The US Commerce Department issued an interim rule on November 16, 2026 requiring export licenses before companies can send AI chips or servers to data centers anywhere in the world if that infrastructure may end up providing computing power to China or other embargoed countries. The rule targets remote, cloud-based access as a workaround to existing chip export bans.
The rule, filed under RIN 0694-AK50 by the Bureau of Industry and Security, closes a gap that let companies based outside the United States sell or route access to advanced AI computing power to Chinese firms without those firms ever physically receiving a restricted chip. TechCrunch reported that the rule also covers indirect access routed through go-betweens or brokers, and requires exporters and data center operators to monitor who actually owns and controls their customers to confirm they are not run by anyone from the broader group of 22 embargoed nations, known as the D:5 list, which includes China, Russia, Iran and North Korea along with 18 others.
What the rule actually covers
The licensing requirement applies to data centers or networks capable of supplying more than 12 million total processing performance, a computing-power measure. TechCrunch estimated that threshold at roughly the maximum output of about 750 Nvidia H100 or H200 chips, the workhorse processors behind most large AI models trained since 2023.
Commerce explicitly excluded commercial recommendation systems, such as those used in e-commerce or social media platforms, from the new control. The rule text itself, quoted by TechCrunch, describes the move as “a logical extension of existing controls,” since “access to such computing power can be used to train and run advanced AI models.”
The core concept the rule is built around is remote access. Until now, most US export controls focused on the physical movement of a chip across a border. A company could legally buy server capacity built around US-made processors that sat inside a data center in a third country, then sell computing time on that infrastructure to a buyer in China over the internet, without ever shipping a chip there directly. The interim rule treats that kind of remote compute access as equivalent to a direct chip export, and requires the same licensing before it happens.

Five more countries now need licenses too
Beyond the existing D:5 embargo list, the rule adds new licensing requirements for exports to Malaysia, Singapore, Indonesia, Thailand and Brazil. Commerce said these countries “present a higher-than-normal risk of diversion,” meaning chips shipped there could be rerouted to restricted buyers. For these five countries specifically, the license requirement does not apply when the chips are destined for a company based in the United States.
What this replaces
The interim rule formally rescinds the Biden administration’s AI Diffusion Rule, issued in January 2025, which had set country-by-country caps on advanced chip exports and restricted transfers of certain powerful AI models. The Trump administration said in May 2026 that it would discard that framework, arguing it placed unnecessary restrictions on American companies and allied nations. The new rule replaces those country tiers with the compute-threshold and ownership-monitoring approach described above.
How much time companies have to comply
Data center operators already providing the covered services have 90 days from the rule taking effect to obtain the required licenses. Commerce did not specify penalties for operators who miss that window in the portions of the rule reported so far.
Why the timing matters
The rule lands days after Nvidia co-founder Jensen Huang resigned amid a dispute over an alleged Chinese spy linked to one of the company’s customers, Megaspeed, which the US government has reportedly considered adding to a restricted entity list over suspected large-scale chip purchases intended to supply compute to Chinese businesses. TechCrunch also cited Reuters reporting that Chinese AI companies have been rushing to lock in overseas computing capacity in anticipation of tighter US restrictions, the exact pattern this rule is designed to close off.
Frequently asked questions
Does this rule stop Nvidia from selling chips to US allies?
No. The rule does not add new restrictions on sales to US-based companies, and allied-country buyers are unaffected unless they operate a data center above the 12 million TPP threshold that could route compute to a D:5-listed country.
Why were Malaysia, Singapore, Indonesia, Thailand and Brazil added if they are not on the D:5 list?
Commerce said these five countries carry a higher-than-normal risk that chips or compute shipped there get diverted onward to a restricted buyer, so exporters now need a license for shipments to those countries specifically, unless the end buyer is a US-based company.
What happens to data centers that do not get a license in time?
Operators already providing the covered compute services have 90 days from the rule’s effective date to secure a license. The reporting reviewed for this story did not specify a penalty structure for operators who miss that window.



Leave a Reply