US tightens AI chip export rules to restrict remote computing access by Chinese entities
Published: 06:13 18 Nov 2026 AEDT
The US has tightened restrictions on advanced artificial intelligence chips and computing systems, extending export controls to remote access to computing capacity by Chinese entities and other restricted customers.
The Commerce Department’s Bureau of Industry and Security (BIS) issued an interim final rule requiring licences for covered chips and systems used by data centres providing more than 12 million aggregate total processing performance, equivalent to 4.2 GB200 NVL72 racks, for companies headquartered in China, Macau and other US arms-embargoed countries.
The restrictions also cover intermediaries knowingly providing computing services for restricted entities, with existing arrangements required to obtain licences within 90 days of the rule taking effect.
Shipments of advanced computing products to Indonesia, Malaysia, Singapore, Thailand and Brazil will also require licences, although exemptions apply to certain US-parented customers and qualifying US-headquartered end users.
BIS has exempted computing used exclusively for commercial recommendation systems, subject to reporting and due diligence requirements.
The rule also strengthens customer screening obligations for chip exporters, foundries and outsourced assembly and testing providers, including checks on direct and indirect ownership.
BIS is rescinding its previous AI Diffusion framework, including its country-tier system and model-weight controls, with the new rule taking effect upon publication in the Federal Register.
The restrictions could create additional revenue headwinds for Nvidia and AMD, as remote computing access by Chinese customers faces tighter controls, according to the accompanying analyst assessment.
The assessment estimates that risk channels could account for as much as 20% of Nvidia’s revenue, with GPU-based AI architectures considered more exposed than custom chips designed for US hyperscalers.
It also warns that the measures could affect access to computing resources for emerging AI companies using open models, adding to uncertainty surrounding data-centre investment and AI financing.