Malaysia has told data centre operators to comply with new US restrictions on advanced artificial intelligence chips, putting its fast-growing data centre industry under closer regulatory scrutiny. Trade Minister Johari Abdul Ghani said operators using US-origin graphics processing units (GPUs) must follow American export-control requirements, including licensing and customer due-diligence rules.

The announcement comes as Malaysia seeks to strengthen its position as a major Southeast Asian data centre hub, attracting nearly $100 billion in investment since 2021. The new US measures target the potential diversion of advanced AI chips to restricted markets, particularly China, including through overseas cloud and data centre services. Malaysia says it will inform all 117 approved data centre projects about the requirements once it receives official clarification from Washington.

Key takeaways

  • Malaysia says all data centre operators using advanced US-origin AI chips must comply with US export-control rules.
  • The country has attracted nearly $100 billion in data centre investments since 2021, according to Reuters.
  • Malaysia has 117 approved data centre projects, including facilities already operating and projects still under construction.
  • The US has introduced new licensing requirements for advanced chip shipments to Malaysia, Singapore, Thailand, Indonesia and Brazil.
  • The rules also restrict certain overseas data centres from providing substantial computing capacity to companies headquartered in China and other restricted jurisdictions.
  • Malaysia is seeking clarification from Washington on licensing requirements and third-party access to computing services.
  • The government says it has not yet received documentary evidence proving that Malaysian data centre tenants illegally re-exported chips.
  • The new rules could increase compliance costs, delay GPU deployment and influence future data centre investments.

Malaysia tells data centres to comply with US AI chip restrictions

Malaysia’s Trade Minister Johari Abdul Ghani has warned data centre operators that they must follow US regulations if they want to continue using advanced American AI chips.

Speaking on November 18, 2026, Johari said Malaysia remained open to foreign investment but expected companies operating in the country to respect the export-control requirements attached to US-origin technology.

The statement follows Washington’s announcement of fresh restrictions on advanced semiconductor shipments and the use of computing infrastructure by customers in restricted markets.

Malaysia is particularly important because it has become a major destination for data centre investment in Southeast Asia.

The country offers access to regional markets, industrial infrastructure and growing digital demand. Johor, in particular, has attracted significant interest from international cloud providers and data centre developers.

However, the expansion has also brought increased scrutiny over whether advanced AI chips installed in Malaysian facilities could be accessed by customers who are restricted from purchasing the hardware directly.

Johari said the government would communicate the requirements to all 117 approved data centre projects once the United States provides official clarification.

Some projects are already operational, while others remain under construction.

He also said Malaysia was communicating with US authorities, including the US Embassy and trade representatives, to understand the rules and their implications for local operators.

What are the new US AI chip rules?

The new measures were introduced by the US Department of Commerce’s Bureau of Industry and Security (BIS).

They are designed to prevent restricted companies, particularly those headquartered in China and other countries subject to US arms embargoes, from accessing advanced American computing technology through third-country data centres.

Previously, US export controls primarily focused on restricting the physical sale or transfer of certain advanced chips to designated destinations and entities.

However, the expansion of cloud computing has created another route to access computing power.

A company that cannot legally purchase an advanced AI accelerator may still attempt to rent computing capacity from a data centre located in another country.

The new rules address this issue by placing additional obligations on chip exporters, data centre operators and other parties involved in supplying computing services.

The regulations include two important elements.

1. New licensing requirements for advanced chip shipments

The US has introduced destination-based licensing requirements for certain advanced computing semiconductors shipped to five countries:

  • Malaysia
  • Singapore
  • Thailand
  • Indonesia
  • Brazil

The requirements cover exports, re-exports and certain transfers within these countries.

This means companies involved in bringing covered advanced chips into these markets may need US export licences.

The rule includes a conditional exemption for shipments destined for companies headquartered in the United States, or whose ultimate parent companies are headquartered there, provided specified reporting requirements are met.

Other transactions may require additional scrutiny and case-by-case licensing decisions.

For Malaysia, this creates a new compliance requirement for businesses planning to install advanced US GPUs in local data centres.

2. Restrictions on remote access to advanced computing power

The second part of the rule addresses overseas data centres that provide computing capacity to restricted customers.

Under the new framework, certain data centre arrangements involving companies headquartered in China, Macau and other US arms-embargoed jurisdictions can require a licence if they provide a significant volume of covered computing capacity.

This is important because the restrictions are not limited to the physical movement of chips.

They also consider how the computing power generated by those chips is used and who receives access to it.

For example, a company headquartered in China may be unable to purchase a particular advanced US GPU directly. However, it could potentially rent access to a server equipped with that GPU from a data centre operating in Malaysia.

The new restrictions are designed to prevent this type of arrangement from being used to circumvent US export controls.

Why Malaysia is under greater scrutiny

Malaysia’s position as a rapidly growing data centre destination has made it strategically important in the global AI infrastructure supply chain.

The country has attracted substantial investment from international technology companies, cloud providers and data centre developers.

Its location provides access to Southeast Asian markets, while infrastructure development in Johor has benefited from demand connected to Singapore’s technology ecosystem.

However, the same growth has raised questions about the movement and use of advanced AI hardware.

The United States is concerned that restricted customers could access advanced computing technology indirectly through facilities located in countries that are not subject to the same direct chip-sale restrictions.

Malaysia is among the countries now facing additional scrutiny under the new US framework.

The government has said it wants to remain an attractive investment destination while ensuring that companies comply with the regulations governing sensitive technology.

Johari said Malaysia would not allow data centres to operate outside applicable rules simply because the chips were physically installed within Malaysian territory.

He also stressed that the United States, as the producer of the chips, should conduct appropriate due diligence on buyers and communicate relevant information to Malaysian authorities.

Malaysia investigating alleged AI chip diversion

The latest announcement comes amid scrutiny of whether advanced Nvidia chips have been diverted through Southeast Asian data centre operations.

One company attracting attention is Megaspeed International, a Singapore-based cloud computing company whose Malaysian operations have faced scrutiny over allegations involving access to advanced AI chips.

The allegations concern whether third-party infrastructure and computing arrangements could have been used to provide restricted Chinese customers with access to advanced US technology.

Megaspeed has denied wrongdoing and said it takes its US compliance obligations seriously.

Malaysia’s Trade Minister Johari said investigations were continuing, but the government had not received documentary evidence proving that data centre tenants had traded or illegally re-exported the chips.

He said authorities would need to examine customs declarations, shipping manifests, the parties involved and the actual movement of GPUs before determining whether violations occurred.

This distinction is important.

The existence of an investigation does not establish that a company has violated export-control rules.

Malaysia’s government has said it wants evidence before making accusations or taking enforcement action.

The investigation also highlights the challenge of monitoring AI infrastructure. A GPU can remain physically installed inside a data centre while its computing power is accessed remotely by customers in different countries.

Regulators therefore need to examine both hardware movements and the commercial arrangements governing access to computing services.

What the new rules mean for data centre operators

The new US restrictions could make it more complicated for Malaysian data centre operators to purchase, install and operate advanced AI accelerators.

Companies may need to provide more information about their ownership structures, customers, end users and the purpose for which computing capacity will be used.

Due diligence requirements are particularly important.

Operators and exporters may need to establish whether customers are directly or indirectly owned or controlled by entities in restricted jurisdictions.

This could increase the amount of documentation required before a transaction is approved.

The main implications include:

AreaPotential impact
GPU procurementAdditional licensing requirements for covered chips
Customer verificationMore detailed know-your-customer checks
Ownership disclosuresGreater scrutiny of direct and indirect ownership
Cloud servicesRestrictions on certain remote computing arrangements
Project timelinesPossible delays in installing or activating GPUs
Compliance spendingHigher legal, administrative and monitoring costs
International partnershipsGreater importance of transparent ownership and customer relationships

The impact will vary depending on the ownership of the data centre, the origin of the hardware, the intended customers and the nature of the computing services.

Facilities operated by qualifying US-headquartered companies may benefit from certain exemptions, subject to reporting conditions.

Other operators may face more extensive licensing and verification requirements.

Why the rules could affect Malaysia’s $100 billion data centre ambitions

Malaysia has attracted nearly $100 billion in data centre investments since 2021, according to Reuters.

This investment reflects growing demand for cloud computing, artificial intelligence, enterprise data storage and digital services across Southeast Asia.

Data centres are capital-intensive facilities. Developers spend heavily on land, construction, electricity infrastructure, cooling systems, networking equipment and computing hardware.

However, access to advanced GPUs has become increasingly important as companies expand AI services.

A data centre may have sufficient land, electricity and cooling capacity but still struggle to attract AI customers if it cannot obtain the necessary computing hardware.

The new licensing requirements therefore introduce another variable into investment decisions.

Developers must consider not only the availability of electricity and land but also whether their hardware procurement arrangements and customer relationships satisfy US regulations.

If licensing procedures take longer, some projects could experience delays in deploying AI computing capacity.

That could affect project economics, particularly when developers have already committed significant capital to buildings and infrastructure.

However, the new rules do not automatically stop data centre construction or prohibit every AI workload in Malaysia.

Their impact depends on the specific chips, ownership structures, customers and computing services involved.

Could international technology companies reconsider Malaysia?

The new rules could influence how international cloud providers and technology companies structure their investments in Malaysia.

Large US technology companies may be better positioned to navigate the requirements because certain transactions involving qualifying US-headquartered companies are eligible for conditional exemptions.

Other operators may need to undertake more extensive licensing and customer verification procedures.

This could influence decisions about:

  • Which companies operate data centres.
  • Which entities purchase and own advanced GPUs.
  • Where computing capacity is deployed.
  • Which customers can access the infrastructure.
  • How cloud services are contracted and delivered.
  • How international data centre partnerships are structured.

Companies may also become more selective when accepting customers whose ownership structures or ultimate end users are difficult to verify.

The result could be a more compliance-intensive AI infrastructure market.

For Malaysia, the challenge will be to maintain investment momentum while demonstrating that its data centre ecosystem can operate as a trusted destination for advanced computing technology.

Singapore, Thailand, Indonesia and Brazil face similar restrictions

Malaysia is not the only country affected by the new US measures.

Singapore, Thailand, Indonesia and Brazil are also covered by the destination-based licensing requirements for certain advanced computing chips.

These countries have different roles in the global technology supply chain, but all are important markets for data centre investment and digital infrastructure.

The United States is seeking to prevent advanced chips from reaching restricted customers through intermediary markets.

This could create a broader compliance challenge for cloud providers operating across several countries.

For multinational companies, the rules may require consistent customer verification and end-use monitoring across different jurisdictions.

The measures could also encourage governments in Southeast Asia to strengthen their own export-control systems and cooperate more closely with US authorities.

Malaysia has already signalled that it wants additional clarity before communicating the full requirements to all approved projects.

What happens to China’s access to overseas AI computing?

China is a central consideration behind the new restrictions.

US export controls have limited China’s ability to acquire certain advanced American AI accelerators directly.

However, access to computing power can also be obtained through cloud services and overseas data centres.

A company may not own the physical GPU but can still use its computing capabilities remotely.

This creates a potential route around restrictions designed around hardware ownership and physical shipment.

The new rules attempt to close that gap by examining who receives the computing service.

For Chinese AI developers, this could make it more difficult to obtain large amounts of advanced US computing capacity through overseas infrastructure.

For data centre operators, it creates additional responsibility to understand their customers and the ultimate users of their services.

The restrictions could also influence the economics of AI cloud services in Southeast Asia, especially for providers that depend on international customers.

The bigger picture

The latest US restrictions show that the global AI infrastructure market is increasingly shaped by geopolitics and technology export controls.

Advanced GPUs are no longer treated simply as commercial computing products. Governments increasingly view them as strategic technologies because they can support advanced AI development, scientific research and military-related applications.

That is changing the responsibilities of companies operating data centres.

Previously, a data centre provider could focus primarily on supplying electricity, cooling, networking and computing capacity to customers.

Now, operators may also need to understand the ownership, location and intended use of the computing services they provide.

For Malaysia, the challenge is balancing its ambition to become a regional AI infrastructure hub with the need to maintain compliance with US restrictions on advanced chips.

The nearly $100 billion investment pipeline shows the scale of the opportunity. But the ability to obtain advanced GPUs and serve international customers under changing export-control rules will be just as important as physical infrastructure.

Looking ahead

Malaysia will continue discussions with US authorities to clarify the licensing requirements and the rules governing data centre services provided to third parties. The government plans to inform all 117 approved data centre projects about the requirements once official guidance is received. Operators will need to review their hardware procurement, customer verification, ownership disclosures and cloud service arrangements to determine whether additional licences are required.

The broader impact will depend on how strictly the rules are implemented and how quickly companies can adapt. Malaysia’s data centre industry may continue attracting investment, but projects involving advanced US chips could face more detailed checks and longer approval processes. The development also signals a wider change in the global AI industry: access to computing power is increasingly determined not only by cost and technical performance but also by export controls, customer identity and geopolitical considerations.

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.