers, and as a reason to spend more on domestic alternatives. In the long run, that may matter more than any single purchase: the chip war is not only about who gets today’s best chips, but about who can make tomorrow’s.

Frequently Asked Questions

What is the Nvidia H200?

The H200 is one of Nvidia’s most powerful AI chips, used to train and run advanced artificial intelligence systems. It carries 141 gigabytes of high-bandwidth memory and can move data at 4.8 terabytes per second. It is part of Nvidia’s Hopper family of data-centre chips.

Which university bought the chips?

Northwestern Polytechnical University, based in Xi’an in north-western China. It is a member of China’s Seven Sons of National Defense group of defence-linked universities and has close ties to the People’s Liberation Army.

How much did the university pay?

It paid 874,700 yuan, about $123,000, for two H200 chips, according to a procurement record dated 21 October on the university’s website.

Did Nvidia sell the chips directly to the university?

No. The chips were bought through a Chinese supplier, Shaanxi Bosuan Network Technology. The records give limited information about where the chips came from or how they reached the supplier.

Why is the university not allowed to buy these chips?

The university has been on the US Commerce Department’s Entity List, the American trade blacklist, since 2001, over alleged involvement in missile proliferation and military research. Exports of controlled US technology to the university are subject to a policy of denial in most cases.

Did the university only buy two chips?

No. The records also show five contracts this year for rented computing capacity, including two seven-month contracts from the university’s cybersecurity school for access to servers containing eight H200 chips each, plus rented access to servers with Nvidia A100 and A800 chips.

Are H200 chips banned from China completely?

No. Since January 2026, the Trump administration has allowed some H200 exports to China on a case-by-case basis, with safeguards meant to stop blacklisted entities and military end-users from accessing them remotely. But licences for blacklisted institutions are normally refused.

What happens now?

The US Commerce Department may investigate the suppliers involved, and Congress is expected to raise the case when Nvidia chief executive Jensen Huang appears for a briefing on 18 November. The report is also likely to fuel debate about whether export controls on AI chips can be enforced effectively.

Conclusion

A blacklisted Chinese university with close military ties bought two of Nvidia’s most powerful AI chips — and rented access to many more — despite more than two decades of US export controls meant to stop exactly that. The purchase, documented in the university’s own procurement records, does not show Nvidia breaking the rules, and it does not show what the chips were used for. But it does show that the controls are leaking, through local suppliers and rented computing power, at a moment when Washington is already asking hard questions about Nvidia and its chief executive. How the US responds — and whether it can close the gaps — will shape the next chapter of the technology rivalry between the world’s two largest economies.

Sources

This article is for general information and education only. It is not legal or financial advice.

Findivoza