Chinese AI projects worth $3.2B still rely on restricted Nvidia hardware

The Hidden Infrastructure

A recent review by Global Banking & Finance Review has uncovered a massive scale of Chinese AI investment that appears to bypass the spirit of US export controls. By examining more than 150 procurement notices, investment plans, and corporate disclosures, researchers identified projects worth more than 23 billion yuan ($3.2 billion) that center on restricted Nvidia processors.

These projects are not limited to the major tech giants in China's wealthy coastal cities. The data shows that Nvidia technology is reaching into remote western provinces, local government-backed ventures, and various universities. This suggests that the demand for high-performance computing is widespread across different sectors of the Chinese economy.

The reported value of these projects includes planned investments and signed contracts, as well as completed installations. It is important to note that these figures cover the entire infrastructure, such as buildings and power systems, rather than just the cost of the Nvidia chips themselves. This paints a picture of a country still building out the physical backbone for AI using American components.

Nvidia's Market Reality

The situation presents a confusing contradiction between official company statements and independent data. Nvidia has stated that US export controls have effectively shut it out of the Chinese AI data-centre market. Consequently, the company has excluded China data-centre computing revenue from its financial forecasts due to regulatory uncertainty.

However, other reports offer a different perspective on the actual flow of hardware. An IDC report cited by Reuters estimated that Nvidia might still account for 34% of Chinese AI-accelerator shipments in the first half of 2026. These two figures measure different things, but they highlight the difficulty of tracking how restricted hardware actually reaches end users.

The discrepancy may be due to how the hardware is being moved or integrated into the supply chain. While Nvidia may no longer be selling directly to the largest firms, the demand for its specific processors, such as the H100, H200, and H800, remains a cornerstone of Chinese computing plans. This creates a complex environment for both the company and the regulators trying to enforce limits.

Leadership and Scrutiny

The landscape for Nvidia has become even more volatile following the resignation of co-founder Jensen Huang. His departure comes amid a scandal involving a businesswoman with suspected intelligence ties to Beijing. This development could weaken Nvidia's ability to lobby for eased restrictions as US lawmakers push for even tighter controls.

While the board launched an independent investigation into the matter, they stated they had no reason to believe company secrets were compromised. There has been no public finding that Huang violated any export-control laws. Nevertheless, the scrutiny has intensified the pressure on the company to navigate a minefield of international regulations.

The allegations involve a cloud partner, Megaspeed, which has denied illegally diverting chips to China. While the specific projects identified in the Reuters review were not linked to this company, the broader atmosphere of investigation is creating significant risk. For Nvidia, the challenge is balancing its dominance in the global AI market with the increasing difficulty of operating in a fractured geopolitical environment.

The Enforcement Dilemma

The ongoing situation highlights a growing debate over how to effectively manage technology exports. Analysts suggest that the focus should shift from who is buying the chips on the surface to who is ultimately using the computing power. This shift in perspective could change how the US government monitors the flow of high-end AI hardware to China.

Since 2022, the US has tightened restrictions on advanced processors and even modified chips designed specifically for the Chinese market. The Trump administration also moved to review H200 export licenses individually, subject to specific conditions. These layers of bureaucracy are intended to slow down China's progress in artificial intelligence.

However, the existence of billions of dollars in the pipeline suggests that the infrastructure is already being planned and funded. Operators of these facilities face risks of supply disruptions or costly redesigns if they are forced to switch to domestic chips. The tension between these existing plans and new regulations will likely define the next phase of the global tech trade war.