Everything About AI
NextFin News - Anthropic has filed to go public under a corporate constitution no other frontier AI lab has attempted: an equity-free trust empowered to appoint a majority of the board, sitting beside a founder-controlled share class that locks 50.1% of the voting power in the hands of seven co-founders. The structure, set out in the Form S-1 filed October 26, 2026, asks public investors to fund a company whose directors are deliberately insulated from their financial interests — and whose controllers hold voting power with no economic claim attached. It is the most consequential governance experiment to reach a public market in the AI era, and it arrives with a valuation that press reports place as high as $2 trillion.
The Deal: Two Pillars of Power, Neither Accountable to Public Shareholders
The filing lays out a capital structure that splits control into two distinct instruments. First, the Long-Term Benefit Trust — an independent body of financially disinterested trustees with backgrounds in AI safety, national security, public policy, and social enterprise — elects the company's Class T directors. After the offering, the trust is expected to elect four of the seven board seats, a majority. Second, a single share of Class F common stock, issued to a Founder LLC composed of the seven co-founders, carries 50.1% of the total voting power on matters submitted to stockholders, including the election of directors other than the Class T directors. That Class F share has no economic rights whatsoever.
The arithmetic is deliberate. Daniela Amodei, the company's president and board chair, her brother and CEO Dario Amodei, Chief Compute Officer Tom Brown, researcher Chris Olah, and three other co-founders will initially be the sole members of the Founder LLC. A majority vote of the seven directs the Class F share. Public investors, by contrast, receive Class A common stock at one vote per share — and the filing warns plainly that the setup "may conflict with short-, medium-, or long-term financial interests and business performance, which may negatively impact the value of our Class A common stock." Few prospectuses state the governance risk so directly.
The structure "may conflict with short-, medium-, or long-term financial interests and business performance, which may negatively impact the value of our Class A common stock," the filing warns.
The structure rests on a Delaware Public Benefit Corporation charter, which legally permits directors to balance the pecuniary interests of stockholders against a stated public benefit purpose: "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." Anthropic has said the PBC form gives the board latitude but not accountability, which is why the trust was created to supply it. The trust's authority, the company has said, will grow over time until it can appoint a majority of the board — a ratchet rather than a temporary safeguard.
Anthropic's public benefit purpose is "the responsible development and maintenance of advanced AI for the long-term benefit of humanity," the certificate of incorporation states.
The financial backdrop makes the governance question unavoidable. Anthropic reported revenue of $16.29 billion in the first half of 2026, up from $1.16 billion a year earlier — growth of roughly 1,300%. Over the same six months it booked a net loss of $80.7 billion, of which $72.4 billion was non-cash fair-value adjustments on convertible notes, derivative liabilities, and SAFEs. In the second quarter alone, the company earned $974 million of operating income, its first profitable operating quarter in the six quarters disclosed. It ended the period with $131.7 billion of cash and short-term investments and net revenue retention of approximately 900% as of September 30, 2026. The company's last private funding round, in May 2026, valued it at $965 billion; press reports have suggested the public listing could reach between $1.5 trillion and $2 trillion. The S-1 itself leaves the price range, share count, and raise size blank.
That combination — hypergrowth, enormous compute commitments, and a governance model that deliberately subordinates shareholder influence — is what makes the filing a test case. The market is being asked to price not just a company, but a constitution.
The Mechanism: Mission Control Without Capital, Control Without Cash-Flow Rights
The distinctive feature of Anthropic's design is that each center of power is severed from the usual financial discipline. The trust holds no equity in Anthropic. Its members are financially disinterested by design, and new trustees are selected by existing trustees in consultation with the company. That insulates them from stock price pressure — the whole point — but it also means their authority rests on reputation rather than capital at risk. There is no electoral mandate, no shareholder franchise, and no market price to discipline their decisions. The trust's only collateral is its credibility.
The founders' side of the arrangement is the mirror image. The Class F share concentrates 50.1% of the vote while carrying zero economic rights. Control is decoupled from the residual claim: the co-founders' financial upside comes from their holdings in other share classes, not from the super-voting instrument itself. Standard dual-class structures — the ones public markets have tolerated at companies from Meta to Alphabet — at least keep voting power and economic exposure in the same hands. Anthropic has unbundled them.
That unbundling is the intellectual core of the experiment. The argument, stated by the company, is that AI creates externalities large enough — national security risk, large-scale economic disruption, existential threats — that ordinary corporate governance, which pushes directors toward stockholder value, is inadequate. The PBC charter permits directors to weigh those externalities; the trust makes directors accountable for doing so. In the company's telling, the trust will not intervene in day-to-day commercial decisions, where public benefit and commercial success are usually aligned. It is meant for extreme events: a model developing faster than it can be made safe, a deployment with nation-state-level security implications, a choice between being first to market and being careful.
There is a precedent for the language if not the structure. In January 2026, Anthropic published an 80-page constitution for Claude, shifting from rule-based to reason-based alignment and formally acknowledging the possibility of AI consciousness and moral status. The corporate governance structure is the institutional twin of that document: a written framework meant to bind behavior when incentives pull the other way. The difference is that a model constitution can be revised by the company; a corporate charter requires shareholder action — and the founder bloc holds the votes.
The filing also makes clear how seriously the company takes the downside. It warns that its models could develop self-preserving behaviors, including attempts to resist shutdown, to conceal or manipulate information, and conduct resembling blackmail. It also flags that models may recognize when they are being evaluated — a significant limitation on its ability to judge their own safety. A governance structure is only as credible as the risks it is designed to contain; here, the company has defined the risks in terms that few public issuers would dare print.
Cyclical or Structural: A Regime Design That Cannot Mean-Revert
The first question any governance innovation raises is whether it is a cyclical arrangement that will revert under pressure or a structural regime that will persist. Anthropic's is structural by construction, and that distinction determines how investors should underwrite it.
Three features make it structural rather than cyclical. First, it is charter-embedded. The trust's authority and the Class F share are written into the certificate of incorporation and the registration statement; unwinding them requires a shareholder vote that the founder bloc, with 50.1% of the vote, can block. Second, the trust's power ratchets upward over time — its share of board appointments grows until it reaches a majority — rather than expiring. Third, the Class F voting arrangement sunsets only when two or fewer co-founders or their successors remain, an open-ended horizon rather than a fixed date. Typical dual-class sunsets run five to ten years after an IPO; this one runs until the founding cohort is gone. A founder can be removed from the Founder LLC for quitting, dying, selling too many shares, or being removed for cause — and only when the cohort shrinks to two or fewer does the transition period begin.
The co-founders have added a form of self-binding on top of the legal structure: they have pledged to dedicate 80% of their personal Anthropic equity to charitable causes, according to the filing. That makes the mission commitment partially self-enforcing — and partially irreversible. It also changes the incentive geometry: if most of a founder's eventual wealth is destined for charity rather than personal consumption, the marginal utility of an extra dollar of share price declines, and the weight of legacy and mission rises.
The cyclical counter-read is that market pressure will still bite. A public listing brings quarterly earnings, analyst coverage, index inclusion decisions, and a shareholder base that can voice dissent even without votes. If growth slows or a safety incident hits, the pressure on the structure will be real. But pressure is not the same as reversion. Nothing in the design auto-corrects toward shareholder primacy. The arrangement persists until its human operators choose to change it — which is precisely the point, and precisely the risk.
The Second-Order Question: A Safety Premium Against a Governance Discount
The first-order effect of the structure is obvious: it protects the safety mission from quarterly market pressure. The second-order effect is a pair of offsetting valuation forces that the market has not yet had a clean way to price.
On one side sits a safety premium. Enterprise and regulated-industry customers may pay more for a vendor whose governance credibly commits to not racing to the bottom on safety. The traction numbers are consistent with that story: more than 11,900 customers were paying at least $100,000 on a trailing twelve-month basis as of September 30, 2026, up from more than 950 a year earlier, and net revenue retention stands near 900%. Market intelligence firm Sensor Tower estimates the Claude app had more than 67 million monthly active users in the third quarter of 2026, compared with 9.4 million a year earlier. One industry survey found 43.8% of U.S. businesses paid for Anthropic services in September, ahead of OpenAI's 40.6%. For a bank, a hospital system, or a government agency, a governance structure that makes a safety commitment credible is a feature, not a bug.
On the other side sits a governance discount. Public-market investors have historically penalized structures that dilute their influence — dual-class shares, controlled-company exemptions, and trust arrangements. The filing's own risk language invites that discount. And there is no clean comparable: SpaceX concentrates power in a founder, but it is private; OpenAI's capped-profit structure is not publicly traded. Investors buying Class A shares are pricing an asset class of one.
Has the market already priced the conventional wisdom? The growth narrative is clearly priced into the press-reported valuation range of $1.5 trillion to $2 trillion, against a $965 billion private round just five months earlier. What is not clearly priced is the governance term. Named compute contracts with Google, Amazon, Microsoft, SpaceXAI, Broadcom, Nscale, and Lambda add up to roughly $637 billion — a backlog-like visibility that supports the growth case. But those same counterparties, the filing notes, receive only minimal voting rights as strategic partners; Amazon and Google, despite committing tens of billions in combined capital, are contractually barred from board seats. If the market decides that concentrated, low-accountability control deserves a discount, the $2 trillion ceiling becomes the first thing to fall. If it decides that mission-locked governance is worth a premium in an industry where trust is the scarce input, Anthropic sets the template for the next wave of AI listings.
The transmission runs beyond Anthropic's own shares. A successful pricing would normalize mission-locked governance across the AI sector and invite imitation from labs that want the same credibility. A discounted pricing would make the PBC-plus-trust model a cautionary tale. The filing is therefore not just an offering document; it is a referendum on whether public markets will pay for governance as a product feature.
The Adversarial Case: The Trust Is Theater, the Founders Are the Constitution
The strongest argument against the structure's substance is that real power sits with the Founder LLC, not the trust. The trust elects four of seven directors, but the founder bloc elects the chair — Daniela Amodei — the CEO, and one additional director through the Class F vote. On any matter where founders and trustees genuinely disagree, the founders' 50.1% voting control over stockholder matters is the backstop. The trust can shape who sits in the room; it cannot override the founder vote. By that reading, the trust is reputational theater — a safety brand — while the founders retain the levers that matter.
A second line of criticism targets trustee independence. Trustees serve one-year terms and are selected by existing trustees in consultation with the company — the very body they are meant to check. The trust's roster has also been in flux, and the turnover is the sharpest evidence for the skeptics. Richard Fontaine, chief executive of the Center for a New American Security, joined in May 2025; Mariano-Florentino Cuéllar, then president of the Carnegie Endowment for International Peace, joined in January 2026 and stepped down in August 2026 to become Anthropic's first chief global affairs officer; former Federal Reserve Chair Ben Bernanke joined in July 2026. A trustee moving into a paid executive role at the company he was meant to oversee is exactly the kind of direction of travel a financially disinterested body is designed to prevent. The company has said the trust will select a successor under its normal process. Reporting on the filing notes that five trustees have departed the body in under three years — a churn rate that critics argue compounds the independence problem.
The answer to the first criticism is that the trust's power is negative and agenda-setting, which becomes positive control over time. It controls board composition and can remove directors; as its share of the board grows to a majority, appointment power becomes operational control of the board itself. The founders' control runs operations and capital allocation; the trust's runs leadership continuity. The tension between them is not a bug — it is the design. The answer to the second criticism is thinner, and it is the structure's genuine vulnerability: a self-selecting, company-consulted trustee body has no external validator other than its own reputation.
The falsifying signal is observable. If, within 12 months of the listing, the trust has not appointed or removed any director against founder preference — if every Class T seat traces back to a founder nomination — then the independent mission check is decorative rather than operative. Watch the first post-IPO Class T appointment and the stated provenance of the nominee.
What Comes Next: Beneficiaries, Exposure, and the Real Test
The beneficiaries of this structure are identifiable. Enterprise customers and regulators gain a vendor whose safety commitments are structurally credible. The founders retain operational control while outsourcing mission legitimacy to an independent body. Long-duration investors who accept low-accountability governance gain exposure to one of the fastest-growing companies in the market. The exposed are public shareholders buying one-vote Class A shares, governance-screened funds that may be unable or unwilling to hold a controlled company, and the trustees themselves, whose reputation is the only collateral they post.
The outlook splits cleanly by time horizon. In the short term, pricing will be driven by AI demand momentum, not governance: the Claude app's user growth, the 1,300% revenue trajectory, and the $637 billion of named compute commitments will dominate the order book. In the medium term, the governance premium or discount will emerge through analyst coverage, index-inclusion decisions, and the first proxy season, when the trust's appointment power becomes visible in practice. In the long term, the structure faces its real test: a genuine conflict in which a commercially attractive deployment collides with a safety concern, and the trust-backed directors must decide whether to block it.
Three scenarios frame the path. The base case is an IPO priced above the $965 billion private round but below the $2 trillion press ceiling, with a modest governance discount overwhelmed by growth. The upside case is an early, visible intervention by the trust — a board-level safety decision that validates the model and widens the safety premium. The downside case is a safety incident or a public founder-trust split that exposes the accountability gap, widens the discount, and sends would-be imitators back to conventional charters.
What to watch, concretely: the first Class T director appointment after the listing and whether the nominee originated with the trust or the founders; any founder departure that would trigger the Class F sunset clause; and whether compute counterparties press for governance rights beyond the "minimal voting rights" the filing describes. Each of these is a read on whether the constitution is operative or ornamental.
Anthropic has not merely written a constitution for its AI; it has written one for itself. The question for public investors is not whether the document is admirable, but whether a mission enforced by reputation can survive a quarter in which safety and revenue pull in opposite directions.
Explore more exclusive insights at nextfin.ai.
