US Plans Worldwide AI Chip Export Controls for Nvidia, AMD

The Trump administration is drafting rules that would require US government approval to ship Nvidia and AMD AI chips to any country in the world, including close allies like the UK, Germany, and Japan. The proposal, first reported by Bloomberg on March 5, 2026, would turn the US Commerce Department into a gatekeeper for every large-scale AI infrastructure buildout on the planet.

What Are the Proposed AI Chip Export Controls?

The draft rules would expand existing China-focused chip restrictions into a global licensing system. The Bureau of Industry and Security (BIS) has written regulations that create a three-tier framework based on shipment volume.

Under 1,000 GPUs gets a standard review — the lightest touch, similar to existing export paperwork. Medium-volume deployments require preclearance before a company can even apply for a license. Orders of 200,000 GPUs or more would require formal certifications from government officials in the host country.

For context: 200,000 Nvidia H100s at current pricing represents roughly $6-10 billion in hardware. That is the scale of a single hyperscaler data center build. Saudi Arabia, UAE, and Japan have all announced projects at that scale in the past 18 months.

Why Markets Dropped Immediately

Nvidia fell 1.8%, AMD declined 2.2%, and Micron dropped 3.4% on the day the Bloomberg report landed.

The financial exposure is real. Nvidia's data center revenue reached $115 billion in fiscal 2025, and the overwhelming majority came from outside the United States. Nvidia had already halted H200 exports to China before these broader controls were proposed — this proposal would extend that friction to every country. Any system that adds government approval to international GPU sales hits the core of the business.

The Trump administration has been careful to say these rules are not an export ban. The stated goal is to keep US companies as preferred suppliers while giving Washington control over who gets the largest AI buildouts. Whether that framing survives contact with regulatory process remains to be seen.

China Responds With a Subsidy Program

Within days of the Bloomberg report, China announced a new AI and semiconductor subsidy program aimed directly at reducing dependence on Nvidia and AMD hardware. The initiative doubles down on self-sufficiency goals and accelerates funding for domestic chip alternatives.

This pattern has repeated since 2022. The US restricts, China funds domestic alternatives. Each cycle tightens the global supply chain and raises costs for everyone not in China or the US.

What This Means for Developers and Infrastructure Teams

Small deployments stay largely unaffected. Orders under 1,000 GPUs face the lightest regulatory burden, which covers most individual company deployments.

The friction concentrates at the top. National AI programs, hyperscaler expansions, and large enterprise clusters in the 10,000 to 200,000+ GPU range face the most uncertainty. A data center project with a 9-month build timeline could add several months of approval process if these rules take effect as drafted.

Procurement teams at companies building AI infrastructure outside the US should start tracking BIS comment periods now. The rules are not final and could change significantly or be scrapped entirely. But planning as if they will pass is the conservative posture.

Key Takeaways

• 1,000 GPUs — the threshold below which standard export review applies
• 200,000 GPUs — the scale requiring host country government certifications
• 1.8% — Nvidia share price drop on the day of the Bloomberg report
• 2.2% — AMD share price drop on the same day
• For developers: Orders under 1,000 GPUs face minimal new friction; teams planning large GPU clusters outside the US should model approval delays into procurement timelines
• What to watch: BIS comment period and final rule publication, expected mid-2026; whether allied nations (UK, EU, Japan) push back on being subject to the same framework as adversaries