The Take — A share you can sell beats a fund you can't sue

In the space of about 30 hours this week, Anthropic promised the public 5 percent of itself and Elon Musk promised every American child one SpaceX share. Musk's version is worth roughly a seventh as much. It is also, structurally, the better offer — and I say that as someone who does not believe he will deliver it.

That is the argument. The fight over AI's wealth is not being settled on size. It is being settled on standing: whether the public ends up holding something it owns, or something a company administers on its behalf and can stop administering whenever the headlines improve.

What Anthropic actually committed

On October 18 Anthropic said it would place 5 percent of its shares in the Anthropic Public Benefit Fund, a wholly owned LLC, subject to stockholder approval. The board can add up to 2 more percentage points for every 50 percent increase in the company's value above a baseline it has not disclosed. Add the co-founders' standing pledge to give away 80 percent of their personal equity and an employee match that about half the staff uses, and roughly 20 percent of Anthropic is now described as dedicated to public benefit. At the $1.5 trillion valuation some investors are anchoring on, per Reuters, the opening 5 percent is about $75 billion and the combined 20 percent is about $300 billion. As our deep dive on the fund laid out, the number is real; what surrounds it is unwritten.

Here is the shape in one sentence: no independent trustees, no beneficiaries with legal standing, no published criteria for who gets money, a ratchet that grows at the board's discretion, and founders who keep 50.1 percent of voting power through a separate Founder LLC. The fund owns the economics of a fifth of Anthropic and cannot vote those shares against the company, because it is the company.

The contrast inside Anthropic's own structure is telling. The company already runs a five-member Long-Term Benefit Trust whose members own no equity but appoint most of the board, as the Wall Street Journal described. That is a governance device: it constrains the company from outside the shareholder register. The Public Benefit Fund is the opposite kind of object — an accounting destination, not a counterweight.

Why ownership beats stewardship

A share is a claim. It can be sold, priced daily, inherited, and — in the foundation-owned model that built Novo Nordisk, Bosch and the Wallenberg sphere — voted. In those structures the charitable vehicle holds a large stake in the operating company and sits in the boardroom with a legal duty to something other than the next quarter. Anthropic's fund takes the economics of that model and leaves the seat empty.

Musk's post, published hours after Anthropic's announcement, named the alternative in one line: "Next year, I will give one of my SpaceX shares to every kid in America. This will be the first of many." CNBC put the value at about $10.5 billion; across roughly 72 million American children, that is around $146 of stock each. It is a rounding error against $75 billion, and it is still the only proposal this week that ends with a person holding an asset rather than a promise about an asset. A family can sell a share. A family cannot sell a sentence in a press release about a wholly owned LLC whose distribution policy arrives "in the coming months."

That is the standard I would hold both pledges to, and it is the standard Anthropic's fund fails. Not because $75 billion is too little — it is an astonishing sum — but because every mechanism that turns the sum into money for actual people is controlled by the party that owes it. A pledge whose growth is discretionary and whose payouts are discretionary is a reputation instrument with a charitable caption. It is at its most valuable the week before an IPO, which is exactly when Anthropic is having it.

The case for Anthropic's version, fairly stated

Musk's pledge is, on inspection, barely a pledge. There is no entity, no custodian, no filing, no stated tax treatment, no confirmation that SpaceX's board or shareholders have approved anything, and no plan for the custodial accounts that 72 million minors would require. "The first of many" is unfalsifiable by construction. Handing a child a single illiquid share of a rocket company is also a poor answer to the specific harm in question: a 54-year-old logistics manager whose scheduling work agents absorbed does not need $146 of SPCX. She needs a wage, or a funded route to a different one.

Anthropic's fund, by contrast, can write cheques at a scale no share distribution can match. It can pay for retraining, for research, for direct cash to specific populations — Anthropic's own example. The need is not hypothetical: the share of junior technologists in Anthropic's own hiring fell to 9 percent as agents absorbed the work, a trend we dug into last week. It is backed by a company with $16.8 billion in third-quarter revenue, up from $11.5 billion the quarter before, per CNBC. And the personal half of the commitment is the credible half: founders pledging 80 percent of their own equity, and roughly half the staff donating into the same fund, is a costly signal that no amount of corporate accounting can fake. If your worry is that AI's gains will pool inside a handful of cap tables, this is the largest voluntary attempt anyone has made to un-pool them.

Why the take still holds

Because scale does not cure standing, and the counter-case proves it rather than answering it. Everything I just conceded about Musk's unseriousness is true, and none of it makes Anthropic's vehicle accountable. The strongest thing you can say about a company-controlled fund is that this company, today, means well. Anthropic's own risk factors — about 80 pages of them ahead of a listing valued in the trillions — exist because "this company, today" is not a durable guarantee.

There is also a pattern worth naming. Corporate giving of this kind tracks reputational pressure: it peaks when the public is watching and the float is being priced, and it thins when both pass. The founders' personal pledge survives that cycle because it is theirs. The fund survives it only as long as the board chooses to keep funding it. That is not cynicism about Anthropic specifically; it is what a structure with no external beneficiaries does over time, in every industry that has tried it.

And note what the disclosure schedule tells you. A $75 billion commitment announced two weeks before an IPO, with governance and distribution rules deferred to "the coming months," is asking the public to price the gift before reading its terms. Investors would not accept that from a supplier. Neither should the people the gift is meant for.

What would change my mind

Four things, all cheap to publish. Independent directors with a written duty to the fund's mission rather than to Anthropic. A disclosed valuation baseline and a distribution formula, released before the shares are marketed. A published voting policy — ideally a binding commitment not to vote the stake with management. And money out the door before the lockup expires, because grants funded in 2027 are evidence and grants funded in 2029 are a prospectus illustration.

If those land, the fund becomes an institution and most of my objection evaporates. If instead, by the middle of next year, Musk's giveaway is still a post with no custodian and no transfers, then direct ownership was a slogan, and the duller, slower, company-run vehicle wins by default — which would be a sad result for a debate that started with $75 billion on the table.

Should AI's public stake come with votes, or is a well-run fund enough? Write to us via our contact page.