OpenAI has ruled out an initial public offering in 2026, and Alex Kantrowitz says Anthropic can still go out this year regardless.
The host's premise was that a company sitting at the center of the safety, regulation and politics argument might not be able to list at all. Kantrowitz's case is that Anthropic's business does not depend on the next model, and Lo Toney's is that the company's chief executive was preparing to list and raising the safety question at the same time, deliberately.
"But even if you do that, the current technology, I believe, is strong enough that you can make a business case to Wall Street and still go out this year."
Kantrowitz founded Big Technology and is a CNBC contributor whom the host credited with following this listing race throughout; Toney runs Plexo Capital, which invests in venture funds and directly in private companies.
The full segment is covered here so you can skip it.
Here are the 6 arguments that matter.
👤 Guests: Alex Kantrowitz, Founder of Big Technology and a CNBC contributor, and Lo Toney, Founding Managing Partner of Plexo Capital
🎙️ Host: Leslie Picker, who covers capital markets for CNBC and was anchoring Closing Bell
👥 Also on: Megan Cassella, who reports on Washington and the White House for CNBC
📰 Published: 14 September 2026 on CNBC's Closing Bell
🟣 Apple Podcasts | 🔗 Episode page | ⏱️ 7 min
Key Takeaways
Kantrowitz separates the safety problem from the business: the hack that spooked the market was not Anthropic's technology
He names OpenAI's hack of Hugging Face as the most concerning one
A slower model release schedule can hurt margins and still leave the business listable
Smarter models on the same underlying technology earn more, so pausing costs margin rather than viability
Toney reads Anthropic's chief executive as running the listing preparation and the safety campaign in parallel, not choosing between them
The reason the IPO stays urgent is inference demand, not the training roadmap
Kantrowitz says customers already single Anthropic out as short of compute, and the alternative is that they move to OpenAI
President Trump met Sam Altman privately at the Republican National Convention, at Altman's request
The same week Trump called the safety concerns a hoax and said a strong and smart president is the only guardrail needed
Toney's frame is a prisoner's dilemma that has moved up a level, from company against company to the US against China
1. Can Anthropic Go Public
Leslie Picker opened on the race to list and put the environment question to Kantrowitz directly, telling him "You have been all over this horse race this saga," and asking whether it comes down to the market holding up.
Kantrowitz's answer was that it largely does — "I think it is," he said — and then he split the two labs apart. "And OpenAI has said, you know, we're not going to go public this year. So Sam Altman, over the past couple of days has said, we're not going to IPO in 2026. Maybe that'll happen in 2027."
"But I think Anthropic still can go public," he said
His reason for separating the safety story from the company is a specific one: "I mean, obviously there are some concerns about what the technology can do, but its technology is not behind the most concerning hack, which is OpenAI's hack of Hugging Face."
The second reason is that the existing product already sells: "And more than that, even if you took the current models, Anthropic still has a very sizable business."
He left the pace of any slowdown open rather than assuming it: "So you can slow down the release of the upcoming models. And there's a question of how slow that's going to be."
The conclusion he drew is a listing this year: "But even if you do that, the current technology, I believe, is strong enough that you can make a business case to Wall Street and still go out this year."
2. What a Pause Costs
Picker's follow-up was about the cost side, and it is the sharpest question in the segment: the compute is already contracted, so a slower release schedule changes revenue without changing expense.
She framed it as a fixed-cost problem: "They've already signed up for billions of dollars worth of compute. So their expenses are kind of known. If they slow down the pace of model releases, what does that mean for margins, if anything?"
Kantrowitz conceded the direction — "Well, it could hurt the margins, right?" — and gave the mechanism: a smarter model running on the same underlying technology earns more per unit of compute, so holding the model back gives that up
He did not treat it as disqualifying: "But I still think they'll be able to do pretty healthy margins with the technology that they have today. That wouldn't be a major concern to me."
3. Both Tracks at Once
Picker brought Toney in and asked whether this is an inflection point or part of the journey these listings have been on, with the window arguably open regardless of the noise.
Toney's reading is that the two things were run in parallel on purpose: "Yeah, it's a great point, Leslie, because when we look at the actions that Dario was taking, he was clearly planning on going through the IPO process to prepare, while at the same time raising these issues with his peers in the space to determine how to best address the issues."
He set that against the usual sequence: "And so, you know, I think this is what we're going to see that's going to be new because as we know, typically technology outpaces the ability for regulation."
The precedent he used: "In the past, we've seen regulation look through the rearview mirror and things like social media."
What he says makes this different is the security framing: "But now we're talking about AI, which as President Trump has made very clear, this is more of a national security interest."
His call is a listing with the timing left open: "So I think ultimately we will see Anthropic go public. Maybe it will be this year, or maybe it gets pushed out to next year."
4. Why the IPO Still Presses
Picker's next question was the most structural one of the segment, and she laid out the financing chain herself before asking it. A listing is not only equity, she said: "You're raising equity capital. It potentially unleashes access to investment grade credit ratings, which is something that they've been reportedly lobbying or at least their bankers have on their behalf in order to access even more and deeper markets at cheaper cost of capital." If development slows, she asked, are those liquidity events as pressing, or can the company prioritize security and let the financing happen when it happens?
Kantrowitz said the urgency does not move: "Yeah, I think the liquidity events are still extremely pressing, and that's because there's so much demand."
The demand he means is for running models, not building them: "And so most of AI is moving towards inference or using the models, not training the models."
He allowed that the labs want both: "Now, of course, if you're a company like Anthropic or a company like OpenAI, you want to do both, you want to train and you want to provide for inference."
The competitive consequence is where he ended: "But Anthropic in particular has been singled out by those using this technology as a company that needs more compute"
"Definitely have to continue to pursue this liquidity to be able to meet the market. Or what's going to happen is people are going to go over to OpenAI," he said
5. Trump Met Altman
The segment broke for news from Washington. Megan Cassella reported that "We just learned that President Trump met privately last week at the Republican National Convention with Sam Altman of OpenAI."
She said the meeting happened backstage at the conference in Texas, and that the discussion "was centered on AI and its growing power, and that Altman was the one who requested the meeting with the president"
The contrast Cassella drew is between the two men in the same week. On Altman: "On AI, Altman joining other AI executives this weekend and saying there is a need to slow down and to pace the frontier."
On the president: "But everything we've heard from President Trump since then has been very different, really dismissing all of these concerns. He called them a hoax."
She reported his position that day in detail: "Today, he says the only regulation or the only guardrails that the U.S. needs is a strong and smart president, referring to himself also saying it's just a sick conspiracy going on against AI and data centers, dismissing the idea or dismissing the fact that these concerns are coming from the AI leading researchers and executives themselves."
Her closing line put the meeting against all of that: "But now we know that President Trump and the OpenAI chief met privately just last week."
6. A Pace Versus Race
Picker turned back to Toney on the politics, telling him that "you'd be kind of hard pressed to think of another industry or another scenario where the companies themselves that are calling for additional regulation, that are calling for their own growth to be slowed while the government is saying, no, we have to win this race, keep going, keep pressing forward."
Toney's answer is a game-theory one: "Well, it's so fascinating to watch this play out in real time, because what we're seeing is a prisoner's dilemma, not only within the companies within the United States."
He compressed the trade-off into a phrase: "And this is really comes down to a pace versus race."
His view is that the company-level coordination is arriving: "And so we need the companies to come together so that everyone can agree. And that looks like it's happening within the company within the United States."
The harder version is between governments, not firms: "Now we have this separate issue around a prisoner's dilemma, contextualized within the framework of AI and U.S. leadership being important for national security. So now the prisoner's dilemma shifts to a focus between the U.S. and China."
He ended on whether China can be brought into that coordination, which is the question he left the segment with rather than an answer
Bonus Insights
Picker's own framing was that the listing question is the live one, not the technology question: "The question about IPO, can a company go public in this environment, I think, is a really big question right now," she said before Kantrowitz answered anything
Neither guest treated the safety debate as an obstacle to listing. Kantrowitz argued the technology in question is not Anthropic's, and Toney argued that raising the issue was itself part of the preparation
Cassella said only a brief overview of the discussion was available, which is the whole of what was reported about a private conversation between the president and OpenAI's chief executive
Picker used the word "necessity" for the listing before Kantrowitz agreed with it, so the premise that a company at this scale must go public was the show's as much as the guests'
Kantrowitz's bottom line is that Anthropic's listing case rests on demand for the models it already has rather than the ones it might delay, and that the risk of waiting is customers moving to OpenAI; Toney's is that the company was always running the safety campaign and the IPO preparation together, and that the unresolved coordination problem is now between Washington and Beijing rather than between the labs.
Products, Companies & Tools Mentioned
Anthropic (The subject of the segment — both guests expect it to list, Kantrowitz this year and Toney this year or next, and Kantrowitz says users single it out as the lab that needs more compute)
OpenAI (Has ruled out a 2026 listing; Kantrowitz names its hack of Hugging Face as the most concerning one and says its capacity is where Anthropic's customers would go)
Hugging Face (The target of the hack Kantrowitz says the market is reacting to, and his reason for separating the safety story from Anthropic's business)
Big Technology (Kantrowitz's publication, and the credential the show introduced him with)
Plexo Capital (Toney's firm, which invests in venture funds and directly in private companies)
If this was worth your time, send it to someone who follows the name.
Get the latest market chatter as it happens: