US Tightens AI Chip Export Rules as China Subsidiary Loophole Comes Under Fire
Washington is moving to close a major gap in its AI chip export controls, a loophole that has reportedly allowed Chinese companies to keep accessing advanced processors from NVIDIA and AMD through overseas subsidiaries.
For months, the question of how China continues to obtain high-performance AI hardware has been a major concern for U.S. officials. Advanced GPUs are essential for training large artificial intelligence models, and chips from NVIDIA and AMD remain among the most sought-after components in the global AI race. Despite export restrictions aimed at limiting China’s access to cutting-edge computing power, companies with ties to China have continued finding ways around the rules.
Now, the Trump administration appears ready to take a tougher stance.
According to an update from the U.S. Commerce Department, licensing requirements will now apply more strictly to advanced AI chips destined for entities that are headquartered in China, even if those entities operate from another country. In practical terms, this means foreign subsidiaries of Chinese companies may no longer be able to purchase restricted AI processors as easily as before.
The move targets a workaround that emerged after the administration chose not to enforce the Biden-era AI Diffusion framework in May 2025. At the time, U.S. officials said their focus would be on limiting the global spread of Huawei’s Ascend AI chips while also preventing American AI hardware, particularly NVIDIA GPUs, from being used to train Chinese AI models.
However, that approach left room for Chinese companies to use foreign-based subsidiaries to buy advanced AI accelerators. These overseas units could acquire chips outside mainland China and then use them to support the AI development goals of their parent companies.
A chip industry source cited in media reports claimed that hundreds of thousands of advanced AI chips from NVIDIA and AMD may have reached Chinese-linked subsidiaries over the past year. If accurate, that scale would highlight just how difficult it has been for the U.S. to control the flow of high-end AI hardware in a fast-moving global supply chain.
The latest Commerce Department action suggests Washington is shifting from broad policy goals to more direct enforcement. By focusing on where a company is headquartered rather than only where it is physically located, U.S. officials are attempting to prevent Chinese firms from using international branches as a backdoor into restricted technology markets.
This is not the first sign of a tougher approach. Earlier in 2025, U.S. officials were reportedly examining the role of front companies in countries such as Thailand and Malaysia, where businesses may have been used to route AI chips toward China. At the time, no formal advisory had been issued, but the concern was already clear: advanced GPUs were still finding paths into restricted hands.
The issue has also reached the criminal enforcement stage. Federal prosecutors recently charged the co-founder of Super Micro Computer and two associates in connection with an alleged scheme involving roughly $2.5 billion worth of NVIDIA GPUs. Authorities claim the chips were diverted to China through a network of front companies, with orders placed under misleading arrangements and shipments allegedly concealed through stealth packaging methods.
The broader stakes are significant. AI chips are now central to national security, economic power, and technological leadership. The ability to train advanced AI models depends heavily on access to the most powerful GPUs and accelerators. For the United States, restricting China’s access to these chips is part of a wider effort to maintain an edge in artificial intelligence and limit potential military or strategic uses of American technology.
For NVIDIA and AMD, the tightening rules add another layer of complexity to an already challenging market. China has historically been a major demand center for AI and data center chips, but U.S. export controls have forced chipmakers to redesign products, seek licenses, and navigate shifting regulatory boundaries.
The new restrictions could also increase pressure on Chinese firms to accelerate domestic alternatives. Huawei’s Ascend chips have already become a focal point in China’s push to reduce reliance on U.S. technology. If access to NVIDIA and AMD hardware becomes even more limited, Chinese companies may have stronger incentives to invest in homegrown AI processors, software ecosystems, and supply chains.
Still, replacing top-tier U.S. GPUs is not easy. NVIDIA’s hardware advantage is closely tied to its software ecosystem, especially tools used by AI developers and data centers. AMD is also pushing deeper into the AI accelerator market, making its chips another important target for export control scrutiny.
The latest U.S. move shows that the AI chip battle is far from over. As Washington closes one loophole, companies and governments will continue adapting to the changing rules. What remains clear is that advanced AI processors have become one of the most strategically important technologies in the world, and access to them will remain a central flashpoint in U.S.-China tech competition.
