Five Percent, Set Aside

Anthropic announced on October 18 that it plans to allocate 5% of its equity to a new Anthropic Public Benefit Fund, a stake worth roughly $48 billion at the company’s reported $965 billion valuation, according to CNBC and the company’s own announcement. The allocation requires shareholder approval, and the board would gain authority to add two percentage points to the fund for every 50% increase in the company’s value above an undisclosed baseline. The proposal arrives weeks before a public offering reportedly set to begin formal marketing the week of November 9.

The first thing to hold onto is what the number is not: $48 billion is equity value at a private valuation, not a cash donation, and it becomes real only in shares that are themselves worth what the market says on the day the market gets to say it. What the number definitely is, is a claim about structure. A company about to sell stock to the public is proposing to ring-fence a twentieth of itself for purposes that are not the shareholders’ — and asking those same shareholders to vote for it.

What a Public Benefit Corporation Keeps Back

Anthropic is already a public benefit corporation, a corporate form that writes a stated public purpose into the charter alongside the fiduciary duty to owners. The fund is that promise made financial: a dedicated equity pool intended to fund public-benefit initiatives, including donations responding to the social and economic effects of AI. The escalator clause — two more points per 50% rise in value — converts the company’s own success into a rising public share, automatically, without a new vote each time.

A charter says what a company is for. A fund says what it owes — and this one is denominated in the company’s own stock.

That design has a precedent logic and a criticism logic. The precedent: mission-locked structures have historically been used to keep founders’ control aligned with stated purpose through an IPO’s dilution. The criticism: equity set aside for “social and economic effects” of the company’s own product is a company deciding, in advance and in its own favor, how much harm is enough to pay for — and the baseline against which the escalator measures is deliberately undisclosed.

The Timing Is the Story

The announcement lands one week after Anthropic reported $16.8 billion in quarterly revenue and a second straight adjusted-profitable quarter to prospective investors, and days before formal IPO marketing is reported to begin. Prospectuses are documents of disclosed risk; this one now also carries a disclosed obligation, voted on by the people being asked to buy in. For underwriters it is a marketing line — a slice of a $965 billion valuation committed to the public benefit before anyone buys. For regulators reading the same week’s news, it is a company volunteering a number while the Federal Trade Commission is collecting fifty pages of questions about what its products actually do.

What Approval Means

Shareholders approving the fund are not being asked to believe a forecast; they are being asked to accept a smaller claim on the company’s upside in exchange for a structure that outlasts them. Whether $48 billion of equity ends up funding anything at the scale its valuation implies depends on the IPO’s final price, the baseline nobody outside the boardroom knows, and the votes. What is already settled is the shape: the largest AI IPO yet will carry a public-benefit ledger on its balance sheet from day one, sized in advance, growing by formula — an answer to a question the industry has so far answered mostly with blog posts.