Tech & AI
US Reportedly Drafts Rule to Stop Chinese AI Firms Renting Nvidia Chips Abroad
Commerce is said to be adding seven countries, from Singapore to Brazil, to chip-control lists as soon as this month.
HedgeWeekly Tech & AI Desk ·
The US Commerce Department is drafting a rule to stop Chinese AI companies from using advanced Nvidia chips through data centres in third countries, The Information reported on 11 November, citing two people familiar with the matter. Existing export controls cover the physical chips but not remote access, meaning renting computing power in a foreign facility. The rules are not final and could be announced as early as this month, according to the reports.
The draft would add seven countries to the chip control list: Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa, according to South Korea's Seoul Economic Daily, which summarised the report. Imports of Nvidia's Blackwell-class processors or newer would need US approval, and importers would have to report the end user and intended use and accept continuing oversight. Chinese groups including ByteDance, Alibaba and Tencent have been accused of reaching Nvidia hardware through data centres in Southeast Asia.
Legal authority is contested. A Baker McKenzie lawyer told The Information that Commerce lacks clear statutory power to regulate remote access, according to AI Weekly. The House passed a bill giving the Bureau of Industry and Security that authority 369 to 22 in January, but the Senate has not acted. Seoul Economic Daily said Singapore accounted for 18% of Nvidia's billed revenue in fiscal 2025 while only 2% of revenue came from chips physically delivered there. The US government has not confirmed the draft.
Why it matters: the rule would target a main route by which Chinese firms obtain AI computing power, and billing hubs such as Singapore may see demand shift. Investors should watch whether the rule is published, whether it covers existing capacity or only new contracts, any legal challenge, and how regional cloud providers and Nvidia respond. Chip-export limits remain a key swing factor for Nvidia and its memory suppliers.