Southeast Asia’s fast-growing data center industry could face greater regulatory scrutiny and investment uncertainty following new US restrictions on advanced artificial intelligence (AI) chip exports to five countries, including Malaysia, as Washington seeks to prevent restricted technology from reaching Chinese-linked companies through overseas computing facilities.
BMI, a unit of Fitch Solutions, said the latest US Commerce Department ruling would extend export controls beyond the physical shipment of advanced chips to include companies accessing the computing capacity provided by those chips. The new measures could create a growing divide between data center projects serving Chinese-linked customers and those supporting US technology companies, with the former facing greater licensing requirements and compliance risks.
The US Commerce Department released the new ruling on Nov 16, 2026, requiring licenses for shipments of specified advanced semiconductors to Indonesia, Malaysia, Singapore, Thailand and Brazil.
Separately, data centers outside the US that provide more than 12 million aggregate total processing performance (TPP) of computing capacity to companies headquartered in mainland China, Macau or other restricted destinations will also require US licenses.
The threshold is estimated to be equivalent to roughly 750 Nvidia H100-class chips, with computing capacity aggregated across linked data center sites.
The measures are intended to close potential loopholes that allow companies in restricted countries to access advanced US computing technology through overseas data centers without physically importing the chips into their home markets.
The new requirements do not change existing restrictions on direct chip sales to China but extend oversight to the ultimate users of computing capacity. The ruling is scheduled to take effect on Nov 19, 2026, upon its publication in the Federal Register.
BMI said chip and server suppliers, including Nvidia and Advanced Micro Devices (AMD), would face the most immediate compliance obligations.
They will need to obtain licenses for shipments of specified advanced chips to the five designated countries, unless the transactions qualify for an exemption involving US-headquartered customers and the required monthly reporting.
Suppliers must also obtain information from data center operators about the identity and ownership of their tenants, including customers accessing computing capacity through their facilities.
Data center operators that refuse to provide the necessary information will be treated as requiring a license unless the compliance concerns can be resolved.
The requirements also extend to foreign semiconductor foundries and packaging companies, which must treat advanced chips designed by companies in the five designated countries or other restricted jurisdictions as controlled unless the chip designers have received the necessary approval or authorization.
BMI said the additional requirements would increase the importance of customer due diligence across the AI infrastructure supply chain, particularly in identifying the ultimate ownership and control of businesses accessing computing capacity.
The US Bureau of Industry and Security (BIS) expects the new framework to generate about 100 additional license applications annually, although approval rates remain uncertain as applications will be assessed on a case-by-case basis.
Chinese-linked projects face greater risks
BMI expects projects built around Chinese-based tenants to face the greatest exposure to the new restrictions, while developments serving US technology companies are relatively better protected by the exemption provisions.
Existing data center operators supplying computing capacity above the prescribed threshold to Chinese-headquartered customers will have 90 days from the rule’s effective date to obtain the necessary licenses.
Assuming the ruling takes effect on Nov 19 as scheduled, the compliance deadline would fall on Feb 17, 2027.
The research house expects the new requirements to influence the allocation of computing capacity in overseas data centers, with affected customers potentially shifting towards smaller, fragmented and more extensively documented computing arrangements rather than being completely shut out of access.
BMI said license decisions would also take into account the effectiveness of export controls implemented by the five affected countries, potentially encouraging their governments to strengthen domestic regulatory frameworks.
Singapore, for instance, has indicated that it will explore trusted access arrangements for its companies.
The research house noted that the new controls could be calibrated rather than absolute, as the licensing requirement applies only when data centre operators exceed the prescribed computing threshold for restricted customers.
Computing capacity used solely for commercial recommender systems is excluded, provided the operator fulfils the relevant reporting requirements.
The restrictions also apply specifically to overseas data centres, meaning computing capacity located within the US falls outside this particular licensing requirement.
Further changes possible, but compliance remains key
Looking ahead, BMI expects the substance of the new restrictions to remain largely intact, although some details could be revised following the public consultation period, which closes in January 2027.
There is also growing pressure from the US Congress for tighter semiconductor export controls, with three chip-control measures reportedly included in the Senate version of the fiscal 2027 defence bill.
However, BMI identified two potential developments that could soften the implementation of the restrictions.
These include the scheduled meeting between US President Donald Trump and Chinese President Xi Jinping in Shenzhen on Nov 19, 2026, and the Jan 10, 2027 deadline for the existing US-China trade truce.
Both developments could provide Washington with incentives to extend trade concessions or ease some restrictions as part of broader bilateral negotiations.
BMI noted that the Trump administration had previously demonstrated flexibility by allowing case-by-case licensing for Nvidia’s H200-class chips in January 2026.
Nevertheless, it cautioned against assuming that similar concessions would be extended to the latest controls.
For the first quarter of 2027, the research house recommends that businesses plan on the basis of full compliance by the Feb 17 deadline, treating any potential regulatory relief as an upside rather than an expected outcome.
Overall, the new US framework adds another layer of regulatory complexity to Southeast Asia’s data centre investment landscape, with ownership structures, customer relationships and access to computing capacity becoming increasingly important considerations for operators, investors and technology suppliers.

