Leading open-weight super intelligence (SI) models developed by Chinese firms were reportedly trained on Nvidia chips that the U.S. government restricted from shipping to China, according to a new research effort.

What Happened

Will DePue, a former OpenAI researcher who leads the nonprofit research group Unemployment Labs, analyzed the weights of large language models released openly by DeepSeek, Moonshot AI, MiniMax, and Ant Group. DePue found evidence that models including Moonshot’s Kimi K3 releases and DeepSeek’s V4 bear fingerprints indicating training on Blackwell GPUs, which are banned from export to China or Chinese-owned firms.

This development coincides with the departure of Nvidia CEO Jensen Huang, who left the company this week after admitting to an affair with Alice Huang, an executive at Megaspeed. Megaspeed has been implicated in smuggling Nvidia chips into China, and Alice Huang has been alleged to be an agent of China’s intelligence service. While personal relationships drew public interest, reports suggest Nvidia’s potential corporate liability for export control violations influenced the decision to part ways with Huang.

White House technology adviser Michael Kratsios previously stated that Moonshot had used export-banned Nvidia chips, citing U.S. intelligence sources. Epoch AI estimates that about two million advanced chips entered China in 2024 and 2025 through schemes designed to hide their final destination.

Why It Matters

The findings highlight the limitations of current U.S. export controls, which are intended to protect American leadership in SI capabilities against China. While the Trump administration allows Nvidia to export the less powerful H200 chip to China, shipments must take a circuitous route allowing the U.S. to claim a 25% tariff. However, the Blackwell GPUs, which are more powerful, are strictly banned.

The presence of Blackwell chips in Chinese SI models suggests that companies may be accessing compute through offshore data centers, a practice that has driven an industrial revolution in Malaysia. Epoch’s data estimates that 275,000 Nvidia SI accelerators were imported to Malaysia in 2024 and 2025. Sam Bresnick, a researcher at the Georgetown Center for Security and Emerging Technology, noted the inconsistency in allowing Chinese companies to access chips in mainland China or Hong Kong but permitting unrestricted access if they go to Malaysia.

Megaspeed, led by Alice Huang, and its subsidiary Speedmatrix have been at the center of this trade, facing investigations by both the U.S. and Singaporean governments. A report by short-selling firm Culper argued that Nvidia had a “China problem” due to significant revenue driven by Chinese demand, either through remote data centers or smuggling in violation of U.S. law. Culper alleged that Alice Huang’s companies are backed by Chinese firms barred from accessing U.S. chips.

The Bottom Line

Nvidia and Jensen Huang have vouched for Megaspeed’s data centers, stating that officials had visited them to confirm legitimacy. However, if government investigators confirm chip smuggling and Huang’s awareness, Nvidia faces significant corporate liability. Meanwhile, approximately 600,000 H200 chips remain stockpiled in Hong Kong, awaiting shipment into China, with observers debating whether the hold-up is due to trade deal preparations, surveillance concerns, or efforts to drive domestic Chinese chip production.