Key facts
- Northwestern Polytechnical University purchased two Nvidia H200 AI chips in the previous month at a cost of 874,700 yuan, or $123,000, based on an October 21 entry posted on its procurement website.
- The procurement record shows the chips were bought for the university's management school through Shaanxi Bosuan Network Technology, a domestic supplier, rather than directly from Nvidia.
- The university is based in Xi'an and its research covers military aircraft, missiles and satellites.
- Washington's Department of Commerce added Northwestern Polytechnical to the Entity List in 2001 over alleged missile-proliferation activities and military research.
Northwestern Polytechnical University bought two Nvidia H200 AI chips for 874,700 yuan, or $123,000, through Shaanxi Bosuan Network Technology, according to an October 21 entry on its procurement website. The Xi’an institution belongs to China’s “Seven Sons of National Defense” group and researches military aircraft, missiles and satellites. The Commerce Department added the university to its Entity List in 2001 because of alleged missile-proliferation work and military research, and in 2019 tightened rules so most export licence requests for controlled US technology are rejected. The H200 processor is one of Nvidia’s most powerful chips for advanced AI training and execution, and most export licence approvals for blacklisted buyers are denied.
This year, the university obtained rented computing capacity from Chinese companies under five contracts to use H200 chips and other high-end Nvidia processors. Reuters examined 26 procurement and corporate records from Chinese institutions this year and found universities acquiring US-restricted Nvidia chips or renting Nvidia-powered computing services through domestic suppliers. Several of those institutions maintain close ties to China’s defense industry or military research. Military-linked institutions still reach American AI computing power through local vendors, buying processors directly or paying for server time on machines that hold them.
Starting in January, the Trump administration permitted some H200 exports to China case by case while keeping separate restrictions for Entity List organizations such as Northwestern Polytechnical. Lawmakers from both parties in the US Congress called for stronger controls on Chinese access to advanced AI chips, including chip-tracking requirements and more congressional oversight of exports.
The Commerce Department approved H200 chip sales to Chinese companies despite some buyers having ties to restricted firms. The New York Times said the Commerce Department cleared dozens of suspect Chinese firms to purchase and sell Nvidia’s advanced AI chips. New Matrix passed the Commerce Department review for H200 chips, and The New York Times found no apparent activity at its listed address in an industrial park northwest of Beijing. Staff at the nearby property management agency told The New York Times they had no record of New Matrix.
Corporate records from China connect New Matrix’s owners to a major private equity firm added to the US Covered List in 2024. Two analysts speaking to The New York Times said New Matrix might serve as a front for companies connected to the restricted private equity firm. The New York Times described the New Matrix chip sale as part of a pattern causing concern among technology experts inside and outside the government. Cybernews said there appear to be blind spots in Washington’s chip export controls.
In early 2026, President Donald Trump let Nvidia sell H200 chips to China if US supply was sufficient and Nvidia contributed 25% of revenue. The Commerce Department had to check the companies involved for national security before Nvidia could sell any H200 chips. Baidu, BYD and Tesla Shanghai already received permission to buy H200 chips, The New York Times reported.
Reuters reported earlier this year, citing anonymous sources, that Alibaba, ByteDance and Tencent won approval to buy more than 400,000 H200 chips. The Chinese military has previously been linked to Baidu and BYD.
Huang said continued American sales to China would maintain US technology leadership and keep Chinese customers reliant on Nvidia technology. Some members of Congress said the chips could speed up Chinese AI development for weapons design, intelligence work and military modernization. Republican Senator Kevin Cramer of the Senate Banking Committee called for Huang to step down as CEO because of his relationship with Alice Huang, who faces accusations of working for Chinese intelligence, and the extortion risk that creates.
Cramer told Politico late on Tuesday, “The potential for extortion is so tremendous that something as sensitive as Nvidia chips, I just think he’s got to go.” Three Republican committee chairmen asked to meet with Huang on November 18 about his reported contacts with an alleged Chinese intelligence operative. On Sunday, Huang acknowledged being unfaithful to his wife.
On Sunday, Nvidia’s board started an independent investigation into whether Huang, a 63-year-old Taiwan-US dual national, was targeted by a foreign intelligence operation after reports connected him to Alice Huang, formerly of a Singapore-based data centre company. Huang helped found Nvidia more than thirty years ago. The company said it sees no reason to believe American national security or confidential information was compromised.
IDC reported that Nvidia’s data center AI chip shipments to China fell 44% year over year in the first half of 2026. US export restrictions restrict sales of Nvidia’s most advanced AI chips to China, while production constraints put deliveries of compliant alternatives at risk. Nvidia received licenses for limited H200 chip shipments to China, but Chinese government restrictions prevented delivery of the full permitted amount.
H200 shipments made up less than 1% of quarterly revenue for Nvidia’s data center segment. Nvidia took a $400 million charge in the first half of fiscal 2027 for excess H200 inventory and related purchase commitments. Taiwan Semiconductor Manufacturing, based in Taiwan, makes Nvidia’s H200 chips. US export licenses require Nvidia to send H200 chips to the United States for inspection before delivery to China, and Nvidia says it cannot pass on the 25% US import tariff to customers.
Huawei held 28% of Chinese data center AI accelerator shipments in the first half of 2026, IDC said, narrowing the gap with Nvidia’s 34% share. Nvidia said limits on selling to China allow rivals to build larger developer and customer networks, making it hard to recover lost market share even if export rules loosen.
Reuters reported that Chinese orders top 6 million cards, while Nvidia’s worldwide output in 2027 may not top 3 million cards. Nvidia expects commercial deliveries in the first quarter of 2027 while giving priority to manufacturing capacity for its more advanced Rubin platform. Making more China-export-compliant chips could shift capacity away from other chips rather than add new sales.
Nvidia’s revenue increased 106% year over year to $96.2 billion in fiscal 2027’s second quarter, which ended on July 26, 2026. Nvidia’s third-quarter guidance of $108 billion leaves out data center compute revenue from China. Chinese data center chip sales do not factor into Nvidia’s near-term growth.
Restored Chinese data center sales could add revenue beyond Nvidia’s current guidance. Shareholders should distinguish between the possibility of restored Chinese sales and actual profit. Access to Trump gives Huang a way to influence US policy but does not guarantee sales in China. Reuters reported that ByteDance, Alibaba and Tencent placed reported orders worth more than $35 billion at list prices for Nvidia RTX PRO 5500 graphics cards, though these are not realized revenue.
Huang said in June 2026 that Trump sometimes phones him late at night about jobs, manufacturing and national security, according to the Associated Press. Huang attended Trump’s September 24 state dinner for Chinese President Xi Jinping, and Reuters said that summit produced no advance in AI-related disputes.

Join the discussion