Commerce Is Finalizing a Rule to Stop Chinese Firms Renting Nvidia Chips Abroad
The Information reported on 11 November 2026 that Commerce is finalizing a rule to curb Chinese AI firms' remote access to advanced chips in third countries.

On this page
Key takeaways
- The Information reported on 11 November 2026 that the Commerce Department is finalizing a rule to curb Chinese AI firms' remote access to advanced chips in third countries.
- Reports name seven additional countries, including Singapore, Malaysia and Thailand, though the full list is unconfirmed and the rule has not been published.
- Operators in the covered countries would need export licences to bring in Nvidia Blackwell-class or more advanced chips, according to the reports.
- Data centres run by U.S. companies could be exempt if they block remote access from arms-embargoed countries and report local customers.
- The rule would close a gap in controls that cover physical chip shipments but not rented compute in overseas data centres.
The U.S. Commerce Department is in the final stages of a rule that would extend semiconductor export controls to remote access, according to The Information, which reported the plan on 11 November 2026, citing two people familiar with the matter. The rule has not been published or announced.
Related reading: FCC Tells California to Stop Blocking AT&T's Copper Retirement · California Subpoenas Nvidia Over Whether Its Chips Reached China.
The target is a gap that has been visible since the first AI export controls. Washington can restrict the shipment of advanced chips to China, but it has had far less leverage over a Chinese company that rents time on those same chips in a data centre in Singapore or Malaysia. The chip never crosses a Chinese border. The compute does.
According to the reporting, Commerce's Bureau of Industry and Security would require export licences for operators in a set of additional countries to bring in Nvidia Blackwell-class or more advanced chips. The rule could be announced this month. Investing.com, relaying The Information's story on 12 November, said Commerce was consulting other government agencies before finalizing the text. CASRAI's summary of the reporting notes that no rule text, Federal Register notice or BIS announcement had appeared by 13 November.
What the rule would cover
The reported scope is broader than a single country. Asia Times listed Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa as the countries that would be added. Investing.com named only Malaysia, Singapore and Thailand. The discrepancy is a warning: the country list is the part of the rule most likely to change before publication.
The licensing requirement would apply to Blackwell-class chips and anything more advanced. That is the class of hardware the reported rule is designed to cover, and it is the hardware that Chinese AI developers have been trying to obtain through intermediaries. Older chips would presumably remain outside the new licence requirement, though the reporting does not say where the line would fall.
There is also a reported exemption for data centres run by U.S. companies. Those operators could avoid the licence requirement if they restrict remote access from arms-embargoed countries such as China and Russia, and if they report their local customers and the scale of computing they provide. That structure tries to separate trusted U.S.-run cloud capacity from local operators who might resell access to restricted users.
The exemption is the most consequential design choice. It would let American cloud providers keep serving customers in the covered countries while giving BIS a reporting channel into who is using the capacity. It would also put local data-centre operators at a disadvantage against U.S. hyperscalers, which could reshape the region's cloud market.
Why remote access became the loophole
Export controls are built around the movement of goods. A licence is required to ship a controlled chip to a destination, and the enforcement mechanism is customs, freight forwarders and end-user checks. Cloud computing breaks that model. A customer in China can open an account with a data centre in a third country and rent the same accelerator by the hour.
Chinese AI companies have reportedly been doing exactly that. Tencent, Alibaba and ByteDance have signed offshore data-centre deals for Nvidia capacity, according to the reporting summarised by CASRAI. Forbes reported on 12 November that Chinese firms were rushing to secure such capacity before a new rule landed. The rush itself is evidence that the companies expect the door to close.
The Megaspeed case shows how the physical and remote channels can blur. The Information reported on 10 November that a network tied to Megaspeed executive Alice Huang acquired about $7.5 billion of Nvidia chips and moved roughly 240,000 restricted chips to Chinese firms. Reuters could not independently verify the allegation, and China has denied it. Earlier New York Times reporting linked Megaspeed's Malaysian subsidiary to at least $4.8 billion of AI equipment purchases. Singapore's Ministry of Home Affairs says its investigation into Megaspeed and six other individuals is ongoing. Megaspeed's owner has said the company cooperates with U.S. export-control inspections.
Those are allegations in press reports, not findings or charges. But they explain why the rule is moving now. If chips can be diverted physically, and compute can be rented remotely, then a shipment-based control regime has two holes rather than one.
The fight inside the administration
The rule is not emerging from a consensus. Investing.com, citing Axios, reported that President Trump pushed for tighter controls while Commerce Secretary Howard Lutnick opposed the measure. The White House and Commerce have not confirmed that account.
The disagreement is familiar. Hardliners argue that any remote access to advanced chips undermines the export-control regime, because the compute is the capability, not the piece of silicon. Industry advocates argue that unilateral restrictions will push third countries toward Chinese hardware and weaken the U.S. cloud providers that the exemption is meant to protect.
The legal question is separate from the policy one. Commentators have asked whether BIS has the statutory authority to regulate remote access without new legislation. The Remote Access Security Act passed the House in January 2026 but had not passed the Senate, according to CNBC. If BIS acts without that statute, the rule is likely to face a legal challenge from affected operators or from Congress.
Enforcement would be the harder problem. A licence is only useful if the issuing agency can verify who is using the chips and where. For physical shipments, BIS relies on end-user certificates, audits and intelligence. For remote access, it would need data-centre operators to report their customers and the scale of computing they provide, which is why the reported exemption for U.S.-run facilities includes a reporting condition. Local operators in the covered countries would face the same requirement without the same legal relationship to Washington.
That asymmetry is the rule's practical weakness. A U.S. cloud provider can be compelled to disclose its customer list. A data-centre operator in Malaysia or South Africa can be asked, but the enforcement depends on local cooperation and on the operator's willingness to lose business. The rule can raise the cost of concealment. It cannot make every operator an informant.
China's response
China warned the United States against imposing additional export controls on AI chips after the Huang scandal, according to Al Jazeera, which reported the warning on 13 November. The timing links the rule to the broader collapse of Nvidia's leadership, but the substance is older. Beijing has consistently opposed extraterritorial controls and has threatened retaliation against U.S. companies when Washington tightens the rules.
The warning is a reminder that the rule is a technical export-control decision and a move in a negotiation that includes rare earths, tariffs, agricultural purchases and the treatment of Chinese companies operating abroad. A rule that covers seven countries would give Washington leverage over a large share of the region's data-centre capacity. It would also give Beijing a new grievance.
What the rule would not do
It would not stop Chinese companies from building their own chips. Huawei and Cambricon are producing accelerators for the domestic market, and Chinese labs are increasingly using domestic hardware for inference. A remote-access rule raises the cost of using American hardware. It does not remove the incentive to build substitutes.
It would not cover every third country. The reported list is seven countries. Chinese firms could move capacity to countries outside the list, or to jurisdictions where enforcement is weak. The rule would raise the friction, not eliminate the workaround.
It would not settle the question of existing contracts. The reporting does not say whether operators with signed deals would get a grace period, or whether they would have to reapply for licences immediately. That detail will determine how disruptive the rule is in practice.
It would not resolve the legal authority question. If BIS cannot regulate remote access without Congress, the rule could be struck down or narrowed. The Remote Access Security Act is the cleaner path, but it has not passed the Senate.
Caveats
The rule is not published. The country list, the chip threshold, the exemption terms and the treatment of existing contracts are all reported, and the reports conflict on the list of countries. No Federal Register notice or BIS statement had appeared by 13 November.
The legal basis is also unsettled. BIS has not said it has authority to regulate remote access without legislation, and the Remote Access Security Act has not passed the Senate. A court challenge would delay or narrow the rule.
The Megaspeed allegations are allegations. Reuters could not verify the $7.5 billion figure or the 240,000-chip count, China has denied the claim, and Singapore's investigation has not produced a public finding. They explain the political context, not the legal case.
What to watch
The first test is publication. A Federal Register notice or a BIS announcement would confirm the country list, the chip threshold, the exemption terms and the treatment of existing contracts. Until then, every detail beyond the basic outline is reported, not official.
The second is the defence authorization bill. If Congress attaches remote-access provisions to must-pass legislation, the rule would gain a statutory footing and become harder to reverse. That is the outcome the rule's supporters want.
The third is the Megaspeed investigation. Singapore's probe and any U.S. action against the company or its executives would show whether the allegations in the press reports can be converted into enforcement. A case would also give the rule a concrete justification.
The fourth is the reaction from U.S. cloud providers. If the exemption works as reported, they gain an advantage over local operators in the covered countries. If it is narrower than they expect, they will lobby against the rule. Their position will shape how the final text reads.
The rule is not published. Until it is, the country list, the chip threshold and the exemption terms remain reports rather than policy.
Tags
Frequently asked questions
- Has the Commerce Department published the remote chip access rule?
- No. As of 13 November 2026, no rule text, Federal Register notice or Bureau of Industry and Security announcement had appeared. The reporting rests on sources cited by The Information and outlets relaying its story.
- Which countries would the rule cover?
- Reports conflict. Asia Times listed Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway and South Africa, while Investing.com named only Malaysia, Singapore and Thailand. The final list is unknown until the rule is published.
- What would the rule change for data centres?
- Operators in covered countries would need licences to import Blackwell-class or more advanced chips. U.S.-run data centres could reportedly be exempt if they restrict remote access from countries such as China and Russia and report their local customers.
- Why does remote access matter for export controls?
- Export controls cover the physical shipment of chips. A Chinese company that rents capacity in a data centre abroad can use the same hardware without importing it, which falls outside the current shipment-based rules.








