The Trump administration is considering new restrictions aimed at preventing Chinese artificial intelligence companies from accessing advanced U.S. chips remotely through overseas data centers, as Washington confronts another potential route around controls designed to limit China’s access to cutting-edge computing power.
A small group within the Commerce Department has been developing a replacement for the Biden administration’s discarded AI Diffusion Rule, according to an Aug. 28 The Information report citing four people familiar with the discussions.
One of the proposal’s main targets is Chinese AI companies accessing advanced Nvidia processors housed in data centers outside China, including facilities in Thailand and Singapore. The chips themselves never have to cross the Chinese border: customers can potentially connect to the machines remotely and rent their computing power to train AI models.
That represents a new challenge for an export-control regime traditionally built around tracking physical products.
U.S. lawmakers have increasingly argued that preventing the direct sale of restricted GPUs to China accomplishes little if Chinese companies can obtain essentially the same computing capability through servers located elsewhere.
For the best of our weekly content！
You are now signed up for our newsletter
Check your email to complete sign up
Rep. Josh Gottheimer (D-N.J.) said in June that China and other U.S. adversaries should not be allowed to circumvent export controls by renting technology they are barred from buying. “We can’t let our adversaries — especially China — dodge our export controls by simply renting what they can’t buy,” he said as he introduced bipartisan legislation targeting cloud access, according to the House Select Committee on the Chinese Communist Party.
The emerging cloud restrictions come as U.S. and allied authorities have uncovered a series of alleged schemes involving third countries, shell companies and false shipping information used to move advanced computing equipment toward China.
Restricted AI hardware has repeatedly surfaced in diversion cases
The cloud-computing issue is only the latest front in a broader struggle over China’s access to advanced American AI technology.
On Aug. 25, Taiwanese prosecutors charged nine people in an alleged scheme to illegally send servers equipped with Nvidia’s advanced B300 processors to China. According to the Associated Press, prosecutors said 74 servers reached China after being routed through Indonesia, Japan and Hong Kong, while another 56 were stopped in Taiwan.
Those charged included an Nvidia manager and former employees of server manufacturer Super Micro Computer. Prosecutors alleged that participants established companies and websites and falsified information about where the equipment would ultimately be used. Nvidia and Super Micro said they were cooperating with authorities.
The case follows several U.S. investigations involving similar allegations.
In March, the Justice Department charged a Chinese national and two U.S. citizens with conspiring to illegally obtain millions of dollars worth of export-controlled computer chips and send them to China through Thailand. Prosecutors alleged that the defendants attempted to conceal the chips’ true destination.
Assistant Attorney General for National Security John Eisenberg said the chips represented “the best of American ingenuity” and that the department would pursue violations of U.S. export controls.
Another Justice Department case unsealed in March alleged that three people attempted to divert high-performance U.S.-assembled AI servers to China using false paperwork, dummy servers and complicated transshipment routes.
Prosecutors described the alleged operation as an effort to use “false documents, staged dummy servers” and transshipment schemes to hide China as the ultimate destination, according to the Justice Department.
And in December 2025, U.S. authorities announced that an operation targeting a separate China-linked trafficking network had resulted in the seizure of more than $50 million in Nvidia technology and cash. One defendant and his company pleaded guilty after authorities said they had exported or attempted to export at least $160 million worth of controlled Nvidia H100 and H200 GPUs.
The House Select Committee on the CCP concluded in an April investigation that China obtains advanced AI capabilities through a combination of legitimate purchases and illicit methods.
Its report said China “utilizes sophisticated smuggling networks to acquire restricted AI chips” and recommended tightening both physical export controls and restrictions on remote access to computing infrastructure.
Washington turns its attention from chips to computing power
The increasing use of overseas infrastructure has forced U.S. policymakers to rethink what an AI technology “export” actually means.
When an Nvidia GPU is physically shipped to China, customs records, shipping documents, serial numbers and end-user declarations can potentially be examined.
Remote access is different.
A data center in Thailand or Singapore can legally possess thousands of U.S.-made processors while a customer located elsewhere connects through the internet. The customer is effectively purchasing computing capacity rather than taking possession of the chips themselves.
The Information reported that awareness of the problem increased after White House Office of Science and Technology Policy Director Michael Kratsios raised concerns in July that Chinese AI developer Moonshot AI may have used advanced Nvidia processors located in Thailand.
The Commerce Department has already begun tightening rules covering Chinese companies operating outside China.
In May, the Bureau of Industry and Security issued guidance clarifying that advanced computing products can still require an export license when supplied to a company located outside China if the company or its ultimate parent is headquartered in China or another restricted jurisdiction.
Existing BIS regulations also contain requirements for certain applicants to obtain information about data centers’ “Know Your Customer” procedures and identify remote Infrastructure-as-a-Service users located in China and other restricted countries.
The rule now being developed could go further by requiring overseas operators receiving advanced American chips to verify who is accessing their GPU clusters and how those customers are using the computing power.
But there is a legal complication.
Janet Kim, an export-control attorney at Baker McKenzie, told The Information that it is “widely acknowledged” that the Commerce Department does not currently have clear authority to regulate true remote access.
Congress is attempting to close that gap
The House passed the bipartisan Remote Access Security Act in January. The legislation would explicitly allow U.S. export controls to cover remote access to controlled technology through cloud-computing services.
“The CCP’s AI ambitions are being fueled by its access to American chips housed in data centers located outside of China,” Select Committee Chairman John Moolenaar (R-Mich.) said after the bill passed.
The bill remains under consideration in Congress, while Commerce explores what it can accomplish through existing regulatory authority.
The new rule could be circulated among AI companies and industry groups for feedback as early as September, according to The Information. The proposal could still change or fail to advance.
The broader direction, however, is becoming increasingly clear.
Washington’s technology controls began by attempting to stop the most advanced U.S. chips from physically reaching China. Alleged diversion schemes have demonstrated how equipment can be routed through third countries, while cloud computing creates another possibility altogether: leaving the chips overseas while allowing Chinese companies to use their processing power remotely.
Original article
Comments