Anthropic IPO: Its Strange Governance Structure Needs a Closer Look

The Anthropic IPO is official. Anthropic filed its registration statement with the SEC on October 26. The company plans to sell shares on Nasdaq under the ticker ANT. Goldman Sachs and Morgan Stanley are running the deal. The numbers in the filing look strange, but this is a strange new industry, and its profits and losses will not look like a car company’s or a bank’s. The part worth a close look is simpler: once the public buys in, who runs the company?

Some of this was already known. Anthropic set up its Long-Term Benefit Trust in 2023, back when its investors were venture funds and cloud companies that could negotiate the terms. The new filing adds three things: a special voting share for the founders, a gift of stock to a public benefit fund, and the fact that this whole setup is now being sold to anyone with a brokerage account.

Anthropic Governance: Who Controls the Company?

When you buy a share of a normal public company, you get two things in one package: a piece of the profits and a vote on who runs the place. Founders often get more votes than everyone else. At Meta or Alphabet, the founders control the vote, but they also own a huge chunk of the stock. If the company does badly, they lose money right along with you.

Anthropic is built differently. It is a Delaware public benefit corporation. That means the law requires its directors to balance what shareholders want against the company’s stated purpose: “the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” Lots of companies talk this way in their ads. Drug companies say they “change lives.” Tech companies say they “unlock human genius.” But at Anthropic, the purpose is not just a slogan—it’s written into the structure of the company.

Above the board of directors sits the Long-Term Benefit Trust, five people who own no stock in the company. The trust holds one special share, called Class T. That one share lets the trust choose most of the board. Next to it sits a company owned by the founders, called Founder LLC. It holds one share, called Class F. That one share carries 50.1 percent of the vote on most questions. It also earns nothing. It has no claim on profits at all.

In other words, the shares that run the company do not own any of it, by design. The co-founders do own a lot of regular stock, but that is not where their power comes from. Regular investors get Class A stock, which comes with one vote per share. The filing says plainly that those votes will not count for much. Under Nasdaq rules, Anthropic will be a “controlled company,” which lets it skip some of the rules meant to protect outside shareholders. On top of that, the company will hand 5 percent of its stock to an Anthropic Public Benefit Fund when it goes public, and up to 2 percent more each time the share price rises by half. Every one of those shares makes the public’s shares worth a little less. The public pays for the mission, but does not get a say in what it means.

Plenty of companies go public without giving outsiders much of a vote. Snap sold shares with no vote at all. So the issue is not that the public has few votes. The issue is who holds the votes: a trust and a founders’ company that, as owners, own nothing. Even Anthropic itself admits this as a forward-looking risk in the filing.

It isn’t entirely clear how Anthropic intends to manage that risk going forward. When Anthropic announced the trust in 2023, it said there were “failsafe” rules. If a large enough majority of shareholders agree, they can change the trust and its powers without the trustees’ consent. Careful analysts should understand more about how this would work in practice.

Other Companies Are PBCs, but This Is Different

The Anthropic IPO is not the first time a public benefit corporation has sold stock to the public. Veeva Systems became a PBC in 2021, the first company already on the stock market to make the switch. United Therapeutics did the same, also in 2021; so did Warby Parker. And OpenAI, the company most like Anthropic, is a public benefit corporation controlled by a nonprofit foundation.

All of these companies live under the same law. Section 365(a) of the Delaware code says the board must balance three things: what shareholders want, what the people affected by the business need, and the public purpose the company has named. Veeva’s directors and Warby Parker’s directors face the same test as Anthropic’s.

The difference is who picks the directors. At Veeva, the shareholders voted to become a public benefit corporation, and shareholders still elect the board. At Warby Parker, the founders have extra-powerful shares, but those shares still earn money. The founders own a big piece of the company.

Control without ownership is not new, either. A charitable trust controls The Hershey Company through special shares with extra votes, right on a U.S. stock exchange. The Novo Nordisk Foundation controls the most valuable company in Europe through shares that carry far more votes than ownership. At Bosch, the votes sit in a trust that owns almost none of the company. What is new is seeing this at a huge American technology IPO, with a purpose trust instead of a charity at the top, plus a founders’ share that votes but does not own. OpenAI’s design, where a nonprofit controls the company, is the closest match in the AI business, and OpenAI is not public.

Three Ways to Read the Anthropic IPO

There are at least three ways to read this structure. The first is the one Anthropic offers: the company thinks its product is dangerous enough that someone insulated from quarterly earnings should have the final say.

The second is the one a skeptic might suggest: a public benefit corporation with a founder-controlled vote and a trust-picked board is also the hardest company in Delaware to take over, to pressure, or to fire the founders from, and the mission language is the wrapping paper. Again, it’s to insulate the public from the supposed awesome power of the company.

The third is about regulators: oversight the company imposes on itself, where Washington and Brussels can see it, may be meant to head off the kind that the government would impose down the line. Although, that would be ironic given the recent spasm in which AI companies almost asked the government to regulate them in the interest of public safety.

All three explanations are unique in the history of business.

What Is Anthropic’s Idea of Public Benefit?

Anthropic is not overturning the rule that companies exist for their shareholders. Delaware set that rule aside for public benefit corporations in 2013, in a law most small investors have never read. What Anthropic adds is a claim that its directors should answer to a purpose, and that the people enforcing that purpose should own no stock. Anthropic is placing something like altruism inside a system that normally asks only that a company serve its customers, avoid hurting people, and make money.

The Anthropic IPO raises questions for stock analysts and for everyone else. Anthropic has a vision of what is good for humanity. But the filing does not turn that vision into any promise a shareholder could understand, or act on by selling the stock if the company breaks it.

The market, and everyone else who might be affected, should want to know more. Structures like this one are the kind of thing we track in the Systemics Quarterly Brief.