CNBC Television Sep 17, 2026
With Steve Eisman, host of The Real Eisman Playbook Podcast and former Senior Portfolio Manager at Neuberger Berman
Steve Eisman was asked about the run of AI safety incidents companies have been disclosing, and answered in six words: "I think this is all nonsense."
His reading is not that the warnings are exaggerated. It is that they are strategic. He said the labs have discovered they have no durable advantage, and that a safety panic is the route to regulation they can shape.
"They realize that there are no moats around their business whatsoever, and they're trying to manufacture a crisis that will create regulation"
Eisman made his name betting against subprime mortgage bonds and later ran money as a senior portfolio manager at Neuberger Berman. He now hosts The Real Eisman Playbook Podcast, and told the program he has been trimming rather than adding to his AI exposure.
The full segment is covered here so you can skip it.
Here are the 5 arguments that matter.
Key Takeaways
Eisman sees no evidence AI has reached general intelligence, and thinks the evidence points to it never arriving
The safety disclosures are, in his reading, a bid for regulation the labs can shape into a duopoly
He says open-weight models are taking big market share, which is what removed the moat
The whole chain, from chips to hyperscalers, rests on Anthropic and OpenAI staying healthy — and he calls OpenAI the weaker of the two
He floated Nvidia's purchase of Hugging Face as an insurance policy protecting the ecosystem from a lawsuit
1. The Terminator Talk
A host put it to Eisman that companies keep announcing their AI systems are behaving in ways they did not anticipate, and that the spectacle of a firm saying it does not know what it built is itself unusual. Eisman rejected the premise outright.
He said there is no evidence any system has reached artificial general intelligence, and that most of what he sees points to it not happening — or not for many years. The doomsday framing, in his account, is doing a different job than the one it advertises.
"I think this is all nonsense." — Steve Eisman
"I think that there's something else completely going on here." — Steve Eisman
2. Manufacturing a Crisis
The alternative explanation Eisman offered is competitive rather than technical. He said the economics of selling ever-larger amounts of model output have stopped working, and that freely available open-weight models are taking meaningful share. A company with no moat, on his reading, has an incentive to invite rules that only it can afford to meet.
"And what I think is happening is that token maxing is over. The open weight models are taking big market share." — Steve Eisman
"They realize that there are no moats around their business whatsoever, and they're trying to manufacture a crisis that will create regulation" — Steve Eisman
One of the hosts noted the argument puts Eisman on the same side as David Sacks, and joked that the two of them should have dinner over a good bottle of wine.
3. The Chain Rests on Two
Asked whether he accepts the list of bubble markers commentators are applying to AI, Eisman said his view there is more qualified. His actual concentration worry is structural: he traces the entire AI supply chain back to two private companies.
"But I think the issue is that at the end of the day, the entire AI chain, from Nvidia to the hyperscalers to Anthropic and OpenAI all depends on the future health of Anthropic and OpenAI, because there's such a huge percentage of the entire chain." — Steve Eisman
"And between the two, OpenAI is the weaker company. I think that's one reason why they postponed their IPO." — Steve Eisman
He noted he had made the same argument on his previous appearance and had not moved off it.
4. The Hugging Face Deal
A host raised the reported incidents in which OpenAI and Anthropic agents got into systems they should not have, and asked why nobody has been sued, suggesting liability for what an agent does might be the lever regulators actually need. Eisman said the liability already exists and no new law is required to create it.
"I'm sure they're liable now. I mean, I don't think that you need to pass laws to make them more liable." — Steve Eisman
He then offered what he labeled the conspiratorial reading of Nvidia's acquisition of Hugging Face: that part of what Nvidia bought was the removal of a potential plaintiff. He added that the damages from such a suit might have been close to nothing in financial terms, and that other bidders were in the process, so a higher Nvidia price could have been paying for protection of the wider ecosystem rather than for the asset.
Against that, a host pointed out that Hugging Face's own chief executive had said on the same program that he went to Jensen Huang because he was worried and wanted a buyer with deeper pockets — a version of events in which the motive sits on the seller's side.
5. Could Both Be True?
The last question tested the middle ground: the labs have been publicly worried about safety for the better part of a decade, so it is possible the concern is sincere and regulatory capture is merely the byproduct, in which case the motive may not matter because the outcome is the same either way.
Eisman did not claim to know the motive. He answered with a test instead.
"My response is really postpone your IPO" — Steve Eisman
Bonus Insights
He confirmed he has taken some of his AI exposure off the table since his last appearance, and has not been adding
He drew a distinction between how he talks about markets generally and how he talks about this: on most subjects a host described him as measured, and here he was not
On the bubble-checklist genre of article, he said his own view is more nuanced than a yes or no, and redirected to the concentration question instead
Eisman's bottom line is that the risk worth watching in AI is not what the models do but what happens to the two companies the rest of the chain is priced against.
Products, Companies & Tools Mentioned
OpenAI and Anthropic (The two companies he says the whole chain depends on; he calls OpenAI the weaker, and points to its postponed IPO)
Nvidia (Top of the chain, and the buyer whose price he suggests included an insurance element)
Hugging Face (The acquisition target, and the party that might otherwise have sued over agent intrusions)
Neuberger Berman (Where Eisman was a senior portfolio manager before moving to podcasting)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: