Procurement records examined by reporters show that, over the past year, four Chinese universities purchased server systems built by Super Micro Computer that were fitted with Nvidia A100 artificial intelligence accelerator chips. Among those buying the systems were two institutions with established links to the People’s Liberation Army (PLA).
The documents do not make clear how those servers were sourced or the supply chain steps that led from U.S.-made components to deployed systems on Chinese campuses. The purchases come against a backdrop of U.S. policy measures, beginning in 2022, that prohibit the sale of certain Nvidia chips - including the A100 - to China because of concerns that powerful AI processors could enhance military capabilities.
San Jose, California-based Super Micro broke into headlines last week after U.S. authorities brought criminal charges against three people connected to the company. Prosecutors allege those individuals aided a scheme to smuggle at least $2.5 billion worth of U.S. AI technology to China. The company itself was not named in the indictment and has described the matter as an elaborate scheme perpetrated by the individuals charged.
Previous reporting has shown that Chinese universities have acquired restricted chips in servers manufactured by Super Micro and other vendors. The persistence of such acquisitions, particularly at institutions linked to the PLA, is likely to raise alarm among some U.S. lawmakers who view access to advanced AI compute as a potential enabler of military research.
In response to the recent indictments, two U.S. senators have asked U.S. Commerce Secretary Howard Lutnick to consider pausing all export licenses that allow advanced Nvidia AI chips and server systems to be shipped to China or to intermediaries in Southeast Asia. The senators framed their appeal as a precautionary step while investigations continue.
Beyond the four confirmed purchases, procurement checks of publicly available tender documents for 2025 and earlier this year show two additional universities - one with a military linkage among them - sought to obtain Super Micro systems with similar GPU configurations. Those documents do not establish whether the attempted purchases were completed.
Super Micro declined to comment on the details contained in the procurement records. Nvidia said it continues to work closely with customers and with the U.S. government to ensure compliance as export regulations have expanded. Requests for comment sent to China’s commerce ministry and the U.S. Commerce Department did not receive responses.
A contentious regulatory environment
Uncertainty persists over the precise volume of U.S. AI chip sales that Washington and Beijing will consider acceptable. The U.S. government has taken steps to restrict certain high-end accelerators, while also approving the sale of others under specific conditions. For example, the Trump administration last year approved sales of Nvidia’s H200 chip to China, subject to conditions that could limit the quantities sold.
Beijing has signaled a desire to support domestic chipmakers by discouraging purchases of U.S. AI chips, even as sources say Nvidia obtained Beijing’s approval to sell the H200. As of now, those reports have not been accompanied by confirmed sales of the H200 into China.
Critics of sales of advanced AI hardware warn that broader access to such compute could accelerate research efforts that, in turn, might support military-related applications. Jacob Feldgoise, senior data research analyst at Georgetown University’s Center for Security and Emerging Technology, warned that expanded access to AI chips "may in turn help improve China’s weapons design and testing, military planning and logistics, autonomous weapons systems, or surveillance." The prospect of such downstream effects is a central concern driving scrutiny in Washington.
Documented purchases
Examples in the procurement records include a March 16 notice from Beihang University, a Beijing-based institution included among China’s so-called "Seven Sons of National Defense" and noted for work in aerospace and defense research. The notice records the procurement of a machine-learning workstation built on a Super Micro platform configured with four Nvidia A100 chips.
Another notice, dated in July, shows Harbin Institute of Technology (HIT) moved forward with a Super Micro system populated with eight Nvidia A100 accelerators. HIT is also counted among the "Seven Sons" and has documented involvement in missile, satellite and robotics work. Reuters’ requests for comment to Beihang University and HIT went unanswered.
Both universities appear on a U.S. export blacklist designed to make it difficult for U.S. firms to obtain licenses to sell to them. The presence of restricted hardware in their tender records underlines tensions between export-control policy and on-the-ground procurement activity.
Implications for markets and sectors
The procurement activity touches multiple sectors - higher education, defense research, and technology hardware - and has prompted lawmakers to call for tighter export scrutiny. How regulators and companies respond in the weeks ahead could influence licensing practices for advanced AI processors and the flow of server systems into China and the region.
At present, key details remain unclear: which parties ultimately supplied the systems recorded in procurement notices, the chain of custody for hardware components, and whether additional purchases have occurred beyond those visible in public tender documents. Those outstanding questions are central to ongoing policy and enforcement debates.