The Trump administration floated one of the most aggressive semiconductor export control proposals in history — then pulled it within days. The draft rule, which would have required US government approval before Nvidia, AMD, or any other American chip company could sell AI chips almost anywhere outside the United States, was quietly withdrawn on Friday after an interagency review. The episode reveals deep tensions within the administration over how far to push chip controls, and what the costs of getting it wrong might be.
The proposed rule was sweeping in a way that went well beyond anything the Biden administration had attempted. Biden's "AI diffusion rule" tiered countries by trust level and restricted the most advanced chips to a narrower set of partners — an approach the Trump administration had already rescinded last May as too complicated. The new draft would have replaced that with something simpler and more aggressive: a blanket requirement for a US export licence for AI chip sales to essentially every country except a small set of close allies. Nvidia's global business — the majority of which now comes from outside the United States — would have operated under ongoing government approval for every major sale.
The industry reaction was swift and pointed. Nvidia has spent the past year watching its China sales evaporate despite receiving approval to sell H200 chips to Chinese customers with a 25% revenue surcharge attached — a deal that generated zero revenue because Beijing blocked the imports on its end. The lesson many in the industry drew was that aggressive export controls create costs and friction on the American side without reliably preventing Chinese access to advanced AI capability, since Chinese labs can develop their own chips or access compute through alternative routes.
The withdrawal doesn't mean the underlying policy debate is resolved — it means the administration couldn't reach internal consensus on it. The Commerce Department, which manages export controls, was pushing for tighter restrictions. Other parts of the administration, more attuned to the economic impact on US semiconductor companies and the diplomatic friction with allied countries that would have needed to seek US permits to buy chips, pushed back. The interagency review concluded without agreement, and the rule was pulled.
What's left is genuine uncertainty. The Biden-era diffusion rule is gone. The replacement proposal is gone. Nvidia and AMD are currently operating under a patchwork of country-specific restrictions that don't amount to a coherent global framework. That ambiguity is uncomfortable for companies trying to plan data centre deals and supply agreements that run years into the future. It's also uncomfortable for allies who have been told — repeatedly, by successive administrations — that American chip access is a feature of the partnership, only to find the terms shifting with each policy cycle.
The episode also exposes a fundamental tension in US chip policy. The goal is to maintain American AI leadership by preventing adversaries from accessing the compute needed to catch up. But American semiconductor companies derive much of their revenue from selling globally, and that revenue funds the research and manufacturing investment that keeps American chips ahead. A policy that significantly constrains those sales doesn't just hurt Nvidia's stock price — it potentially undermines the industrial base the policy is trying to protect. Getting that balance right has defeated several administrations now, and the withdrawal of this draft suggests it's defeated this one too, at least for the moment.