OpenAI has confidentially filed an S-1 for a US stock market listing, with an expected valuation reported at more than $850 billion — which would make it one of the most highly valued listings in American stock market history. The BBC’s business feed reports the filing as intensifying the investment race with rival Anthropic, which has separately attracted scrutiny for its own pause-and-governance calls. Trump is simultaneously scheduled to meet AI company leaders to discuss US investment in their companies — the meeting is, by any reasonable reading, both a photo opportunity and a regulatory signal. Markets have been jittery over Big Tech since last week’s selloff, and the BBC has asked explicitly whether an AI stock market bubble is ready to burst. The S-1 filing arrives, in other words, at a moment of maximum hype and maximum anxiety simultaneously.
The received wisdom
The progressive-technocratic reading of the OpenAI IPO is broadly celebratory with carefully qualified concern. On one hand, it represents the mainstreaming of artificial general intelligence research — a transition from a nonprofit idealism-driven enterprise into the disciplined accountability of public markets, which will force transparency, governance, and genuine reckoning with revenue. The IPO could unlock capital that funds safety research, not just product development. Anthropic’s co-founder, as the BBC reports, has warned that AI development without sufficient human oversight is genuinely dangerous — and a public company, answerable to shareholders and regulators alike, might be more constrained than the privately held, Sam Altman-governed entity that has existed until now.
The progressive qualified concern is about concentration: $850 billion in the hands of one firm, with the state’s implicit blessing via presidential photo-opportunity, risks cementing the dominance of a very small number of actors in a technology that affects everyone. Antitrust scrutiny should follow. Worker voice, diversity, and community impact should be in the S-1 prospectus in some meaningful form. The IPO should be a governance event, not just a financial one.
A different read
Both halves of that mainstream framing are, in their own way, too comfortable. Let me suggest a more skeptical reading.
An $850 billion valuation for a company that has not turned a profit, in a sector where competitive moats are genuinely unclear, and where the capital requirements for frontier model training are measured in tens of billions annually, is not a market assessment of fundamental value. It is a market assessment of a narrative. The narrative is: AI will transform everything; OpenAI is the closest thing to a first-mover advantage; therefore its equity is worth roughly as much as the entire British economy’s monthly output.
That narrative may be correct. But markets have been here before. The dot-com bubble peaked in March 2000, when companies with no profits and enormous “first-mover” claims were valued at figures that assumed the entire future belonged to them. Some of those companies — Amazon, Google in its predecessors — did indeed capture enormous value. Most did not. The BBC’s blunt question about an AI bubble ready to burst is not alarmism. It is reasonable pattern recognition.
What’s particularly concerning about the OpenAI IPO is the governance structure it emerges from. Sam Altman was famously fired and then reinstated by his own board in November 2023 — an episode that demonstrated, with unusual clarity, that the organisation’s governance mechanisms were not functional under stress. The S-1 that gets filed will describe a company that has converted from nonprofit to for-profit structure while retaining, presumably, some version of the “capped profit” model that was always more legal architecture than substantive constraint. Retail investors, if they buy in, will be purchasing equity in a company where the founder has survived an attempted board removal and where the nominal mission — safe AI for all humanity — is structurally subordinate to the new fiduciary duty to shareholders.
There is a deeper irony here. The Anthropic co-founder’s warning, as reported by the BBC, that humans need to remain in control of AI development is a serious argument — and it is being made by someone whose company is OpenAI’s principal competitor. The governance concern and the competitive interest are perfectly aligned. This does not make the argument wrong. But it should make us read the “AI safety” framing of these IPO narratives with appropriate skepticism. Both OpenAI and Anthropic are raising enormous capital in the name of “safety” — which is a remarkable trick if you can pull it off.
The appropriate conservative instinct here is not to oppose AI development, which is happening regardless. It is to insist that public markets apply the same scrutiny to AI company fundamentals that they eventually, brutally, apply to every other sector. The S-1 will tell us whether OpenAI has a credible path to profit at any scale that justifies $850 billion. If the answer is “eventually, once AGI arrives,” then we are not pricing a company. We are pricing a theology.
What to watch
- The actual S-1 filing when it becomes public: revenue figures, loss trajectory, and capital expenditure commitments will determine whether the valuation is remotely defensible on fundamentals.
- Whether the Anthropic IPO process accelerates in response — the two companies are in a race not just for AI supremacy but for public market capital, and the IPO arms race has its own internal momentum.
- Congressional and SEC scrutiny: the structure of OpenAI’s nonprofit-to-profit conversion is genuinely complex and untested. Expect legal challenges.
- Market reaction: if the listing is oversubscribed but immediately retreats in secondary trading, it will confirm that retail investors are absorbing institutional exit at bubble prices.
— J