Anthropic Wants $2 Trillion From a Market That Hasn't Seen the Books

Anthropic investors expect the company to go public in October at a valuation of $2 trillion or more, according to reporting from the Financial Times that has been picked up across the financial press since August 13. A filing could come as early as late August. Morgan Stanley, Goldman Sachs, and JPMorgan are reported to be leading the offering.

If it prices there, it is the largest initial public offering in history — larger than SpaceX, which listed in June at roughly $1.77 trillion and itself displaced Saudi Aramco's 2019 record.

Two things about that sentence deserve attention. The first is the number. The second is where the number came from.

The number is investor speech, not company speech

Anthropic has not confirmed a valuation. Reporting explicitly describes the IPO details as still under discussion, with the valuation not formally fixed — and notes that senior executives reportedly have not settled on a range even in private conversations.

The $2 trillion figure is what backers expect, told to reporters by backers. That is a specific kind of number. It is the number that makes existing positions look good, that anchors the conversation before a roadshow, and that costs nothing to say because nobody has to defend it in a prospectus.

None of which makes it wrong. It makes it unaudited. The gap between an investor's expectation in August and a price set by an underwriting syndicate in October is where most of the interesting information lives, and right now that gap is entirely unobserved.

The revenue case

The case for the valuation is growth, and the growth is real. Anthropic's backers expect annualized revenue to reach $100 billion to $120 billion by year-end — more than ten times its level at the start of 2026.

A 10x revenue year at that absolute scale has approximately no precedent in enterprise software. If the year-end figure lands in that band, a $2 trillion valuation is roughly 17–20x forward revenue. For comparison, that is a multiple public markets have historically paid for high-growth infrastructure businesses with durable margins, not an obviously insane one.

The word doing the work in that paragraph is if.

Annualized run rate is a rate, not a result. It captures a moment and multiplies it. In a business where a handful of enterprise contracts and API customers can move the aggregate substantially, and where the largest customers are themselves venture-funded companies buying inference at a loss, the run rate is a measure of current demand rather than of contracted revenue. A prospectus will show cohort retention, gross margin after compute cost, customer concentration, and the actual GAAP loss. Those are the numbers that determine whether the multiple holds.

The sequence

The listing calendar matters here. Anthropic filed confidentially with the SEC after closing a $65 billion Series H at a $965 billion valuation in May. OpenAI confirmed its own confidential S-1 days later, at a reported $852 billion. SpaceX priced in June. Vantage Data Centers is exploring a listing near $100 billion.

The AI boom's private capital phase is converting to public equity, all at once, in a single twelve-month stretch. That is not a coincidence of individual company timelines. It is what happens when private markets reach the limit of what they can absorb and the only remaining pool of capital large enough to keep funding the buildout is the public one.

Which means the October window is not just Anthropic's test. It is the first properly priced public read on whether the frontier-lab business model — enormous compute costs, enormous revenue growth, unresolved unit economics — clears at trillion-dollar scale.

If it prices at $2 trillion and holds, the private valuations behind it are validated and the next round of capital gets cheaper for everyone in the sector. If it prices lower, or prices there and breaks, the mark-to-market runs backward through every AI position on every balance sheet that holds one.

What a doubling in three months implies

Anthropic's May round valued it at $965 billion. An October listing at $2 trillion is a 107% increase in roughly five months, achieved without a public market ever having tested either figure.

That can be justified. Revenue growing 10x in a year justifies a great deal. Anthropic's enterprise position is genuinely strong, its coding products have unusual attachment, and the constraint on the business has consistently been compute supply rather than demand.

But a doubling on private marks, followed immediately by the largest IPO ever attempted, is a sequence that asks public investors to accept the private market's arithmetic at face value and then pay a premium on top of it. The prospectus is the first document in this whole chain that carries legal liability for being accurate.

That is the thing worth waiting for. Not the leaked target, not the banker chatter, not the comparison to Aramco. The S-1 — the gross margins after compute, the concentration, the cash burn, and the growth rate stated in a form that someone can be sued over.

Until then, $2 trillion is a hope with a date attached.