The US Commerce Department has moved to tighten rules on sales of advanced semiconductors to companies headquartered in China, regardless of where those firms operate. The new guidance is designed to block a route that had allowed some Chinese AI companies to obtain high-end Nvidia chips through overseas subsidiaries.
Previously, the gap in policy was largely geographic. A company headquartered in China that could not buy restricted chips directly at home might still try to purchase them through a related entity in another country. The latest guidance changes that approach by tying export-license requirements to the location of a company’s headquarters rather than the address of the purchasing entity.
The Commerce Department posted the guidance on Sunday. It extends licensing requirements to advanced chips sold to any entity that is China-headquartered, even if that entity is based elsewhere. The change is aimed at future shipments, not equipment that has already been delivered.
The rule affects some of the most powerful chips on the market, including Nvidia’s Rubin and Blackwell processors and AMD’s MI350x. Those products are widely seen as critical for AI training and high-performance computing, making them central to the broader US effort to limit China’s access to frontier semiconductors.
One industry source familiar with supply chains told Reuters that hundreds of thousands of advanced chips may have reached Chinese-linked entities abroad during the period when the loophole was open. That estimate underscores the scale of the issue the government is trying to address, though the exact volume is not publicly known.
The policy shift follows a series of changes to US chip controls over the past several years. Washington began restricting China’s access to advanced semiconductors in 2022 and has expanded those controls repeatedly since then. But enforcement has faced persistent workarounds, including the use of third-country subsidiaries and, in some cases, smuggling.
The current guidance also reflects the history of the so-called AI Diffusion rule, a broader framework finalized in the final days of the Biden administration. In May 2025, the Trump administration said it would not enforce that rule, leaving some uncertainty around sales to overseas units of Chinese companies for nearly a year. The new guidance now resolves that ambiguity.
While the move is significant, it stops short of ordering existing data centers to stop using the chips they already have. It also does not halt servicing for advanced computing equipment such as servers. That means the measure is intended to slow future exports rather than force immediate shutdowns or claw back hardware already in use.
The decision fits a familiar pattern in US export policy. Each new restriction seeks to narrow the channels available to China’s technology sector, but companies and intermediaries have often found ways around earlier controls. Recent enforcement actions have also shown that Washington is still pursuing alleged evasion schemes. Prosecutors have separately brought a case accusing a Thai company of helping route Nvidia chips to Alibaba.
Nvidia and AMD did not immediately comment on the guidance.
For Chinese firms, the change narrows one of the more straightforward legal paths to advanced US chips. For Washington, it is another attempt to make sure the latest controls follow the parent company, not just the shipping destination.