Briefly
- Reuters recognized greater than $3.2 billion in Chinese language AI tasks specifying Nvidia chips topic to U.S. export controls.
- The determine covers general mission funding—not confirmed chip purchases or shipments.
- Nvidia nonetheless holds a considerable share of China’s AI chip market regardless of Huawei’s enlargement.
Chinese language builders are planning and constructing billions of {dollars} price of AI knowledge facilities round Nvidia processors restricted by Washington, in line with a Reuters investigation printed Friday.
After analyzing greater than 150 procurement notices, funding plans and company data, the news agency recognized tasks price over 23 billion yuan ($3.2 billion). Many specify Nvidia’s H100, H200, or H800 chips, which give the computing energy used to coach and run superior AI fashions.
However $3.2 billion is the whole worth of the tasks, together with buildings, electrical energy and different infrastructure. The paperwork don’t set up that each one the processors had been obtained, or that any export legal guidelines had been damaged.
What the paperwork reveal
One proposed computing middle in Fujian province is valued at 6.2 billion yuan ($873 million). Its official investment listing specifies H100 servers however remains to be looking for financing companions.
In Shaanxi, a state-owned engineering group gained a contract linked to a 3 billion yuan middle naming H100 and H200 chips. A 2.26 billion yuan mission in Jilin proposes combining Nvidia {hardware} with Chinese language processors.
Interior Mongolia has one other proposed growth specifying 625 H200 servers. In the meantime, authorities in Wuxi reported that an H800 computing cluster started working in March 2025. Working amenities and proposed investments usually are not proof of the identical stage of chip entry.
Nvidia’s processors can perform hundreds of calculations without delay, making them priceless for AI coaching. Its extensively used software program additionally makes switching to rival techniques extra difficult and doubtlessly costly.
The US started limiting superior AI chip gross sales to China in 2022, citing nationwide safety considerations. In January 2026, the Commerce Division moved H200 export functions to case-by-case review, though gross sales nonetheless require licenses. H100 shipments stay extra tightly restricted.
That leaves Chinese language builders with a dilemma: design amenities round acquainted Nvidia gear they could wrestle to accumulate, or spend money on different processors that may require software program adjustments and extra engineering.
China’s different is rising
An IDC report cited by Reuters estimated Nvidia accounted for 34% of China’s AI accelerator shipments within the first half of 2026, in contrast with 28% for Huawei. These figures cowl the broader chip market, not confirmed shipments of restricted fashions.
Chinese language AI developer Z.AI has already released an image-generation model educated solely on Huawei chips, demonstrating that aggressive techniques needn’t rely upon American {hardware}.
Nonetheless, the brand new mission data present how deeply Nvidia stays embedded in Chinese language AI infrastructure plans. Additionally they add strain on U.S. officers dealing with congressional scrutiny over how superior chip export guidelines are enforced.
Each day Debrief E-newsletter
Begin on daily basis with the highest information tales proper now, plus authentic options, a podcast, movies and extra.
Be a part of 72,000+ readers
You might also like
More from Web3
Movie2K Defendant Could Hand Over 40,000 Bitcoin to End Trial—and Stay a Billionaire
In short A Leipzig choose has supplied to finish the trial of the alleged Movie2K operator if he fingers over …
Scattered Spider Hacker Sentenced to 6 Years Over $11.6M in Stolen Crypto
Briefly Tyler Buchanan, a 24-year-old Scot, was sentenced to 6 years in U.S. federal jail for a scheme that stole …
Why Microsoft Thinks Smaller, Cheaper AI Can Beat Claude and GPT
Briefly Microsoft says MAI Code 2 delivers 96% of Claude Opus 5.5's coding efficiency at 5% of its value, although …





