U.S. Weighs Rule to Block China’s Remote Access to Nvidia AI Chips

The U.S. Department of Commerce is weighing a rule to restrict Chinese AI companies’ remote use of advanced Nvidia chips in data centers across seven countries, aimed at closing a route around direct-shipment limits.

The proposal would require licenses for shipments of Blackwell-class or newer chips to Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa. Buyers would have to disclose each chip’s end user and purpose and accept ongoing monitoring.

U.S. companies operating data centers in those countries could avoid the license requirement by blocking remote access from China, Russia and other countries subject to U.S. arms embargoes. They would still need to report local customers and computing capacity.

The department is consulting with other federal agencies, and the rule could be announced as soon as November. Existing U.S. controls govern chip shipments, while rentals of computing power at third-country data centers have remained outside that framework. Tencent, Alibaba and ByteDance have signed overseas data center deals to train models on Nvidia’s leading chips.