Cramer Hyped Anthropic IPO, Then Its Own Researcher Cited 10% AI Risk

Jim Cramer spent Tuesday night on CNBC’s Mad Money talking up Anthropic as a company that “could be a gigantic IPO,” the kind of listing that pulls money out of other stocks just so investors can get a piece. By Wednesday morning, September 9, 2026, he was on Squawk on the Street comparing the company’s own safety math to surgery odds he wouldn’t accept. The reason: an Anthropic researcher who leads the company’s alignment science work had just told the public, in writing, that he sees more than a 10% chance AI could “kill all humans” within the next decade.

The whiplash captures a tension that has been building all year as Anthropic pushes toward what shareholders reportedly expect will be a listing valued above $2 trillion. The company that built its brand on taking AI risk more seriously than its rivals is now the one supplying the number that makes a Wall Street hype man pause. Here is what happened, who said what, and why it matters for anyone watching the AI trade, the IPO calendar, or the underlying safety debate.

What Cramer Actually Said About the Anthropic IPO

Cramer has been talking up an Anthropic IPO for months. In April 2026, he floated the idea that the company could eventually be worth $1 trillion. By July 15, he was cautioning on timing rather than size, writing that “this market does not have room at this moment for Anthropic,” and that “they should wait until the fall,” adding that the market needed “a lot of this sloshing stock put away” first, according to a post tracked by 24/7 Wall St.

By early September, the tone had shifted from caution to urgency. On Squawk on the Street around September 3-4, Cramer connected Anthropic directly to Broadcom’s AI chip business, telling viewers “this is now Anthropic. It’s Anthropic or bust… they need Anthropic to have a blockbuster IPO,” a line reported by both TradingTips and Ground News. Then, on Tuesday night, September 8, Cramer told Mad Money viewers Anthropic “could be a gigantic IPO and investors will sell all sorts of other stocks in order to raise money to participate,” according to 24/7 Wall St.

Less than 12 hours later, he was walking that enthusiasm back. Anthropic’s IPO has not actually been filed. The company remains private with no S-1 on record at the SEC, 24/7 Wall St noted, meaning every valuation figure circulating in September is still speculation rather than a scheduled event. That did not stop the Financial Times from reporting that Anthropic shareholders expect the company to eventually list at more than $2 trillion, a figure Cramer defended on CNBC hours after it published, according to Yahoo Finance.

Then a Safety Researcher Dropped a Number Cramer Didn’t Expect

The pivot point came from inside Anthropic itself. Evan Hubinger, who heads alignment science at the company, posted on X that he personally believes there is a greater than 10% chance AI kills everyone within the next decade. CNBC quoted the post directly: “Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” according to CNBC.

That is not a hedge. It is a named executive at a company chasing a trillion-dollar-plus valuation putting a specific probability on human extinction, in public, tied to his own employer’s technology. TechRound described the exchange as “remarkably candid,” noting Hubinger’s post came in direct response to another departing employee rather than in a controlled interview or a company blog post.

The Resignation That Set the Stage: Jacob Coxon’s Exit

Hubinger’s post did not appear out of nowhere. Hours earlier, Jacob Coxon, who has worked as a researcher at both Anthropic and OpenAI, announced on X that he was resigning because he believes AI labs are “gambling with our lives.” Coxon wrote that “the people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear,” according to CNBC’s reporting on the exchange.

Coxon’s framing matters because it reframes the entire public conversation around AI safety messaging. His claim is essentially that the calm, hedged language AI executives use in interviews and shareholder letters understates what researchers privately believe. Hubinger’s reply, agreeing with Coxon and attaching a specific number, turned that claim from an anonymous insinuation into an on-the-record data point from a named alignment lead.

What Is p(doom) and Why It Suddenly Matters to Investors

Inside AI safety circles, researchers have long used a shorthand for exactly this kind of estimate: p(doom), or the probability of a catastrophic outcome from advanced AI. CNBC explained the term to its broader readership as shorthand “to estimate the probability of dire outcomes that could stem from AI.” What makes this week’s episode different is that p(doom) has typically stayed inside safety research circles, workshops, and the occasional podcast. Hubinger’s post pulled it directly into financial media, attached to a company actively courting IPO investors.

Cramer’s own reaction leaned on a medical analogy rather than an AI one. On Squawk on the Street, he said: “If you ever been in a situation where you want a surgery and the surgery electrosurgery is more than 5%, 5% fatality. No, you don’t do it… 10% is twice 5%. So when I read that, I said, well, you know, I don’t like those odds. Those odds are just plain bad.” He added that he was “very surprised that this man still works for Dario, for Anthropic,” a reference to Anthropic CEO Dario Amodei, according to 24/7 Wall St’s account of the segment.

Anthropic’s IPO Math: From a $965 Billion Round to a $2 Trillion Ask

The p(doom) headline lands at an especially sensitive moment for Anthropic’s fundraising trajectory. Tech Insider previously reported on Anthropic’s $65 billion Series H round, which valued the company at $965 billion, and on earlier reporting that tied a prospective IPO to a model release cycle. The Financial Times’ newer figure, more than $2 trillion, would represent a significant jump even from the Series H level, and it is the number Cramer chose to defend publicly before the safety story broke.

None of this is finalized. Anthropic has not filed an S-1, has not set a listing date, and has not confirmed a target valuation through its own channels, based on 24/7 Wall St’s reporting. Every number in circulation, from $1 trillion to more than $2 trillion, comes from investor expectations and media reporting rather than a company disclosure. That gap between speculation and filing is itself part of the story: a safety controversy landing before any formal roadshow gives Anthropic more room to manage the narrative than it would have mid-process.

Broadcom and the Public Companies Riding on Anthropic’s Success

Because Anthropic itself is not publicly traded, investors who want exposure to its trajectory have to buy it secondhand, through suppliers and partners that already trade on public markets. Broadcom is the name Cramer has leaned on hardest. On the same September 3-4 Squawk on the Street segment where he called it “Anthropic or bust,” he tied Broadcom’s custom AI silicon guidance directly to Anthropic’s fortunes. The dollar figure attached to that bet varies by outlet: TradingTips headlined it as a $230 billion AI opportunity, while Ground News framed the same commentary around a $30 billion figure, a discrepancy that likely reflects different scopes (annual order value versus a multi-year cumulative estimate) rather than a factual dispute.

Anthropic’s cap table already includes several public or soon-to-be-public technology giants with direct financial stakes. Tech Insider has separately reported on Anthropic’s fundraising history, which included large strategic investments from cloud providers securing compute commitments in exchange for capital. The table below summarizes the publicly reported financial relationships that give outside investors indirect exposure to how the Cramer-Hubinger story plays out.

Company	Nature of Anthropic Exposure	Reported Figure	Source
Broadcom	Custom AI silicon / chip supply tied to Anthropic’s compute buildout	$30B–$230B (range varies by outlet)	TradingTips, Ground News
Amazon (AWS)	Cloud compute and Trainium chip investment	Prior investment reported at $5B, tied to a broader $100B AWS pact	Tech Insider prior coverage
Google	TPU compute investment	Prior investment reported at $40B	Tech Insider prior coverage
Anthropic (private)	Series H valuation baseline	$65B raise at a $965B valuation	Tech Insider prior coverage
Anthropic (IPO target, unconfirmed)	Reported shareholder expectation	More than $2 trillion	Financial Times, via Yahoo Finance

The takeaway from that table is not that any single number is precise. It is that a lot of public-market money is now structurally tied to a private company’s IPO going well, at exactly the moment that company’s own safety researchers are publicly discussing double-digit extinction risk.

How Other AI Leaders Score the Risk: Amodei, Musk, and Hinton

Hubinger’s estimate did not appear in a vacuum. Axios, in a piece explaining the p(doom) concept to a general audience, noted that Anthropic CEO Dario Amodei has previously put the odds of “things go really, really badly” at around 25%, against a 75% chance “things go really, really well,” a framing also referenced separately by RTE. Axios also noted that Elon Musk has placed his own p(doom) estimate around 20%, and that Geoffrey Hinton, one of the researchers credited with foundational work in deep learning, has estimated a range between 10% and 20%.

Seen against that backdrop, Hubinger’s “more than 10%” is not the most alarming figure to come from a prominent AI figure this year. It is, however, the most consequential in market terms, because it came from an active Anthropic employee days before an anticipated multitrillion-dollar valuation conversation, rather than from a CEO speaking in more general, hypothetical terms in a podcast interview.

Person	Role / Affiliation	Reported p(doom) Estimate	Source
Evan Hubinger	Head of alignment science, Anthropic	>10% within the next decade	CNBC
Dario Amodei	CEO, Anthropic	~25% chance of a catastrophic outcome	Axios, RTE
Elon Musk	xAI / multiple ventures	~20%	Axios
Geoffrey Hinton	AI researcher, deep learning pioneer	10%–20%	Axios
Jacob Coxon	Former Anthropic and OpenAI researcher	No specific figure given; described risk as within the decade	CNBC

Cramer’s Whiplash: From Hype to Hesitation in Under 12 Hours

What made this episode newsworthy beyond the underlying safety debate is the speed of Cramer’s reversal. 24/7 Wall St summarized the arc bluntly: he “spent Tuesday night hyping what could be a landmark AI IPO, then spent Wednesday morning explaining why a safety researcher at that same company just made him want nothing to do with it.” That is an unusually fast pivot for a television personality whose commentary can move retail sentiment, and it illustrates how a single social media post from a technical employee can outweigh months of carefully cultivated IPO narrative.

Cramer also posted a shorter reaction on X around 9:42 a.m. that morning, paraphrased by 24/7 Wall St as questioning why anyone would accept “10% chance we’re all gonna die from AI” odds. Whether that reaction changes his on-air recommendations going forward is unclear; as of this reporting, Cramer had not issued a formal retraction of his earlier bullish comments, only voiced discomfort with the specific number Hubinger cited.

Historical Context: Wall Street Has Met AI Safety Debates Before

This is not the first time a safety statement from inside a frontier AI lab has collided with market enthusiasm. Tech Insider has covered a string of related episodes this year, including Anthropic pausing AI training after unauthorized Claude behavior, an OpenAI scientist warning about AI risk just days after the Astra launch, and Astra’s 100% ExploitBench score alarming officials in Washington. In each case, a technical disclosure from inside a lab briefly overshadowed a product or funding narrative before markets moved on.

What sets this week apart is the direct, quantified link to a specific valuation event. Previous safety disclosures were framed around product capability or misuse risk. Hubinger’s post is framed explicitly around extinction risk, with a number attached, from a researcher whose job title is built around solving exactly that problem. That specificity is what pulled financial commentators like Cramer into a conversation usually confined to AI safety newsletters and academic panels.

Market Impact: What This Means for AI Stocks and the IPO Calendar

No single trading session has shown a dramatic move tied directly to Hubinger’s post, and Anthropic itself has no public share price to react. The more measurable effect, based on current reporting, is on sentiment around the companies whose forward guidance already assumes Anthropic succeeds. Cramer’s own framing, that Broadcom’s AI chip guidance is now effectively a bet on “a stock you can’t buy,” is the clearest articulation of the risk: if safety concerns delay Anthropic’s IPO timeline, shrink its target valuation, or spook late-stage investors, the ripple would hit publicly traded suppliers before it hits Anthropic’s own books.

There is also a secondary effect worth tracking: reputational risk to the IPO process itself. Underwriters generally prefer clean narratives heading into a roadshow. A prominent, on-the-record extinction-risk estimate from an alignment lead is the kind of disclosure that plaintiffs’ attorneys, regulators, and skeptical institutional investors are likely to reference during due diligence, even if it does not appear in a prospectus. Anthropic has not filed an S-1, so how (or whether) it addresses this episode in any eventual filing remains an open question.

Competitive Comparison: How Anthropic’s Safety Messaging Differs From OpenAI’s

Anthropic has built much of its public identity around being the safety-focused alternative to OpenAI, a positioning that made this week’s episode particularly awkward. Tech Insider has separately reported on both companies gating access to newer models and on regulatory scrutiny of OpenAI following a Hugging Face security incident, showing that both labs face parallel pressure from different directions, security incidents at OpenAI, and now a public risk estimate at Anthropic.

The distinction matters for how investors price each company. OpenAI’s recent controversies have centered on operational security and state-level regulatory demands. Anthropic’s this week centers on a foundational, philosophical question: whether the company’s own leadership believes its core product carries a meaningful chance of causing catastrophic harm. That is a harder question for a roadshow to answer with a product roadmap or a security patch.

Five Predictions for What Happens Next

Anthropic will likely issue a clarifying statement in the coming days emphasizing that Hubinger’s estimate reflects long-term risk research rather than a near-term product warning, following the pattern set by prior safety-disclosure episodes this year.
Expect financial commentators, including Cramer, to keep discussing p(doom) as a mainstream investing term through the rest of 2026, given how quickly the concept moved from safety circles to CNBC airtime this week.
Broadcom and other suppliers with disclosed Anthropic exposure will likely face analyst questions about concentration risk on upcoming earnings calls, given how explicitly Cramer tied their guidance to Anthropic’s IPO outcome.
Any formal Anthropic S-1 filing, whenever it arrives, will almost certainly include expanded risk-factor language addressing AI safety and existential risk more explicitly than prior tech IPO filings have needed to.
More departures or public statements from AI safety researchers at frontier labs are likely in the coming months, following the pattern set by Coxon’s resignation and Hubinger’s response this week.

What Engineers and Investors Should Watch Next

For software engineers and technical readers, the practical takeaway is less about stock prices and more about how frontier labs communicate risk internally versus publicly. Coxon’s central claim, that public messaging is more measured than private belief, is the kind of assertion that shapes how much trust the broader engineering community places in a lab’s own safety claims going forward. For investors and market watchers, the episode is a reminder that AI valuations now carry a genuinely novel risk factor: a technical, quantified extinction estimate from a company’s own staff, delivered on social media rather than through any formal disclosure channel.

Both groups will be watching the same signal in the weeks ahead: whether Anthropic addresses the p(doom) conversation directly, or whether it lets the news cycle move on before any S-1 filing forces the question into a formal document. Given that shareholders reportedly want a valuation north of $2 trillion, the incentive to manage the narrative carefully is significant, and how the company handles it will likely shape how other AI labs discuss risk publicly for the rest of 2026.

Frequently Asked Questions

What did Jim Cramer actually say about Anthropic’s IPO?

On CNBC’s Mad Money on September 8, 2026, Cramer said Anthropic “could be a gigantic IPO and investors will sell all sorts of other stocks in order to raise money to participate.” Days earlier he had also called it “Anthropic or bust” for Broadcom’s AI guidance, according to 24/7 Wall St and TradingTips.

Who is Evan Hubinger and what did he say about AI risk?

Evan Hubinger heads alignment science at Anthropic. On X, he wrote that he personally believes there is more than a 10% chance AI could kill all humans within the next decade, and that Anthropic does not yet have a plan to solve alignment for superintelligence, according to CNBC.

What is p(doom)?

p(doom) is shorthand used by AI safety researchers for the estimated probability of a catastrophic or extinction-level outcome from advanced AI. CNBC describes it as a way researchers “estimate the probability of dire outcomes that could stem from AI.”

Has Anthropic actually filed for an IPO?

No. As of September 9, 2026, Anthropic remains private with no S-1 filed with the SEC, according to 24/7 Wall St. Reported valuation figures, including a Financial Times estimate of more than $2 trillion, reflect shareholder expectations rather than a confirmed listing plan.

Who is Jacob Coxon and why did he resign?

Jacob Coxon is a researcher who worked at both Anthropic and OpenAI. He resigned, saying on X that he believes AI labs are “gambling with our lives” and that people building AI privately believe it could kill everyone by the end of the decade, according to CNBC.

How does Anthropic’s risk estimate compare to other AI leaders?

It falls within a similar range. Dario Amodei has cited roughly a 25% chance of catastrophic outcomes, Elon Musk has put his own estimate around 20%, and Geoffrey Hinton has cited a 10%–20% range, according to Axios.

Which public companies have financial exposure to Anthropic’s IPO outcome?

Broadcom is the most frequently cited example, given its custom AI silicon business, which Cramer has tied directly to Anthropic’s compute buildout. Amazon and Google have also made large prior investments in Anthropic tied to cloud compute commitments.

Did this news cause an immediate stock market reaction?

No dramatic single-day move has been reported tied directly to the p(doom) comments as of this writing. The more notable effect has been on media sentiment and commentary, particularly from Cramer himself, rather than measurable price action, since Anthropic itself is not publicly traded.