A new investigation by Bloomberg has found that several Chinese universities with close ties to the country’s military and defense industry are seeking access to NVIDIA’s advanced H200 artificial intelligence chips, underscoring the continuing challenge facing U.S. export controls designed to keep cutting-edge technology out of military hands.
According to procurement records reviewed by Bloomberg, at least seven Chinese universities linked to the People’s Liberation Army (PLA) or China’s defense sector have sought access to NVIDIA’s H200 processors, currently among the most powerful AI chips still permitted for sale to approved Chinese customers under U.S. regulations.
RELATED: Propaganda in the AI Era: New Study Warns Biased Training Data Shapes AI Answers
Seeking access
Beihang University and Northwestern Polytechnical University are among China’s most prominent defense-linked universities, with longstanding ties to military research and weapons development programs. Both institutions have been placed on U.S. export control lists because of concerns that their research contributes to China’s military modernization efforts.
Bloomberg reported that Beihang University’s School of Cyber Science and Technology has sought to rent computing resources powered by NVIDIA H200 chips. Similar procurement records indicate that Northwestern Polytechnical University’s cybersecurity program has also pursued access to H200-powered computing capacity.
Success
You are now signed up for our newsletter
Success
Check your email to complete sign up
The investigation found that the effort extends beyond a handful of institutions. Since 2025, more than 25 Chinese universities and laboratories connected to the military or defense industry have either used or attempted to obtain NVIDIA AI chips, including earlier-generation products such as the A100, A800, H100, and H20.
At least six of those institutions have been placed on U.S. Commerce Department export control lists because of their involvement in programs related to missile development, nuclear technologies, and other sensitive military projects.
NVIDIA’s H200 Chips
While procurement records reveal active efforts to obtain H200 computing resources, Bloomberg noted that it remains unclear whether the institutions have successfully secured access to the chips or how they intend to use them.
The findings come amid growing concern in Washington over the military applications of advanced artificial intelligence systems. Though many of the projects identified involve relatively modest computing resources, often a single server equipped with eight NVIDIA GPUs, experts say even limited access can be significant.
Michael Deng, a technology analyst with Bloomberg Economics, noted that a server containing eight H200 chips would be powerful enough to modify and adapt existing open-source AI models for military applications, including autonomous weapons development and cyber warfare research. Such systems would still significantly outperform most domestically produced Chinese alternatives.
The limits of export controls
The report also highlights the ongoing technological gap between leading U.S. chipmakers and China’s semiconductor industry. Ryan Fedasiuk, a researcher at the American Enterprise Institute and former U.S. State Department official, said Chinese firms such as Huawei and Semiconductor Manufacturing International Corporation (SMIC) continue to lag behind their American competitors in both performance and production capacity.
As a result, institutions across China’s research sector remain eager to obtain NVIDIA hardware despite Beijing’s push to build a self-sufficient semiconductor industry.
Bloomberg’s investigation found that organizations are pursuing a variety of methods to gain access to advanced computing power. Some seek chips through third-party intermediaries, while others rent processing capacity from overseas data centers rather than purchasing hardware directly.
The findings also reveal that several sanctioned universities are already operating substantial NVIDIA-powered systems. Public records show that Beijing University of Posts and Telecommunications operates a supercomputing center equipped with 144 NVIDIA A800 chips. The university was added to the U.S. Entity List in 2020 because of its involvement in advanced weapons and weapons-system research.
Harbin Institute of Technology, another member of the Defense Seven Sons and a university previously sanctioned by the United States for its role in missile-related programs, has publicly disclosed the deployment of NVIDIA H100 chips and sought access to H20 resources in 2025.
Washington pushes for tighter safeguards
Meanwhile, the University of Science and Technology of China has reported operating a supercomputing system containing 200 NVIDIA A100 chips. The institution has previously faced U.S. sanctions because of its involvement in China’s quantum technology and nuclear research programs, adding new concerns to how Washington balances commercial access to U.S. technology with national security concerns.
House Foreign Affairs Committee Chairman Brian Mast recently emphasized that preventing advanced American AI chips from reaching Chinese military end users remains a top priority. “If it’s kids playing video games, I don’t care at all,” Mast said. “But when it comes to actual weapons systems, actual militaries and real casualties, Congress pays very close attention.”
The concerns come as the U.S. Commerce Department recently issued additional guidance tightening restrictions on advanced AI chip exports. The updated rules clarify that Chinese companies cannot circumvent export controls through overseas subsidiaries, requiring entities headquartered in China to comply with U.S. licensing requirements even when purchasing chips abroad.
