The Trump administration is preparing to roll out a comprehensive new set of export controls targeting artificial intelligence chips and data centers located outside the United States. The policy is explicitly designed to prevent foreign companies, particularly those based in China, from circumventing existing U.S. trade restrictions by leveraging remote access to high-performance AI infrastructure hosted in third-party countries.

Under the terms of the new interim rule issued by the Commerce Department, companies will be required to secure specialized government licenses to export, reexport, or transfer advanced AI chips and servers to data centers anywhere in the world if those facilities could potentially provide computing power to entities in China or other restricted nations.

These stringent remote-access regulations are specifically aimed at companies headquartered within the United States’ D:5 group of embargoed nations. This restrictive classification encompasses China, Russia, Iran, North Korea, and 18 additional countries subject to strict U.S. national security oversight.

For some time, the federal government has enforced regulations mandating that domestic chipmakers obtain a formal license before shipping advanced AI hardware directly to any business domiciled in, or owned by interests within, this designated group of nations. The latest regulatory action builds directly upon that foundation.

“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.”

The introduction of these rules follows closely on the heels of high-profile leadership turbulence within the semiconductor industry. Days prior to the announcement, Nvidia co-founder Jensen Huang resigned from his post amid a sweeping corporate scandal involving an alleged Chinese intelligence operative linked to one of the prominent chipmaker’s primary customers. According to reports, federal authorities had sought to place that specific customer, Megaspeed, onto a special government entity list over suspicions that the firm was aggressively acquiring thousands of Nvidia processors to supply computational resources to mainland Chinese enterprises, thereby bypassing regulatory roadblocks.

In the wake of these developments, industry reports highlighted a frantic rush by Chinese artificial intelligence firms to secure massive overseas computing capacity before Washington could tighten the regulatory loop.

To address these vulnerabilities, the new rule establishes strict quantitative and operational thresholds. It applies to any data center or interconnected network configuration capable of supplying a total processing performance exceeding 12 million TPP, which serves as the established export-control metric utilized by the U.S. government to measure aggregate computing capability. For context, 12 million TPP roughly equates to the maximum computational ceiling of approximately 750 high-end Nvidia H100 or H200 accelerators.

Furthermore, the framework aggressively targets indirect access facilitated through brokers, intermediaries, or shell organizations. Under the directive, exporters and data center operators must maintain rigorous oversight of their customer networks, constantly auditing ownership structures and operational activities to verify that the underlying resources are not being utilized by any entity linked to the D:5 group.

In addition to broad remote-access provisions, the interim rule establishes new licensing hurdles for chip exports, reexports, and transfers destined for Malaysia

, Singapore, Indonesia, Thailand, and Brazil. The Commerce Department justified the inclusion of these nations by pointing out that they present a notably higher-than-normal risk of technology diversion. However, the policy carves out an explicit exemption for shipments explicitly destined for companies headquartered in the United States.

Existing data center operators whose current service models fall under the purview of the new regulations are being given a 90-day grace period from the effective date of the rule to acquire the necessary compliance licenses.

Notably, the restrictions exempt commercial recommendation systems, such as the everyday algorithmic engines widely deployed in e-commerce platforms and consumer social media applications.

As part of this regulatory overhaul, the administration is formally rescinding the previous administration’s AI Diffusion Rule, which was enacted in January 2025. That prior framework sought to impose rigid country-level export limits on advanced AI chips alongside restrictions on the cross-border transfer of powerful AI models. Shortly after taking office, the Trump administration signaled its intention to dismantle that framework, arguing that it imposed unnecessary administrative burdens and economic friction on American technology enterprises and key international allies.

“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 text of the rule emphasizes. “Access to such computing power can be used to train and run advanced SI models capable of supporting military- and intelligence-related activities that could threaten U.S. national security or foreign policy interests.”

By Basiran