Waiting for good IPO

This article is an on-site version of our Unhedged newsletter. Premium subscribers can sign up to get the newsletter delivered every weekday. Standard subscribers can upgrade to Premium here, or explore all FT newsletters

Good morning. Crash averted. US tech stocks opened the week up, led by a rally in chipmakers. More evidence that there’s always someone willing to buy the dip. Meanwhile, Anthropic pledged 5 per cent of its shares to a “public benefit fund” to support research in AI safety and biomedicine. At current valuations, that’s $48bn, but as I write below, there’s a cloud hanging over that figure. Not to be outdone, Elon Musk said he would give SpaceX shares to children. It’s a nice idea — getting the youth into investing early and all that — but let’s not forget the Doge cheques that never arrived. Are these ideas the beginnings of an AI wealth fund or a sign of the US’s slide towards some kind of benevolent plutocracy? Email us: unhedged@ft.com.

Anthropic IP-no?

This was supposed to be the year of blockbuster initial public offerings. Elon Musk’s SpaceX went first, raising a record $85.7bn at an almost $2tn valuation. But as the summer dragged on, the timeline for the others slipped. Now the market is skittish and some companies are delaying or pulling their plans to list.

Let’s leave OpenAI aside. The ChatGPT maker was slow to focus on commercialisation and since June, at least, we’ve known it was unlikely to list this year. But Anthropic — maker of Claude — pushed its listing from September/October to November, giving it a very narrow window between the US midterm elections and Thanksgiving to pull off what could be the biggest stock market debut in history. Will it happen? (On Thursday, you can ask this and other questions about the AI trade of our San Francisco colleagues George Hammond and Richard Waters.)

Polymarket is pricing a 60 per cent chance that the Anthropic listing will go through in November (the site is also taking bets on how soon the company will declare bankruptcy). And at least publicly the AI lab has given no reason to doubt that it will go public. It is speaking to investors and trying to build momentum ahead of the formal bookbuilding process. But at the same time, market volatility is squeezing the window in which it could happen, and Anthropic, led by chief executive Dario Amodei, is facing growing questions about legal liability for its agents’ actions. Public markets may simply be unwilling or unable to price that risk until the regulatory treatment of AI start-ups is clarified. (The comptroller of New York City, who oversees the city’s $325bn pension fund, has also called on the company to pause its listing while it addresses the existential risks of its technology.)

On the financials, Anthropic was the first of the major AI labs to report a profitable quarter, although that may have been on the basis of less than standard accounting. More recent reporting has suggested the company’s growth, while still phenomenal, may be slowing. As our colleagues reported last week:

. . . SemiAnalysis, a research firm, projected sales grew to roughly $74bn on an annualised basis ...Those estimates are lower than more bullish predictions during the summer, when a number of Anthropic’s investors told the FT the company would far surpass $100bn in annualised revenue by year-end.Joey Brookhart, an analyst covering AI labs at SemiAnalysis, said investors had been lowering their projections for Anthropic’s annualised revenue over the past three months.

. . . SemiAnalysis, a research firm, projected sales grew to roughly $74bn on an annualised basis ...

Those estimates are lower than more bullish predictions during the summer, when a number of Anthropic’s investors told the FT the company would far surpass $100bn in annualised revenue by year-end.

Joey Brookhart, an analyst covering AI labs at SemiAnalysis, said investors had been lowering their projections for Anthropic’s annualised revenue over the past three months.

The company is also facing higher borrowing costs as the Federal Reserve raises interest rates and appetite for AI infrastructure bonds may be souring. Over the weekend, the FT reported that Anthropic offered Blackstone a 10 per cent coupon on $9bn of new debt, as part of a broader $60bn chip-financing package. This is high for a company that hopes to eventually receive an investment-grade credit rating and raises questions about the cost of financing its growth. Until it releases its prospectus, we simply won’t know how sustainable these commitments are.

There is a tonne of investor enthusiasm for AI exposure. But it looks different than it did at the start of the year. Take the examples of Nvidia and CoreWeave — two sides of the AI infrastructure build-out. One is able to mostly self-finance its growth through current revenues. The other is reliant on debt — and it has the price volatility to show for it. Investors want AI exposure, but they’re scrutinising the cost of building it.

We’re not saying Anthropic is CoreWeave. But in the AI lab’s need for external financing it has more in common with the neocloud provider than it does with Nvidia. It has built a large business of paying customers, but its growth also depends on access to financing and compute from a small group of partners. Reuters reported that nearly a quarter of its 2025 revenue came from two customers, while the company has disclosed at least $518bn in infrastructure commitments over roughly a decade.

But there’s an important reason to believe the AI lab will push through despite these challenges: there may be few better opportunities for the current owners to cash out. IPOs are usually seen as a way for groups to raise capital to finance investment. But the better way to think of these kind of listings — of companies that have experienced high growth and may be reaching their steady state — is as exit events for a small number of early investors, as the venture capitalist Paul Kedrosky put it on a recent episode of the Ezra Klein podcast. When early investors sell existing shares, they transfer risk from the current owners to retail and institutional investors.

Research suggests IPO delays are often a way for groups to wait out volatility or seek better market timing. It’s not a stretch to say that condition has been met. Some of those companies return to the market later, sometimes at lower valuations; others remain private or are acquired. A full withdrawal — a long shot — would be a signal that investors weren’t willing to buy at the price or on the terms Anthropic’s owners wanted. In the unlikely event it does delay, the immediate consequence for the start-up might not be terrible. It could wait. But it would be a sharp indicator to the market that the wild enthusiasm around the AI trade has peaked.

One good read

Central banker diary stalking

FT Unhedged podcast

Can’t get enough of Unhedged? Keep following Rob Armstrong and Katie Martin’s incisive and unfiltered analysis on the Unhedged podcast, a twice-weekly, 15-minute lively dive into the latest markets news and financial headlines. New episodes every Tuesday and Thursday.

Recommended newsletters for you

Due Diligence — Top stories from the world of corporate finance. Sign up

The AI Shift — John Burn-Murdoch and Sarah O’Connor dive into how AI is transforming the world of work. Sign up