Washington is moving to stop Chinese companies from renting the computing power they are barred from buying.
Key Points
- The US Commerce Department issued a rule on Monday requiring licences for advanced AI chips and servers sent to data centres that may supply computing power to firms in China and other restricted countries.
- The rule covers facilities able to deliver more than 12 million total processing performance, which TechCrunch estimates is roughly the output of about 750 Nvidia H100 or H200 chips.
- Shipments of advanced semiconductors to Indonesia, Malaysia, Singapore, Thailand and Brazil will also need licences, with US-headquartered companies exempt.
- The move follows the resignation of Nvidia co-founder Jensen Huang and scrutiny of Singapore data-centre firm Megaspeed.
- The rule takes effect when published in the Federal Register; data centres already serving covered customers have 90 days to obtain licences.
The United States moved on Monday to close a gap in its chip export controls, requiring licences for advanced AI chips sent to overseas data centres that rent computing power to companies in China.
The rule, issued by the Commerce Department’s Bureau of Industry and Security, targets a route that has let Chinese firms use American chips without buying them: renting time on the same hardware in a data centre abroad. Washington has barred direct sales of its most advanced AI chips to China for years. Until now, renting them from another country sat outside those controls.
What happened
The interim final rule requires a licence to export, re-export or transfer advanced AI chips and servers to any data centre, anywhere in the world, that may supply computing power to a company headquartered in China, Macau or one of the other countries on the US restricted list, known as Country Group D:5. That group also includes Russia, Iran, North Korea and 18 other countries.
The licensing line is set at scale. It applies to a data centre, or a network of linked data centres, able to deliver more than 12 million aggregate total processing performance, the export-control measure of computing power. TechCrunch estimates that figure is roughly the maximum output of about 750 Nvidia H100 or H200 chips. That estimate is the publication’s own reading of the rule, not a number in the rule itself.
The rule also reaches indirect access, through intermediaries or brokers, and requires exporters and data centre operators to check who their customers are and who owns them. It does not cover commercial recommendation systems, such as those used in e-commerce or social media.
Separately, the rule adds a new licensing requirement for shipments of advanced semiconductors to five countries: Indonesia, Malaysia, Singapore, Thailand and Brazil. The Commerce Department said those destinations carry a higher-than-normal risk that chips will be diverted onward. Chips bound for US-headquartered companies in those countries are exempt, provided exporters file monthly reports.
The rule also formally drops the Biden administration’s AI Diffusion Rule from January 2025, which would have sorted countries into tiers for chip access. The Trump administration said in May it would discard that framework, arguing it placed needless limits on US companies and allies.
In its statement, the Commerce Department said the rule was written at the direction of the Treasury Department and the National Security Council. “The rule is designed to ensure that existing restrictions on access to sensitive American technology cannot be circumvented through remote computing arrangements involving data centers in third countries,” the department said.
Why it matters
The timing is tied to the scandal that cost Nvidia its chief executive. Jensen Huang, the chipmaker’s co-founder, resigned last Thursday after admitting an affair with Alice Huang, a former executive at Megaspeed, a Singapore data-centre company whom US officials suspect of working for Chinese intelligence. The US government believes Megaspeed bought thousands of Nvidia chips to supply computing power to Chinese businesses, getting around the export controls. Megaspeed denies sending chips to China, and China denies the allegation. Reuters reported that Megaspeed has not been found to have violated the rules.
Pressure on the Commerce Department had been building before the rule appeared. Senator Maria Cantwell, the top Democrat on the Senate Commerce Committee, wrote to Commerce Secretary Howard Lutnick in a 15 November letter, accusing the department of failing to stop the flow of AI chips to China. Treasury Secretary Scott Bessent said last week that the Treasury, the Pentagon, the State Department and the US Trade Representative’s office had favoured putting Megaspeed on a trade blacklist for months, while the Commerce Department, which runs the list, had not.
The five countries named in the rule are now caught between two large markets. Malaysia and Singapore host much of the region’s data-centre capacity, and both have drawn heavy investment from US cloud and chip companies. Jennifer Lind of Dartmouth College told Fortune the administration wants Malaysia and Singapore to stay inside the US AI ecosystem without becoming a route for Chinese firms. Joshua Kurlantzick of the Council on Foreign Relations told Fortune the controls “could be disastrous” for Malaysia. Fortune reported no official response from Beijing.
Reuters reported last week that Chinese AI companies were already rushing to lock in computing capacity overseas, fearing tighter restrictions from Washington. The rule now puts a licence in front of that capacity.
What happens next
The rule takes effect when it is published in the Federal Register. Nikkei reported it could take effect as early as 19 November. Once it does, data centres already serving covered customers will have 90 days to obtain licences, and the public will have 60 days to comment on the rule.
The House passed the Remote Access Security Act by 369 votes to 22 in January, but the Senate has not yet scheduled a vote on it.
Investors will get their first read on the commercial fallout on Tuesday, when Nvidia reports results after the US market closes, with Colette Kress serving as interim chief executive. The rule also lands as President Donald Trump travels to Asia for talks with Chinese President Xi Jinping this week.
Sources: Reuters, TechCrunch, Nikkei Asia, Fortune, TechTimes. Published: 17 November 2026, Kuala Lumpur time.



