American regulators are facing a new headache: Chinese artificial intelligence companies are increasingly using a scheme that allows them to circumvent export restrictions on advanced Nvidia chips. Instead of physically importing processors into the PRC, firms rent computing power in data centers located in third countries, such as Malaysia or Thailand, and gain remote access to them.
The mechanics of the "gray" scheme
Since 2022, Washington has been consistently tightening rules on the supply of high-performance AI chips to China. However, current regulations primarily govern the export of physical equipment, leaving a legal loophole for cloud computing. Chinese technology corporations have taken advantage of this.
The essence of the scheme is simple: servers with Nvidia chips remain in Southeast Asia, but their computing power is used to train models by Chinese developers. Based on industry data, my estimates suggest that giants such as Alibaba and ByteDance are among those employing this practice. The U.S. Bureau of Industry and Security has already begun a systematic study of these gray schemes, compiling lists of countries through which both physical traffic of prohibited equipment and virtual access to it flow.
Kimi K3 as a catalyst for discussion
This topic gained particular resonance after the release of the open-source model Kimi K3 by Moonshot AI. In July, the director of the White House Office of Science and Technology Policy, Michael Kratsios, stated that the startup allegedly purchased servers with Nvidia GB300 and had access to them in Thailand. He provided no public evidence, but the very fact of such accusations shows that Washington is seriously concerned about technology leakage.
Notably, Moonshot AI is also accused of using Anthropic Fable's developments in creating its model, which the Chinese side categorically denies. Kimi K3 itself, containing 2.8 trillion parameters, has demonstrated impressive results in benchmarks, indirectly confirming access to powerful computing resources.
Legislative response and Nvidia's position
To close this loophole, the House of Representatives has already approved the "Remote Access Security Act." The document proposes extending export controls not only to physical shipments but also to the use of technologies via the internet and cloud services. The bill now awaits consideration in the Senate.
Nvidia, for its part, opposes expanding the restrictions. The company rightly points out that current rules deliberately allow the creation of cloud infrastructure outside controlled jurisdictions, and its partners undergo strict vetting. The manufacturer has already suffered significant losses: licensing H20 chips in April 2025 cost the company approximately $5.5 billion, and the Chinese market has effectively been lost to it as its second-largest commercial market.
My analysis: This situation demonstrates a fundamental problem with modern export controls — in the era of cloud computing, the physical border ceases to be a barrier to the spread of technology. While the U.S. tries to close one loophole, China, with its vast financial resources and engineering talent, will find another. Moreover, such restrictions only stimulate the development of domestic semiconductor production in the PRC, which in the long term could cause greater damage to the American economy than the technology leakage itself.