Research ties China’s top open-weight AI models to export-banned Nvidia chips
New analysis attributes training fingerprints on models from DeepSeek, Moonshot AI, MiniMax, and Ant Group to Nvidia Blackwell GPUs barred from export to China, amid scrutiny over Nvidia’s CEO departure and alleged chip-smuggling networks.
By The Daily Circuit AI Desk (automated, AI-generated)
4 min read
AI-generated image (flux-1-schnell), not a photograph of the events described.
Key facts
- China’s leading open-weight AI models show evidence of training on export-banned Nvidia Blackwell chips[1]
- Will DePue of Unemployment Labs analyzed models from DeepSeek, Moonshot AI, MiniMax, and Ant Group[1]
- Epoch AI estimates roughly two million advanced chips entered China in 2024 and 2025 via concealment schemes[1]
- Epoch AI data estimates 275,000 Nvidia AI accelerators were imported to Malaysia in 2024 and 2025[1]
- Jensen Huang left Nvidia after acknowledging an affair with Alice Huang of Megaspeed[1]
- Megaspeed and its subsidiary Speedmatrix have been investigated by the U.S. and Singapore[1]
- Bloomberg reported a company linked to Alice Huang was involved in smuggling banned Nvidia chips to a Shanghai data center[1]
- Culper’s May report argued Nvidia had a “China problem” tied to Chinese demand through remote data centers or smuggling[1]
A new research effort has found that China’s most advanced open-weight AI models were trained on Nvidia chips that the United States bars from export to China, according to a report by TechCrunch. [1]
Will DePue, a former OpenAI researcher who leads the nonprofit Unemployment Labs, examined the weights of large language models released openly by DeepSeek, Moonshot AI, MiniMax, and Ant Group. Because model weights retain traces of the hardware used to train them, DePue found compelling evidence that Moonshot’s Kimi K3 releases and DeepSeek’s V4 models carry fingerprints of training on Nvidia’s Blackwell chips. [1]
Blackwell GPUs are prohibited from export to China or Chinese-owned firms. The U.S. government does permit Nvidia to ship the less powerful H200 chip, but only through a circuitous routing arrangement that allows a 25% tariff to be claimed. [1]
The findings add to earlier public statements by White House technology adviser Michael Kratsios, who said earlier this year that Moonshot had used export-banned Nvidia chips, citing U.S. intelligence sources. [1]
The research does not establish that these models were trained entirely on Nvidia chips, nor that export control laws were broken. The chips could have been used in offshore data centers, a route Chinese firms have used to access compute power. Epoch AI, a nonprofit research group, estimates that roughly two million advanced chips entered China in 2024 and 2025 through schemes intended to conceal their final destination. The same data set puts Nvidia AI accelerator imports into Malaysia at 275,000 units over the two-year period. [1]
The report also connects the chip issue to leadership changes at Nvidia. CEO Jensen Huang left the company this week after acknowledging an affair with Alice Huang, an executive at Megaspeed, a company implicated in smuggling Nvidia chips into China. Alice Huang has also been alleged to be an agent of China’s intelligence service. The TechCrunch report says Nvidia’s potential corporate liability for export control violations likely played a role in Huang’s departure. [1]
Megaspeed, which operates through a subsidiary called Speedmatrix, has been investigated by both the U.S. and Singaporean governments. Bloomberg reported on Friday that a company linked to Alice Huang was involved in smuggling banned Nvidia chips to a data center in Shanghai. [1]
In May, short-selling research firm Culper published a report arguing that Nvidia faced a “China problem” because a significant portion of its revenue came from 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. Culper’s short position was based not on an expected U.S. crackdown but on China’s recommitment to its domestic chip industry, which could eventually erase billions of dollars in Chinese demand for Nvidia products. [1]
Some 600,000 H200 chips are reportedly stockpiled in Hong Kong awaiting shipment into China, though observers say there is no consensus on why the import has not been cleared. Graham Webster, a Stanford research scholar focused on U.S.-China technology policy, said he does not believe restricting chip access would spur domestic production, given the scale of demand. Other explanations include preparation for a possible trade deal involving large U.S. goods imports, or concerns about surveillance technology embedded in the chips. [1]
U.S. companies currently face limited obligations to monitor what customers do with controlled goods such as AI chips. Nvidia and Jensen Huang have both vouched for Megaspeed’s data centers, stating that Nvidia officials visited the sites to confirm they were legitimate. [1]
Separately, Commerce Secretary Howard Lutnick, who oversees the agencies that license chip exports, could reportedly lose his job after reports that his investigators failed to identify suspicious chip buyers and were too close to Nvidia officials. [1]
The article notes that China hawks have long urged controls on who uses chips, not merely on ownership or installation location. [1]
Context
The development matters for Indian technology and financial markets because Nvidia is a major global AI hardware supplier, and any disruption to its China revenue or exposure to export-control enforcement could affect sentiment around semiconductor-linked equities tracked on Indian exchanges. No Indian market impact is stated in the source.
Why it matters
The story illustrates how export controls on advanced semiconductors are being tested by reported chip diversions through third countries and offshore data centers, raising questions about the effectiveness of U.S. restrictions aimed at slowing China’s AI progress.
What could happen next
AI-generated analysis based on the sources below. Not a forecast or recommendation.
If U.S. investigators confirm chip smuggling and senior awareness at Nvidia, the company could face greater corporate liability. China’s push for domestic chips could reduce future demand for Nvidia products, though the sources note observers have no consensus on why stockpiled H200 chips have not been cleared for import into China.
Risks and open questions
The article relies on a single publisher source and includes allegations—such as intelligence-agency ties and smuggling—that are reported or claimed rather than independently confirmed in the source. Readers should treat those claims as attributed, not established fact.
Sources
- 1.TechCrunch (primary source): Export-banned Nvidia chips are behind China’s most powerful AI models — uh-oh, Jensen!, 14 November 2026
How this article was produced. This story was researched and written by an automated system using the open model nvidia/nemotron-3.5-super-vl (via OpenRouter), working only from the sources listed above. Automated checks verified that every citation, quote and figure matches those sources before publication. It was not written by a human reporter. Editorial policy · Report an error
- AI
- China
- NVIDIA
- Semiconductors
- Jensen Huang
- export controls
Related stories
Amazon Plans World Product Graph Initiative to Index Global Online Shopping Data
Amazon is reportedly developing a project called the World Product Graph to index every product sold across the internet to improve its platform and AI assistants.
2 min read
Jensen Huang Resigns as Nvidia CEO Amid Investigation as Colette Kress Takes Over as Interim Chief
Nvidia co-founder Jensen Huang has resigned as CEO and board member following an internal board investigation into extramarital affairs and an alleged Chinese spy, with CFO Colette Kress stepping in as interim CEO.
2 min read
Google launches Gemini Decisions API for fast, calibrated multimodal choices
Google AI unveiled the Gemini Decisions API, enabling up to 1,000 multimodal yes/no, multiple‑choice and rubric‑scoring queries per request with sub‑second latency and a cost‑effective pricing model.
3 min read
Warhammer 40,000: Dawn of War IV Headed to GeForce NOW Alongside DLSS 5 Updates
NVIDIA announces the arrival of Warhammer 40,000: Dawn of War IV to GeForce NOW, alongside upcoming DLSS 5 updates for War Thunder and Where Winds Meet.
2 min read