Chinese AI Projects Worth $3.2 Billion Are Built Around Restricted Nvidia Chips

Reuters said it examined more than 150 Chinese procurement notices, investment plans, construction contracts and corporate disclosures. The projects it identified are designed around Nvidia's H100, H200 and H800 chips. Its estimate covers planned investments, signed contracts and finished installations, and includes buildings, power systems and other infrastructure as well as chips.

The findings sit uneasily with Nvidia's own account. In an August filing the company said US export controls had effectively shut it out of China's data centre computing market. Jensen Huang, who resigned as chief executive on Thursday, said last year that Nvidia's share of the Chinese market had fallen from 95% to zero, and the company leaves China data centre revenue out of its forecasts, citing regulatory uncertainty. Operators of the projects now face possible delays, supply disruptions or expensive redesigns around domestic chips if Washington tightens controls further, according to the review.

The projects range from wealthy coastal technology hubs to remote western provinces. One example Reuters cited is a facility in Gansu province, in China's northwest, that was expected to install thousands of Nvidia chips before the end of the year. The news agency said it was unclear whether those processors ever arrived. Some of the facilities it identified are already operating and others are under construction.

Reuters did not link any of the projects to Megaspeed, the Singapore-based Nvidia cloud partner at the centre of the scandal that led to Huang's departure, or to the allegations involving Alice Huang, a former executive linked to Megaspeed whom US authorities suspect of intelligence ties to Beijing. Megaspeed denies illegally diverting Nvidia chips to China. Nvidia's board has opened an independent investigation and has said it has no reason to believe company secrets or US national security were compromised. No public finding has said Jensen Huang broke export control laws.

The review does not show how restricted hardware reached Chinese buyers. Washington began restricting Nvidia's most advanced AI processors in 2022 and later extended the curbs to chips Nvidia had modified for China, including the H800. In January, the Trump administration moved to review H200 export licences for eligible Chinese customers case by case, subject to conditions. An IDC estimate cited by Reuters in October put Nvidia's share of Chinese AI accelerator shipments at 34% in the first half of 2026, though Reuters noted that the shipment data and its own project tally measure different things.

Huang was one of the industry's loudest voices for easing export restrictions, and Reuters said his departure could weaken Nvidia's lobbying position just as lawmakers in both parties push for tighter controls. On Friday, Chinese AI stocks fell after House Speaker Mike Johnson said lawmakers were converging on limits to China's access to advanced chips, and several congressional committees are seeking testimony from Huang and Commerce Secretary Howard Lutnick about how chip sales were approved.

Sunny Cheung, an analyst at the Jamestown Foundation in Washington, told Reuters the findings showed that "Washington must stop focusing solely on who buys the chips on the surface" and should look at who ultimately uses the computing power. China's commerce ministry and Nvidia did not immediately respond to Reuters' requests for comment.

The projects identified so far are a small slice of China's AI spending, but they give investigators a concrete trail. Huang has been asked to sit for a closed-door interview with House committees on December 1.