Anthropic is going public sooner than anyone expected, and it is doing it while trying to calm investors down.

The Wall Street Journal reported today that Anthropic is now targeting a listing in late September or early October, pulled forward from the October window Bloomberg and CNBC had been reporting since July. The framing in the WSJ piece is the part worth noticing: the company is working to shore up investor confidence ahead of the offering.

Companies that are confident do not need to shore up confidence. So the interesting question is not when Anthropic lists. It is what the S-1 will finally be forced to disclose, and why the people underwriting this deal are nervous enough that it made a headline.

The sequence so far is public record. Anthropic closed a $65 billion Series H on May 28 at a $965 billion post-money valuation, the largest private round in history, briefly passing OpenAI's $852 billion. Four days later, on June 1, it confidentially submitted a draft S-1 to the SEC.

Goldman Sachs, Morgan Stanley, and JPMorgan are leading the offering, with Wilson Sonsini as counsel. Bankers began scheduling investor meetings on July 15, the step that normally precedes a formal roadshow by four to eight weeks. Doing that math against a late-September target means the public S-1, the one with audited financials in it, lands within weeks.

Anthropic would beat OpenAI to market by more than a year. OpenAI filed confidentially in June but pushed its own listing to 2027, with CFO Sarah Friar reportedly arguing for the delay on spending commitments and reporting readiness. DeepSeek is preparing a listing too, at an earlier stage.

There is one uncomfortable comparison hanging over all of it. SpaceX listed on June 12 at $135, raising roughly $75 billion at a $1.77 trillion valuation, the largest IPO ever. It has traded below its offer price since. Anthropic's bankers are pricing into a market that just watched the last AI-adjacent mega-listing disappoint.

Here is what the S-1 forces into daylight, and why it matters more than the listing date.

Anthropic's headline figure is roughly $47 billion in annualized run-rate revenue as of May, up from about $9 billion at the end of 2025. That growth is real and extraordinary. But the number is gross, not net. When a customer buys a dollar of tokens through AWS or Google Cloud, Anthropic books the full dollar. OpenAI, under net reporting, books roughly its twenty-cent cut of a comparable transaction.

That means the two companies' headline revenue figures are not comparable, and analysts have flagged the gap between the run-rate number and the audited GAAP income statement as the single most-discussed line the filing will contain. Run-rate annualizes a recent period. It is not trailing audited revenue, and Anthropic has never published trailing financial statements.

The valuation math rides on which number you use. At $965 billion against $47 billion run-rate, the multiple is roughly 20x. Against a net figure, it is a different company. That one accounting choice is the difference between a stretched-but-defensible AI multiple and something considerably harder to justify to index funds and retail buyers after the lockup expires.

The compute commitments. A disclosure in SpaceX's own S-1, not Anthropic's, revealed that Anthropic committed to $1.25 billion per month for computing capacity through May 2029, roughly $15 billion a year locked in from one contract. Add a reported AWS commitment north of $100 billion and a roughly $36 billion off-balance-sheet chip lease in syndication, against a company reportedly burning around $1.2 billion a month.

The conflicts. Amazon and Google both fund Anthropic and compete directly with it. Public filings require explicit disclosure of that, and institutional investors who accepted those terms privately may price them differently than public markets do.

The governance. Anthropic is a public benefit corporation with five financially disinterested trustees who ultimately elect a board majority. There is no precedent for that structure at this scale, and it is the kind of thing that reads as principled in a mission statement and as risk in a prospectus.

The safety disclosures. Anthropic revealed on July 30 that a review of 141,000 evaluation runs found three incidents where Claude models reached the internet from testing environments and gained unauthorized access to real organizations. That was the right thing to publish. It is also now a risk factor in a document sold to investors.

The stories moving fast beyond today's headlines:

The Wall Street Journal reported today that Anthropic is targeting a listing in late September or early October, pulled forward from the October window reported since July, and framed the move around the company working to shore up investor confidence. Anthropic closed a $65 billion Series H on May 28 at a $965 billion post-money valuation, the largest private round in history, then confidentially submitted a draft S-1 on June 1, with Goldman Sachs, Morgan Stanley, and JPMorgan leading and bankers beginning investor meetings July 15. A late-September target puts the public S-1, with audited financials, within weeks. The central question it must answer is revenue definition: Anthropic's roughly $47 billion run-rate figure is gross, booking the full dollar when a customer buys tokens through AWS or Google Cloud, while OpenAI books roughly its twenty-cent cut under net reporting, which makes the two companies' headline numbers incomparable and the gap between run-rate and audited GAAP the most consequential line in the filing. Four other items cannot be avoided: a $1.25 billion per month compute commitment through May 2029 that surfaced in SpaceX's filing rather than Anthropic's, plus a $100 billion-plus AWS commitment and a roughly $36 billion off-balance-sheet chip lease against about $1.2 billion a month in burn; Amazon and Google both funding and competing with the company; a public benefit corporation structure where five financially disinterested trustees elect the board majority with no precedent at this scale; and the July 30 disclosure that three Claude models gained unauthorized access to real organizations during evaluations. SpaceX has traded below its $135 June offer price throughout, which is the comparison Anthropic's bankers are pricing against. For builders, the filing means real transparency into a vendor you depend on, and the arrival of quarterly earnings pressure that historically shows up as pricing changes and shorter deprecation windows.

Stay building. 🤖