AI-assisted research analysis by FutureKnowledge. Sources checked 17 November 2026. This is not investment advice.
Reviewed by FutureKnowledge: figures, dates and attributions were checked against the linked sources before publication.
The US Commerce Department on 16 November issued a rule requiring export licences for advanced AI chips shipped to five countries, Indonesia, Malaysia, Singapore, Thailand and Brazil, to stop Chinese firms renting American compute in overseas data centres. Chip stocks fell sharply, and Asian markets followed on Tuesday.
What the US AI chip remote-access rule does
The Bureau of Industry and Security (BIS) described the measure as an interim final rule on remote access to advanced chips and servers. It said the rule was “written by and at the direction of” the Treasury Department and the National Security Council. Its stated aim is to make sure existing restrictions cannot be sidestepped through remote computing arrangements in third countries.
The rule text published by BIS has two main parts:
- Data centres serving Chinese firms. A licence is needed to export covered chips to data centres that remotely supply significant compute to companies headquartered in Country Group D:5 (which includes China) or Macau. The threshold is more than 12,000,000 units of “aggregate TPP”, a measure of total processing performance, with linked facilities counted together.
- Five named countries. Exports of covered chips to Indonesia, Malaysia, Singapore, Thailand or Brazil now need a licence. The rule calls them countries with a “higher-than-normal risk of diversion”.
Existing providers that already serve such customers must obtain a licence within 90 days of the rule’s effective date. The effective date is the date of publication in the Federal Register, followed by a 60-day comment period. Bloomberg reported on Monday that the rule had not yet been posted there, so the 90-day clock had not started.
The rule exempts end users headquartered in the United States that meet reporting requirements. Fortune noted that this would likely cover large US cloud groups such as Microsoft, Oracle and Amazon. Licences will be reviewed case by case, according to Bloomberg.
Why Washington moved now
The rule follows the scandal that led to Jensen Huang’s resignation as Nvidia chief executive last week. Huang admitted an affair with a former executive of Megaspeed, a Singapore cloud company and Nvidia partner. Reuters reported that Megaspeed is under scrutiny over the possible diversion of chips to China. Megaspeed has denied sending chips to China, according to Fortune. Bloomberg reported that Alibaba gained access to Nvidia chips in Malaysia through Megaspeed via a Singapore shell company.
Bloomberg quoted Treasury Secretary Scott Bessent as saying “Commerce had the pen” while other departments pushed for the measure, and Trump said he had no plans to replace Commerce Secretary Howard Lutnick, The Next Web reported. The timing matters diplomatically: Trump is meeting ASEAN leaders in Manila and is due to meet Chinese President Xi Jinping on Thursday.
Market reaction: Nvidia and chip stocks
Monday’s selloff was broad. Nvidia shares closed down 4.12%, Micron fell 7.62% and Intel 7.02%, according to StockTitan market data. The Philadelphia Semiconductor Index dropped as much as 5.5% during the session, Bloomberg reported. The Nasdaq 100 tracking fund QQQ fell 2.67% and the S&P 500 fund SPY 1.90%. Alibaba’s US-listed shares dropped more than 3.5%.
The selling carried into Asia on Tuesday. Reuters reported that South Korea’s KOSPI slid about 2% after the open, Taiwan fell 1.4% and Japan’s Nikkei 1%. This adds to the pressure FutureKnowledge described when Asian chip stocks slid before Nvidia’s results. Nvidia reports quarterly results after the US close on Tuesday, 17 November. Options implied a move of about 7.8%, according to a UBS strategist quoted by Reuters.
How the rule affects Asian data centres and India
The most direct exposure is in Southeast Asia, which has built a large data-centre industry partly on demand from Chinese technology groups. Lowy Institute’s Robert Walker told Bloomberg the rule would pose “a major headwind” to data-centre investment, with Singapore, Malaysia and Thailand most exposed. The Council on Foreign Relations’ Joshua Kurlantzick told Fortune that for Malaysia the controls “could be disastrous”. FutureKnowledge looked at those risks earlier in Malaysia’s $94.7 billion data-centre boom.
Singapore said it was studying the measure carefully, Brazil said it was monitoring it closely and Indonesia said it was reviewing it and seeking industry input. Malaysia and Thailand did not comment, and Nvidia did not respond to Bloomberg’s request for comment.
India is not one of the five named countries in the rule text. That means Indian data-centre projects do not face the new country licence requirement, although any facility anywhere that supplies large volumes of compute to China-headquartered firms would fall under the first part of the rule. None of the sources reviewed reports any shift of investment toward India so far.
Who is more or less exposed
- More exposed: data-centre operators in the five named countries with Chinese tenants; chip and server suppliers whose sales lean on Southeast Asian billing; Chinese AI developers that depend on rented overseas capacity. The Institute for Progress estimated, as cited by Bloomberg, that at least 30% of China’s advanced AI compute in 2027 could come from overseas rentals and smuggling.
- Less exposed: US-headquartered cloud providers that meet reporting conditions, and data-centre markets outside the five countries without large China-headquartered customers.
Scenarios to watch
The table below sets out conditional scenarios, not predictions.
| If… | Then a possible outcome is… | Signal to watch |
|---|---|---|
| The rule is published in the Federal Register soon and licences are granted sparingly | Southeast Asian data-centre projects tied to Chinese demand slow or are redesigned | Publication date; first licence decisions |
| Licences are approved widely with monitoring conditions | The market impact fades and the rule acts mainly as a compliance burden | Operator statements and capex plans in Malaysia and Thailand |
| Trump and Xi agree a broader trade understanding this week | Implementation could be softened or delayed | Statements after Thursday’s meeting |
| Nvidia’s guidance reflects the rule’s costs | Chip stocks stay under pressure across the US and Asia | Nvidia’s results and commentary on 17 November |
What to watch next
- Federal Register publication, which starts the 60-day comment window and the 90-day licence deadline for existing providers.
- Nvidia’s quarterly results and any comment from interim CEO Colette Kress on Southeast Asian sales. See our coverage of the debate over Nvidia’s next CEO.
- Formal responses from Malaysia, Thailand and China.
Bottom line
The new US AI chip rule targets the main route Chinese firms have used to reach advanced American compute since direct exports were blocked: renting capacity in nearby countries. Its real force will depend on the publication date, how strictly licences are granted and what comes out of this week’s Trump-Xi meeting. For now, it has made Southeast Asia’s data-centre boom harder to finance and added to pressure on chip stocks just before Nvidia reports.
Method note: This analysis compared the BIS press release and interim final rule text (primary sources) with reporting by Fortune, The Next Web, Bloomberg (via Yahoo), Reuters (via Business Recorder) and StockTitan market data. Where reports differed, we used closing data and the rule text.