The Trump administration is considering sweeping new export controls that could require U.S. approval for nearly all global shipments of advanced artificial intelligence chips made by American firms such as Nvidia and AMD. Draft regulations described in multiple reports would expand today’s country-based limits into a worldwide licensing system, giving Washington broader authority over where major AI computing projects can be built and under what conditions.
The proposal targets powerful “AI accelerators” that are widely used to build and run modern AI systems. Several reports describe these chips as central to the current AI boom, with major tech companies buying them in large volumes for data centers that power services like ChatGPT and Gemini.
What the draft rules would do
Reports describe draft regulations that would make the U.S. government a gatekeeper for exports of advanced AI processors, extending oversight well beyond the current set of restricted destinations. One account says current U.S. export restrictions mainly apply to about 40 countries considered high risk, while the draft under discussion would broaden licensing to cover shipments to nearly any country.
The same set of reports emphasizes that the idea is not presented as a total ban on exports. Instead, the draft framework is described as a way to maintain U.S. influence over how global AI infrastructure grows, including large facilities used to train and operate AI models.
How approvals could work
Several outlets describe a tiered approach tied to the size of chip orders. Smaller purchases—such as shipments of up to 1,000 of Nvidia’s latest GB300 GPUs—are described as likely to face a simpler review process, with the possibility of exemptions in some cases.
Bigger deployments would face more scrutiny. Reports say companies planning large AI clusters could be required to obtain pre-clearance before they can apply for export licenses, and they may face compliance conditions. Examples mentioned include operational transparency requirements, disclosure of business activities or business models, and possible U.S. government site visits to data centers, depending on the specifics of the project.
For extremely large buildouts—described in one report as clusters powered by around 200,000 GB300 GPUs operated by a single company within one country—approvals could shift into direct government-to-government arrangements. In that scenario, approvals are described as potentially depending on national-security assurances and commitments tied to investment in U.S.-based AI infrastructure.
Market and industry reaction
Financial markets reacted after the draft rules were reported. One report said Nvidia shares fell as much as 1.9% and AMD declined about 2.3% in trading, reflecting investor concern that tighter export controls could complicate global sales.
Some reporting also notes that the policy discussion is ongoing and could change. One outlet said officials across federal agencies have been providing input, and it reported that the Commerce Department’s Bureau of Industry and Security, Nvidia, and AMD had not responded to requests for comment about the potential impact.
Concerns raised abroad
International reaction described in the coverage varies sharply depending on the source. A report in China’s state-backed Global Times framed the idea as an attempt to extend U.S. “long-arm jurisdiction” and quoted a Chinese analyst arguing it could face opposition, including from U.S. allies, because it would affect other countries’ AI ambitions and commercial interests.
That same Global Times report said the proposed permitting approach was being discussed as China steps up efforts to boost technology self-reliance. It highlighted a public appeal co-authored by executives and scholars from China’s semiconductor industry calling for a coordinated national push to build a homegrown equivalent of lithography giant ASML during China’s 15th Five-Year Plan period (2026–30), citing tightening U.S. technology restrictions and limits on access to advanced lithography tools.
Meanwhile, other reporting focuses less on political reaction and more on practical consequences: how quickly licenses would be granted, what conditions would be attached, and whether the approval process would slow the construction of major data centers outside the United States. In that telling, the central uncertainty is not whether AI chips can be sold at all, but how the rules would shape where the world’s biggest AI computing clusters end up being built.
