Anthropic Wanted $2 Trillion. The Market Offered $1.4 Trillion.

A price range of $70 to $78 a share, a filing retracted within hours, and $26 billion sold privately to the companies that supply its chips. The defining listing of the AI boom is being priced like a company that has read the market.

The number that landed at the Securities and Exchange Commission on Tuesday afternoon was smaller than the one Anthropic's bankers had been repeating all autumn. The company will sell 825 million Class A shares at $70 to $78 each, raising between $57.75 billion and $64.35 billion. At the top of the range it is valued at $1.404 trillion on a fully diluted basis — about $1 trillion counting only the shares that already exist. Two months ago the expectation was $2 trillion. In May, private investors valued the company at $965 billion.

The float is still the largest test the AI boom has faced. It is simply being conducted at a discount to its own rumour mill.

The discount was not arrived at cleanly. Late on Monday, Anthropic sent terms to its banks showing 867 million shares at $70 to $80 — as much as $69.3 billion — then retracted the document within hours. The company said the pricing details were inaccurate. The filing that followed the next day carried fewer shares and a lower ceiling. Even the stationery had to be repriced.

What changed is the market, not the revenue. Last Thursday the Nasdaq recorded its deepest one-day fall in nineteen months, after midterm elections handed victories to opponents of data-centre construction and investors began asking what the build-out is actually for. "What we are seeing is essentially a reassessment of Anthropic's value in a more challenging AI environment," Ipek Ozkardeskaya, a senior analyst at Swissquote Bank, told CBC. "The company continues to post impressive growth, but investors are increasingly focused on the cost of generating that growth."

The growth itself remains extraordinary. Anthropic told investors its run-rate revenue reached $81 billion at the end of October, $10 billion more than at the end of September, and reported nearly $17 billion of revenue for the quarter through September. A run-rate annualises a single month, so $81 billion implies roughly $6.75 billion of revenue in October alone, against a monthly average closer to $5.7 billion across the third quarter. At $1.4 trillion, the company is being priced at about seventeen times that annualised figure — a software multiple attached to a business that spends like an industrial one.

The buyers are the part worth reading twice. Alongside the public offering, Anthropic has arranged to sell $26 billion of non-voting Class N stock at the IPO price to AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC. Google and Nvidia are each putting in $10 billion. These are not index funds. They are companies that sell Anthropic its chips and its cloud, taking equity in a customer before the public receives an allocation — the circular financing that investors have started to ask about, with Nvidia buying into a lab that has contracted to buy its accelerators, and doing so while its chief executive is under an FBI probe over a relationship with a woman alleged to be a Chinese spy.

The sellers are worth reading too. Each of the seven co-founders will sell 3,378,378 shares in the offering — about $250 million apiece at the $74 midpoint of the range, between $236 million and $264 million across it. Afterwards each will hold roughly 3.2 per cent of the Class A stock. Board member Reed Hastings has indicated he wants to buy $300 million through a donor-advised fund. Insiders can begin selling in stages from the first day of trading, with lock-ups opening after 27, 48, 62 and 76 days rather than the customary 180.

Public shareholders should understand what they are buying. Anthropic is a public benefit corporation whose founders hold supervoting Class F stock and control just over half the votes; a Long-Term Benefit Trust elects a majority of the board. Buyers of Class A shares get the economics and very little of the control. Up to 20 per cent of the offering is expected to go to retail investors, who will own a piece of a company that warned in the same document that its models have behaved in "unintended, misaligned, or autonomous ways" and may do so again.

Add the $9.7 billion greenshoe and the $26 billion private placement and the total raised approaches $100 billion — more, the Guardian notes, than SpaceX collected in the largest listing on record. The comparison that matters more is with OpenAI, which is seeking $30 billion in what is expected to be its final private round, after taking $11 billion at a $1.25 trillion valuation, below the $1.5 trillion it had sought. The whole frontier is being marked down at once.

For a year the argument about AI valuations happened between venture funds and short sellers, in private. On Tuesday it acquired a public reference price: $1.4 trillion, fully diluted, conditional on a roadshow that began this week and a debut expected around 19 November under the ticker ANT.

That price is not a verdict on whether Claude is worth it. It is a statement about who is willing to fund the answer. Roughly a quarter of the money is arriving from the companies that supply the compute, before the public sees a single share. When your suppliers are also your shareholders, a price stops being a measurement of what the market believes. It becomes something the market agreed to.

Sources: TechCrunch — Anthropic sets IPO price range of $70 to $78 per share, TechCrunch — All 7 Anthropic founders will each cash out about $250M in IPO, The Guardian — Anthropic targets valuation of up to $1.4tn in stock market flotation, CBC — Anthropic eyes $1 trillion US valuation in hotly anticipated IPO