The bankers and lawyers running Anthropic’s initial public offering had told investors to expect the company’s registration documents as soon as this week. The calendar has moved. Anthropic now plans to file its prospectus in late September and to begin its roadshow no earlier than mid-October, according to people familiar with the matter. The company still hopes to list its shares before November’s midterm elections, the people said.
The shift, reported by Reuters on Friday, is measured in weeks rather than months. People close to the deal said the timing was never as fixed as the market assumed, and the extra days give auditors room to finish their work and banks time to settle a syndicate that has drawn nearly every major Wall Street firm.
That syndicate has been one of the offering’s quiet dramas. Morgan Stanley is expected to serve as lead underwriter, according to people familiar with the matter, with Goldman Sachs in line to act as the stabilization agent that would support the stock in its first days of trading. JPMorgan, Citigroup and Barclays are expected to take senior roles. The competition for seats reflects the prize: an artificial-intelligence company whose name investors have chased since it first signaled an intention to go public.
Expectations around the deal have run in two directions at once. Investors and analysts have at various points floated valuations above $2 trillion, a figure that would put Anthropic among the largest companies ever to list in the United States. The company’s roughly $65 billion in annualized revenue, while far beyond what most software firms achieve at any age, sits below what the most aggressive bulls imagined. The distance between those two numbers is the subject the roadshow will have to settle.
Anthropic was founded in 2021 by Dario Amodei and his sister Daniela, two former OpenAI researchers, and builds the Claude family of models sold to businesses through direct contracts and cloud providers. Amazon and Google are among its backers. A revenue run rate of roughly $65 billion would make it one of the largest private software companies in the world, and the listing is widely seen as a test of how far public investors will go in valuing frontier labs whose spending on data centers and researchers runs far ahead of their income.
The timing adds to the pressure. OpenAI released its newest flagship model, GPT-6 Astra, only days before Reuters reported the schedule change, and the rivalry between the two companies has moved from research papers into product releases and, soon, investor pitches. Anthropic will market itself to the same buyers, the same analysts and the same fund managers who are trying to decide which of the big AI builders deserves a durable premium.
Public markets in 2026 have been warm to artificial-intelligence stocks but picky about price, analysts said. Companies with revenue to show have drawn strong demand; those asking investors to pay for decades of future profits in a single valuation have found thinner receptions. An offering of Anthropic’s size will test both instincts at once, because the company has real sales and real costs, and the two are far apart.
The schedule leaves little margin for error. A filing in late September followed by a roadshow in mid-October compresses the marketing period before the election, and people close to the deal said any further delay could push the listing into a post-election calendar that bankers describe as less hospitable. IPO windows can close as quickly as they open, and a company this large has few chances to pick its moment.
Fund managers will bring their own questions to the roadshow. How much of the revenue comes from a small number of large customers, and how durable are those contracts? How does the economics of selling models directly compare with selling through cloud partners who also compete with Anthropic? Banks will use the meetings to test a price range, and the company will have to balance a debut that rewards its earliest investors against a first-day pop that would suggest it left money on the table.
The mechanics reflect the size of the deal. The choice of a separate stabilization agent suggests underwriters expect the stock to attract enough volume, and enough volatility, to justify the insurance; that firm would step in to buy shares if the price wobbled in the first sessions, a tool most offerings never need to use. The roster of banks alone tells investors how much fee revenue the deal promises.
The election calendar adds its own texture. American markets have handled listings around votes before, but underwriters prefer clean windows, and a company presenting itself as the future of computing does not want its debut tangled in the noise of a campaign season. Targeting the weeks before the election gives Anthropic a defined moment and a reason to hold to it, though it also means the company has no slack if anything else goes wrong.
For all the machinery arrayed around the offering, the number that matters most has not been set. Bankers say the valuation will be decided in the final days, when the order book is full and investors have put money behind their enthusiasm. The roadshow table, not the spreadsheets circulated beforehand, is where Anthropic’s price will be found.
