The story of AI giant Anthropic in three numbers

When Anthropic this week published the official paperwork in advance of what could be a record-breaking initial public offering (IPO), it set off a rush to fillet the draft prospectus for clues.

However, three numbers that don’t appear in the filing highlight some of the biggest open questions about both Anthropic’s business and the wider state of artificial intelligence (AI) at a moment when anxiety on Wall Street has been climbing.

One of those questions is: How fast is Anthropic growing in the weeks immediately before its share sale? The filing leaves a space for the company to report its revenue run rate in October, a figure that could play a surprisingly important role in determining sentiment among investors.

This extreme timeliness does not usually factor into an IPO, but the whiplash caused by recent events has made it a highly relevant detail.

The company’s short-term growth has already been through a wild swing this year. Henry Blodget, who made his name as an analyst during the dotcom bubble, calculated that month-to-month increases in Anthropic’s revenue went from an unheard-of 88 per cent in February this year to just 4 per cent in September.

[ Anthropic says its AI technology ‘could pose catastrophic or existential risk to humanity’ ]

Much of the story of AI in 2026 can be seen in those figures. Anthropic’s surge in growth on the back of the coding agent Claude Code set the industry alight early in the year. More recently, a slowdown has raised questions about how deeply AI use is reaching into business life and the effects of growing competition.

The Claude Code revenue take-off was a sign that Anthropic had found “product/market fit”, the moment when a technology meets a particular need and demand takes off. But beyond coding, Anthropic and its rivals are still searching for ways to unlock a much bigger market.

The recent slowdown is also a reminder that, unlike traditional software companies, little of Anthropic’s revenue is tied to regular subscriptions. In the first half of this year, 83 per cent came from consumption, making it highly dependent on short-term usage.

The summer lull may have been partly caused by seasonality, a factor that Anthropic notes in its filing. A growth rebound in October would certainly ease the worries. But seasonality is normally seen in mature businesses, not a hypergrowth company.

A second question from the filing: How likely is it that Anthropic’s giant bet on the future will pay off? The filing leaves an empty space for how much it has committed to spending on future computing capacity. There are details of more than $650 billion (€580 billion) in deals with other companies, but some estimates put the total at closer to $1 trillion.

These commitments, stretching out to 2045, are large even by the standards of an industry in the midst of a historic boom. Ratings agency Moody’s recently put the total off-balance-sheet commitments of the five largest hyperscalers at $2.8 trillion. That figure is the value of all the leases, purchase commitments and contracts they have signed for the data centres needed to support their own AI services and sell capacity to other companies – including Anthropic. Set against that figure, Anthropic’s bet of as much as $1 trillion is a meaningful number.

Tech start-ups often make bold claims about the huge long-term opportunities they see. They don’t usually put their money where their mouth is in quite such an extreme way.

The third question prompted by this week’s filing: Are recent price cuts fuelling a boom that will carry the technology deeper into personal and working life, or a signal that a ruinous price war has set in?

Anthropic said one of its latest models, launched last month, completed AI tasks 40 per cent more cheaply than its predecessor. To ease worries that this might suggest a damaging level of price deflation, it pointed to what happened when it cut pricing for a model launched last November: revenue growth in the following quarter rose from 48 to 132 per cent.

Using something that happened nearly a year ago may have less bearing on the current dynamics of the AI market. This year has seen rapid improvements in the performance of lower-priced models.

There is plenty to marvel at in Anthropic’s filing. It shows a company that has notched up unprecedented growth after only three years, with profit margins and cash flow notably stronger than most analysts had expected. What comes next – both for Anthropic and the rest of the industry – is less clear. – Copyright The Financial Times Limited 2026