Despite the Trump administration’s stricter export controls on advanced semiconductor technology of the next generation, at least $1 billion worth of advanced AI chips of Nvidia found their way to China through sophisticated networks of smugglers within three months of the controls.
The prohibited chips–such as Nvidia’s top-of-the-line B200, H100, and H200 models–are the exact same chips used by the world’s most sophisticated artificial intelligence systems on the planet today. They are so strategically valuable that the U.S. government specifically singled them out for export controls to keep them out of Chinese hands, lest they be used for AI research and possible military uses.
The smuggling networks were remarkably professional and bold. Rather than transferring individual chips from one side of the border to the other, smugglers coordinated whole server racks of the illicit processors and had them sent directly to Chinese data centers. This allowed Chinese companies to bypass U.S. regulations entirely in order to obtain the exact hardware they needed for their AI research.
China’s Thriving Black Market for Prohibited AI Chips Defies US Embargo
By May 2025, the demand had become so intense that Chinese vendors ceased to deny their business. They started openly marketing these prohibited chips to AI-oriented data center operators, selling the prohibited hardware as any other product even as the U.S. embargo continued.
Industry experts estimated the number of such prohibited Nvidia chips to be between 10,000 and hundreds of thousands that found their way into China in 2024 alone. The scale of the operation helped China’s fast-expanding AI and data center infrastructure as U.S. policymakers attempted to crack down on the growth.
The smuggling operations involved more than simple border crossings, too. They were embedded in sophisticated international networks, as demonstrated when Singaporean authorities arrested people involved in suspected theft of $390 million servers that were ultimately bound for China.
Nvidia, stuck between government limits and huge market demand, attempted to play down the success of the smuggling operations. The company claimed that it is technically difficult and economically unviable to build data centers using smuggled chips, as Nvidia only offers service and support for products sold through official channels.
The Trump administration and U.S. Commerce Department put numbers to the extent of the issue and took steps to investigate new legislation aimed at improving AI chip tracing. The moves were made amid growing fears that current export controls were failing to deliver.
The Unintended Consequences of US Chip Export Controls on China
It all got more complicated again when Trump reversed himself in July 2025 and issued a license to allow Nvidia to resume selling some chips namely the H20 model to China. It connected chip commerce with the negotiating over rare-earth minerals, showing how deeply semiconductor policy was getting mixed into wider trade negotiations.
The sheer scale of chip smuggling has exposed fundamental weaknesses in the functioning of export controls in practice. Where there is such vast demand for such a highly sought-after product as cutting-edge AI chips, determined buyers and sellers will find ways around the government’s restrictions.
These events have actually defeated what U.S. policymakers had desired. Rather than slowing down China’s AI development, the export controls potentially only helped establish a profitable black market while not being able to stop Chinese businesses from obtaining the technology they desired.
The case has touched off intense controversy among lawmakers regarding national security and how to impose integrity in semiconductors supply chains. Some are urging even stronger controls and better enforcement, while others question whether such controls can even work in this era of globalization.
China Revenue, AI Chips, and US Sanctions
For Nvidia, the stakes are tremendous. The company’s China business generated $17 billion in revenue last fiscal year nearly 13% of the company’s overall revenue. To lose that market altogether would be a devastating economic blow, forcing the company to come up with a way to continue selling to Chinese consumers without running afoul of U.S. law.
It is a classic example of the challenges that the government faces in controlling trade in much-sought-after technology. The $1 billion of illicit chips is only a fraction of the global AI chip market, but it is a fair reflection of the rapidity with which sophisticated networks are able to evolve to meet demand when official means are cut off.
As artificial intelligence technology continues to become the focal point of economic competitiveness and national security, anticipate these tensions between policy goals and market forces to grow further.
