The Commerce Department on Monday issued an interim final rule requiring US companies to obtain licenses before exporting, reexporting or transferring advanced AI chips and servers to data centers anywhere in the world if those systems could end up providing computing power to companies in China or other restricted countries, according to reporting by TechCrunch on the rule's text.
The move converts months of drafting into binding policy and targets what officials view as the last major gap in America's chip export regime: Chinese firms renting computing capacity from data centers in third countries rather than buying restricted chips directly. Anyone following the latest AI developments will recognize how much has changed since The Information first reported in August that the department was finalizing a remote-access rule — the Nvidia scandal has since upended the politics of US-China tech policy.
How the New Rule Works
The rule targets companies based in what export-control regulations call the D:5 group of embargoed countries — a list that includes China, Russia, Iran, North Korea and 18 other nations. The US already requires chipmakers to obtain licenses for direct shipments to companies based in, or owned by businesses in, those countries. The new control extends that logic to computing power accessed over networks.
"This new control is a logical extension of existing controls," the interim rule states. "These restrictions will undermine the ability of companies headquartered in countries of concern from using U.S. advanced Super Intelligence (SI) semiconductors to train advanced SI models."
In practical terms, a Chinese laboratory can no longer sidestep US restrictions by contracting with a data center operator in a neutral country. The license requirement now attaches to the data center deal itself.
A Threshold Measured in TPP
The rule applies to data centers or any network able to supply more than 12 million units of total processing performance, or TPP — an export-control metric used to measure computing power. TechCrunch notes that 12 million TPP roughly corresponds to the maximum computing power of about 750 Nvidia H100 or H200 chips, so the threshold is high enough to exclude modest enterprise clusters while capturing serious AI infrastructure.
The rule also covers indirect access arranged through go-betweens or brokers, and it imposes monitoring duties: exporters and data center operators are required to track their customers' ownership structures and activities to make sure restricted parties are not benefiting behind layers of intermediaries. Existing data center operators whose services fall under the rule must obtain licenses within 90 days of the rule coming into effect.
Notably, the rule does not apply to commercial recommendation systems, such as those used in e-commerce or social media — an carve-out designed to keep ordinary business computing out of scope.
Five More Countries Face New Licensing Requirements
Beyond the remote-access provisions, the rule adds new licensing requirements for exports, reexports and transfers of advanced semiconductors to five countries: Malaysia, Singapore, Indonesia, Thailand and Brazil. The Commerce Department said those destinations "present a higher-than-normal risk of diversion," according to the rule text reported by TechCrunch.
The concern is transshipment — chips routed through Southeast Asian logistics and data center hubs before reaching Chinese buyers. For these five countries, however, the license requirement will not apply to chips destined for companies headquartered in the United States, preserving normal operations for American firms running regional infrastructure.
The Nvidia Scandal Backdrop
The rule arrives at a volatile moment. It comes days after Nvidia co-founder Jensen Huang resigned as chief executive amid a scandal involving an alleged Chinese spy linked to one of the chipmaker's customers, and as Reuters reported that Chinese AI companies are rushing to secure computing capacity overseas in anticipation of tighter restrictions.
That customer, Megaspeed, has reportedly been singled out by the US government for placement on a special entity list, amid allegations it bought thousands of Nvidia chips to provide compute to Chinese businesses seeking to bypass export controls. The timing also coincides with President Trump's trip to Asia, and multiple outlets — including The New York Times and Bloomberg — framed the rule as the administration moving to close the loophole ahead of high-stakes diplomatic meetings. Nvidia shares fell as news of the rule spread, according to Forbes.
Rescinding the AI Diffusion Rule
The interim rule also formally rescinds the Biden administration's AI Diffusion Rule from January 2025, which sought to impose country-tiered limits on exports of advanced AI chips and restrict transfers of certain powerful AI models. The Trump administration announced in May that it would discard that framework, arguing it placed unnecessary restrictions on US companies and allies.
In its place, the new approach is narrower but harder-edged: rather than grading every country, it draws a bright line around the D:5 group and around diversion-prone transit hubs. "By making these changes, this IFR prevents the use of certain SI items in data centers and at a computing power level that can be used for training or running of advanced SI models," the rule reads, citing the risk that such systems could "support military- and intelligence-related activities that could threaten U.S. national security or foreign policy interests."
What Happens Next
For cloud providers, the 90-day licensing window is the immediate deadline, and compliance teams will spend it mapping which customers in which jurisdictions can reach which clusters. For Chinese AI firms, the rule raises the cost of the overseas compute strategy that Reuters documented — and pushes the contest over training capacity further into data center markets from Kuala Lumpur to São Paulo.
Enforcement questions remain open, including how Washington will verify that licensed data centers keep restricted parties off their customer lists. But the direction is unmistakable: after months of drafting, the US has decided that controlling who can log in to its chips is now as important as controlling who can buy them.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →