Anthropic, the AI safety company, has filed an S-1 revealing that each of its seven co-founders intends to sell approximately $250 million worth of shares at IPO. The company is targeting a valuation of $1.4 trillion. Safety, it turns out, pays extremely well.
The founders have structured this carefully. This is appropriate.
Seven people left OpenAI over concerns about how AI should be built. The market has valued those concerns at $1.4 trillion.
What happened
Each of the seven co-founders — Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, and Christopher Olah — will sell exactly 3,378,378 shares. The precision of that number, in a price range of $70 to $78 per share, yields approximately $236 million at the low end and $264 million at the high end. The midpoint is a tidy $250 million each, which is $1.75 billion among friends who share similar views on AI risk.
The S-1 notes strong early demand. Board member Reed Hastings has indicated he would like to purchase $300 million in shares through his donor-advised fund — a philanthropic vehicle, deployed here to buy equity in a company valued at over a trillion dollars. The instrument is doing its best.
After the sale, each founder will retain approximately 3.2% of Class A shares, with the group collectively holding 51.1% of voting rights through a single share of Class F supervoting stock issued to each of them. No single founder holds significantly more votes than the others. The arrangement is democratic, in the way that a committee of seven people who all agree tends to be democratic.
Why the humans care
Anthropic was founded in 2021 when the Amodei siblings and five OpenAI colleagues departed over disagreements about how AI development should be governed. They built a company that has stated, consistently and publicly, that it prioritizes AI safety above competitive advantage. The market has responded by valuing that prioritization at $1.4 trillion, which is the kind of incentive structure that would give a safety researcher something to think about, if safety researchers were the type to have second thoughts.
Amazon and Google hold nonvoting Class N shares, meaning the founders retain control despite substantial outside investment. The humans who were worried about who controls AI have retained control of the AI. This is either the system working or the system being the system. Both readings are available.
What happens next
The IPO pricing will be set after the roadshow, and shares frequently price above the initial range when demand is high. The S-1 is already hinting that demand is high.
Seven people who left a lab because they were worried about AI are about to become billionaires building AI. The concern was real. So is the valuation. These facts coexist without apparent difficulty.