OpenAI files draft IPO prospectus May 22 — the $1 trillion question is whether the for-profit's profitability framing survives the S-1 disclosure regime
OpenAI filed a draft IPO prospectus on May 22, 2026, beginning the public-market process for what would be one of the largest tech IPOs in history. The Fortune cover question reframes it cleanly: the S-1 disclosure regime is going to force the company to publish the unit economics that the private-market narrative has been able to leave implicit since the 2023 Microsoft commercial agreement.
The disclosure questions the analyst community will read the S-1 for: (1) what fraction of revenue comes from each customer cohort (ChatGPT consumer, ChatGPT Enterprise, API direct, Azure-routed), (2) the cost-per-token at current scale and the trajectory through 2027-2030, (3) the actual compute commitment structure with Microsoft, Stargate, and the Cerebras $20B / 750MW deal, (4) the for-profit-vs-nonprofit governance bifurcation post-2024 restructure. Every one of these has been hand-waved in the private-market storyline.
The $1.25B/month SpaceX-Anthropic compute deal disclosed in the same week is the comparison-set reference. OpenAI's compute commitments are larger by an order of magnitude but more diffusely structured. The S-1 will be the first public reckoning of the actual numbers. If they support the $1T-class valuation, the IPO will set a new ceiling for AI-lab public-market pricing. If they don't, the entire frontier-lab valuation stack — including Anthropic's $380B post-money — gets repriced downward, regardless of those companies' own metrics.
Fortune — Big questions OpenAI trillion-dollar IPO filing may answer → · IG — SpaceX OpenAI Anthropic upcoming IPOs 2026 → · LetsDataScience — AI Firms Accelerate Toward Mega IPOs →