Priced by Stature: What a Public Shareholder Actually Buys in Anthropic's IPO

Anthropic's S-1 shows half-year revenue of $16.3 billion, up roughly fourteenfold, and almost none of it helps answer what the company is worth. The price will be set the way prices for companies of a certain stature usually are: backwards, from a number that feels right. The second question is the one the prospectus answers precisely, and it is less comfortable: what a public shareholder gets for that price, in a company whose founders hold 50.1% of the votes and whose board majority is chosen by a trust.

Photo: NASA/Bill Ingalls, public domain via Wikimedia Commons

By Fabian Fach, Founder, Black Flower Capital

On 26 October, Anthropic filed its S-1 for a Nasdaq listing under the ticker ANT. The numbers inside are remarkable by any standard. Revenue for the first half of 2026 came to $16.3 billion, against roughly $1.16 billion in the same period a year earlier. Second-quarter revenue alone was $11.6 billion. The headline net loss for the half, $80.7 billion, is mostly a non-cash charge of about $72.4 billion on convertible notes and similar instruments; the operating loss for the half was a far more ordinary $693 million.

We have written before about the financing loop around the AI labs, about what the filings of their counterparties confirm, and about the risk factors in this very prospectus. This piece asks two narrower questions. How will the market arrive at a price for a company like this? And what exactly does a buyer of the shares own once it has?

A Filing That Cannot Answer the Valuation Question

On the FT's Unhedged podcast this week, Lex columnist John Foley made an argument that sounds flippant and is not: the historical numbers in the S-1 are close to useless for valuing the company. A business that did not exist before 2021 and had no revenue before 2023 changes too fast for its past to say much about its future.

The arithmetic makes the point on its own. Take first-half revenue, double it to an annual pace of about $33 billion, and let it grow fourteenfold a year, as it just did. After three years, annual revenue would be roughly $89 trillion, most of the world's entire economic output. Nobody believes that, which is precisely the problem: the one growth rate the filing documents is the one that cannot continue, and the filing does not say what replaces it. Management's own forecasts, the only forward numbers that might help, are not in the document. They will be discussed on the roadshow.

Everything that actually determines what the company earns sits outside the S-1: how fast customers adopt the products, what competitors charge, whether a price war arrives, what data center capacity costs, how much of the value of AI ends up with the companies that build the models rather than with their users. With growth this steep, moving any one of those inputs slightly moves the valuation enormously.

How the Number Will Actually Be Set

So the price will be found another way. Foley's expectation is that the bankers and the company settle on a target, his guess is around $1.5 trillion, and the analysis is then done backwards to justify it. Reported expectations have ranged from $1.5 trillion to $2 trillion. The last private round valued the company at $965 billion.

The benchmark will not be a discounted cash flow model. It will be a question of stature: is this company as important as SpaceX, which listed in June at about $1.77 trillion and became the largest IPO on record? Is it in the same league as Meta or Alphabet? If the answer feels like yes, the valuation lands near theirs. Foley called this uncomfortably unscientific, and he is right. It is also how very large, very new companies usually get priced, because there is nothing better to price them with.

That has a consequence worth stating plainly. A valuation set by comparison to other giants is only as stable as the comparison. If sentiment toward the reference group moves, the anchor moves with it, regardless of anything Anthropic itself reports in a given quarter. A price derived from stature behaves like a sentiment indicator for the whole group of companies it is benchmarked against.

What the Prospectus Does Answer Precisely: Who Decides

Where the S-1 is unusually precise is governance, and here the answers matter more for a public shareholder than any revenue line.

Anthropic is a public benefit corporation. In a conventional company, directors owe their duty to shareholders, and shareholders can sue if that duty is breached. A public benefit corporation's directors must balance shareholder interests against the interests of those affected by the company's conduct and against a stated public purpose, in Anthropic's case, broadly, the responsible development of advanced AI for the long-term benefit of humanity. A shareholder who believes the company spent too much in pursuit of that purpose has much less to stand on than in an ordinary company.

On top of that sit two centres of control. The seven co-founders will hold 50.1% of the voting power through a founder-controlled entity, even if they sell most of their economic stake. And a separate body, the Long-Term Benefit Trust, holds a special class of stock that lets it elect and remove the directors who make up a majority of the board. Its trustees do not run the company. They choose who oversees the people who do.

In plain terms: the founders control the shareholder vote, the trust controls the board majority, and a public investor buying shares in the IPO controls neither. Outside of rare circumstances, ordinary shareholders will not decide anything of consequence.

Why This Is Not a Criticism of the Mission

It would be easy to read all this as a warning. It is not meant as one. Anthropic built this structure deliberately, years before an IPO was on the table, because its founders take seriously that the technology they build could cause serious harm. A company that says so in its own filing, and then designs its control structure so that a commercial quarter cannot override that concern, is being consistent. Many of the most valuable companies of the last two decades went public with founder control and did well by their shareholders.

The question for an investor is not whether the structure is admirable. It is whether the price reflects it. A buyer at $1.5 trillion or $2 trillion pays for the full economic upside of a company whose decisions, by design, are not primarily made for that buyer. Those are two different things, and the second should come with a discount. A valuation set by stature, benchmarked against companies where public shareholders do have a voice in how the company is run, will not apply that discount on its own.

There is also a question of speed. The checks in this structure, a trust that elects directors and a board that oversees management, work over years. Problems in AI can arrive in hours. The governance is built for the long horizon, which is its purpose, but it is not a mechanism that protects a shareholder from a fast, sharp repricing.

What This Means for a Portfolio

Most investors will not decide whether to buy Anthropic. They will end up owning it anyway. A company valued in the trillions will find its way into the major equity indices, and from there into every portfolio that tracks them. The same index-weighted exposure that already concentrates portfolios in the AI trade will add one more very large position, this time with almost no voting rights attached and a price anchored to the sentiment around its peers.

That is not a reason to avoid the company, and this is not a recommendation about the shares in either direction. It is a reason to know what the position is: a large, fast-growing business whose valuation will move with the stature of the group it is compared to, and whose governance, by deliberate design, answers to its mission first. An allocator who holds it through an index should count it as part of the same AI exposure as the rest of the group, not as a separate source of return.