Anthropic is structuring a dual-class share arrangement for CEO Dario Amodei and the company's six co-founders before it goes public, giving them enhanced voting rights that public buyers will not receive. The move creates a governance wall between the founding team and the institutional investors and retail buyers who will own the majority of the economic exposure after the offering.
The IPO is expected as early as late September and ranks as a candidate for one of the largest public offerings ever. Anyone buying shares in that offering gets no board control — a three-member trust and the supervoting founders govern the company.
The dual-class structure mirrors the mechanism used by Alphabet, Meta and Snap to let founders retain control after dilution. At Alphabet, Class B shares carry 10 votes each versus one vote for Class A, letting Larry Page and Sergey Brin maintain majority voting power long after their combined economic ownership fell below 50 percent. Anthropic's arrangement follows that template.
The strategic logic is direct. Anthropic has taken capital from Google, Amazon and other investors across multiple funding rounds. That outside capital has steadily diluted the founders' ownership stakes. Without a structural fix, the founders would have minority voting power the moment the IPO adds public shareholders on top of existing backers.
The supervoting shares address that dilution problem. They do not change economics — the founders' share of future profits and losses is determined by how many shares they own, not by the vote multiplier. What the structure preserves is decision-making authority: board composition, major strategic pivots, acquisitions and executive appointments.
For public investors, the arrangement is a known trade-off in AI infrastructure deals. The founders argue that long-horizon research and safety priorities require insulation from quarterly earnings pressure and activist shareholders. The counter-argument is that it removes a check on management — if the founders make a bad capital allocation decision, no outside shareholder vote can stop them.
Anthropic competes directly with OpenAI, Google DeepMind and Meta AI for frontier model leadership. Each competitor is either privately held, a subsidiary of a public company, or an open-source project — none faces the same public-market governance scrutiny Anthropic is about to invite. The supervoting structure is Anthropic's answer: tap public markets for capital without surrendering control.
A three-member trust arrangement adds another layer. According to the governance structure being prepared, a trust — not a simple shareholder vote — holds additional authority alongside the supervoting founders. The specific composition and powers are not yet public.
Frontier model training is capital-intensive at a scale that limits serious competitors to a handful of organizations. Anthropic's ability to tap public equity markets — rather than relying solely on strategic investors like Google and Amazon — provides a capital channel those competitors lack. The supervoting structure makes that channel available without surrendering the governance autonomy the founders believe they need to pursue safety-oriented AI development on a multi-year timeline.
What public investors price when they buy Anthropic shares is economic participation in the company's revenue and asset growth, with no meaningful governance voice. That is the same deal Meta offered in 2012 and Google offered in 2004. Both worked out for buyers on the economics. Anthropic is betting its IPO buyers will make the same calculation.
