Chinese companies and institutions are reportedly developing more than $3.2 billion worth of AI computing projects involving NVIDIA processors that face US export restrictions.
A new investigation identified more than 150 procurement notices, investment plans, construction contracts and corporate disclosures suggesting that demand for NVIDIA's advanced AI hardware remains strong across China.
The projects reportedly involve NVIDIA H100, H200 and H800 accelerators, which have been subject to different US export restrictions.
The findings suggest that Chinese organizations are continuing to plan computing infrastructure around NVIDIA technology despite Washington's efforts to limit access to advanced processors.
However, the investigation does not establish that every project has successfully obtained the restricted hardware or that the associated facilities are already operational.
More Than 150 Chinese AI Projects Linked to NVIDIA Hardware
The investigation examined public records associated with computing infrastructure developments across China.
These documents reportedly identify projects involving advanced NVIDIA processors in several sectors, including technology, education and regional government infrastructure.
The reported $3.2 billion figure covers computing projects involving the hardware rather than the confirmed value of GPUs delivered to customers.
This distinction is important because procurement plans and construction contracts do not necessarily establish that the required processors have been purchased or installed.
The projects also extend beyond China's largest technology companies.
Local government supported initiatives, universities and regional computing centers are reportedly among the organizations planning infrastructure built around NVIDIA accelerators.
Their involvement suggests that demand for advanced AI computing resources is spread across different parts of China's technology sector.
Gansu Province Could Receive Thousands of NVIDIA AI GPUs
One example identified in the investigation involves a computing facility in Gansu province in northwestern China.
The facility is reportedly being repurposed to accommodate thousands of NVIDIA processors before the end of 2026.
If completed as planned, it could expand the region's capacity to support AI research and other computational workloads.
However, the investigation does not confirm that the necessary processors have already reached the facility.
The project could also face difficulties if export enforcement becomes stricter or procurement arrangements fail.
Operators may have to redesign parts of their infrastructure around Chinese alternatives if they cannot obtain the NVIDIA hardware originally specified.
Such changes could require additional investment and delay planned deployments.
NVIDIA's Export Restrictions Have Not Eliminated Chinese Demand
NVIDIA has previously indicated that US trade restrictions have effectively reduced its data center market share in China to zero.
However, the newly identified projects suggest that organizations in the country continue to seek access to its processors.
Some Chinese businesses reportedly prefer NVIDIA hardware even when availability is uncertain and delivery schedules are extended.
Major technology companies, including Tencent, Alibaba and ByteDance, have also continued investing in advanced computing infrastructure.
China has domestic semiconductor alternatives, but switching between computing platforms can involve changes to hardware configurations and software environments.
NVIDIA's established software ecosystem remains an important consideration for organizations developing AI applications.
The preference for its processors may therefore continue even as Chinese suppliers expand their own offerings.
Reports Raise Further Questions About GPU Supply Routes
The investigation follows earlier allegations involving shipments of restricted NVIDIA processors through intermediaries and overseas business arrangements.
Some reports have described transactions involving companies operating outside China or through corporate structures intended to conceal the final destination of hardware.
However, the existence of an overseas procurement arrangement does not automatically establish a violation of US export regulations.
Determining whether a particular transaction breached restrictions would require examining the processor model, shipment date, destination and applicable licensing requirements.
The latest findings also arrive shortly after Jensen Huang stepped down as NVIDIA CEO.
His departure has renewed attention on the company's previous approach to the Chinese market and its position on US chip export policies.
Nevertheless, the investigation does not establish a direct connection between Huang's departure and the reported computing projects.
Chinese AI Infrastructure Plans Face Continued Uncertainty
The reported projects highlight the difficulty of assessing the effectiveness of semiconductor export restrictions based solely on announced investments.
Although more than $3.2 billion in computing developments has been identified, it remains uncertain how many facilities will secure the processors needed to operate as planned.
Tighter enforcement could complicate procurement, while limited hardware availability may force organizations to reduce capacity or adopt domestic alternatives.
The findings nevertheless indicate that Chinese institutions continue to pursue advanced AI computing infrastructure built around NVIDIA technology.
For now, the available records demonstrate substantial planned investment, but they do not confirm that all the restricted processors have been delivered or that every project will be completed.
Discussion (0)
Be the first to comment.