ThinkFacility Sign in
  1. Home
  2. News
  3. Open weights

Nathan Lambert says Jensen Huang's exit from Nvidia is a major setback for US open models

Lambert expects Nvidia's money for open models to dry up, and his biggest worry is new state laws.

Part of a running story: Nvidia huang megaspeed briefing (5 parts)

On November 14, 2026, Nathan Lambert posted an essay on his newsletter, Interconnects, about what Jensen Huang's exit from Nvidia does to open AI models in the US. He called it Open-source AI without Jensen. His first line says the resignation is a major setback for the US development of an open model ecosystem.

Nvidia's board announced the resignation two days earlier, on November 12, and named chief financial officer Colette Kress interim CEO. Lambert had said it would be bad news before it happened.

Who
Nathan Lambert, on Interconnects
Published
Nov 14, 2026
What changed
Jensen Huang resigned as President, Chief Executive Officer, and a member of the Board
Interim CEO
Colette Kress, who stays Chief Financial Officer
His fix
a truly frontier open model player in the US

Why Lambert thinks Huang mattered

Lambert's case is about money first. He writes that Nvidia still has a reason to back open models, since a few big customers building their own chips would cut into its profits. But he expects the company's first goal now to be steadying itself and delivering to customers over the next 24 months, and he says Huang's open-model funding "will very likely dry up."

A concrete wall with a green Nvidia logo and the number 2788, with a glass building under a wave-shaped silver roof behind it
Nvidia's Endeavor building in Santa Clara, California. The board named Colette Kress interim CEO on November 12, 2026. Photo: Daniel J. Prostak; Crocodiletiger~commonswiki Crocodiletiger~commonswiki used courtesy of Daniel Prostak, CC BY-SA 4.0, via Wikimedia Commons

The second part is protection. Lambert says Huang was more than a cheerleader and was willing to use his power on open models when trouble started. That's the part he thinks is hardest to replace.

Jensen’s support, and protection, gave the ecosystem time to figure this out.

Open-source AI without Jensen - by Nathan Lambert

He'd made the same call two days before the board's announcement, in one line on X, at 7:12 p.m. Eastern on November 10.

Nathan Lambert@natolambert

Jensen no longer being CEO would be very very bad for the open model cause in the US.

View the post on X

What he's worried about now

Lawmakers come first. Lambert points to a big rise in Democratic funding for candidates who want to regulate the AI industry, and says many of them think open models are unsafe without knowing much about what they're good for. (We covered one side of that money after the midterms, when Public First said the candidates it backed had won.)

My biggest near-term worry is sloppy regulation at the state level, followed by national regulation.

From Open-source AI without Jensen

He also expects a crisis. No major hack has yet been driven by a popular Chinese open-weight model, he writes, but "it is a matter of when, not if." When it does, he wants someone to urge caution as the people he calls open-weight hawks come hunting.

Money alone won't fix it, in his view. He quotes Dylan Patel, CEO of SemiAnalysis, from a ChinaTalk podcast, saying the problem is talent: the strongest researchers mostly work at four big labs and the rest are spread across a hundred startups. Lambert says he agrees completely.

He's hopeful about one US lab. He calls Thinking Machines' Inkling 2 Small the best Western open-weight model in a while, but still far behind Moonshot's Kimi K3.1 Pro. Thinking Machines said it would put the full weights out under an Apache 2.0 license after outside safety testing, which we reported when it launched.

For a new champion, Lambert looks to the hyperscalers, and Zuckerberg is the first name he gives. He doesn't say any of them has offered.

What this means outside the AI industry

An open model is one whose weights anyone can download, so a company can run it on its own machines and tune it to its own work. Lambert's upside case rests on that: he says the open model economy has latched onto ownership of intelligence and enterprise post-training. His downside is cost. Inference startups can't get enough chips because Anthropic and OpenAI are buying so much capacity, he writes, so open models aren't yet the cheap commodity they could be.

The state rules he fears are already arriving. New York's RAISE Act requires large frontier AI developers to register with the state by January 1, 2027, and to report any critical safety incident within 72 hours, according to Governor Hochul's office. The governor's page doesn't say how the law treats a lab that publishes its weights, and Lambert doesn't name New York.

Nobody at Nvidia has said its open-model spending will change. Its November 12 release doesn't mention open models at all, and its board chair, Harvey Jones, says he expects the product roadmap to continue without interruption. Lambert's prediction is his own reading, and I'd treat it that way until Nvidia's spending shows up in its numbers.

Nvidia reports its third-quarter fiscal 2027 results on Tuesday, November 17, at about 1:20 p.m. Pacific, with the earnings call at 2 p.m.

More on Open weights

All Open weights stories