For months, the question hanging over the artificial intelligence industry was not whether Anthropic would go public, but when — and on Monday the Anthropic IPO stopped being a rumour and became a document, as the company behind the Claude chatbot lodged its prospectus, the Form S-1, with the Securities and Exchange Commission.
Table of Contents
The destination is Nasdaq, the ticker ANT, and the prize a valuation somewhere between $1.5 trillion and $2 trillion. Because the filing — made on 26 October, following a confidential draft filed in June — sets neither a price nor a number of shares, the final figure will be set by investor appetite during the roadshow, with mid-November the expected window, the week of 16 November the reported target, and Goldman Sachs and Morgan Stanley said to lead an underwriting syndicate of more than twenty banks.
Anthropic IPO: a loss that grew because the company did
Anyone scanning the prospectus for the bottom line should sit down, because the same six months produced record revenue and a net loss of $80.7 billion, dwarfing the $1.3 billion lost a year earlier — and yet roughly $72.4 billion of that loss never involved money leaving the building at all, being non-cash accounting charges on convertible notes and similar instruments, paper promises whose value swelled precisely because Anthropic kept getting more valuable.
In other words, the better the company performed, the deeper its paper loss appeared — and strip those charges away, and the second quarter alone delivered a first: an operating profit worth $974 million.
Revenue growth with almost no precedent in the filing
If the loss needs explaining, the top line does not, because the filing documents first-half 2026 revenue of $16.29 billion against just $1.16 billion a year before — a leap of roughly thirteen hundred percent that few growth stories can match.
The acceleration came late and fast: the second quarter contributed $11.56 billion of that total, comfortably more than two-thirds of it, and against full-year 2025 revenue of $4.59 billion — itself around twelve times the year before — the trajectory looks near-vertical.
Anthropic IPO: nine billion on compute, six hundred billion in promises
Staying at the front of the AI race is ruinously expensive, and the filing prices it honestly: spending on computing power and infrastructure climbed to almost $9 billion in the first half of 2026, from $2.3 billion a year earlier, to train and run ever larger models.
The future bill is bigger still: cloud agreements worth roughly $450 billion are locked in across the next seven to ten years — including about $111 billion with Google, $110 billion with Amazon Web Services and $31 billion with Microsoft — and with a Google chip lease arranged through Broadcom and further talks valued at up to $150 billion added, total obligations pass $600 billion. Whether that is visionary scale or colossal fixed cost depends on demand — and the prospectus flags that close to a fifth of first-half revenue came from just two customers, either of whom could reshape the growth story by spending less.
Founders keep control, and the warnings stay in writing
Ownership of the future company will not mean control of it, since the filing confirms Anthropic’s seven co-founders will retain 50.1 percent of the voting power through a special share structure, leaving public investors a substantial economic stake but no ability to outvote the people who built the business.
Just as distinctive is the candour: more than ninety pages of the document are devoted to the dangers of the technology itself, as far as existential threats to humanity and scenarios of models resisting being switched off or distorting the information they give — a safety-first creed translated into the formal language of investor disclosure.
What the Anthropic IPO really asks investors to decide
Read one way, the $80.7 billion loss is an accounting shadow rather than a wound, cast by instruments that grow more expensive on paper as the company succeeds; read another, the $600 billion of commitments are real money owed whether AI demand keeps doubling. The roadshow will price which of those two stories investors believe, and whether the second quarter’s operating profit was a turning point or a snapshot.
The Anthropic IPO is therefore a referendum on the whole AI boom: a warm reception near $2 trillion would signal that almost any price can be justified, while a cool one could drain momentum from names as distant as Nvidia and Micron. For ordinary investors, it is the first chance to hold a direct stake in one of the two firms leading the AI race, alongside OpenAI — and the same tension the prospectus lays bare: spectacular growth, spectacular spending and unusually honest warnings, all resting on a single balance sheet.