Anthropic Wants to Put 5 Percent of Its Shares in a Fund That Grows With the Stock. It Has Not Said Who Decides How It Is Spent.
On Sunday, October 18, 2026, Anthropic published a short post titled "Investing in the public benefit." It says the company is committing a portion of its shares to a new entity, the Anthropic Public Benefit Fund. Subject to stockholder approval, 5 percent of Anthropic's shares go in now.
The second sentence is the interesting one. The board may add another 2 percent "for every 50% increase in the company's value above a set baseline." The fund is built to get bigger as Anthropic does.
That is a lot of money on paper. Anthropic's last priced round, in May, valued it at $965 billion after the money. Five percent of that is about $48 billion. Anthropic did not put a dollar figure on the stake, so that number is our arithmetic, not its disclosure.
What Anthropic Actually Committed
The post runs about 440 words. Here is what it pins down, and what it leaves for later.
| Term | What the post says | Still open |
|---|---|---|
| Initial stake | 5% of shares, committed today | Needs stockholder approval |
| Escalator | +2% per 50% rise in value, at the board's discretion | The baseline is not disclosed |
| Legal form | Wholly-owned LLC, not a 501(c)(3) | Who sits on it and who decides |
| Uses | Large donations to nonprofits, direct distributions to specific populations; labor market impacts as one example | Any payout rule or timetable |
| Exclusions | No lobbying or political giving with its Anthropic equity, in the US or abroad | None stated |
| Next disclosure | Structure, governance and valuation milestones "in the coming months" | No date |
Anthropic explains the LLC choice plainly: it wants "adjustability between different forms of giving," and it names direct distributions to specific populations as one possible use. Anthropic has already committed $200 million to studying AI's effect on jobs, and it now says "bolder approaches may be needed."
An Escalator Borrowed From Executive Pay
Anthropic makes the comparison itself. "It is common practice for tech companies to offer escalating compensation awards to their executives," the post says. "We are instead directing that equity to the public benefit."
The best-known recent example is Tesla. On Thursday, November 6, 2025, Tesla shareholders approved, with more than 75 percent support, a package that pays Elon Musk in 12 tranches, the first at a $2 trillion market value and the last at $8.5 trillion. Anthropic's version points the same mechanism at a fund instead of a person.
How fast it climbs depends on a sentence Anthropic has not finished. "Every 50% increase" can be read two ways: steps of half the baseline, or 50 percent on top of the last step. The gap gets wide quickly.
| Maximum fund stake | Value needed if steps are 0.5x baseline | Value needed if steps compound |
|---|---|---|
| 5% | Baseline | Baseline |
| 7% | 1.5x | 1.5x |
| 9% | 2.0x | 2.25x |
| 11% | 2.5x | 3.375x |
| 13% | 3.0x | 5.06x |
TensorFeed calculation. Every step after the first 5 percent is at the board's discretion, so these are ceilings, not commitments.
The baseline matters even more than the step. Investors at Anthropic's Wednesday, October 14 investor day discussed an IPO valuation of about $1.5 trillion, down from roughly $2 trillion, according to Reuters as reported by Investing.com. If the baseline were the $965 billion May mark, a $1.5 trillion listing would already clear the first step. If the baseline is set at the IPO price, it would not. Anthropic has not said which.
Next to the OpenAI Foundation
The obvious comparison is OpenAI's recapitalization, completed on Tuesday, October 28, 2025. The two arrangements are built for different jobs.
| Anthropic Public Benefit Fund | OpenAI Foundation | |
|---|---|---|
| Stake | 5%, pending approval | About 26% |
| Value at announcement | About $48B at $965B (our math) | Around $130B, as reported |
| Upside mechanism | +2% per 50% gain, board discretion | More equity if OpenAI Group's value rises more than tenfold after 15 years |
| Who controls whom | Anthropic owns the fund | The foundation controls the company and appoints its board |
The last row is the real difference. OpenAI's foundation sits on top of the company. Anthropic's fund sits underneath it, as a wholly-owned subsidiary. Anthropic's oversight of its own mission runs through a different body, the Long-Term Benefit Trust, which Anthropic said in 2023 would elect a majority of its board within four years. The fund is a wallet, not a governor.
The 20 Percent Figure
Anthropic also says that with this fund, "roughly 20% of Anthropic shares will be dedicated to the public good." The rest of that comes from two older commitments: the co-founders' pledge to donate 80 percent of their equity, and an employee equity match program in which about half of employees take part.
Subtract the fund and the older pledges come to roughly 15 percent. Those are pledges by individuals plus a company match, not shares the company holds today, and Anthropic gave no breakdown. The 5 percent is the only piece that is new, and the only piece Anthropic itself will own.
Caveats
Nothing has been transferred yet. The 5 percent needs a stockholder vote, and every step above it needs the board. The $48 billion and the escalator table are our numbers built on Anthropic's percentages, not figures Anthropic published. The $1.5 trillion is a level investors discussed, per Reuters, not a price. And the post does not say whether the fund's shares come from new issuance or existing stock, which decides who is diluted.
Our Take
This is a bigger commitment than it looks in a 440-word post. Five percent of a company whose investors are discussing a $1.5 trillion listing would be a $75 billion pool before the escalator adds anything. The escalator is clever, and we like that it borrows the logic of a CEO mega-grant and hands it to the public instead. The ban on using the equity for lobbying, written in the first post rather than added later, is the right instinct about two weeks before the US midterms.
But a fund that Anthropic wholly owns, governed by rules Anthropic has not written yet, is a promise with the hard parts deferred. Three numbers decide whether it is a real counterweight or a line in an S-1: the baseline, the share of the board that is independent, and the minimum the fund must pay out each year. A US private foundation generally has to pay out about 5 percent of its investment assets a year under the tax code. An LLC like this one does not. We would want Anthropic to adopt a floor at least that high on its own.
Timing is the other tell. This fall already holds a founder voting class proposal reported in September, the stockholder vote this fund needs, and an IPO window reported for November. Locking in the fund before the stock trades is smart. Publishing its governance before the stock trades would be smarter.
Three signposts for the next 60 days: whether Anthropic names the baseline valuation and says whether the steps compound; whether the fund's governance includes people who do not work for Anthropic and a minimum annual payout; and whether the stockholder vote and the fund's terms appear in a public S-1 before the listing.
Sources: Anthropic (Investing in the public benefit), TechCrunch, CNBC, The Independent via Yahoo Finance, Investing.com (citing Reuters), NBC News (OpenAI restructuring), Axios via Benton Institute, OpenAI (our structure), CBT News (Tesla vote), IRS (private foundation payouts), Anthropic (Long-Term Benefit Trust).