Restricted Nvidia Chips Still Anchor $3.2B in China AI Projects
Reuters tied $3.2 billion in Chinese AI projects to restricted Nvidia chips, a day after Jensen Huang quit. Here is who is buying them and what is at risk.
Hakkı ALKANPublished 5 min read
Chinese developers, local governments and universities are still planning data centers around Nvidia chips that China is not supposed to get. A Reuters review published on November 13 found more than 23 billion yuan ($3.2 billion) in computing projects and equipment deals in China built around Nvidia's restricted H100, H200 and H800 processors. Demand for Nvidia chips in China, in other words, has not gone away.
Reuters went through more than 150 procurement notices, investment plans, construction contracts and corporate filings. The total covers buildings, power systems and other infrastructure as well as chips, so it is not what was spent on Nvidia hardware alone, and the documents do not show how the chips reached China or, in several cases, whether they arrived at all.
Which Chinese Projects Are Built on Restricted Nvidia Chips?
The H100 and H200 are Nvidia's data center AI accelerators, the chips used to train and run large AI models. The H800 is a cut-down H100 that Nvidia made for China and that Washington later banned as well. All three fall under US export controls, the rules that decide which advanced technology can be sold to which countries.
The projects Reuters named run from coastal Fujian to Inner Mongolia and Gansu in the far west. Most were found in procurement notices and tenders, the public documents Chinese buyers use to invite bids and spell out what equipment they want.
| Location | Reported value | Nvidia chips | Stage |
|---|---|---|---|
| Xianyou county, Fujian | 6.2 billion yuan ($873 million) | H100 | Proposed, seeking commercial partners |
| Shaanxi | 3 billion yuan project | H100, H200 | State-owned Sinomach unit won a contract |
| Jilin | 2.26 billion yuan | H100 plus Chinese chips | April government prospectus, about 80% bank-financed |
| Inner Mongolia | 1.5 billion yuan | 625 H200 servers | Construction reported to have begun in June |
| Gansu | 1.16 billion yuan | About 3,072 H100/H200 | Construction contract awarded in July |
| Wuxi, Jiangsu | Not disclosed | H800 | Operating since March 2025 |
| Fujian (Damingcheng) | 759 million yuan services deal | H800 | Cluster built, contract ended early |
The Gansu project, a retrofitted commercial site, was expected to install its chips before the end of the year, but Reuters said it remains unclear whether they ever arrived. The contract went to a consortium led by a subsidiary of state-owned Shaanxi Construction Engineering, according to the Reuters factbox.
Demand also reaches research labs. Shenzhen University awarded a contract for three H100 processors in September, and US-blacklisted Northwestern Polytechnical University bought two H200 chips a month later, Reuters reported.
Didn't Nvidia Say It Was Out of China?
It did. Nvidia's August filing said it was effectively excluded from China's data center market, the company leaves China data center revenue out of its forecasts, and Jensen Huang said last year that its China share had fallen from 95% to zero.
Yet an IDC estimate cited by Reuters in October put Nvidia at 34% of Chinese AI accelerator shipments in the first half of 2026. Reuters stresses that these numbers measure different things: Nvidia is talking about what it can sell directly, while the documents show what Chinese buyers keep planning to use.
"Washington must stop focusing solely on who buys the chips on the surface and start examining who ultimately uses the computing power, and for what purpose," said Sunny Cheung, an analyst at the Washington-based Jamestown Foundation.
Why Does Jensen Huang's Exit Matter Here?
The review landed a day after Huang resigned as Nvidia CEO on November 12 amid a scandal over his relationship with Alice Huang, a former executive associated with Singapore-based Nvidia cloud partner Megaspeed, whom US authorities suspect of intelligence ties to Beijing. Megaspeed denies illegally diverting Nvidia chips to China, and there has been no public finding that Huang broke export control laws.
Nvidia's board has opened an independent investigation and says it has no reason to believe company secrets or US national security were compromised. Reuters stressed that none of the Chinese projects it identified has been linked to Megaspeed or to the allegations around Huang.
Huang was a vocal advocate of easing export restrictions. Reuters says his departure could weaken Nvidia's lobbying power just as US lawmakers push for stricter controls, leaving Chinese operators exposed to delays, higher costs or forced redesigns around domestic chips.
How Does This Fit With the H200 Stockpile and License Fights?
Washington first restricted Nvidia's top AI chips in 2022, later closed the door on China-specific versions such as the H800, and in January began reviewing H200 export licenses for approved Chinese buyers case by case. That path has been messy. Earlier this month The New York Times found that some H200 license holders had empty offices and ties to blacklisted groups.
Beijing has been slowing things down from its side too. Reuters reported on November 5 that about 600,000 H200 chips were sitting in Hong Kong because China had not approved large-scale imports, a sign of Beijing's reluctance to deepen its reliance on US technology.
Put together, the new review shows the squeeze from both directions. Washington wants to tighten, Beijing wants local chips, and a large part of China's AI buildout is still designed around Nvidia hardware it cannot reliably buy.
What Should You Watch Next?
The next signal comes on November 17, when interim CEO Colette Kress hosts Nvidia's earnings call while the board searches for a permanent chief. Any comment on China demand, H200 licenses or compliance checks will be read closely in Washington and Beijing.
Also watch whether lawmakers cite these projects in the push for tighter controls, and whether Chinese operators redesign more projects around domestic chips. China's commerce ministry and Nvidia did not immediately respond to Reuters' requests for comment.