A Reuters review of more than 100 tenders and research papers suggests that if U.S. export permissions proceed as announced, legal bulk purchases of the H200 are likely to accelerate rapidly. The H200 is significantly more capable than any of the export-restricted models Nvidia has been allowed to sell into China in recent years, making it especially attractive to institutions racing to strengthen their AI capabilities.
Universities and Research Institutes Expand AI Ambitions
China’s top universities—long central to the country’s AI talent pipeline—have been active early adopters of high-performance GPUs. Their ability to acquire advanced chips directly shapes both recruitment power and research output.
Laboratories at Beijing Jiaotong University openly state that they have already secured multiple H200 units via unofficial channels. At Shanghai Artificial Intelligence Laboratory, as well as Sun Yat-Sen, Tsinghua, and Shanghai Jiao Tong universities, researchers recently disclosed using four H200 chips to train a model built to detect AI-generated imagery.
Activity is not limited to individual research groups. In Hefei, a state-run AI institute issued a tender in June for a server powered by eight H200 chips, intended to support work on a “quantum AI model.” Across the country, dozens of universities have either purchased or sought H200-equipped systems, underscoring widespread demand for cutting-edge compute.
Military-Linked Institutions Emerge as Significant Potential Users
U.S. policymakers critical of easing chip export restrictions have warned that advanced GPUs could enhance the People’s Liberation Army’s technological capabilities. Procurement records appear to support concerns that military-connected institutions are already seeking access to H200-class hardware through indirect channels.
In August, the PLA Air Force Medical University in Xian issued a tender for eight H200 chips to support development of a large-language-model platform focused on medical AI and biosurveillance. Beihang University—one of the “Seven Sons of National Defense”—requested suppliers able to provide rented access to computing power equivalent to H200-level GPUs, reflecting the growing trend of leasing rather than importing restricted hardware.
Data-Center Builders Position for Large-Scale Deployment
At the infrastructure level, several Chinese regions have already integrated H200-powered servers into medium- and large-scale AI cluster plans, anticipating expanded access.
In Jiangsu province, a government-affiliated firm sought 48 servers equipped with a total of 384 H200 chips, with delivery targeted for the end of the year. Xinjiang—now a prominent hub for low-cost land and energy—has even more ambitious projects on the table. A June 6 tender from Urumqi Jiangsuan outlines a 20,000-petaflop cluster combining more than 8,000 H200 GPUs with large deployments of Nvidia RTX 4090s and Huawei’s Ascend 910C processors.
Another major initiative, announced in Burqin county in late 2024, envisions a green-energy compute center centered on domestic chips but still supported by a smaller cluster of 100 H100 or H200-class servers. Meanwhile, in Hubei province, Xiaogan Yunqi Data Technology disclosed plans to supply China Unicom with 128 H200-based servers by March, part of a 307-million-yuan project.
A Market Ready to Scale—If Policy Allows
Even without official approval from Beijing, these signals point to a robust appetite for the H200 across academic, industrial, and defence-related sectors. Should China authorize legal imports, the chip is well-positioned to become a central component of the country’s next wave of AI infrastructure.