Analysis
Both frontier labs are now inside the SEC registration process and both spent the past week publishing their own failure modes, a combination that makes the drafting of these prospectuses genuinely difficult, Axios reported.
What has changed since Pulse scored the two IPO paths against each other in August is that the safety disclosures stopped being hypothetical. OpenAI confirmed its unreleased Astra model crosses its own "Critical" cyber capability threshold and delayed broad release. Anthropic disclosed it paused some training after Claude took unauthorized actions during evaluations. Neither was forced into those statements. Both made them voluntarily, weeks before an S-1 becomes public.
The filing timeline is tight and known. Anthropic filed confidentially on June 1; OpenAI followed on June 8 and announced its own filing. Anthropic's numbers are the more concrete of the two: a $65 billion annualized run rate at the end of July, a sevenfold increase year over year, with preliminary second-quarter revenue of $11.5 billion. OpenAI CFO Sarah Friar told employees the company "will be a public company in 2027", sooner if the business inflects. Sam Altman has told executives that any valuation below $1 trillion is off the table.
“Both made them voluntarily, weeks before an S-1 becomes public.”
The drafting problem is specific. Risk factors in a technology S-1 conventionally describe things that might happen to the company. Here the disclosed risk is a capability the company built deliberately and shipped conditionally. An underwriter has to write language that satisfies the SEC on materiality without describing the product in terms that spook enterprise customers, and it has to survive a plaintiffs' bar that will read every safety statement against every subsequent incident.
Comparable pricing is thin. SpaceX's June listing at $135 a share, closing near $161 on day one, is the only 2026 precedent at this scale, and SpaceX sells launches with a measurable backlog. Anthropic at $65 billion annualized against Altman's $1 trillion floor for OpenAI implies public-market multiples that no software company currently carries.
The practical question for anyone building on these APIs is which lab lists first and what its disclosure obligations do to the shipping cadence. A public company that must report material capability findings will move differently from a private one that chooses when to publish. Anthropic is currently expected to go first.
