Nvidia CEO Jensen Huang used his first X post to send a thinly-veiled message to OpenAI and Anthropic: Chinese artificial-intelligence models are here to stay, and you have to compete with them.
As the Trump administration reportedly weighs a crackdown on low-cost, "open-weight" models from Chinese labs, some in the AI industry have defended their overseas counterparts. That was the subtext of an open letter Huang published Friday, joined by signatories such as Palantir Technologies, Microsoft, Meta Platforms, and Dell Technologies.
"Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector," Huang wrote, seemingly addressing U.S. policymakers.
"Open-weight models -- AI models that anyone can download, inspect, modify, and run on their own infrastructure -- are an important part of that foundation because they make advanced AI more accessible, adaptable, and widely available," he added.
The industry erupted into a debate over competitive fairness last week after Chinese lab Moonshot AI released a model called Kimi-K3 that stacked up well against U.S. rivals. Some analysts worried about whether the massive buildout to support Anthropic and OpenAI was justified if open-weight alternatives could push down token prices and squeeze profits.
Neither OpenAI nor Anthropic replied to Barron's requests for comment.
Much of the debate has centered on "distillation," which refers to labs using outputs from competitors to essentially reverse-engineer parts of their own models.
Michael Kratsios, director of the White House Office of Science and Technology Policy, said Wednesday that he had evidence Moonshot distilled Anthropic's Fable model. Treasury Secretary Scott Bessent followed that up with a warning that the government could put sanctions on Chinese firms that "cross the line into [intellectual property] theft."
Moonshot didn't respond to Barron's request for comment.
There isn't a consensus within Silicon Valley over how much distillation went into Kimi-K3, whether it is illegal or unethical, or even what distillation actually is.
"Policymakers should be careful not to conflate legitimate model development techniques with misappropriation," Huang argued, adding that distillation "reflects a long tradition of learning from, building upon, and improving existing technologies."
He urged policymakers to avoid sweeping restrictions and instead apply targeted measures to combat "unlawful efforts to extract value from closed models."
Huang also addressed safety concerns posed by open-weight models, noting that, once released, the models can be modified without the developers' control. At the same time, he said, enterprises need these models to combat potential cyberattacks.
As the CEO of the country's most valuable company by market capitalization, Huang's comments will be welcomed by those in the tech industry who have speculated that OpenAI and Anthropic want to shut out open-weight competitors.
"Americans will not benefit by being forced to buy Anthropic, at whatever price they charge, while the rest of the world hosts open source models," Parker Conrad, CEO of payroll software company Rippling, wrote in a social media post earlier this week.
On the other hand, the likely state-involvement in China's labs makes some in the policy space queasy. In an appearance on the MTS podcast, Peter Wildeford, head of policy at the AI Policy Network, called the distillation part of a "geopolitical strategy" to weaken U.S. labs.
Huang understandably fell somewhere in the middle of those two poles. Nvidia has an incentive to broaden access to AI and widen the demand for its chips, but it is also a supplier of OpenAI and Anthropic.
That said, it is notable that OpenAI and Anthropic didn't appear as signatories on Huang's letter. When it comes to the debate over open-weight models, Nvidia and two of its largest users may not see eye to eye.
Write to Nate Wolf at nate.wolf@barrons.com
This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
July 24, 2026 13:20 ET (17:20 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments