Anthropic preps founder super-votes to lock in control
Anthropic is not just preparing a public float. It is rewriting who gets to tell the founders no once the stock starts trading.
The Information reported Tuesday that Anthropic has been preparing a new class of stock with extra voting power for CEO Dario Amodei and the other co-founders — the first time the Claude maker's leaders would hold that kind of control — to insulate them from outside shareholder pressure. Two people familiar with the matter described the work; the vote multiples themselves have not leaked, and the plans can still change. Anthropic has not commented. The problem the share class is meant to solve is already on the record: the co-founders hold a relatively small slice of the company compared with typical tech founders, so a conventional one-share-one-vote listing would hand real control to the institutions that buy the IPO.
That is the opposite of how most founder-led platforms went public. Google, Meta, and Snap used dual-class stock to keep voting power with people who no longer owned a majority of the economic interest. Anthropic is reaching for the same tool, with a twist that is unique to its setup. The company also plans to keep its existing body of non-shareholder trustees, who already hold a special class of stock that lets them elect a majority of the board. Super-votes for the founders plus a trustee-elected board majority is belt and suspenders. Public-market investors get the economics. The people who built Claude, and a group that does not even have to be shareholders, keep the steering wheel.
The timing is not subtle. We covered the investor meetings earlier this month — Anthropic courts investors ahead of a blockbuster fall IPO — and the listing could come as soon as late September, in what The Information called a candidate for the biggest IPO in history. A fall debut after a first profitable quarter is a financial story. Locking in founder control before that debut is a political one. Every frontier lab that has flirted with a public listing has had to answer the same question: what happens when quarterly earnings collide with a decision to slow a model, refuse a customer, or spend another year on safety work that does not show up in revenue. Anthropic's reported answer is to make sure the people who would take that hit cannot be voted out by the people who would rather not.
There is a cost. Dual-class structures are legal, common, and increasingly resented by index funds that have spent a decade arguing they mute accountability. Pairing them with a non-shareholder trustee majority will look, to some buyers, like a company that wants their money and not their voice. That may be the point. The OpenAI board crisis of 2023 taught every lab what happens when a governance experiment meets a CEO the staff wants back. Anthropic is trying to settle the argument on paper before the S-1 is public and the stock is a referendum. Whether that insulation survives the roadshow — or gets bargained down by buyers who already spent August asking whether frontier-lab growth outlives the compute buildout — is the real test.
What to watch: whether the S-1, if and when it drops, actually discloses the vote multiples — and whether the trustee majority is still intact by the time the shares price.
Should a public AI lab be allowed to sell you the economics and keep the votes? Tell us in the comments.
Sources: The Information · Reuters via CNA · Techmeme
