US closes loophole restricting China's remote access to advanced AI chips

The US Department of Commerce announced on Monday (November 16, 2026) an interim rule to restrict Chinese companies' remote access to advanced artificial intelligence chips hosted in overseas data centers.

The measure aims to close a loophole that allowed circumvention of existing export restrictions. Chinese companies, barred from directly buying the most powerful chips from NVIDIA and other American manufacturers, rented computing capacity from data centers in countries such as Malaysia, Singapore, Indonesia, Thailand, and Brazil.

According to the Bureau of Industry and Security (BIS), the rule requires a license for the export, reexport, or transfer of AI chips and servers to any data center worldwide that provides more than 12 million Total Processing Performance (TPP) to companies headquartered in China, Macau, or other D:5 group countries (including Russia and Iran). This volume roughly equals the capacity of about 750 NVIDIA H100 or H200 GPUs.

Data centers already serving these customers have 90 days to obtain the license after publication in the Federal Register. Additionally, exports of advanced chips to Indonesia, Malaysia, Singapore, Thailand, and Brazil will now require a license, as they are considered higher risk for diversion.

The rule does not apply to commercial recommendation systems, such as those used in e-commerce or social media. It also formally rescinds the Biden administration's AI Diffusion Rule.

The announcement comes days after Jensen Huang's resignation as NVIDIA CEO amid investigations into possible chip diversions via companies like Singapore-based Megaspeed. Official sources describe the measure as necessary to prevent remote access to American technology from being used in military or intelligence activities that threaten US national security.

Sources

This article was produced with the assistance of artificial intelligence and editorially reviewed.

Por Grok