Anthropic, the AI company behind the Claude models, is courting investors for a September-or-October IPO that could be the largest in history. An IPO — initial public offering — is the day a private company first sells its shares to the public. The number on the pitch is a $47 billion revenue run-rate: the company’s current pace of sales, stretched out over a full year.
The number is real and enormous. It is also arriving in the same week the papers were busy cataloging the market’s AI jitters. Healthcare stocks now openly trade as the anti-AI insurance policy — the market has started buying protection against its own favorite theme. That is the mood into which history’s biggest IPO is being marketed.
None of this makes the offering good or bad. It makes it worth reading slowly. So let’s read slowly.
The money machinery
The most instructive story of the week was not about any AI model. It was about financing. Nvidia (NVDA) and a roster of Wall Street’s largest firms — Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs (GS) and KKR — have assembled roughly $500 billion of chip-financing platforms. The structure is simple once you slow it down: special funds buy the AI chips, then lease them to the AI companies that use them — the way airlines lease planes instead of buying them.
Now follow the arrows. Nvidia sells the chips. Nvidia also promises to cover part of the loss if used chips turn out to be worth less than hoped — up to 25% of a project’s cost. Broadcom (AVGO) disclosed a $29 billion promise of that kind, tied to a $35 billion lease package for Anthropic. And AI companies have borrowed $344 billion in bonds this year to fund the buildout. Skeptics call the pattern “circularity lite” — the seller helping finance its own customers — and the market gave the phrase a nod: Nvidia’s stock fell 2.9% on the news.
To be measured about it: leasing expensive equipment is one of the oldest, most respectable ideas in finance. Airlines lease planes. Railroads lease cars. Nobody calls that a bubble. But when the seller of the equipment also guarantees its future value, the seller is holding a piece of the risk that demand stays strong. That is not a scandal. It is a fact that belongs in the math — sitting right next to the $344 billion of bonds, because bonds, unlike enthusiasm, come with due dates.
Sellers pick the moment
Now, the timeless part. Companies do not go public at random. An IPO is priced at the moment the sellers — who know the business best — find most agreeable. That is not cynicism; it is just seating arrangements. The people with the most information choose the date. The people with the least information supply the money. It has always worked this way, and knowing it costs nothing.
A $47 billion sales pace commands genuine respect. Few companies in history have grown this big this fast, and the official filing, when it arrives, will be one of the great business documents of the decade. We intend to read every page of it.
Our three questions
First: how much of the revenue is propped up by the same circular financing that funds the chips? When your customer’s lease is guaranteed by your supplier, the word “demand” needs a footnote.
Second: what must the company keep spending to keep the sales pace running? A run-rate is a speedometer, and speedometers say nothing about fuel.
Third: what price does the offering ask for all of it? Because the difference between a great company and a great investment is always, boringly, the price.
Until those numbers are on paper, enthusiasm is not a strategy — and neither is scorn. The right posture is the oldest one: admiration for the achievement, arithmetic on the terms. One of those is free. The other one is the job.