MAIN CONTENT
When a chip restriction reaches the cloud
The US Bureau of Industry and Security announced an interim final rule on November 16 to address a gap between restrictions on shipping advanced chips and access to those same chips from abroad. Under the framework, certain overseas data centres that provide significant computing capacity to restricted-country customers may need US authorization. The regulation also adds licensing requirements for exports, re-exports and in-country transfers of advanced semiconductors to Singapore, Malaysia, Thailand, Indonesia and Brazil. These countries are not subject to a blanket prohibition: licensing, end-use information and customer due diligence are central to the design.
That distinction matters. A company does not necessarily need to possess an Nvidia processor to train an AI model; it may rent capacity from a remote service provider. Washington argues that without oversight of this arrangement, restrictions at the physical border can be circumvented through cloud access. The new policy therefore extends compliance questions from chip vendors to the operators and customers of computing infrastructure. Singapore’s government acknowledged the rule on November 16, reinforcing that the industry must now interpret legal requirements alongside engineering and commercial ones.
















