U.S. Commerce Department Moves to Close AI Chip Remote Access Loophole, Extending Controls to Seven Countries
nashnova research
The US Commerce Department is finalizing a rule that extends AI chip export-license requirements to seven additional countries including Singapore and Malaysia, targeting the loophole that lets Chinese firms remotely access advanced chips via third-country data centers. The rule could drop as soon as this month.
What is the "remote-access loophole" — and why close it now?
Chinese AI companies cannot buy restricted chips directly, but they can rent compute from data centers in Southeast Asia and use those chips to train models — sidestepping existing export controls entirely.
In plain terms = the chips never cross a border, but the compute does. Current rules govern who buys; they don't govern who logs in remotely.
The new rule's core aim is to bring "who is using the chip" inside the control perimeter and shut down this workaround at the source.
Which seven countries — and why them?
The rule extends license requirements to Singapore, Malaysia, Thailand, Indonesia, Brazil, Norway, and South Africa, covering Nvidia Blackwell-class and above chips.
This means → data centers in these countries that want to import advanced AI chips must identify the end-user and intended use to the Commerce Department and submit to ongoing monitoring.
Saudi Arabia and Vietnam already face similar license requirements; the new rule pushes the same framework to more "middle-ground" nations.
Are US companies affected?
US companies operating data centers in the seven countries are exempt from the license requirement — but only if they block remote access from users in China, Russia, Venezuela, and other US arms-embargoed nations above a certain compute threshold.
They will also face new disclosure obligations. In plain terms = no license needed, but no free pass either — the government wants to know who is using your compute.
An added benefit of expanding the license scope: cracking down on chip smuggling, since every export must now account for its downstream destination.
Why does Nvidia oppose this — and where does the company stand?
Nvidia had lobbied against stricter data-center requirements, worried that restricting overseas customers would hurt chip sales.
The Commerce Department pressed ahead anyway — this reflects Nvidia's declining influence in Washington.
Nvidia's board is reportedly investigating CEO Jensen Huang over personal matters and evaluating whether to replace him. This means → the company's leverage in the policy fight is weakening further.
Will this rule actually land?
The Bureau of Industry and Security (BIS) has been under sustained criticism for moving too slowly on new export controls.
If the rule is published on schedule, it would be one of BIS's rare concrete policy advances on China-facing AI chip controls.
Whether the final text survives stakeholder lobbying without material dilution remains to be seen — in plain terms = the direction is set, but the details could still get softened.
市场有风险，内容仅供研究参考，不构成投资建议。