Both leading AI labs have now confirmed they have taken the first formal step toward a public listing, and neither has committed to when. Anthropic announced a confidential draft S-1 on June 1, 2026, and OpenAI announced its own a week later on June 8, saying timing remained undecided. The genuinely novel part of the story is not price. It is that Anthropic is reported to be preparing a governance structure that would give public investors the economics of the company without proportionate control over it.
Key facts
- Anthropic confirmed a confidential draft S-1 on June 1, 2026; OpenAI confirmed its own on June 8, 2026.
- On August 19, CNBC reported that OpenAI CFO Sarah Friar told employees the company would be public in 2027, or sooner if the business keeps inflecting. This is reporter-sourced, not company-published.
- On August 18, Reuters, citing The Information, reported Anthropic preparing a founder supervoting share class with a listing possibly as soon as late September. Also reported, not filed.
- Anthropic is a public benefit corporation whose board is elected partly through the Long-Term Benefit Trust.
It is worth separating what is confirmed from what is reported, because the two are being quoted interchangeably and they carry very different weight. Both companies have publicly confirmed their own filings; those are company statements. The 2027 timing for OpenAI comes from a CNBC report of internal remarks. The Anthropic supervoting structure and the late-September window come from The Information via Reuters. Reported details from good outlets are worth knowing and are not the same as a document you can read.
A confidential draft S-1 is a registration statement submitted privately to the Securities and Exchange Commission. It starts the regulatory review clock without publishing the financials. Companies file them to keep optionality: you can be nearly ready to list and still walk away, and nobody outside sees your margins in the meantime.
The governance question is where this gets genuinely unusual. Anthropic is a public benefit corporation, a structure that legally permits directors to weigh a stated public mission alongside shareholder returns. On top of that sits the Long-Term Benefit Trust, which holds power over a portion of board seats and exists specifically to keep safety commitments from being overridden by investors wanting faster commercialization.
Now add a reported founder supervoting class -- shares carrying many votes each, held by the founders. Dual-class structures are common in technology listings; Google, Meta and Snap all use versions of them. Stacking one on top of a benefit-corporation charter and an independent trust is not common, and it would mean public shareholders buy exposure to the returns while holding very little say over the decisions.
An analogy: buying into a restaurant where you get a share of the profits but the chef, the landlord and a trust all have to agree before the menu changes, and none of them answer to you.
Whether that is admirable or alarming depends entirely on your priors. The case for it is that safety commitments which evaporate under quarterly earnings pressure were never commitments. The case against is that "insulated from accountability" and "insulated from short-termism" describe the same arrangement, and only time distinguishes them.
One correction is worth making because the data is being widely misused. Ramp's AI Index is circulating as evidence for which lab is winning. Ramp's own methodology page states that the index derives from corporate card and invoice payments across more than $100 billion in annual spend and over 50,000 US businesses. That is a measure of adoption behaviour in a tech-forward sample of American companies. It is not revenue, it does not capture enterprise contracts negotiated outside card and invoice rails, and it should not be quoted as either lab's growth rate.
The honest caveat is that until one of these companies files publicly, everything about their economics is inference. Nobody outside has audited revenue, gross margin, the true cost of compute obligations, or how much of the reported growth is durable. The first public S-1 will be the first time anyone gets a look at the actual numbers behind two years of valuation reporting, and it is entirely possible that document reframes the whole conversation in a direction nobody currently expects.
Related reading on this site: an AGI thesis fund fell 67 percent and took a market maker with it and Stripe is buying the company that keeps score on every model.
Originally published on Ground Truth, where every claim is checked against the primary source.
Top comments (0)