Research suggests the most advanced AI models developed by Chinese firms were trained on Nvidia hardware that the US government has banned from export.
This finding highlights a major challenge for the chipmaker. Nvidia CEO Jensen Huang stepped down this week after admitting to an affair with Alice Huang, a Megaspeed executive implicated in smuggling Nvidia chips into China. Alice Huang has also been accused of acting as an agent for China’s intelligence service.
While public attention focused on Huang’s personal relationships, the company’s potential liability for export control violations likely influenced the decision to end his tenure.
US export controls aim to protect American AI leadership against China, yet Huang previously sought to open that market. Blackwell GPUs are prohibited from export to China or Chinese-owned firms. The Trump administration allows Nvidia to ship the H200, a less powerful chip, provided shipments follow a circuitous route that enables the US to claim a 25% tariff.
Will DePue, a former OpenAI researcher leading the nonprofit Unemployment Labs, examined the weights of large language models released by DeepSeek, Moonshot AI, MiniMax and Ant Group. These weights reflect the chip configurations used during training. DePue found evidence that key models, including Moonshot’s Kimi K3 releases and DeepSeek’s V4, bear fingerprints indicating training on Blackwell chips.
White House technology adviser Michael Kratsios stated earlier this year that Moonshot had used export-banned Nvidia chips, citing US intelligence sources. Epoch AI, a nonprofit research organization, estimates that approximately two million advanced chips entered China in 2024 and 2025 via schemes designed to hide their final destination.
This does not mean these models were trained entirely on Nvidia hardware, nor that export laws were necessarily violated. The chips could have been located in an offshore data centre, a popular method for Chinese firms to access compute that has driven an industrial revolution in Malaysia. Epoch data estimates that 275,000 Nvidia AI accelerators were imported to Malaysia in 2024 and 2025.
However, the stated goal of the export bans is to prevent exactly this scenario: Chinese AI labs training on the most powerful chips. China hawks have long called for stronger controls on who uses chips, not just who owns them or where they are installed.
“If we care about the U.S. tech edge, it doesn’t make any sense to me that you would not allow Chinese companies to access chips in mainland China or Hong Kong and Macau, but then say, actually, if you just go over to Malaysia, you can do whatever you want,” Sam Bresnick, a researcher at the Georgetown Center for Security and Emerging Technology, told TechCrunch this summer.
Megaspeed, led by Alice Huang, has been central to this trade for years through its subsidiary Speedmatrix. The company has been investigated by both the US and Singaporean governments. Bloomberg reported on Friday that a company with links to Alice Huang was involved in smuggling banned Nvidia chips to a data centre in Shanghai.
In May, the short-selling firm Culper published a report arguing that Nvidia faced a “China problem” because a significant share of its revenue came from Chinese demand, either through remote data centres or smuggling in violation of US law. It detailed Megaspeed’s relationship with Nvidia and alleged that Alice Huang’s companies are backed by Chinese firms barred from accessing US chips.
Ironically, Culper’s bet against Nvidia was not driven by an expectation of a US government crackdown. Instead, it focused on the Chinese government’s response to the bans, which was to recommit to its domestic chip industry.
If China succeeds, billions of dollars in Chinese demand for Nvidia chips could disappear. Approximately 600,000 H200 chips are reportedly stockpiled in Hong Kong, awaiting shipment into China. While Chinese companies might want the chips, observers have no consensus on why they have not been cleared for import.
Graham Webster, a research scholar at Stanford focusing on US-China technology policy, said he does not accept the idea that holding the chips back would drive domestic production, given the sheer scale of demand. Other theories for the delay include preparation for a potential trade deal involving a large volume of US goods, or concerns about potential surveillance technology in the chips.
For now, US companies have limited obligations to track what their customers do with controlled goods like AI chips. However, Jensen Huang’s personal and professional entanglements with Alice Huang create a much bigger problem for Nvidia if government investigators confirm that chip smuggling took place and he was aware of it. Nvidia and Huang have both vouched for Megaspeed’s data centres, stating that Nvidia officials visited them to confirm legitimacy.
The US government’s treatment of Nvidia and Megaspeed is also under scrutiny. Secretary of Commerce Howard Lutnick, who oversees the agencies licensing chip exports, could reportedly lose his job after reports that his investigators failed to uncover suspicious chip buyers and were too close to Nvidia officials.
What it means
For people making AI, the practical reality is that the most powerful hardware available to them may still be accessible through complex routing or third-party jurisdictions, regardless of official bans. This limits the ability of US export controls to stop the flow of compute power to foreign labs. The situation also creates significant legal risk for the companies supplying the hardware and their executives.
