Chinese AI Firms Seek Overseas Data Centers to Dodge Nvidia Export Controls

According to recent foreign reports, sources familiar with the matter have revealed that several Chinese artificial intelligence companies are actively seeking to lease overseas data centers to access the computational resources of NVIDIA chips. The primary driver behind this move is concern over a scandal involving NVIDIA CEO Jensen Huang, with fears that the U.S. government may tighten restrictions on the remote use of advanced processors as a result. Although the reports did not detail the nature of the scandal, such concerns are not unfounded, given the strict U.S. controls on high-end semiconductor exports and the sustained pressure on Chinese tech firms. In recent years, the United States has repeatedly expanded the Export Administration Regulations (EAR) and the National Security Council (NSC) lists, placing multiple high-performance GPUs and AI accelerators under restriction and requiring companies to obtain special permits for cross-border usage. If Huang’s conduct is found to negatively impact corporate governance or compliance, NVIDIA’s reputation and the U.S. export licenses for its products may face further scrutiny, thereby subjecting Chinese companies to more stringent import conditions.

NVIDIA plays an indispensable role in the global AI supply chain. Its GeForce RTX series, Tesla series GPUs, and the recently launched Hopper architecture have become core hardware for training and inference in many machine learning models. The U.S. government has placed these chips on the "High-Tech Export Control List" to prevent advanced computing capabilities from falling into the hands of potential competitors. Over the past two years, driven by China’s "New Infrastructure" and "Mass Entrepreneurship and Innovation" policies, domestic demand for high-performance GPUs among AI enterprises has grown explosively. However, China’s local chip manufacturing capacity remains insufficient to meet this demand, particularly for high-bandwidth, low-latency GPUs. U.S. restrictions on exporting NVIDIA chips to China have forced many Chinese companies to seek alternatives, including developing proprietary chips, partnering with TSMC, or leasing overseas data centers.

In this context, leasing overseas data centers has become a critical contingency path for Chinese AI enterprises. By establishing data centers in the United States, Europe, or other countries compliant with export regulations, companies can leverage NVIDIA’s computational advantages without directly importing the hardware. This approach not only mitigates compliance risks associated with export controls but also helps avoid additional restrictions stemming from any credibility crisis involving NVIDIA in the U.S. It is noteworthy that such cross-border computing leases often involve complex contractual and regulatory reviews, particularly regarding compliance with U.S. export control laws (EAR) and international trade laws. If NVIDIA’s reputation suffers further damage, the review process for relevant export licenses may become more stringent, potentially leading to the suspension or revocation of many chips that were previously available.

The impact of this trend on the global AI supply chain cannot be ignored. First, if the U.S. tightens export restrictions on NVIDIA products, Chinese AI enterprises will be compelled to accelerate the development of proprietary chips or seek other suppliers, thereby driving technological breakthroughs in the local semiconductor industry. Second, the increase in cross-border computing leases may pose challenges to U.S. data privacy and security regulations, as data transmitted and processed across borders must meet the compliance standards of multiple nations. Third, if U.S. export controls on NVIDIA tighten further, the global AI research community will face computing bottlenecks, potentially slowing the pace of AI technological progress and intensifying international competition and fragmentation in the AI industry.

For NVIDIA, the scandal involving Jensen Huang may not only damage the company’s reputation in global markets but also trigger stricter U.S. government reviews of its product exports. If the company is deemed to have deficiencies in compliance or governance, it could be added to the "Entity List," thereby losing the right to sell in key markets such as China. This would not only directly affect NVIDIA’s revenue but also weaken its competitive advantage in the AI hardware market. For the United States, maintaining control over exports of high-end AI hardware is a key means of safeguarding national security and technological superiority. If NVIDIA’s reputation is compromised, the government may seize the opportunity to further expand the scope of export controls, potentially adding other critical chips to the restricted lists.

In summary, the potential risks arising from the Jensen Huang scandal have prompted Chinese AI enterprises to turn to leasing overseas data centers to ensure their computing needs. This move not only reflects the deep structural dynamics of U.S.-China tech competition but also highlights the vulnerability of the global AI supply chain under dual political and economic pressures. In the future, as U.S. export control policies evolve and NVIDIA’s corporate governance shifts, the competitive landscape in the AI industry between the two sides of the Taiwan Strait may be further reshaped. (Source: CNA)