Technology news and discussion on TalkFuse.
US plans licences for overseas data centres serving restricted countries
The Trump administration is preparing new export controls on advanced AI chips and servers to stop companies in China and other restricted countries from accessing computing power through data centres abroad. The measure extends existing restrictions on direct chip sales, aiming to limit the use of US technology to train or run advanced AI models that could support military or intelligence activities.
Under the interim rule, licences would be required for exports, re-exports or transfers to data centres worldwide if they could provide compute to restricted countries. It covers facilities or networks above 12 million total processing performance, roughly equivalent to about 750 Nvidia H100 or H200 chips, and includes indirect access through brokers. Separate licence requirements would apply to advanced chips sent to Malaysia, Singapore, Indonesia, Thailand and Brazil, which the Commerce Department considers at higher risk of diversion; existing data centre operators would have 90 days to apply. The rule exempts commercial recommendation systems and rescinds the Biden administration’s 2025 AI Diffusion Rule.
In short
- The US plans to restrict remote access to advanced AI computing.
- The threshold is about 750 Nvidia H100 or H200 chips.
- New chip licensing rules also cover five additional countries.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters argue that chip export limits are ineffective if restricted countries can obtain the same computing power through data centres abroad. Licensing could close that loophole and reduce the risk that US technology enables advanced military or intelligence capabilities, while case-by-case review allows legitimate commercial uses to continue.
The case against
Critics argue that extending controls to overseas data centres and indirect access could burden operators and customers well beyond the restricted countries, creating uncertainty and compliance costs for lawful business. They may also contend that broad licensing rules could disrupt global technology supply chains and weaken US firms’ competitiveness without reliably preventing determined actors from obtaining computing power elsewhere.
Explained simply
US export controls limit where certain advanced computer chips can be sold, because they can provide the processing power used to develop and run artificial intelligence. Data centres are buildings full of computers that rent this capacity to customers, who may access it remotely.
The US Commerce Department oversees export controls, while chipmakers such as Nvidia supply some of the advanced hardware involved. The proposed licensing system would also cover overseas facilities if they could provide computing capacity to countries facing US restrictions, including China, even when access is arranged through intermediaries.
The policy concerns how governments can limit access to technology that may have military or intelligence uses while allowing other commercial services to continue. It also affects data centre operators and chip suppliers in countries that host or buy the equipment.
AI Americas Asia Technology World
Read the full article at the source →
Originally published by TechCrunch as “US tightens AI chip export controls to deter Chinese companies from accessing compute remotely”.