Who Actually Controls Anthropic After Its IPO? The Five Share Classes, Decision by Decision

The prospectus offers the familiar public-company bargain: buy a share, receive a vote. Anthropic's proposed IPO would sell Class A stock, each share carrying one vote. That vote is real. It also arrives inside a structure where a single Class F share would command 50.1% of votes on general shareholder matters, and a separate Class T share would generally appoint most directors.¹ The cap table tells you who owns the economics. It does not tell you which meeting settles a decision.

The filing calls these five types of common stock: A, N, T, F, and E. Strictly, its charter describes four common-stock classes, with A and N as two series of Participating Common Stock. The useful reading is by rights, not lettering. The company filed its S-1 on 26 October 2026; as of this article's date, the offering terms and final share counts remain blank, and the governance provisions are described as taking effect immediately before or upon completion of the offering.¹ Treat every post-IPO statement below as the filed plan, subject to amendment.

The filing's own rights table is the fastest way to leave the cap-table habit behind. A and N participate in the financial upside. T, F, and E mainly allocate decisions, with no dividends and only par value on liquidation. The oddity is intentional: the shares with the most important governance rights are almost economically empty.¹ In most board decks, the word common implies a family resemblance. Here it is a filing category with a very large seating chart.

That table is a map of jurisdiction. Class A and F share the general shareholder ballot, but Class F is weighted to win it if the Founder LLC reaches a decision. Class T gets its own director election. Class E appears when a particular trust question meets a particular trigger. Class N can own a substantial economic stake and remain absent from an ordinary vote. Counting all five as though they cast ballots in the same room produces a confident answer to the wrong question.

The N distinction matters beyond Anthropic. The filing says Amazon and Google would each hold more than 5% of outstanding capital stock after the specified conversions. Their N shares would be nonvoting while held as N, then generally convert to voting A on a qualifying sale or transfer. That can dilute the voting weight of existing A holders within their 49.9% side of a general vote. It does not dilute the F share's fixed 50.1% general-vote entitlement.¹ The investor can have a large cheque and a large economic interest without an ordinary ballot. Anthropic's Public Benefit Fund shares use the same Class N economics, with a different question about who controls their distribution.

The Founder LLC's single F share would carry an aggregate 50.1% of outstanding voting power on matters put to shareholders generally, recalculated at each record date. Future A issuance would increase the number of votes assigned to F to preserve that percentage. That is stronger than an ordinary ten-votes-per-share structure: dilution changes the denominator and the F share adjusts with it.¹ A Class A buyer can vote against the founders' choice for a non-T director or a merger. The buyer cannot assemble a general-vote majority against a unified Founder LLC while F remains in force.

The seven co-founders do not each possess a personal slice of that 50.1%. They would vote inside the LLC, one vote per Continuing Founder, and a majority would instruct the F share on ordinary matters. Initially that means four of seven. If they fail to reach a majority, F would generally follow the proportions of votes actually cast by A holders, subject to the filing's Class A quorum condition. For changes to F rights or the trust, higher internal thresholds apply. Their personally held A shares remain separately votable at their discretion.¹ The public vote becomes consequential if the founders split. The Founder LLC membership and sunset rules deserve their own reading; this is also the question behind what a 51% stake can actually decide, with an extra decision maker inside the majority vote.

Class T supplies the other half of the arrangement. The Long-Term Benefit Trust, through its one T share, would generally elect and remove directors filling a majority of the board's authorized seats. The A and F holders together elect the remainder; T does not vote in that election.¹ Anthropic says the trust and Founder LLC together would be able to elect the whole board. Its prospectus calls the company a Nasdaq "controlled company" because the trust can elect a majority, even though F controls general shareholder ballots. It expects to use the controlled-company exemption for a nominating and governance committee that would not be entirely independent.¹ There is no single control percentage that describes all of this.

The board still manages the business. Anthropic says the trustees do not direct daily operations or set strategy; they choose directors and advise management. That distinction has legal teeth. Under Delaware law, directors manage or direct corporate affairs, and a merger agreement normally needs board approval before the shareholder vote. Anthropic's public benefit corporation status also requires its board to balance stockholder financial interests, affected stakeholders, and its stated public benefit.¹²³ The trust's appointment power is enormous, but it is exercised through directors with their own duties. Calling the trustees the operating bosses would flatter a governance chart at the expense of the actual decision path.

Financing supplies another check on easy labels. The filing would allow the board to issue undesignated preferred stock with approval from 80% of authorized directors, without a further general shareholder vote. While F exists, however, the board could not designate or issue a preferred series without the Founder LLC's affirmative vote or consent.¹ A trust-appointed board majority alone would not clear that particular financing. F's power here is a separate consent right, not merely its 50.1% on a shareholder ballot.

One T seat has a further check. The trust's nomination, removal without cause, or change in designation of the "Specified Class T Director" needs concurrence from the board's nominating and governance committee. If that committee does not concur within 15 days, the trust may withdraw its proposal or seek approval from E holders by a majority of votes cast, with a majority of outstanding E shares present for quorum. The committee is planned to have one T director and two non-T directors.¹ This is a real employee vote, though only after a trust and committee disagreement over one designated seat. The Class E eligibility and redemption rules determine who may still cast it when that dispute arises.

Imagine a buyer offers a premium for Anthropic. The buyer cannot treat F's 50.1% as a signature that transfers the company. For an ordinary Delaware merger, the board must approve the agreement and, where required, submit it to shareholders. Anthropic's filing says mergers, consolidations, and sales of all or substantially all assets require board approval under Delaware law. The trust would have elected a majority of that board. If the transaction then requires a general shareholder vote, the Founder LLC could determine its outcome through F, provided its members reach the required internal instruction.¹² Two gates, two different people with leverage.

That structure can stop a transaction at either gate. The board could decline to approve a deal the founders want. The founders could reject a deal the board approves at the general shareholder stage. Neither inference means the trust has a personal merger veto written on its T share, nor that F can force the board to recommend a transaction. The legal form of a transaction can change which approvals are required, so a real bid requires reading the actual deal documents and applicable law. The filing also has specific provisions for vacancies, board size, and class changes; an activist cannot assume a routine shareholder slate will replace the T directors.¹² The unsolicited-bid analysis follows those gates through a tender offer and a merger.

Class N shows why the economic cap table still matters here. N holders share in proceeds alongside A holders, although they lack an ordinary vote while N. A purchaser of N in a qualifying transfer may instead receive A shares and the votes attached to them. So the economic distribution and the approval chain are separate calculations. A deal model that puts both on one spreadsheet tab should still have different formulas.

It is tempting to assume F's 50.1% can rewrite the trust away. The S-1 describes a narrower and more difficult path. If the trustees favor a change to the trust terms or T rights, approval would require 80% of authorized board members and 80% of trustees then in office. If the trustees oppose while F still exists, a route within the first three years after the offering requires unanimous Continuing Founders, a majority of E votes cast at a 40% E quorum, and a majority of A votes cast at a majority A quorum. Another route available while F exists requires at least two thirds of Continuing Founders, 75% of E votes cast at a majority E quorum, and a majority of A votes cast at a majority A quorum. If F no longer exists and trustees oppose, the filing requires approval of 85% of outstanding A shares. The same threshold structure can redeem T for no consideration.¹

Notice what the threshold does. The founders' general voting majority cannot replace the specific A and E approvals on an opposed trust change. Employees, who have no economic claim through E, can block the change under the applicable route. Public A holders, individually weak on general matters, acquire a separate collective gate. The trustees have an easier path when they agree, but even they need a large board majority. "Mission lock" sounds like an engraving on the wall. The operative text is a series of quorum and consent requirements.¹

Class E would be distributed by a formula: one share for each completed six months of employment, plus one for each A share an employee pledges to donate through the matching program. The filing says timing of issuance will be determined after closing, co-founders are excluded, and E shares cannot be transferred. They would be redeemed for nothing on a staged schedule after ordinary separation, immediately for specified events including termination for cause or service with a direct competitor, and when F disappears.¹ That is a deliberately bounded employee constituency. Do not read it as employee ownership of the trust, and do not ignore the service rules when estimating whose E votes could be outstanding at a future record date.

The F share does have an end condition. The filing says a 450-day transition starts once fewer than three qualifying Continuing Founders remain for 90 consecutive days. F and outstanding E would be redeemed for no consideration at the end of that period. The Founder LLC can also choose to redeem F under its supermajority rule, or F can end if the LLC dissolves. A founder can cease to qualify for several reasons, including resignation, death, incapacity, specified competitive conduct, or falling below a very small share-retention threshold measured against that founder's A holdings at IPO completion.¹ The IPO anniversary alone does not switch the F share off.

T has a different life. The filing says its sunset does not occur automatically from elapsed time or a specified event. Ending T needs the trust-amendment approvals. If F sunsets first and T remains, the trust would still generally elect a board majority, while the general shareholder vote would change without the F block. If T ends first and F remains, the founders would retain their general-vote majority while the board election rules change. Those are different companies from a control perspective, even if the economic cap table barely moves.¹ A founder's favorite phrase, "we have a sunset," is doing a lot of work if nobody asks which share sets with the sun.

The filing also raises the vote needed to amend certain charter and bylaw provisions from an A/F majority before F's sunset to two thirds of A voting power afterward.¹ F disappearing would therefore end one concentrated vote while making some amendments harder for the remaining electorate. A sunset clause can redistribute control and increase a threshold at the same time.

Start with one decision, not with the percentages. A board appointment, CEO replacement, product launch, financing, merger, charter amendment, and trust amendment each has a different approval path. For each one, write a short sequence:

1. Who may initiate it? For a merger, the board approval comes before the usual shareholder vote. For one disputed T seat, the trust proposes, the nominating committee responds, and E may become the tiebreaker.¹²
2. Who chooses the decision makers? In Anthropic's filed plan, T generally appoints most directors; A and F choose the rest. Appointment rights are not the same as instructions to directors on every later vote.¹³
3. Which ballots are separate? The general A/F vote, an A class approval, an E class approval, a trustee vote, and an internal Founder LLC instruction are different ballots. Never add their percentages together.
4. What is the failure rule? A deadlock inside the Founder LLC can send F's votes to the A vote proportions on ordinary matters. A failed higher threshold on a trust amendment means F votes against it. Quorum can decide whether an apparent majority has any effect.¹
5. What survives dilution, transfer, and sunset? N can turn into A on a qualifying transfer. New A shares dilute other A voters but do not dilute F's fixed general-vote percentage. F and T have different end conditions.¹

Then open the actual instruments. The S-1 summarizes the proposed charter, bylaws, trust agreement, stockholders' agreement, and Founder LLC agreement; the operative text and any later amendments govern. Recheck the final prospectus when the offering prices. The IPO proceeds and lockup schedule answer the separate question of when each group can turn its shares into cash. A private-company founder can run the control exercise after a seed round. If your explanation of who controls a sale fits in one percentage, you have probably skipped at least one room.