Haircut No. 2: Anthropic IPO Talk Slides To $1.2 Trillion As AI Rout Crushes Odds Of A November Listing

Keeps going lower.

Three weeks ago we reported that Anthropic IPO investors had marked the valuation from $2 trillion down to $1.5 trillion as the growth slowdown sank in. This morning, Bloomberg reports that the price tag has been cut again: after Thursday's AI bloodbath, some potential backers are now discussing a value as low as $1.2 trillion for the company, a person familiar told the news agency (deliberations are "ongoing," no final decisions have been made, and Anthropic declined to comment).

Put in context: in early October investors were floating a fair value "as high as $2 trillion." The low end of today's range is 40% below that peak, which is a lot of haircut for a company that, according to the WSJ, just pushed its annualized revenue run rate past $80 billion. And for some, even that's not low enough.

"Fair Value," Revised (Again)

The Bloomberg story fills in the timeline. After the company's October briefing on its latest financials, investors told the bankers on the deal that their estimate of a fair valuation had moved to around $1.4 trillion to $1.5 trillion. Then came Thursday, when the Nasdaq 100 plunged 6.4%, wiping out $2.3 trillion in market value, a record for a single day, and that number "shifted again." Which is why prospective investors are now "facing a weekend of uncertainty ahead of the kickoff of formal marketing."

The buy-siders Bloomberg spoke to are all over the map, though the direction of travel is the same:

• BNP Paribas Wealth Management Germany's CIO Stephan Kemper expects the valuation to land at the very low end of the rumored $1.5-$2 trillion range, "maybe some 10% to 15% below that," which works out to roughly $1.3 trillion.
• Gallo Partners' Michael Alfaro said before Thursday's rout that Anthropic needed to target $1.3 trillion to $1.4 trillion "in order to build a quality book." That was before the Nasdaq's worst day since Covid.
• Sycomore's David Rainville: "If they set a valuation toward the bottom of the range, say $1.2 trillion, they shouldn't have a problem finding investors." Translation: at $1.5 trillion, they might.
• And HSBC's Max Kettner thinks around $1 trillion "could be a more comfortable level."

Kettner's reasoning is worth quoting, because it's the one that should worry the underwriters most:

"I don't think we need the headline 'oh my god, it is a trillion dollar IPO,' because I think what's actually more important is the secondary. If we get a dynamic like we had with SpaceX, that you get it on secondary and then stock rips for a couple hours, but then it dumps, in the current environment you don't want to have that."

Kettner isn't alone at $1 trillion. Late on Thursday the WSJ reported that three major investors planning to take part in the offering think the $1.5 trillion valuation will likely need to come down, with one putting it "closer to $1 trillion." Executives and advisers are set to meet in the coming days, the raise could shrink from the planned up to $100 billion, and one investor floated a delay because it may be hard to convince institutions "burned in the recent selloff" to take a big AI risk again. At about the same time, Reuters reported that potential investors doubt Anthropic can start its roadshow next week and still list in November, something we discussed in "Is Anthropic About To Delay The Biggest IPO Ever?"

Here is how the "fair value" has evolved in the space of five weeks:

Note the dashed line: at $1 trillion, the IPO would price barely 4% above the $965 billion post-money of May's Series H. So much for the "biggest IPO ever" step-up.

$80 Billion Of ARR... And It Still Doesn't Clear

The awkward part is that the company itself is still growing fast. The WSJ reported this morning that Anthropic's ARR topped $80 billion at the end of October, against some expectations in the mid-$70 billion range, which probably helped this morning's premarket bounce in Nvidia and the memory names. Badly needed good press, and maybe a sign the August-September slowdown we flagged after investor day wasn't the start of a stall.

Some napkin math (approximate): at $2 trillion, buyers were being asked to pay 25x that $80 billion run rate. At $1.25 trillion, the midpoint of the new range, it's about 16x. In other words, the multiple has compressed by roughly 40% while the run rate went up. That's what a derating looks like.

On a trailing basis the math is still eye-watering. Against the roughly $35 billion Anthropic booked over the past four quarters, $1.25 trillion is about 35x sales. As Fortune noted this week, citing Jay Ritter's IPO data, anything above ~$1.41 trillion would mean more than 40x trailing revenue, a level reached by just 14 of 4,110 IPOs since 1980. And that's before you get to the $895 billion in committed spend disclosed in the amended S-1, roughly $70 billion of which is due in the next 12 months, or the October revenue line the filing left conspicuously blank.

Which brings us to the real problem: the company's numbers are no longer the only thing being priced. After Tuesday's Democratic sweep, the market is also pricing permitting risk on the data centers Anthropic needs to hit those numbers. Goldman estimates that the states likely to add restrictions hold about 45% of planned US capacity, and Anthropic's own Texas campus was already stuck on a $15 billion bridge loan before the election.

Polymarket Got There First

As usual, the prediction markets didn't wait for the Bloomberg scoop. Yesterday morning, with the Nasdaq rout just starting, we pointed out that the odds of an Anthropic listing were tumbling:

(The "by Nov 25" contract in that chart had slid to 69% from 86% a day earlier.) They kept tumbling. As of this morning, Polymarket gives Anthropic a 54% chance of listing by Nov 25 and 60% by Nov 30, down from 86% for both before the rout. Even the odds of an IPO at any point in 2026 have dropped from 93% to 73%:

The valuation markets moved even faster. Polymarket's "What will Anthropic's IPO valuation be?" contracts now make $1.25-$1.50 trillion the favorite at 43%, up from about 5% a week ago. The $1.50-$1.75 trillion bracket, which was a ~65% favorite right after the midterms, has collapsed to 32%, and the odds of anything $1.75 trillion or above are down to 16%. Bloomberg's $1.2-$1.3 trillion range straddles the bottom of the new favorite bracket, and the "below $1.25 trillion" tail has quietly gone from about 1% to 7%:

Nothing says "orderly price discovery" like the favorite outcome changing three times in a month.

1998 Or 2000?

Over at Goldman, the trading desks are split on what Thursday's crash means for the AI trade, and by extension for whoever ends up buying Anthropic stock. Goldman Delta One head Rich Privorotsky, who admitted "what I got wrong" on AI just one day earlier, posed the question everyone is asking in his morning note (available to pro subs):

"Is this a 1998 style liquidity crisis and buying opportunity, or March 2000 and the beginning of a structural technology unwind? I lean toward 1998. Russia's default/LTCM were the external shocks but US growth remained intact. Once liquidity returned and the Fed eased, markets recovered rapidly... The uncomfortable parallel with 2000 is the sheer scale of spending and uncertainty around returns. Are we liquidating a sound growth story or discovering its economics are wrong? I lean toward the former, but demand, utilization and returns need to hold."

Privorotsky lists four concerns behind the selloff: i) insufficient capital to finance the buildout, ii) insufficient revenue to justify the investment, iii) insufficient power and permitting capacity, and iv) an unsustainable aggregate spending trajectory. The market, he argues, "has latched onto number three, arguably the least important." Maybe. But the Anthropic IPO is a referendum on all four at once: it is a $100 billion capital raise (#1), priced off a run rate (#2), by a company that needs permitted, powered data centers (#3) to service nearly $900 billion in commitments (#4).

And Goldman's own credit strategists are hardly reassuring on #1. In her latest Global Credit Trader, "AI-Related Supply: Outlining the Directional Risks"  (also available to pro subs), Amanda Lynam notes that hyperscaler bond issuance has already hit $296 billion this year, well above Goldman's $250 billion full-year forecast, and warns that between now and early December the risks to spreads are "firmly tilted towards more widening." That is the backdrop against which bankers are trying to place what would be the largest equity raise in history, which is something we have been warning about since last October, when AI quietly became the largest sector in the bond market.

Meanwhile, the people who were supposed to anchor the book are licking their wounds. Goldman's derivatives desk says clients "had aggressively dumped protection" into Wednesday's post-midterm relief rally, leaving the market underhedged for Thursday's reversal, and tech-focused hedge funds have just wiped out their entire 2026 gain. Not exactly the crowd that will be clamoring for a $100 billion allocation next week.

And Then There's OpenAI

Anthropic isn't the only frontier lab finding out what the AI unwind does to private marks. Last Friday, the FT reported that OpenAI had slashed its funding target by $10 billion to about $20 billion at a $1.25 trillion valuation, a week after it trimmed the price from $1.4 trillion following a "thud" from Morgan Stanley's wealthy clients. Now the WSJ reports that OpenAI has secured just $11 billion of the $30 billion it originally targeted, and that's at the adjusted $1.25 trillion valuation.

Here's the punchline: just last week, the FT's explanation for OpenAI's struggles was that Anthropic's IPO was "absorbing both investor capital and focus." Well, if Anthropic prices at the low end of Bloomberg's range, the "hot" public listing would value the company below the $1.25 trillion OpenAI is struggling to raise money at privately. One of those two marks is wrong, and probably not the public one.

Size Down, Price Down... Or Wait Until 2027?

So what are Anthropic's options? Per Bloomberg, the company and its selling shareholders could start marketing with a conservative size and valuation and upsize on demand, the Cerebras and Figma playbook. Bloomberg helpfully adds that both "are now trading well below their IPO prices." The "less palatable option" is to delay: IPO filings must be public for at least 15 calendar days before formal marketing, and some issuers wait weeks or months, updating their quarterly financials along the way. For Anthropic, that would probably mean a Q4 refresh and a 2027 listing.

Our take: the bulls have a point that the business is growing (an $80 billion run rate is no small number), and Privorotsky may well be right that this is a liquidity event rather than a demand collapse. But his own 1998 analogy ended with the Fed easing. This time, Fed speakers Barkin and Hammack spent Thursday openly arguing for higher rates while the Nasdaq had its worst day since March 2020. In that environment, the more likely path is a third haircut or a delay, not a triumphant upsizing. As IG's Alexandre Baradez put it, "maybe it's better to go to the market when the valuation excesses have already been purged."

The question is whether they have been, or whether $1.2 trillion is just this week's number. For now, everyone wants to be in the room for the biggest IPO ever; fewer and fewer want to pay for the privilege.

More in Privorotsky's "Liquidation. 1998 or 2000." note and Goldman's Global Credit Trader, both available to pro subs.