China’s Demand for Restricted NVIDIA AI GPUs Remains Strong Despite U.S. Export Controls
China’s appetite for advanced NVIDIA AI GPUs appears far from slowing down, even as U.S. export restrictions continue to limit official access to the company’s most powerful data center chips. A new investigation suggests that Chinese organizations are still building major computing projects around restricted NVIDIA processors, raising fresh questions about how effective the current trade controls really are.
According to Reuters, an independent review of procurement notices, investment plans, construction contracts, and corporate disclosures found more than 150 projects in China tied to NVIDIA AI hardware. Together, these projects are valued at over $3.2 billion and involve chips such as the NVIDIA H100, H200, and H800, all of which are highly sought after for artificial intelligence training, cloud computing, and high-performance data center workloads.
The findings suggest that demand for NVIDIA’s advanced AI chips remains intense across China, despite the company previously saying that U.S. sanctions had effectively reduced its data center market share in the country to zero. While official sales channels have been restricted, the investigation indicates that Chinese buyers may still be finding ways to obtain or plan around NVIDIA hardware through alternative routes.
What makes the situation especially notable is that these projects are not limited to China’s largest technology companies. The review found NVIDIA AI GPUs linked to local government-backed initiatives, universities, regional computing centers, and other public or semi-public projects. This points to a broader national push to secure high-performance computing power for artificial intelligence development.
One reported example involves a facility in Gansu province in northwestern China, which is said to be undergoing changes that could allow it to house thousands of NVIDIA processors by the end of 2026. If completed, projects like this could strengthen China’s domestic AI infrastructure at a time when access to cutting-edge chips has become a major point of tension between Washington and Beijing.
Still, the situation remains uncertain. Analysts note that project documents do not always confirm whether the listed NVIDIA GPUs have actually arrived at the planned locations. Some projects may be proposals, planned purchases, or construction efforts built around the expectation that chips will eventually be obtained. If enforcement becomes stricter, operators may be forced to redesign systems around domestic Chinese processors, which could increase costs and delay deployment.
The report also highlights a key challenge for U.S. policymakers: restricting advanced chips on paper is easier than fully controlling global supply chains. NVIDIA’s AI GPUs are compact, extremely valuable, and in high demand, making them difficult to keep out of restricted markets entirely. Smuggling networks, third-party resellers, shell companies, and overseas intermediaries can complicate enforcement.
Chinese companies also appear reluctant to rely solely on domestic alternatives. Major AI players have shown strong interest in NVIDIA hardware because of its performance, software ecosystem, and widespread support among AI developers. Even when delivery times are long or availability is uncertain, many buyers still appear willing to wait for NVIDIA GPUs rather than switch to less mature options.
The continued pursuit of NVIDIA H100, H200, and H800 chips underscores how central these processors have become to the global AI race. For China, access to advanced GPUs can determine the speed and scale of large language model training, scientific computing, autonomous systems, and cloud-based AI services. For the United States, controlling that access has become a strategic priority.
The growing evidence of China-based projects built around restricted NVIDIA AI GPUs suggests that export controls may be slowing access, but not eliminating demand. As artificial intelligence becomes increasingly important to national competitiveness, the battle over high-end chips is likely to remain one of the most closely watched issues in the global technology industry.
