US Commerce rule requires licences for AI chips in data centers serving Chinese firms
A BIS rule of 16 November 2026 requires licences for AI chips sent to data centers renting large compute to Chinese firms, and to five more countries.
By Berkant A.Published
Tech AI LatestThe US Commerce Department's Bureau of Industry and Security issued an interim final rule on 16 November 2026 to stop Chinese companies renting advanced AI chips through data centers abroad. Supplying any data center that gives such companies more than 12 million aggregate TPP of compute now needs a licence, as do chip shipments to Indonesia, Malaysia, Singapore, Thailand and Brazil. The rule also formally scraps the AI Diffusion Rule.
Key facts
- Released: 16 November 2026 by the Bureau of Industry and Security as an interim final rule (RIN 0694-AK50)
- Threshold: licence needed to supply a data center giving covered customers more than 12 million aggregate TPP, counting linked sites
- Covered customers: companies headquartered in, or with an ultimate parent in, Macau or Country Group D:5, which includes China
- New country licences: Indonesia, Malaysia, Singapore, Thailand and Brazil, reviewed case by case
- Existing data centers serving covered customers: 90 days to obtain a licence
- Also: AI Diffusion Rule regulations formally rescinded
What the rule does
The US Department of Commerce's Bureau of Industry and Security (BIS) released an interim final rule on 16 November 2026 titled Addressing Remote Access to Advanced Semiconductors. BIS says the rule was written by and at the direction of the Treasury Department and the National Security Council, and that its aim is to make sure existing restrictions on sensitive American technology cannot be circumvented through remote computing arrangements involving data centers in third countries.
Until now, US export controls on AI chips mainly governed where the hardware was shipped. A company headquartered in China could not buy the most advanced chips, but it could rent their computing power from a data center in another country. The new rule targets that gap. According to coverage of the rule text by The Next Web and Tech Times, an exporter now needs a licence when it knows that chips, software or technology are going to a data center that provides, or will provide, more than 12 million aggregate TPP (total processing performance) to a covered customer. Capacity across digitally or physically linked data centers counts toward the threshold.
Covered customers are companies headquartered in Macau or in a destination in Country Group D:5, which includes China, or whose ultimate parent is. Data centers already providing that level of compute to such customers have 90 days to obtain a licence. The rule also adds a licence requirement for advanced chips shipped to Indonesia, Malaysia, Singapore, Thailand and Brazil, with applications reviewed case by case, and it exempts qualifying commercial recommendation systems such as those used by online retailers and social media services, NPR reports.
The Megaspeed case behind it
The rule follows weeks of scrutiny of Megaspeed, a Singapore-based company with facilities in Malaysia that is suspected of selling remote access to US AI chips to Chinese companies. NPR reports that Megaspeed and its Malaysian affiliate bought about $4.8 billion of NVIDIA AI chips and other hardware over three years, and that Commerce and the Justice Department have confirmed a criminal investigation. Megaspeed denies wrongdoing.
The case reached the top of NVIDIA. The company's board opened an inquiry into reports about chief executive Jensen Huang on 8 November, and Huang resigned on 12 November, with chief financial officer Colette Kress named interim CEO. NVIDIA has not commented on the rule in the coverage reviewed here.
What changes for cloud providers and chip buyers
The burden falls on exporters and data center operators, who now have to know who their tenants are and where those tenants' parent companies are based. Cloud and colocation providers in Southeast Asia, which have been building capacity for regional and Chinese customers, face the biggest change: shipments to Singapore, Malaysia, Thailand and Indonesia need a licence regardless of who the end customer is. Brazil is the only country outside Asia on the list. For buyers, a licence requirement adds time and paperwork to every order, even when the end customer is not a covered company.
The rule also formally rescinds the regulations that implemented the Biden-era AI Diffusion Rule, which set country tiers for chip exports and was never enforced. The current administration had announced its withdrawal in May 2025. The move fits the administration's wider supply chain push, including the Pax Silica declaration signed with nine partners a day earlier.
What to watch next
As an interim final rule it takes effect on publication in the Federal Register, with comments accepted afterwards; the publication date and comment deadline were not confirmed at the time of writing. Open questions are how BIS will check tenant ownership across linked sites and how quickly it can process licence applications from the five named countries. More export control coverage is on the AI policy hub.
What changed
| Before | US rules controlled where advanced AI chips were shipped, but Chinese firms could rent their compute from data centers in third countries without a licence. |
|---|---|
| After | Supplying a data center that gives Chinese and other D:5 companies more than 12 million aggregate TPP needs a licence, and shipments to five more countries need one too. |
| Who it affects | Cloud and data center operators abroad, chip exporters such as NVIDIA, and Chinese AI companies renting compute overseas |
Questions people ask
- What is the US remote access rule for AI chips?
- An interim final rule from the Bureau of Industry and Security, released on 16 November 2026, that requires a licence to supply chips to data centers giving companies from China and other countries of concern large amounts of remote computing power.
- Which countries now need a licence for AI chip shipments?
- The rule adds Indonesia, Malaysia, Singapore, Thailand and Brazil, with applications reviewed case by case.
- What happened to the AI Diffusion Rule?
- The new rule formally rescinds the regulations that implemented it. The rule was never enforced, and its withdrawal was first announced in May 2025.
Sources
- BIS Issues Rule Addressing Remote Access to American Super Intelligence Technology (Bureau of Industry and Security, 16 Nov 2026)
- U.S. tightens controls on remote access to AI technology amid Nvidia fallout (NPR, 16 Nov 2026)
- US Moves to Curb China's Remote Access to Nvidia's AI Chips (Bloomberg, 16 Nov 2026)
- After Nvidia scandal, Commerce Department moves to curb remote access to AI chips (CBS News, 16 Nov 2026)
- US curbs China's remote access to AI chips through overseas data centres (The Next Web, 16 Nov 2026)
- US moves to block China's remote access to AI chips amid Nvidia scandal (Al Jazeera, 17 Nov 2026)
Reference pages
- NVIDIA company
- US AI chip export controls policy
Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.
Comments
Loading comments.





