China's Nvidia-Based AI Projects Worth $3.2 Billion, Reuters Finds

Chinese projects built around Nvidia's AI chips exceed 4.6 trillion won ($3.2 billion) in value despite U.S. export controls, according to a Reuters review published on the 14th. The resignation of former Nvidia Chief Executive Jensen Huang has clouded the outlook for any easing of those restrictions, raising concerns that chip supplies for the projects could be disrupted.

Reuters examined more than 150 documents, including procurement notices, investment plans, construction contracts and corporate disclosures. It found projects and equipment contracts based on Nvidia chips worth more than 23 billion yuan, about 4.6 trillion won.

The figure does not represent spending on Nvidia chips alone. It covers investment in data center infrastructure such as buildings and power equipment. It also spans everything from completed projects to those still at the contract-signing or investment-planning stage.

Which chips are involved?

AI data center projects using Nvidia's H100, H200 and H800 chips are under way across China, Reuters said. All three are subject to U.S. export restrictions. The H200 has one narrow opening: it has been eligible for shipment to China since early this year, if it clears case-by-case U.S. government review and receives an export license.

The operators behind the projects vary widely:

• China's large technology companies
• Startups backed by local governments
• Universities
• Regional computing centers

What do the projects look like?

In Xianyou County, Fujian province, a 6.2 billion yuan (about 1.2 trillion won) AI data center is under construction. The project specified plans to use Nvidia's H100 chips when it solicited partner companies.

In Jilin province, a 2.26 billion yuan data center project is also under way. It will use Nvidia's H100 alongside Chinese-made chips. The developer plans to finance about 80% of the total investment through bank loans.

In Gansu province, a contract was signed to build a computing center installing about 3,072 Nvidia chips, including H100 and H200 units. Reuters could not confirm whether those chips have actually entered China. In Wuxi, Jiangsu province, by contrast, an H800-based computing facility has been in operation since last year.

What does this change for the market?

Nvidia has maintained that U.S. export controls have effectively shut it out of China's AI data center market. The review tells a different story: demand for Nvidia technology remains substantial as China expands its AI infrastructure, Reuters said.

Huang had consistently argued for easing U.S. export restrictions on chip sales to China. He said last year that Nvidia's market share in China had fallen to zero from 95%. Nvidia has also excluded revenue from data center chips in China from its earnings guidance, citing regulatory uncertainty.

With the U.S. Congress pressing for tighter export controls on advanced chips, and Huang now out of the company, uncertainty surrounding China's AI projects is expected to grow. Delays or halts in Nvidia chip supplies could push back projects and force design changes to accommodate Chinese-made chips.

"U.S. authorities should not simply focus on who is buying the chips on the surface, but should investigate who ultimately uses the computing power and for what purpose," said Sunny Cheung, an analyst at the Jamestown Foundation, a U.S. think tank.

China's Ministry of Commerce and Nvidia did not respond to Reuters' requests for comment on the findings.