China is moving forward with a colossal plan to build a network of AI data centers powered by over 115,000 high-end NVIDIA GPUs, including restricted H100 and H200 models, in open defiance of ongoing US export bans. This initiative underscores Beijing’s determination to advance its AI infrastructure and technological capabilities, even as Washington seeks to limit access to cutting-edge American chips.
Ambitious Scale: 36 Data Centers Across Western China
The project involves constructing 36 to 39 data centers in China’s western regions, such as Xinjiang and Qinghai. The centerpiece is a state-owned facility in Yiwu, Xinjiang, which alone is set to house approximately 70% of the total GPUs—around 80,500 units. These remote locations are chosen for their abundant renewable energy resources and cooler climates, which are ideal for large-scale computing operations.
Sourcing Restricted Chips: Trade Loopholes and Third Parties
Despite the US ban on exporting advanced NVIDIA AI chips to China, Chinese companies appear to be leveraging alternative procurement channels. Reports suggest chips may be acquired through third-party suppliers in Southeast Asia or via less formal networks, although the exact mechanisms remain unclear and are the subject of skepticism among US officials. There is no public evidence that China has already secured the full quantity of chips, but the scale of construction and investment approvals indicates strong confidence in their ability to obtain them.
Strategic and Global Implications
The planned infrastructure, while still smaller than the US’s AI compute capacity, would significantly boost China’s position in global AI development. The data centers are expected to support foundational large-language models and other advanced AI applications, with major Chinese tech firms and state support driving the project.
This development has raised concerns in Washington, with fears that such computing power could enhance China’s military and technological edge. The US initially restricted NVIDIA chip sales to China in 2022 to prevent such advancements, but the effectiveness of these controls is now being questioned as China appears to find ways around them.
Domestic Alternatives and Future Outlook
While China is investing in domestic chip manufacturing, such as Huawei’s AI chips, these alternatives have yet to match the performance of top-tier NVIDIA GPUs. For now, China’s AI ambitions remain closely tied to acquiring foreign hardware, by any means available.
“Chinese firms aim to install more than 115,000 Nvidia Corp. AI chips in some three dozen data centers across the country’s western deserts. Operators in Xinjiang intend to house the lion’s share of those processors in a single compound — which, if they can pull it off, could be used to train foundational large-language models like those of Chinese AI startup DeepSeek.”
As construction continues and global scrutiny intensifies, the world will be watching to see if China can realize this unprecedented AI infrastructure buildout—and how the US and its allies may respond.