New Research Claims China's Most Powerful AI Models Were Trained on Export-Banned Nvidia Chips

type

Post

status

Published

date

Nov 14, 2026

slug

summary

New Research Claims China's Most Powerful AI Models Were Trained on Export-Banned Nvidia Chips

Some of the most capable open-weight AI models coming out of China appear to have been trained on Nvidia chips that the U.S. government has restricted from reaching its largest geopolitical rival. That is the central claim of a new research effort, and it lands at a moment when the chipmaker's China business is under extraordinary scrutiny.

Blackwell fingerprints in open weights

Will DePue, a former OpenAI researcher who leads the nonprofit research group Unemployment Labs, analyzed the weights — the unique learned structure — of large language models that DeepSeek, Moonshot AI, MiniMax, and Ant Group have released openly. Weights still reflect the configurations of the chips used to train a model, so they can serve as a kind of forensic record of the hardware behind it.

On that basis, DePue found what he describes as compelling evidence that the most closely watched models — including Moonshot's Kimi K3 releases and DeepSeek's V4 models — carry the fingerprints of training on Nvidia's Blackwell chips.

The export rules make that finding awkward. Blackwell GPUs are banned from export to China or to Chinese-owned firms. The Trump administration does allow Nvidia to ship the less powerful H200 to China, but only along a circuitous shipping route that lets the U.S. claim a 25% tariff.

A pattern, not a one-off

The new analysis is not the first indication of a problem. Earlier this year, White House technology adviser Michael Kratsios said, citing U.S. intelligence sources, that Moonshot had used export-banned Nvidia chips. Epoch AI, a nonprofit research organization, estimates that about 2 million advanced chips entered China in 2024 and 2025 through schemes designed to hide their final destination.

There is an important caveat: this does not mean the models were trained entirely on Nvidia chips, or that export-control laws were necessarily violated. The chips could have been located in an offshore data center — a route that has proven popular among Chinese firms seeking compute, and one that has driven an industrial revolution in Malaysia. Epoch's data estimates that 275,000 Nvidia AI accelerators were imported to Malaysia in 2024 and 2025.

But that workaround cuts against the ostensible goal of the export bans, which is precisely to stop Chinese AI labs from 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.

A CEO exit with export-control stakes

Nvidia CEO Jensen Huang left the company this week after admitting to an affair with Alice Huang, an executive of 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. While the CEO's relationships with a customer and two employees drew the most public interest, Nvidia's potential corporate liability for export-control violations likely factored into its decision to part ways with him.

U.S. export controls are intended to protect American leadership in AI capabilities against China, yet Huang had long sought to open up the Chinese market for Nvidia.

The Megaspeed connection

Megaspeed, the company led by Alice Huang, has sat at the center of this trade for several years through its subsidiary Speedmatrix, and has been the subject of investigations by both the U.S. and Singaporean governments. On Friday, Bloomberg reported that a company with links to Alice Huang was involved in smuggling banned Nvidia chips to a data center in Shanghai.

In May, the short-selling research firm Culper published a report arguing that Nvidia had a "China problem," because a significant share of its revenue was driven by Chinese demand — either through remote data centers or through smuggling in violation of U.S. law. The report detailed Megaspeed's relationship with Nvidia and alleged that Alice Huang's companies are backed by Chinese firms that are barred from accessing U.S. chips.

The bigger bet: China builds its own chips

Ironically, Culper's wager against Nvidia was not built on the expectation of a U.S. government crackdown. Instead, it focused on the Chinese government's response to the bans: a renewed commitment to its domestic chip industry. If China succeeds, billions of dollars in Chinese demand for Nvidia chips could dry up.

Meanwhile, some 600,000 H200 chips are reportedly stockpiled in Hong Kong, awaiting shipment into China. Chinese companies might want the chips, but observers have no consensus on why they haven't been cleared for import. Graham Webster, a research scholar at Stanford who focuses on U.S.-China technology policy, said he doesn't buy the idea that holding the chips back would drive domestic production, since demand for chips is so large.

Whether or not the weights analysis holds up, the underlying dilemma remains: as long as compute can flow through offshore data centers and intermediaries, controls on who owns chips will keep colliding with the question of who actually uses them — while China's own push to replace Nvidia only raises the stakes.

Source: TechCrunch, https://techcrunch.com/2026/11/13/export-banned-nvidia-chips-are-behind-chinas-most-powerful-ai-models-uh-oh-jensen/