Chinese AI firms seek Nvidia computing power in Southeast Asia

Seeking a computing workaround

Chinese artificial intelligence companies are actively looking for ways to secure Nvidia-powered computing capacity outside of China. This effort follows ongoing US export controls that restrict the shipment of advanced AI processors to the Chinese market. By using overseas facilities, these firms hope to maintain their development pace without importing restricted hardware directly.

Current regulations allow Chinese companies to rent processing power from overseas cloud services, provided they follow certain export-control requirements. These services enable developers to train and run AI models remotely without physically owning the underlying chips. However, the ease of this remote access has prompted a shift in how US authorities view international trade.

Reports from Tech Wire Asia indicate that several Chinese cloud providers and smaller AI developers are already in talks with Southeast Asian operators. These discussions aim to secure the high-performance computing needed for large-scale model training. While the specific companies and amounts of capacity have not been disclosed, the trend suggests a significant shift toward regional hubs.

The Remote Access Security Act

US lawmakers are currently considering the Remote Access Security Act (RASA) to address these remote access loopholes. The legislation, which passed the House of Representatives in January 2026, aims to expand government oversight of advanced computing technology. It targets access through overseas cloud services rather than just the physical movement of hardware across borders.

If passed, RASA would authorize the US Commerce Department to regulate remote access to controlled technology through network connections. The Congressional Budget Office estimates that the compliance costs and lost revenue from this proposal could exceed US$206 million in the first year. The bill would give the Bureau of Industry and Security the power to issue licenses and impose penalties for violations.

The act does not automatically ban Chinese companies from renting overseas capacity. Instead, the specific restrictions would depend on the regulations and licensing requirements established once the bill becomes law. This represents a move toward a more granular control system that monitors how data and computing power flow across the internet.

Malaysia as a regional hub

Southeast Asia, particularly Malaysia, is becoming a primary destination for the infrastructure required to support these AI workloads. The Malaysian Investment Development Authority reported that global tech giants like Amazon, Google, Microsoft, and Oracle have announced over US$16 billion in data center projects. Much of this investment is concentrated in the Johor region near Singapore.

Malaysia is specifically attracting these projects because it can support high-density computing systems that require advanced cooling technology. For example, Bitdeer Technologies Group is developing a 67MW facility in Malaysia designed for Nvidia GB300 NVL72 systems. This facility will use liquid cooling to manage the heat generated by rack-scale GPU systems used for heavy AI workloads.

Other operators are also expanding their footprint in the region to meet growing demand. ST Telemedia Global Data Centres recently broke ground on a 22-acre campus in Johor with a development potential of 120MW of IT capacity. These projects illustrate how the region is positioning itself as a critical node for the global AI supply chain.

Compliance and scrutiny

The move toward overseas computing has already drawn scrutiny from US authorities regarding how cloud providers manage their customers. Megaspeed, a Singapore-based cloud company, faced investigation over suspicions that it helped Chinese customers obtain restricted Nvidia processors. The company has denied these allegations and stated that compliance is a fundamental priority for its operations.

Current Export Administration Regulations already require applicants for certain advanced computing licenses to disclose information about their remote customers in restricted jurisdictions. They must also provide details on their customer-screening procedures to prevent unauthorized remote access. These rules apply to specific licensing arrangements rather than every type of cloud service available today.

The Bureau of Industry and Security clarified in May 2026 that data center operators following these rules are not required to stop using affected equipment just because of new guidance. However, the introduction of RASA would provide additional statutory authority to regulate these connections. This creates a complex environment for cloud providers who must balance global demand with increasingly strict trade laws.

What comes next for AI trade

The next phase of this story will depend on whether the Remote Access Security Act successfully moves through the Senate and becomes law. If it passes, the US government will begin establishing the specific licensing requirements that could affect how Chinese firms use Southeast Asian data centers. This could lead to a more complicated and costly process for companies seeking high-end computing power.

For the AI industry, this means that the geographical location of a data center is becoming as important as the hardware inside it. Companies will need to navigate a web of international regulations to ensure they can access the chips necessary for their models. This may lead to a fragmented landscape where some regions are easier to access than others.

Investors and developers should watch for the specific licensing rules issued by the Commerce Department. These rules will determine the actual cost and feasibility of using overseas cloud services for AI training. As the infrastructure in Malaysia and other parts of Southeast Asia comes online, the tension between trade policy and technological demand will likely intensify.

Try it

Watch

Whether the Remote Access Security Act passes the Senate and what specific licensing rules are created.

Which specific Chinese AI firms are signing the largest overseas computing contracts.

How Malaysia's data center capacity will impact the cost of AI training for global firms.