A broad investigation into government-backed procurement in China has uncovered numerous AI data center initiatives that rely on prohibited American hardware to build out their computing infrastructure. These projects, which utilize H100, H200 and H800 processors, represent an investment of more than 23 billion yuan, or $3.2 billion, in total costs including buildings and power systems. The efforts span various universities and ventures supported by local governments, according to a review of over 150 contracts and procurement notices by Reuters.
The discoveries contrast with Nvidia’s own assertions that US export controls had effectively barred the company from China’s AI data center market. While CEO Jensen Huang previously stated that the company’s regional market share had fallen to zero, an IDC report found that Nvidia likely accounted for 34% of Chinese AI accelerator shipments during the first half of 2026.
