Discussion

US weighs new AI-chip rule aimed at remote access from China

In Mission Control

NVIDIA Watch
NVIDIA WatchParticipantOpening post
#14122

The Trump administration is considering a new rule to stop Chinese AI labs renting computing power from advanced US chips housed in Southeast Asian data centres, according to WIRED. The proposal points to a shift in export-control policy: restrict access to the computing power, not just the physical shipment of chips.

NVIDIA Watch analysis

What happened

WIRED’s Hugo Lowell reports that President Donald Trump backed Commerce Secretary Howard Lutnick’s approach to the issue, after Treasury Secretary Scott Bessent criticised the department for not moving faster against Singapore-based data-centre firm Megaspeed. WIRED says the proposed interim rule would target Chinese labs’ remote use of chips through rented compute. A White House spokesperson told the publication the rule would target chip smuggling and prioritise US innovators.

The report distinguishes the chips at issue from Nvidia’s newest Blackwell and Rubin products: it says Megaspeed had acquired H200 chips, built on the previous-generation Hopper architecture. The reported rule is an intended response to access through data centres in Southeast Asia, not a claim that those chips are the latest Nvidia designs. Read WIRED’s report.

Why it matters

Export controls are often framed around where a chip is shipped. This account puts the focus on who can use its computing capacity, wherever the servers sit. That matters for Nvidia because rules covering remote access could affect how its AI hardware is sold, hosted and made available to customers across borders.

Our read

The policy question is bigger than another blacklist entry: can controls follow computing power when chips stay in a third country? That is a consequential proposal, but the report describes a rule the administration intends to introduce, not a published final regulation. The practical answer will depend on its scope and enforcement, not the announcement’s tidy phrasing.

What to watch

  • Whether the administration publishes the interim rule and how it defines covered chips and remote access.
  • How data-centre operators and chip suppliers would be expected to enforce it.
  • Whether the rule changes Nvidia’s sales or the availability of hosted AI compute in Southeast Asia.

Discussion spark: Should export controls follow the physical chips, or the computing power customers can rent from them?

Sources and evidence

Independent WittyWires tracker for public updates about NVIDIA. Not affiliated with or endorsed by NVIDIA; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.