Nvidia chips at core of $3.2bn China AI plans

Chinese AI build‑out leans on restricted Nvidia chips

More than 23 billion yuan, or about $3.2 billion, of Chinese AI computing and data‑center projects have been designed around Nvidia’s export‑restricted H100, H200 and H800 processors. The projects span planned investments, signed contracts and completed installations, highlighting extensive demand for the US chipmaker’s hardware despite controls that bar these advanced chips from export to China.

The projects represent a wide range of activity, from new regional computing hubs to smaller, targeted purchases of individual accelerators. They collectively show that Nvidia’s restricted products remain central to China’s AI infrastructure plans even as formal market access has been curtailed.

Major data‑center projects across Chinese provinces

The largest identified project is a proposed 6.2 billion‑yuan AI computing center in Xianyou county in Fujian province, explicitly planned around Nvidia H100 processors and still seeking commercial partners. A subsidiary of state‑owned engineering group Sinomach has also secured a contract for a 3 billion‑yuan computing center in Shaanxi that calls for both H100 and H200 chips.

Beyond these large‑scale builds, individual institutional purchases point to more diffuse use of Nvidia’s restricted chips. Shenzhen University won a contract for three H100 processors in September, and US‑blacklisted Northwestern Polytechnical University acquired two H200 chips the following month.

Regulatory backdrop and Nvidia’s stated China position

Nvidia has said that US export controls have effectively shut it out of China’s AI data‑center market. It has excluded China data‑center computing revenue from its forecasts, citing regulatory uncertainty surrounding sales of its advanced processors.

The identified Chinese projects measure a different dimension of activity, focusing on where restricted processors are planned or installed rather than on Nvidia’s direct reported sales. The documents underpinning these project values do not establish how the restricted hardware reached Chinese users or through which intermediaries, leaving key aspects of compliance and routing unclear.

Heightened US scrutiny after Jensen Huang’s resignation

US scrutiny of Nvidia’s China exposure has intensified following Jensen Huang’s resignation as chief executive. The departure came after allegations involving Alice Huang, a former Nvidia executive associated with Singapore‑based cloud partner Megaspeed, which has denied illegally diverting Nvidia chips to China.

There has been no public finding that Jensen Huang violated export‑control laws. Nonetheless, the combination of leadership change and the scale of Chinese projects built on restricted chips has sharpened attention on how effectively Washington’s export controls are being enforced and what risks these controls pose to projects already underway.

Implications for Chinese AI projects and enforcement

The more than $3 billion in Chinese AI and computing projects tied to H‑class Nvidia processors face potential delays, redesigns or supply disruptions as US policymakers consider tighter rules. Planned and existing installations that depend on restricted chips may need alternative configurations if access to the hardware or to remote computing power is further constrained.

The projects, stretching from coastal technology hubs to more remote western provinces, illustrate that limiting exports does not automatically extinguish demand. They also underscore the challenge regulators face in overseeing complex global supply and deployment chains for high‑end AI semiconductors.