Seven founders pledged 80% of their equity. The structure means it costs them nothing they care about. Anthropic cofounders · Class F share · Founder LLC
- What they did
- Each cofounder pledged 80% of personal Anthropic equity to philanthropy. Separately, a single share of Class F stock, held by an LLC owned by the seven, carries 50.1% of shareholder voting power and no economic rights. A cofounder stays in the LLC while holding at least 2% of the Class A shares they held at IPO.
- The document
- Form S-1, 26 October 2026. The filing calls the pledges “voluntary, revocable expressions of each cofounder’s philanthropic intent” that “do not constitute legally binding obligations.” Timing, amounts and recipients are undecided.
Consensus read
An extraordinary giving commitment — 80% of personal equity, materially ahead of the Giving Pledge’s 50%, made by seven people at once and before the money is even liquid. Add the Public Benefit Fund’s 846.8 million shares (5% of fully diluted) and this is among the largest founder-philanthropy structures ever committed at an IPO.
Steelmanned: an early, public, specific commitment does raise the reputational cost of reneging, even when it is legally revocable. That is a real constraint, and it is the constraint most large pledges actually run on.
Contrarian read
Under a conventional dual-class structure, giving away 80% of your stock dilutes your votes proportionately. That is what makes founder philanthropy a genuine trade.
Here it does not. The vote is attached to one share with no economic rights, held through an LLC whose membership test is 2% of IPO-date Class A — a floor 80% giving does not breach. The filing states the effect in its own risk factors: as cofounders give stock away, their economic interest shrinks while their voting control stays the same.
This is the first very large founder-giving commitment that can be honoured in full without transferring a single basis point of control.
Evidence that discriminates
The consensus read treats generosity and control as in tension. These facts are inconsistent with that tension existing:
- The Class F share decouples them by construction: one share, 50.1% of the vote, zero economics. Votes cannot follow donated shares out of the door, because the votes were never in those shares.
- The 2% membership floor survives the pledge. Honouring 80% leaves 20% — ten times the threshold at which a cofounder exits the Founder LLC. The pledge was sized well clear of the only line that matters.
- The S-1 acknowledges the asymmetry in its own risk disclosure — shrinking economics, constant control. The company is not hiding this; the coverage simply reports the pledge without it.
- It has already happened once, in the table. Dario Amodei’s 4.6m donor-advised-fund shares: ownership gone, votes retained under proxy agreements. That is the mechanism running live, in a disclosed number, before the IPO.
Inference, labelled: that the structure was designed so the pledge would be costless in control terms is my inference. What the filing establishes is that this is the effect.
What would falsify me
If the final charter ties the Class F share to a minimum economic ownership threshold that honouring 80% would breach — a sunset on the control share triggered by the giving itself — then giving really would cost control and this reading is wrong.
Second, cleaner test: if within the first year a cofounder executes an irrevocable transfer that also surrenders the vote — a charitable remainder trust, a binding pledge agreement, shares given without a retained proxy — rather than the donor-advised-fund-plus-proxy pattern already visible in the table, the claim weakens sharply.
Pattern Benjamin Mann, whom the filing says is “often referred to as a cofounder,” is not a cofounder for S-1 purposes: not in the Founder LLC, not in the ownership table. Whatever else the Class F structure is, membership in it is narrower than the public understanding of who founded the company.