Restricted Nvidia chips named in $3.2B of Chinese AI data center plans: Reuters
In brief
- Reuters reviewed more than 150 procurement notices, investment plans and corporate records.
- Projects worth over 23 billion yuan ($3.2 billion) named Nvidia H100, H200 or H800 chips.
- The $3.2 billion figure covers buildings, electricity and infrastructure, not chips alone.
- Reuters said the documents don't establish that chips were obtained or export laws broken.
- Nvidia held 34% of China's accelerator shipments versus Huawei's 28%, per IDC data cited by Reuters.
What the records show
Many of the projects specified Nvidia's H100, H200 or H800 chips (the hardware used to train and run advanced AI models). LeoDex News didn't review the underlying documents, so everything below comes from Reuters' findings as reported by Decrypt.
One proposed computing center in Fujian province was valued at 6.2 billion yuan ($873 million). Its official investment listing specified H100 servers, and the project was still seeking financing partners. In Shaanxi, a state-owned engineering group won a contract linked to a 3 billion yuan center naming H100 and H200 chips. A 2.26 billion yuan project in Jilin proposed combining Nvidia hardware with Chinese processors, while a proposed development in Inner Mongolia specified 625 H200 servers. Authorities in Wuxi reported that an H800 computing cluster began operating in March 2025.
There's an important caveat.
Reuters said the documents didn't establish that all the processors were obtained or that any export laws were broken. Operating facilities and proposed investments also don't indicate the same level of chip access.
Export rules and the software problem
The United States began restricting advanced AI chip sales to China in 2022, citing national security concerns. In January 2026, the Commerce Department moved H200 export applications to case-by-case review. H200 sales still required licenses, though, and H100 shipments remained more tightly restricted.
Software is part of it. Nvidia's widely used software made switching to rival systems more complicated and potentially expensive, which left Chinese developers with a choice: design around Nvidia equipment they may struggle to acquire, or invest in alternative processors that can require software changes and additional engineering.
Market share and pressure in Washington
An IDC report cited by Reuters estimated Nvidia accounted for 34% of China's AI accelerator shipments in the first half of 2026, versus 28% for Huawei. Those numbers cover the broader chip market (not confirmed shipments of restricted models).
Domestic alternatives aren't standing still. Decrypt reported that Chinese AI developer Z.AI released an image-generation model trained entirely on Huawei chips.
The project records also add pressure on U.S. officials, who are facing congressional scrutiny over how advanced chip export rules are enforced.
Frequently asked questions
Is the $3.2 billion figure the value of Nvidia chips headed to China?
No. According to Reuters, as reported by Decrypt, $3.2 billion is the total value of the projects, including buildings, electricity and other infrastructure. It isn't the value of the chips alone.
Did Chinese developers actually obtain restricted Nvidia chips?
Reuters said the documents don't establish that all the processors were obtained or that any export laws were broken. Operating facilities and proposed investments also don't indicate the same level of chip access, though Wuxi authorities reported an H800 cluster began operating in March 2025.
Which Nvidia chips does the US restrict for China?
The US began restricting advanced AI chip sales to China in 2022. In January 2026, the Commerce Department moved H200 export applications to case-by-case review, but H200 sales still required licenses and H100 shipments remained more tightly restricted.
