The US Commerce Department has issued an interim final rule blocking companies from China and other restricted countries from renting computing power on advanced AI chips in third-country data centres. Exports of advanced chips to Malaysia, Singapore, Indonesia, Thailand and Brazil now also need a separate licence.
According to Korea's Chosun Ilbo, Taiwan's United Daily News and outlets including TechCrunch, the Bureau of Industry and Security (BIS) on Nov. 16 (local time) released an interim final rule (IFR) targeting ‘remote access’ to advanced AI chips and servers. The United Daily News summed it up as a ‘restriction on remote access to computing power.’ Where earlier US chip controls focused on stopping physical shipments, the new rule looks at who uses the compute, wherever the chips sit.
광고 문의 · 300×250Under the rule, a licence is required to export, reexport or transfer advanced chips to any data centre that may supply compute to firms headquartered in China, Macau or other D:5 arms-embargoed countries, or whose ultimate parent is based there. The threshold is more than 12 million units of total processing performance (TPP), which TechCrunch estimated at roughly 750 Nvidia H100 or H200 chips. Capacity supplied through linked data centres or via brokers is also covered.
Five countries face extra licensing; existing centres get 90 days
The rule names Malaysia, Singapore, Indonesia, Thailand and Brazil as carrying a ‘higher-than-normal risk’ of diversion and requires separate licences for advanced chip exports there. It is not a blanket ban but case-by-case review, and US-headquartered companies are exempt if they meet reporting requirements. Overseas data centres already serving restricted customers must obtain licences within 90 days of the rule taking effect to keep operating.
Recommendation systems used by e-commerce and social media are excluded, but generative AI that produces text, images, video or code on request is not. The rule takes effect on publication in the Federal Register, followed by a 60-day comment period. It also formally rescinds the Biden-era AI Diffusion Rule, which had set country-level caps.
Markets reacted quickly. According to The Next Web, Nvidia shares fell as much as 2.9% on Nov. 16, and the Philadelphia Semiconductor Index dropped as much as 5.5% at one point. Joshua Kurlantzick of the Council on Foreign Relations warned the controls could be disastrous for Malaysia, which has been courting data centre investment.
Megaspeed controversy as trigger; Beijing has decried ‘abuse of export controls’
The move follows allegations that Megaspeed, a Singapore-based Nvidia customer, bought large volumes of Nvidia chips to supply compute to Chinese companies; Megaspeed denies wrongdoing. Commerce said the rule was written at the direction of the Treasury Department and the National Security Council, amid criticism of Commerce Secretary Howard Lutnick from lawmakers and Treasury over enforcement.
No official Chinese response to the new rule had been confirmed. When Washington restricted chip purchases by overseas Chinese subsidiaries in June, however, China's Ministry of Commerce accused the US of ‘abusing export controls and disrupting the global semiconductor supply chain.’ Washington says the rule extends existing policy to control compute that could train advanced AI models for military or intelligence use, and the gap is likely to become another point of friction ahead of the Nov. 19 summit in Shenzhen.