On June 9, OpenAI announced that it had submitted a confidential S-1 filing to the SEC, the U.S. securities regulator—eight days after Anthropic, which had formalized the filing of its own draft document on June 1. OpenAI’s announcement is deliberately nonchalant: “We recently submitted a confidential S-1. We expect it to leak, so we’re announcing it ourselves. We haven’t decided on the timing.” Behind the tone lies a major shift: the two labs that defined the AI decade are, almost simultaneously, taking the path toward a public listing.
What a confidential S-1 is — and is not
A confidential filing allows a company to have its IPO registration statement reviewed by the SEC out of public view, and to iterate on the regulator’s comments without exposing its financials or risk factors. It is not a decision to go public: the timeline remains open, and neither OpenAI nor Anthropic has set one. But the rule is unambiguous on one point: before the offering can be marketed to investors, the document will have to be made public, financial statements included.
What the documents will have to reveal
That is the real stakes of this sequence. For years, the economics of AI labs have been readable only through unverifiable figures: private funding rounds, annualized revenue figures disclosed without audit, and valuations negotiated bilaterally. A public S-1 will close that blind spot. It will have to include audited financial statements—revenue, losses, cash position—the risk factors, including compute commitments contracted from cloud providers, the capital structure and governance rights, and executive compensation.
For companies whose capital needs are dictated by the cost of training and inference, the off-balance-sheet commitments section—how many billions of contracted compute, over what period, with whom—will likely be the most closely scrutinized part of the document.
Why now, and why eight days apart
The near-simultaneity of the filings is no coincidence of administrative timing. The two labs are funded in the same private markets, from investors whose capacity is not unlimited, for compute needs that run into the tens of billions. The public markets are the only capital reservoir at a larger scale. And the first one to list will set the valuation benchmark for the other—neither can allow the other to define that benchmark alone.
What to watch
Three things, in order. The publication of the documents themselves, which will come weeks before any listing and will provide the first audited look at the real economics of the labs. The treatment of the unusual governance structures at both companies, which institutional investors will have to assess. And the reaction of compute providers and strategic partners, whose agreements will, for the first time, appear as enforceable risk factors.