The largest initial public offering ever attempted has moved a month. Anthropic now plans to list in November rather than October, people familiar with the matter told the Journal, so that it can show investors third-quarter numbers after rival OpenAI shipped its newest model, Astra, in September. Investors expect a valuation around $2 trillion and a raise of up to $100 billion. Both would top the records SpaceX set in June.
The people involved say the decision was made before a former Anthropic researcher’s public warning set off this month’s argument about whether AI is moving too fast. Chief Executive Dario Amodei was the first of the three big AI executives to call for a slowdown; Sam Altman and Elon Musk echoed him. Prospective investors are expected to ask what a slower rollout means for a company whose backers expect more than $110 billion of annualized revenue by year-end. The company’s advisers argue existing models generate enough revenue that a slower frontier changes little. One venture investor told the Journal that going on the record about safety now may protect the company later, once shareholders can sue.
What the calendar looks like now
OpenAI has made clear it will not list until 2027 and is discussing financing at more than $1.2 trillion. So the fall supply is Anthropic, and Anthropic is now a November event. For a sense of scale: the S&P 500’s entire utilities sector is worth less than the $2 trillion this offering is expected to value one company at, and $100 billion of new stock is roughly what the whole U.S. IPO market raised in its best full year.
The reference case is trading in the stock table. SpaceX (SPCX), which listed in June at $2.1 trillion after its first day, closed Friday at $152.71, down 1.36% on the day and roughly a third below its $225.64 high. It was the third most active stock on the exchange. Elsewhere in this edition a couple who spent fourteen years accumulating SpaceX stock explain what they did with the windfall; the stock has fallen a third since they made the plan.
Holman Jenkins, on the opinion page, makes the case that the IPO cannot come soon enough — that OpenAI and Anthropic would be better run by a broad shareholding public than by “their highly ideologized and oddball founders,” and that the real risk to IPO investors is the industry being swallowed by the national-security state. That is an opinion, and this desk does not adopt it. The facts in the same column are worth having: these companies are about to be answerable to quarterly earnings, and quarterly earnings do not care about the frontier.
Our read
The September letter listed “the two big private AI labs preparing to list as early as this fall” as a reason for caution, because a great deal of new stock arriving into thin October books drains liquidity from everything else. The stock is now arriving into thin November books, the week of the vote, which the letter already treats as the far edge of the drop window. That is not better. It is the same weight, moved to the end of the plank.
For the portfolio, nothing changes: the AI build-out is owned through the platforms that spend the money — Microsoft (MSFT), Alphabet (GOOGL), Meta (META) — and the builders that get paid regardless, not through the frontier labs, which cannot be owned yet and which will be priced, when they can, by a market that has to absorb $100 billion in one bite. SpaceX at a third off its high is the honest preview of what a record IPO looks like a quarter later.