
AI-generated illustration
AI Chip Export Rules Tighten for Remote Access
The new AI chip export rules from the US Commerce Department arrived on 16 November 2026, and they target a gap critics have pointed at for months: renting advanced chips remotely. The interim rule requires licenses for large data centers that serve companies in China, and for advanced chip exports to five more countries.
It takes effect once the Federal Register publishes it. A 60-day public comment period follows. The coverage I read does not include an Nvidia response yet.
What do the new AI chip export rules require?
The rule extends existing chip controls to remote access. Until now, the limits mostly covered physical shipments. Now they also cover compute rented through data centers in other countries.
- Remote access license. Data centers anywhere need a license if they serve companies headquartered in China, Macau or other restricted countries above 12 million aggregate TPP. Reports equate that to roughly 750 Nvidia H100 or H200 chips. Existing data centers in that position get 90 days to apply.
- Five new countries. Exporting advanced chips to Brazil, Indonesia, Malaysia, Singapore and Thailand now needs a license. Commerce says these places carry a higher than normal risk of diversion. US-headquartered companies get a carve-out, with conditions.
- Tougher checks. Foundries, packagers and data center operators have to look harder at who owns their customers.
- Old rule gone. The Biden-era AI Diffusion Rule, which was never enforced, is formally rescinded.
- Exemption. Recommendation systems used by online shops and social platforms are not covered.
Why did Commerce act now?
The timing follows Jensen Huang’s resignation as Nvidia CEO on 12 November. Reports tie that exit to a scandal involving a former executive at Megaspeed, a Singapore-based company accused of funneling Nvidia chips to China. Megaspeed denies sending chips there.
CBS News says the exit raised fears that China could reach Nvidia hardware by renting compute in Southeast Asia. Markets reacted fast. Nvidia shares fell as much as 2.9% on the news, and the Philadelphia Semiconductor Index dropped as much as 5.5%.
What are experts saying about the rule?
Reaction is mixed. Chris McGuire of the Council on Foreign Relations called it a major step forward, but said loopholes remain. Joshua Kurlantzick, also at CFR, warned the controls could be “disastrous” for Malaysia, which wants to host data centers for Chinese firms.
Hanna Dohmen of Georgetown’s CSET made a different point. Changing rules will not stop smuggling without enforcement. Several bills are also pending in Congress, including the Remote Access Security Act, which passed the House in January and awaits the Senate.
What should readers watch next?
Four things matter from here:
- Federal Register publication. That sets the effective date and starts the 90-day and 60-day clocks.
- Nvidia’s response. None appears in the reports so far.
- Southeast Asia. Malaysia and Singapore data center operators are the most exposed.
- Congress. Lawmakers may add their own limits.
If you rent GPU compute from a provider in one of the five countries, check its notices for license changes. The rule aims at advanced AI data center chips. None of the reports I read mention consumer graphics cards.
This is a developing story.