Washington has closed what it calls the remote-access loophole – the trick of renting compute abroad instead of importing chips at home. On November 16, the Commerce Department’s Bureau of Industry and Security issued an interim final rule (RIN 0694-AK50) that expands AI chip export controls to five new countries and, for the first time, reaches overseas data centers that quietly supply large volumes of computing power to China-linked customers.
The rule requires US licenses to export, re-export or transfer advanced semiconductors to Brazil, Indonesia, Malaysia, Singapore and Thailand – jurisdictions BIS deems present a “higher-than-normal risk of diversion” to China. Chinese firms have generally been barred from buying the most advanced Nvidia and AMD chips directly; the market’s workaround was renting capacity in third-country data centers. That workaround just got a lot more expensive.
Data centers worldwide now face the license regime
The second half of the rule is the sharper instrument. Data centers anywhere in the world must obtain a US license to provide significant volumes of advanced-chip compute to companies headquartered in China, Macau, or other countries in the BIS D:5 group (China, Iran, Russia, North Korea). Operators get 90 days to secure licenses to keep serving such customers, and the rule obliges data centers, foundries and chip exporters to run “robust due diligence” – including checks on whether customers are owned, directly or indirectly, by parties on US restricted lists.
Carve-outs exist: the five-country license requirement does not apply to chips destined for US-headquartered firms or their overseas subsidiaries (subject to reporting), and generally exempts data centers operated by US companies in those countries.
Why now: Megaspeed, Nvidia and a policy vacuum
The timing is not subtle. The rule lands days after Nvidia’s board accepted Jensen Huang’s resignation amid reports involving Singapore-based cloud firm Megaspeed, now under scrutiny for alleged illegal diversion of advanced AI chips, and follows a Commerce-Justice joint statement on the probe. It also fills a vacuum left when the Biden-era “AI diffusion” framework was announced in January 2025, then left unenforced – a hiatus that allowed billions of dollars in restricted Nvidia chips to reach data centers used by Chinese AI companies.
BIS framed the move plainly: existing restrictions “cannot be circumvented through remote computing arrangements involving data centers in third countries.” For an industry that treated offshore rentals as the fine print of the chip wars, the fine print just became the headline.
