Project approval documents show that in the 4th quarter of 2024, local governments in Xinjiang and in neighboring Qinghai province green-lit a total of 39 data centers that intend to use more than 115,000 Nvidia processors.
All of the companies stated in their investment plans that they aim to obtain Nvidia’s H100 or H200 GPUs. This year Nvidia debuted a new, more advanced model — dubbed the Grace Blackwell — that is banned along with the H100 and H200 from export to China without a US government license.
One of the largest projects involves a company ultimately controlled by Nyocor Co (金开新能), a Tianjin-based energy firm mainly engaged in solar and wind power. It proposes to build a data center powered by 625 H100 servers, one of the banned Nvidia models. It would start with 250 servers in the first phase. That's 2,000 H100 chips. Tender documents show the Nyocor project has started installing servers and other equipment at the data center building, and has asked China Bester Group (中贝通信), a Hubei-based IT company, to supply the hardware. Unlike the investment approval documents, which explicitly state the company wants to use H100s, the tenders don't specify whether the installed servers run on Nvidia chips or some alternative. The amount of the investment was not disclosed.
Nyocor is selling its computing power to Infinigence AI (无问芯穹), one of the largest AI infrastructure companies in China. The company has raised one billion yuan since creation.
In order to complete all of the 39 projects as outlined, companies would need to figure out a way to purchase more than 14,000 data servers or 115,000 Nvidia H100 or H200 chips, both banned for China-based entities. These chips would be worth billions of dollars based on black market prices in China.
In documents for 27 other data center projects approved in Xinjiang and Qinghai last year, companies stated they would source more than 9,000 Nvidia servers that would include up to 72,000 Nvidia H100 or H200 chips.
China Energy Investment Corporation, the Beijing-based state-owned conglomerate and one of China's largest green energy investors, plans to install 644 servers powered by more than 5,000 H100 or H200 chips with its two newly-created subsidiaries.
11 projects did not disclose the exact number of servers they would use, but based on their planned computing power, they would need around 5,000 H100 or H200 servers 
40,000 H100 or H200 GPU chips.
7 Xinjiang projects that aim to use those processors had started construction or won open tenders for AI computing service as of June 2025. One operator says it’s already using advanced hardware facilities to support cloud access to DeepSeek’s R1 model. Still, the provincial projects’ description of their intended computing capabilities may be somewhat aspirational: local party officials try to signal to Beijing that they are working toward national priorities, but Chinese companies frequently launch initiatives that are never completed.
Around 70% of computing power planned by the identified projects is in a single compound set up by the local government in Xinjiang.
DeepSeek and other Chinese AI startups have already expressed interest in collaborating with the data center projects in Xinjiang.
Meanwhile, Chinese companies continue to build their data centers, a sign they expect to receive AI chips from somewhere.
Two such construction projects were approved by the Qinghai government in Dec 2024, with a total investment of 13.5 billion yuan.
The companies applying for construction permits for both projects were founded in that same month. China’s company registry services show both entities can be traced by shareholding data to the same group of controlling companies: one real estate firm in Qinghai named Qinghai Borong Group and one AI tech company in Sichuan called Chengdu Qingshu Technology.
bloomberg.com/graphics/2025-
40,000 H100 or H200 GPU chips.
7 Xinjiang projects that aim to use those processors had started construction or won open tenders for AI computing service as of June 2025. One operator says it’s already using advanced hardware facilities to support cloud access to DeepSeek’s R1 model. Still, the provincial projects’ description of their intended computing capabilities may be somewhat aspirational: local party officials try to signal to Beijing that they are working toward national priorities, but Chinese companies frequently launch initiatives that are never completed.
Around 70% of computing power planned by the identified projects is in a single compound set up by the local government in Xinjiang.
DeepSeek and other Chinese AI startups have already expressed interest in collaborating with the data center projects in Xinjiang.
Meanwhile, Chinese companies continue to build their data centers, a sign they expect to receive AI chips from somewhere.
Two such construction projects were approved by the Qinghai government in Dec 2024, with a total investment of 13.5 billion yuan.
The companies applying for construction permits for both projects were founded in that same month. China’s company registry services show both entities can be traced by shareholding data to the same group of controlling companies: one real estate firm in Qinghai named Qinghai Borong Group and one AI tech company in Sichuan called Chengdu Qingshu Technology.
bloomberg.com/graphics/2025-