China is constructing a network of data centres in Xinjiang to advance its AI capabilities, despite US export restrictions. The project involves installing over 115,000 Nvidia AI chips, including restricted models, to power these facilities. These data centres will train large language models, rivalling those developed elsewhere.
Located in remote regions like Xinjiang and Qinghai, these facilities leverage local resources and cooler climates. A significant portion of the computing power will be concentrated near Yiwu in Xinjiang. The acquisition of restricted chips remains unclear, with possibilities including third-party suppliers or alternative channels.
This initiative highlights China's determination to become a leader in AI, potentially circumventing US export controls. The scale of the project signals Beijing's strategic intent to establish a major AI hub, raising concerns about potential implications for technological competition.
