The United States is finalising a sweeping new rule that would stop Chinese artificial intelligence firms from renting time on advanced Nvidia chips housed in data centres abroad, closing the “cloud loophole” that let Beijing’s developers tap American computing power without buying a chip. The Commerce Department’s Bureau of Industry and Security is preparing to extend its semiconductor export controls to seven additional countries, The Information reported on Wednesday, November 11, 2026, citing two people familiar with the matter.

It would be the biggest tightening of American chip controls since the export rules were first imposed, and it follows the Megaspeed scandal, in which a network tied to a suspected Chinese intelligence asset allegedly funnelled billions of dollars’ worth of Nvidia chips to China.

Key Points

• The US Commerce Department is finalising a rule to curb Chinese AI firms’ remote access to advanced AI chips through data centres in third countries.
• Seven countries would be added to the controls: Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa.
• Importers there would need an export licence for Nvidia Blackwell-class chips or anything more advanced.
• US-run data centres would be exempt if they block remote access from arms-embargoed nations such as China and Russia and report local customers and computing scale.
• The rule could be announced this month, with publication in the Federal Register expected in late November.

What Happened

The Information reported on November 11, citing two people familiar with the matter, that the Bureau of Industry and Security is finalising a rule to close the gap letting Chinese AI developers rent computing time on powerful chips in data centres outside China instead of buying them. Selling advanced chips directly to China is restricted, but renting access from a server hall in Singapore or Malaysia has sat in a grey area.

The draft would expand the controls to Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa. Importers there would need an export licence for Nvidia Blackwell-class chips or more advanced hardware. It could be announced as soon as this month, with Federal Register publication expected in late November, though no official text has been published yet. US-run data centres there would be exempt if they block remote access from arms-embargoed nations such as China and Russia and report local customers and computing scale; local and Chinese-linked operators would need a licence per shipment.

Background: The Megaspeed Scandal Behind the Rule

The urgency comes from the scandal shaking Nvidia and Washington. The Information reported on November 10 that a network tied to Alice Huang and her sister Huang Yu acquired about $7.5 billion of Nvidia chips and moved roughly 240,000 restricted chips to Chinese firms, while the New York Times reported that Megaspeed’s Malaysian subsidiary Speedmatrix bought at least $4.8 billion of AI equipment between June 2024 and February 2026.

Huang, a former Megaspeed executive whom US intelligence suspects of working for China, is at the centre of widening investigations. Nvidia chief executive Jensen Huang resigned on Thursday, November 12, after revelations of an affair with Alice Huang, and the Justice Department is probing whether entities linked to her helped smuggle Super Micro products to China. Oracle has cancelled a five-year, roughly $7 billion leasing deal with Tencent after US officials raised concerns.

How the Nvidia Chip Export Controls Would Work

The licence requirement is the heart of the rule. Any importer in the seven countries wanting Blackwell-class chips or better would have to apply to the Bureau of Industry and Security, which could scrutinise the end user, the installation site and who may log in remotely. Today a shipment to a data centre in Malaysia or Thailand can clear customs with little visibility into whose workloads it will run.

The exemption for American operators protects the global footprint of US cloud firms, but it comes with conditions: block remote access from China, Russia and other arms-embargoed nations, and disclose local customers and computing capacity. Local and Chinese-linked operators would need a licence for each shipment, blocking quiet cluster-building for Chinese developers.

Reactions: Trump Pushes Harder, Lutnick Pushes Back

The rule is being shaped against a split inside the administration. Investing.com, citing Axios, reported that President Donald Trump pushed for tighter controls while Commerce Secretary Howard Lutnick opposed the measure. Lutnick has agreed to testify before the Senate Banking Committee on December 10 about the department’s role in the Nvidia affair, and former Nvidia chief executive Jensen Huang is due to testify on December 3.

On Capitol Hill, senior lawmakers are pushing three chip-export bills, the AI Overwatch Act, the Chip Security Act and the MATCH Act, to be attached to the National Defense Authorization Act. Their sponsors argue the Megaspeed affair shows rules alone, without enforcement and verification, are not enough.

Why It Matters

Advanced AI chips are the choke point of the global AI race. Training a frontier model needs thousands of the most powerful processors running together for weeks, and Nvidia’s Blackwell-class hardware is the standard rivals are measured against. Whoever controls access to that computing power largely controls who can build the most capable AI systems.

The cloud loophole mattered because it let that control be bypassed without a single chip crossing into China: a Chinese firm needed only an account with a data centre in a third country. The reported scale, about $7.5 billion of chips and roughly 240,000 restricted units, suggests the loophole was a major channel, not a marginal gap.

What Happens Next

The next milestone is publication: an announcement as soon as this month, with Federal Register publication expected in late November. Until the official text appears, key details, exactly which chips are covered, how the US-operator exemption will be policed and what counts as a reportable customer, remain unconfirmed.

Frequently Asked Questions

What is the cloud loophole?
It is the gap that let Chinese AI firms rent computing time on advanced Nvidia chips in data centres abroad, even though selling those chips directly to China is restricted. The new rule aims to close it.

Which countries would the new rule cover?
Seven additional countries: Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa. Importers there would need an export licence for Nvidia Blackwell-class chips or more advanced hardware.

Would American data centres be affected?
US-run data centres there would be exempt, provided they block remote access from arms-embargoed nations such as China and Russia and report local customers and computing scale.

What is the Megaspeed scandal?
A network tied to Alice Huang and her sister Huang Yu is accused of acquiring about $7.5 billion of Nvidia chips and moving roughly 240,000 restricted chips to Chinese firms.

When will the rule take effect?
It could be announced as soon as this month, with interagency clearance and Federal Register publication expected in late November. No official rule text has been published yet.

Related coverage: Jensen Huang agreed to testify before Congress in the Nvidia China spy probe; the Justice Department is probing an alleged Super Micro smuggling link to Alice Huang as Nvidia faces an antitrust inquiry; Bessent confronted Lutnick over Megaspeed chips.