American regulators are seriously concerned about a new scheme that allows Chinese AI developers to circumvent export restrictions on advanced Nvidia chips. Instead of directly importing processors into the PRC, companies rent computing power in data centers outside the country, keeping the physical hardware in Malaysia, Thailand, and other Southeast Asian nations, while accessing it remotely.
The Mechanics of the Loophole
Since 2022, Washington has been consistently tightening the rules on supplying the most powerful AI chips to China. However, the restrictions primarily target the physical export of hardware, leaving a significant legal gray area for remote access to these capabilities. Chinese giants, including Alibaba and ByteDance, are actively exploiting this by renting servers with Nvidia GPUs in Southeast Asia. In my assessment, this is not just a one-off trick but a systemic strategy that allows Chinese companies to maintain access to cutting-edge technology without formally violating the letter of the law.
The U.S. Bureau of Industry and Security is currently trying to close this gap. The agency is compiling lists of countries through which banned chips could physically reach the PRC, as well as states where Chinese firms gain remote access to them. The case of Alibaba is telling: the company works with Singapore-based Megaspeed, which in turn is linked to a structure in the Cayman Islands controlled by the Chinese giant. The agreement itself is structured to avoid directly violating export rules, making it legally difficult to challenge.
Kimi K3 as a Catalyst for Discussion
The topic of sanctions evasion resurfaced after the release of the open-source model Kimi K3 from Moonshot AI with 2.8 trillion parameters. In July, the director of the White House Office of Science and Technology Policy, Michael Kratsios, stated that the startup may have rented servers with Nvidia GB300 in Thailand to train its models, although he provided no public evidence. Notably, Moonshot AI had previously been accused of distilling capabilities from Anthropic's Fable, which the Chinese side categorically denied.
A similar story previously emerged around DeepSeek. In June 2025, the U.S. State Department claimed that the startup attempted to use shell companies in Southeast Asia to bypass controls. Nvidia then countered that DeepSeek used legally acquired H800 chips. This pattern—renting cloud capacity through third-party jurisdictions—is becoming the new norm for China's AI sector.
Legislative Response and Nvidia's Position
The U.S. Congress is proposing a radical solution—the Remote Access Security Act. The document, already approved by the House of Representatives in January, extends export controls to the use of technology via the internet and cloud services. If the law is passed by the Senate, restrictions would apply not only to the physical supply of chips but also to the very possibility of utilizing their computing power from China.
Nvidia, for its part, opposes further tightening. The company emphasizes that current rules deliberately allow the creation of cloud infrastructure outside controlled jurisdictions, and its partners undergo rigorous vetting. The manufacturer also warns that export bans have already effectively deprived it of its second-largest market and play into the hands of competitors from other countries. Recall that in April 2025, the U.S. imposed licensing requirements on the H20, costing Nvidia $5.5 billion, and in January 2026, it allowed shipments of the more powerful H200 to select Chinese clients under certain conditions.
My take: This struggle is a classic arms race in the AI sphere. While the U.S. tries to close cloud loopholes, China is already adapting, seeking new jurisdictions and schemes. The question is not whether the PRC can gain access to chips, but how quickly Washington can rewrite the rules of the game without destroying the global technological ecosystem. The bill is just a first step, and its effectiveness will depend on international cooperation, which remains uncertain for now.