What the New Rule Actually Says
On November 15, 2026, the U.S. Commerce Department's Bureau of Industry and Security (BIS) issued an Interim Final Rule (IFR) aimed at preventing restricted nations from accessing advanced American AI technology. The core of the rule requires a U.S. license
for data centers anywhere in the world to provide significant computing power to companies headquartered in or whose parent company is based in China, Russia, Iran, and other countries under a U.S. arms embargo. This move fundamentally changes the focus of export controls from the physical shipment of chips to regulating access to the computational power the chips provide, a concept the government now refers to as 'Super Intelligence' or SI. Essentially, it treats renting AI computing power as equivalent to exporting the chip itself.
Closing the Cloud Computing Loophole
Previously, U.S. export controls focused on blocking the physical sale of high-end AI chips to countries like China. However, a significant loophole remained: companies in restricted nations could simply rent processing power from cloud providers or data centers in third-party countries. This allowed them to train powerful AI models on advanced U.S. technology without ever importing the hardware. The new rule directly targets this workaround. It mandates that data center operators must know their customers and their ultimate ownership to ensure they aren't indirectly providing services to entities from restricted countries. The stated goal is to stop American technology from being used to develop advanced AI models that could support military and intelligence activities threatening U.S. national security.
Who Is Immediately Affected?
The rule has wide-ranging implications. Major US cloud providers like Amazon Web Services, Google, and Microsoft, along with other data center operators globally, will face significant new compliance burdens. They must now implement stricter 'know-your-customer' protocols to verify the identity and ownership of their clients. The rule also creates new licensing requirements for exporting advanced chips to Brazil, Indonesia, Malaysia, Singapore, and Thailand, countries identified as having a 'higher-than-normal risk of diversion'. An exemption exists for chips sent to facilities in these countries if the end-user is a US-headquartered company, but reporting requirements will apply. Existing data centers currently providing services to restricted companies have a 90-day grace period to obtain the necessary licenses to continue their operations.
Why an 'Interim Final Rule'?
An Interim Final Rule (IFR) is a regulatory tool used by federal agencies to make a rule effective immediately, without the standard public comment period beforehand. This is typically done when an agency finds 'good cause' that the normal process would be impracticable or contrary to the public interest, often in situations involving national security or emergencies. In this case, the rule took effect upon publication to prevent a rush to exploit the loophole before it closed. While the rule is now active, the BIS will accept public comments for 60 days. Based on this feedback, the agency may make adjustments before issuing a final, permanent version of the rule.
The Broader Geopolitical Context
This rule represents a significant tightening in the ongoing tech rivalry between the U.S. and China. It also formally rescinds a previous AI-related framework from the prior administration, known as the AI Diffusion Rule, which the current administration had declined to enforce. The announcement came shortly after the resignation of Nvidia's CEO amid scrutiny over the company's ties to a Singapore-based data center suspected of diverting chips to China, a situation that put a public spotlight on the remote access loophole. The policy was reportedly championed by the Treasury Department and the National Security Council, highlighting the interagency consensus on the need to act decisively. The move signals a clear strategy shift: controlling access to computation is now as critical as controlling the hardware itself.
















