Deep Dive — Anthropic's Public Benefit Fund: 5% now, 20% pledged

If you searched "Anthropic public benefit fund" to find out what it actually is, the short version: on October 18 Anthropic said it will commit 5 percent of its shares to a new wholly owned entity, subject to stockholder approval, and its board can add another 2 percentage points for every 50 percent increase in the company's value above a baseline it has not disclosed. Combined with the co-founders' standing pledge to give away 80 percent of their equity and an employee donation match that about half of staff use, roughly 20 percent of Anthropic is now dedicated to public-benefit purposes — the equity that would otherwise have funded escalating executive awards, per Anthropic's announcement.

Put an IPO price on those percentages and the size becomes concrete. Reuters reported on October 15 that some current and prospective Anthropic investors are now anchoring on $1.5 trillion for the November listing, below the roughly $2 trillion they had expected weeks earlier. At $1.5 trillion, the fund's opening 5 percent is about $75 billion of stock, and the combined 20 percent Anthropic describes is about $300 billion. At the older $2 trillion expectation, the same 5 percent is about $100 billion. Kevin Roose of The New York Times noted that if the company lists at a $1 trillion valuation, the pledge could amount to "the largest commitment of philanthropic capital ever made," and called the trillion-dollar case probably conservative.

So the money is real and the scale is unprecedented. Everything about how it is controlled, when it can be spent, and what it means for shareholders is still being written.

What was actually committed

The vehicle is an LLC, wholly owned by Anthropic — deliberately not a 501(c)(3) nonprofit, a donor-advised fund or a foundation with independent trustees. Anthropic's stated reason is flexibility: it wants "adjustability between different forms of giving." The fund can make large donations to existing nonprofits, or "direct distributions to specific populations," a phrase that sent parts of the internet straight to universal basic income. Labor-market disruption is the example Anthropic itself offers, building on the $200 million it has already committed to studying AI's effect on work. The company concedes the reasoning: AI "may generate enormous economic value," but it "could also cause instability for many workers," and given the pace of progress, "bolder approaches may be needed."

Two constraints are in the announcement. None of the equity held by the fund will be used for lobbying or political giving in the United States or abroad, and the fund is structured so its stake grows automatically with the company — the ratchet is the design, not an afterthought. The board may add 2 points per 50 percent valuation increase. From a $1.5 trillion baseline, hitting $3 trillion would add 2 points, roughly another $60 billion, at the board's discretion.

What is missing from the announcement is almost everything an outside party would want to check. The valuation baseline is undisclosed. The governance of the fund is undisclosed — who runs it, whether it has independent directors, whether it can vote its shares, and what the criteria are for a distribution. Anthropic says it will share those details "in the coming months," along with a public-benefit report on progress against its mission.

What 5 percent is worth, and why the number moves

The fund's firepower is denominated in a share price nobody has set yet. Anthropic told investors at its October 14 investor day that third-quarter revenue reached $16.8 billion, up about 46 percent from $11.5 billion in the second quarter, with an adjusted profit — the first time the company has shown both growth and positive adjusted earnings to a pre-IPO audience, as CNBC reported. The listing is planned for November, after the US midterms, with share marketing possibly starting the week of November 9, and the figures have moved in both directions: earlier expectation of $2 trillion, then Beltway-sourced talk of $1.5 trillion once the numbers were on the table.

That gap is not a detail for the fund. It is the fund. A pledge expressed as a percentage of shares converts to dollars only at pricing, and the disclosed arithmetic sits inside a range where the same 5 percent is worth anywhere from about $36 billion at a $725 billion floor to $100 billion at $2 trillion — before anyone knows whether the shares, once traded, hold the number printed on the cover of the prospectus.

The second thing a percentage hides is liquidity. The fund receives equity, not cash. To pay grants it either sells shares into the market — which, at 5 percent of a company about to be the largest listing in history, is an overhang every institutional buyer will price — or it holds and funds grants by borrowing against the stake, or it waits. Anthropic has not said which. The employee side of the pledge has already generated real numbers: The Information reported that Anthropic's match of employee charitable gifts passed $660 million in the six months through March, a program that will scale with the IPO and dilute the same shareholders the fund dilutes.

Why now: the midterms, the layoffs, the listing

The timing is doing work that the structure cannot. Lise Buyer, a partner at the IPO advisory firm Class V Group, told CNBC the fund is both "mandatory to prepare for an initial public offering" and "beneficial from an image perspective" — it would read favorably if the money funds retraining or helps displaced workers find other jobs. Her next sentence is the honest one: "Of course, it is also to burnish their reputation at a time when AI is coming under significant scrutiny."

That scrutiny has numbers behind it. A Pew Research Center survey in August found 71 percent of US adults expect AI to mean fewer jobs over the next two years. IMF Managing Director Kristalina Georgieva said last week that roughly 40 percent of jobs worldwide will be augmented, downgraded or eliminated by AI, with the figure reaching 60 percent in advanced economies, Reuters reported. Anthropic's own public evidence points the same way: the share of junior technologists in its hiring fell to 9 percent as agents absorbed the work, a trend we unpacked in Anthropic cut junior hires to 9% — who trains the experts now?, and one that follows a path we traced earlier in AI is deleting the first rung, not the senior job: the entry-level tasks go first, and the ladder changes shape before the profession notices.

Anthropic is listing after an election in which AI's effect on work is a live campaign issue, as a public benefit corporation whose prospectus devotes about 80 pages of its 261-page main body to risk factors. A fund that can send money to displaced workers is a defensible answer to the question every regulator, senator and investor will ask in November. It is also, structurally, the company choosing the shape of its own remedy rather than having one imposed.

The governance catch

A wholly owned LLC is not a trust. It has no independent trustees, no external beneficiaries with standing, and no separate board of the kind that governs a foundation. Anthropic controls the entity that holds the shares, exercises the board discretion built into the ratchet, and defines the categories of people who might receive distributions. Economically the equity moves toward public purposes; in governance terms nothing that constrains the company has changed. The fund cannot act as a counterweight shareholder, because it is the company holding itself.

That is the shape of nearly every criticism attached to the announcement. David Sacks, the White House AI and crypto czar, put the structural point in one line: "You've put 20% of your equity in an LLC. Whether it serves the public good remains to be seen." Elon Musk's reply was that Anthropic should give the shares directly to the public. Vijar Kohli argued the pledge pre-emptively dilutes voting rights and distorts the shareholder base ahead of listing — a real concern for the investor buying into the float, because anything that moves economics away from public holders without a corresponding governance transfer changes what an IPO share is. Brad Gerstner of Altimeter called the commitment "an extraordinary gift" while flagging the wealth-concentration problem that comes with any vehicle holding tens of billions in a single stock, and said he hoped a meaningful portion of AI's wealth arrives as a direct dividend instead. Travis Kalanick's version was shorter: "Behold, the EA infinite money glitch."

Some of the reaction was simpler than that — a running reply was "thanks for the tax write off." The tax argument is a partial one: donating appreciated stock inside a donor-controlled structure is deductible in ways cash gifts are not, though the company is still giving up the economics, so the deduction is a discount on a real transfer, not a way of keeping the shares. The deeper version of the objection is about substitution. If the largest AI companies answer job displacement with philanthropy they design and control, then the redistribution of AI's gains is a corporate policy rather than a public one — voluntarily, at a scale and pace the company chooses, with no obligation to continue it after the founders move on.

It is worth remembering what the alternative architecture looks like, because it exists. In the older foundation-owned model — the Novo Nordisk Foundation and its holding company, Bosch's, the Wallenberg sphere — the charitable vehicle owns a large stake in the operating company and votes it, sitting in the boardroom as a permanent shareholder with a legal duty to something other than the next quarter. Anthropic's fund gets the economics without that seat. The same split shows up in Anthropic's own cap table: as Reuters reported on September 28, its IPO filing gives the seven co-founders 50.1 percent of voting power through a new Founder LLC. The public gets the upside of the shares. The founders keep the votes.

We laid out the structure question when the trustee plan was first priced in September, in Anthropic's $2 trillion IPO turns its trustee experiment into a public-market question: the market's job now is to decide what those safeguards are worth when the shares trade. The Public Benefit Fund adds the largest number yet to that test.

What to watch

Shareholder approval and the baseline. The 5 percent is contingent, and the valuation baseline that triggers the ratchet is the single number that determines whether the fund grows faster or slower than the company. Watch whether it lands near the IPO price — meaning each new milestone is cheap for the board to hit — or well above it, which would make the ratchet decorative.

The governance document. When Anthropic publishes the fund's structure, three details decide whether the criticism sticks: whether it can vote its shares, whether it has directors who are not company officers, and who sets the distribution categories. An LLC with an independent board and a published mandate is a different object from an LLC that is a line item inside the corporate entity.

The first money out. Whether the fund funds anything before the shares are liquid. A commitment that only converts to grants after the lockup expires is a promise; a hospital wing or a retraining program funded in 2027 is evidence.

The IPO itself. Marketing possibly from November 9, pricing after the midterms, and the overhang question every bookrunner will have to answer: does a 5 percent holder that intends to spend need to sell, and does the market get told when. Reuters' drift from $2 trillion to $1.5 trillion in expectations is the clearest sign that buyers are already doing their own arithmetic on what Anthropic is worth with a large, illiquid, charitably bound block of stock inside it.

The competitive response. OpenAI is raising $30 billion at a $1.4 trillion pre-money valuation, per Bloomberg — a bridge round in place of a listing after ruling out a 2026 IPO — and it is not carrying a comparable pledge. If the fund's reception is favorable, the incentive for every frontier lab to publish its own version of a public stake grows, and the industry standard shifts from disclosure to transfer. That would be the most durable effect of Sunday's announcement, and it would arrive without a single vote in Congress.

Is a company-controlled fund the right vehicle for AI's labour-market bill, or should the shares come with votes? Write to us via our contact page.