Chinese artificial intelligence firms are circumventing U.S. export restrictions on advanced Nvidia semiconductors by renting computing power remotely through data centers in Southeast Asia, a strategy that exposes a critical weakness in America's technology containment strategy. While the U.S. government has successfully blocked direct shipments of Nvidia's most powerful chips, including the GB300 model, to China, companies like Moonshot AI, DeepSeek, and Alibaba have found a legal workaround: they simply access the chips' computing power from overseas facilities instead of owning the hardware themselves.

How Are Chinese Companies Accessing Banned Chips Legally?

The answer lies in a fundamental gap between what U.S. export controls actually regulate and what they leave unprotected. The current regime focuses exclusively on the physical ownership and transfer of semiconductor hardware. It does not restrict remote, cloud-based access to those same chips when they sit in data centers outside U.S. borders.

"The U.S. export control regime controls physical AI chips. It does not cover remote access to those chips," explained Cassia King, senior researcher on the Compute Policy team at the Institute for AI Policy and Strategy.

Cassia King, Senior Researcher, Institute for AI Policy and Strategy

This distinction matters enormously. When Moonshot AI released its Kimi K3 model in July, White House official Michael Kratsios publicly accused the company of using Nvidia's GB300 chips through a facility in Thailand. The accusation was notable not because the access was illegal, but because it highlighted how easily Chinese firms could achieve what export controls were designed to prevent: training advanced AI models on cutting-edge U.S. semiconductor technology.

The arrangement works like this: instead of purchasing physical chips, Chinese hyperscalers including ByteDance, Alibaba, and Tencent rent computing capacity from cloud providers operating in countries like Thailand, Malaysia, and Japan. ByteDance, for example, reportedly worked with Singapore-headquartered cloud provider Aolani to access Nvidia compute power in Malaysia, according to reporting from the Wall Street Journal.

What Companies Are Exploiting This Loophole?

The companies taking advantage of this gap represent some of China's most ambitious AI developers. Recent months have seen a wave of new Chinese AI models achieving impressive performance benchmarks, and experts say remote access to advanced compute is a key factor in their rapid progress.

• Moonshot AI: Released the Kimi K3 model in July and was specifically called out by White House officials for accessing GB300 chips through a Thai data center facility.
• ByteDance: The company behind TikTok has reportedly accessed Nvidia compute remotely through Aolani's infrastructure in Malaysia, according to sources familiar with the arrangement.
• Alibaba and Tencent: Both major Chinese tech firms have reportedly accessed computing power from Nvidia chips remotely via facilities across Thailand, Malaysia, and Japan.
• DeepSeek: Recently released new AI systems that have scored well on performance benchmarks, with access to overseas compute cited as a contributing factor.

When asked about these arrangements, Aolani told CNBC that it works "with a global and diversified customer base spanning customers from North America and Asia" and that "any permitted access to our services, infrastructure or technology is fully compliant with all applicable regulations". The company emphasized that clients "do not have ownership, potential future claim or physical access to the chips that power our solutions".

Aolani

Why Is This a National Security Concern?

U.S. policymakers view this loophole as a direct threat to American technological dominance. Nvidia's effective monopoly over the most powerful AI chips has made export controls a cornerstone of Washington's strategy to maintain its advantage over China in the AI race. The entire premise of those controls is to deny China the ability to train frontier AI models using advanced U.S. semiconductors.

"The point of chip export controls is to deny China the ability to train frontier AI using advanced U.S. chips," noted Michelle Nie, a visiting fellow in technology and national security at the Center for a New American Security.

Michelle Nie, Visiting Fellow in Technology and National Security, Center for a New American Security

The concern is not merely theoretical. Southeast Asia is experiencing a data center boom as companies race to build infrastructure for advanced computing. According to data compiled by DC Byte, there are 31 planned data centers of 100 megawatts or larger across Malaysia, Indonesia, and Thailand, compared to just two today. Real estate company JLL estimates that global data center capacity could roughly double to 200 gigawatts by 2030. This expansion creates an expanding network of potential access points for Chinese firms seeking to circumvent U.S. restrictions.

What Legislative Solutions Are Being Considered?

Congress is aware of the problem and has begun drafting solutions. A proposed piece of legislation called the Remote Access Security Act, or RASA, seeks to expand U.S. export controls to include remote cloud-based access to critical hardware and software. The bill passed the House of Representatives in January but has yet to pass the Senate.

However, even if RASA becomes law, significant hurdles remain. The legislation would give the U.S. government the authority to regulate remote access, but it would still need to create specific rules governing which types of computing power are covered and how to enforce restrictions.

"The challenge will be in making a rule that's effective and enforceable. Policymakers will need to decide what compute is covered, who should be prohibited from remotely accessing the compute, and how to implement a robust know-your-customer scheme," said Cassia King.

Cassia King, Senior Researcher, Institute for AI Policy and Strategy

Industry pushback is also expected. Cloud providers would bear significant compliance burdens under any new regime, including customer verification requirements and ongoing monitoring of who accesses their systems. Michelle Nie warned that "cloud providers would bear the compliance burden of any KYC and customer verification requirements mandated by the bill".

Michelle Nie

If the Trump administration decides to move forward, the Bureau of Industry and Security could potentially push through a rule quickly, possibly within "a matter of days" with White House support, according to experts. But the real challenge lies not in speed, but in crafting rules that are both effective at preventing Chinese access and practical for global cloud providers to implement.

How to Understand the Export Control Challenge

• Physical vs. Remote Access: Current U.S. export controls regulate the sale and transfer of physical semiconductor hardware but do not restrict remote, cloud-based access to those chips when located outside U.S. territory.
• The Compliance Gap: Cloud providers operating in Southeast Asia can legally offer computing services powered by restricted Nvidia chips to any customer, including Chinese firms, as long as the customer does not own the physical hardware.
• The Enforcement Problem: Even if new legislation passes, regulators must decide which types of computing power to restrict, which customers to prohibit, and how to verify customer identity and intended use across international borders.
• The Infrastructure Expansion: Data center capacity in Southeast Asia is growing rapidly, creating more potential access points for Chinese companies seeking to train advanced AI models outside U.S. jurisdiction.

The loophole reveals a fundamental tension in modern technology policy: export controls were designed for a world where advanced computing meant owning physical equipment. Today's cloud-based infrastructure has made that assumption obsolete. As Chinese AI firms continue to release competitive models and the U.S. considers its next move, the question is whether policymakers can close the gap before remote access becomes the primary pathway for Chinese AI development.

" }