Technology

Home  •  Blog  •  US Moves to Shut China's Cloud B...

US Moves to Shut China's Cloud Back Door to AI Chips, Rattling Southeast Asia's Data Centre Boom

A new Commerce Department rule requires licences for AI chips headed to five countries and for data centres renting compute to Chinese firms. Asian chip and data centre stocks slumped.

Glowing AI microchip inside a red barrier with a holographic padlock in front of data centre server racks and a world map

Washington's new rule targets rented computing power, not just shipped chips, and markets from Kuala Lumpur to Taipei felt it within hours.

The United States has issued a sweeping new rule aimed at stopping Chinese companies from tapping advanced American AI chips through data centres in other countries. The interim final rule, released on Monday, November 16, by the Commerce Department's Bureau of Industry and Security (BIS), also requires export licences for top-end chips shipped to Indonesia, Malaysia, Singapore, Thailand and Brazil. By Tuesday, shares of data centre and chip supply-chain companies across Southeast Asia had tumbled, according to Reuters.

What the new rule does

For years, US export controls focused on where chips physically went. The gap, as Washington now frames it, is that a company in China does not need to import a chip if it can rent the computing power from a data centre somewhere else. In its announcement, BIS said the rule is meant to ensure existing restrictions "cannot be circumvented through remote computing arrangements involving data centers in third countries."

The rule has several moving parts, according to an analysis by law firm Nixon Peabody:

  • A worldwide data centre control. A licence is now needed to ship advanced chips and servers if the exporter knows they are headed to a data centre anywhere in the world that provides more than 12 million units of "aggregate TPP" (total processing performance) to companies headquartered in China, Macau or other countries in the US "D:5" group. Capacity spread across "linked" data centres counts toward the threshold.
  • Know-your-customer pressure. Exporters are expected to obtain details of a foreign data centre's customer checks. A refusal to share that information is treated as a red flag.
  • A 90-day clock. Data centres already supplying significant compute to such companies must obtain a licence within 90 days of the rule taking effect if they want to keep doing so.
  • Five "diversion-risk" countries. Shipments of covered chips to Indonesia, Malaysia, Singapore, Thailand and Brazil now need a licence, reviewed case by case. Companies headquartered in the US, or whose ultimate parent is, are exempt if the exporter files monthly reports.

The rule also formally rescinds the Biden-era "AI Diffusion Rule" and its country tiers. It is expected to be published in the Federal Register and take effect on November 19, with comments due 60 days later, the Nixon Peabody alert says. Notably, the rule uses the term "super intelligence" in place of "artificial intelligence," in line with a recent executive order.

An unusual author: Treasury and the NSC

One detail stood out to trade lawyers. BIS stated that the rule was "written by and at the direction of the Department of Treasury and the National Security Council," rather than drafted by BIS itself in the usual way. Nixon Peabody read this as a sign of "significant interagency pressure" to tighten chip controls, and warned that more action may follow. The rule also says Commerce will monitor for countries that "tolerate evasion" and could respond with import restrictions, tighter licensing, or Entity List designations.

The timing is not lost on anyone. As Reuters reported, the rule arrived days after Nvidia co-founder Jensen Huang resigned as CEO following reports of an affair with a suspected Chinese spy who worked at Singapore-based cloud firm Megaspeed, an Nvidia partner that has come under scrutiny over possible illegal diversion of AI chips to China.

ADVERTISEMENT

Markets react: a sharp selloff across Asia

Investors moved quickly. Reuters reported that Malaysian data centre infrastructure firms MN Holdings and Critical Holdings each plunged nearly 20% on Tuesday, while YTL Power, an Nvidia AI infrastructure partner, fell nearly 10%. In Singapore, ASML supplier Frencken slid 3.91% and chip-testing firm AEM dropped 2.9%.

The damage spread to suppliers of circuit boards, cooling systems and power equipment in Japan, Taiwan, South Korea and China, with Taiwan's Delta Electronics among several names down more than 5%. Alibaba lost 5% and Tencent 3%. One notable winner was SMIC, China's biggest contract chipmaker and a pillar of Beijing's push for home-grown alternatives, which jumped 7%.

The stakes for the region are large. Malaysia has approved 385.7 billion ringgit (about $95 billion) in data centre-related investment since 2021, while Thailand has 42 data centres representing 3,400 MW of IT load, Reuters noted, citing official figures.

What governments are saying

Singapore's trade and communications ministries said they were studying the measure, engaging affected companies and contacting US authorities about its scope. "We take our international reputation seriously and do not condone companies using their association with Singapore to circumvent or violate the laws of other countries," the joint statement said, as quoted by Reuters.

Beijing was blunter. Chinese foreign ministry spokesperson Guo Jiakun called the rule "a typical act of economic coercion and long-arm jurisdiction" and said China "will take necessary measures" to protect its interests, according to the New Straits Times and China News Service.

Why it matters

Analysts quoted by Reuters described the rule as a shift in who carries the compliance burden. "The regulatory burden has fundamentally shifted from hardware manufacturers to cloud service providers and enterprises," said Neil Shah of Counterpoint Research, adding that losing remote access creates a "severe short-term compute bottleneck" for Chinese tech giants but could accelerate decoupling over the long run. Saxo's Charu Chanana warned that operators reliant on Chinese customers could see weaker utilisation, pricing pressure and delayed expansion.

There is also a question of whether controls on hardware can keep pace with other methods. In a separate Reuters investigation published the same day, a review of more than 300 Chinese documents found Chinese entities increasingly using "distillation" — training smaller domestic models on the outputs of leading US models. The practice is legal and cannot fully replicate a frontier model, but Reuters noted it could undercut the effectiveness of chip curbs.

What happens next

The key dates are close. If publication lands on November 19 as expected, data centres serving Chinese-headquartered clients will have 90 days to secure licences, and shipments already in transit under the old rules have 30 days to complete. Companies will be watching for guidance on how the 12 million TPP threshold and "linked data centre" definition will be applied in practice, and whether Washington adds more countries to the diversion-risk list. For Southeast Asia, the near-term question is whether its AI boom shifts toward US hyperscalers and "trusted" providers, or simply slows down.

Verified Sources & References
TAGS: #AI Chips #Export Controls #China #Southeast Asia #Data Centres #Nvidia
SHARE THIS STORY
Entro

Entro

Staff Writer & Senior Contributor

Curating in-depth stories, analysis, and ideas for Enterteno readers. Committed to editorial accuracy, independent reporting, and original perspective.