A Reuters review of over 100 public tenders and academic papers finds that many of China’s elite universities and research institutes, including institutions like Tsinghua University and Shanghai Jiao Tong University, already hold H200 chips for AI-model training and research. In one case, researchers used four H200s to train a model that detects AI-generated images.
Beyond academia, data-centre developers across regions such as eastern Jiangsu and western Xinjiang plan large-scale deployments: some tenders call for hundreds of H200 GPUs to power petaflop-scale AI compute clusters. Military-affiliated institutions are also on the list of H200 recipients, with documents indicating that a medical university linked to the armed forces sought servers fitted with H200 chips for training large language models.
Because export of high-performance chips like H200 has been restricted, many uses are facilitated through grey-market purchases or by renting server time on H200-equipped machines — effectively sidestepping official export constraints.
This picture shows that, while formal trade restrictions remain, demand and workaround channels have already enabled significant deployment of advanced AI hardware in China.