TL;DR: Anthropic's proposed capital structure separates economic ownership, general shareholder voting, board-election authority, and limited employee governance rights. The initial S-1 describes five common-stock classes with materially different roles. A percentage of economic shares therefore cannot be assumed to equal the same percentage of votes or director-selection power. Build a rights inventory by decision type, then read the conversion, transfer, redemption, and sunset provisions that can change that inventory over time.
The specimen cabinet has five drawers
The cover of Anthropic's October 26 public filing describes Class A, Class N, Class T, Class F, and Class E common stock. The proposed public offering concerns Class A. The other classes are not interchangeable versions of the same investor claim. Some carry economic rights without ordinary voting power; others carry governance functions without economic rights.
That structure makes the ordinary sentence owns part of the company incomplete. A reader needs to ask what the ownership entitles the holder to receive, which decisions the holder can influence, and under what conditions those rights change. Counting certificates alone cannot answer those questions.
The controlling source is the proposed structure in the initial Form S-1, reviewed here as of October 28, 2026. The offering and associated changes are described conditionally because filing is not completion. This article is an interpretation aid, not a substitute for the certificate, agreements, or transaction-specific legal advice. Anthropic Form S-1, cover and Description of Capital Stock
We inspect each drawer, then run a hypothetical ownership calculation. The example uses small invented share counts and omits contingent instruments deliberately. It is not a reconstruction of Anthropic's unresolved final share count or an estimate of its offering valuation.
Drawer A: an economic share with one ordinary vote
Class A is the class proposed for the public offering. The filing describes one vote per share on matters submitted to stockholders generally, within a structure that also assigns special rights to other classes. The one-vote description is accurate but insufficient by itself to establish the holder's influence over the company.
A person holding one hundred Class A shares can count those votes. To calculate their percentage of general voting power, the person also needs the other votes entitled to participate in that decision. To calculate economic ownership, the person needs the relevant economic-share denominator. Those denominators can differ.
The investor should also distinguish a vote from an assured outcome. A minority holder can have a meaningful formal right while lacking the power to determine a contested result. Influence may depend on coordination, the proposal, class-specific approvals, and the structure of the board. The existence of a ballot is not the same as control.
The practical evidence card for Drawer A records economic participation, ordinary votes, class-specific protections, and transfer conditions. It does not infer guaranteed dividends, future appreciation, or a right to direct daily model development. Those are different questions from whether the share carries a general vote.
Drawer N: similar economics with a different voting role
The filing describes Class N's rights as identical to Class A's except for voting and conversion, with general voting absent except where law requires it. It also describes conversion to Class A generally on transfer to an unaffiliated third party. The reader should retain the conditions rather than shorten the description to always nonvoting or always the same as Class A.
This class demonstrates why economic ownership and governance participation need separate columns. An economically substantial holding can have little ordinary voting power under the relevant terms. The absence of a vote does not make the economic claim disappear, and the size of the economic claim does not create a vote the instrument does not provide.
Transfer can change the voting denominator without changing total economic shares. If a specified number of Class N shares converts into Class A when transferred under the applicable terms, the underlying economic interest may persist while the allocation of ordinary voting participation changes. The model should identify the event that causes the change.
For an analyst, this is a reason to maintain a conversion scenario separately from the historical capitalization. A proposed transfer is not an observed conversion. The actual conditions, exceptions, and timing belong to the controlling documents, not to a generic assumption about how nonvoting shares usually work.
Drawer F: control is not measured by the certificate count
The proposed Class F share is a single share held by a founder entity. The filing assigns it voting power equal to fifty-point-one percent of the outstanding voting power on matters submitted to stockholders generally, while providing no economic rights. Its role therefore cannot be understood by saying there is only one such share.
A fixed percentage is different from a fixed number of votes. As the relevant ordinary voting base changes, the votes assigned to Class F can change to maintain the specified proportion under the terms. Issuing additional Class A shares need not dilute that percentage in the way a reader might expect from an ordinary one-vote capital structure.
The absence of economic rights also matters. The Class F share itself should not be counted as another ordinary economic slice merely because it is called common stock. The founders can hold economic interests through other instruments, but that is a separate inquiry. A governance instrument and the holders' broader wealth are not the same object.
The decision card identifies which matters fall within general voting and which have separate rules. It also points to the founder entity's internal arrangements and the Class F transition provisions. Those provisions affect how the right is exercised and eventually ends; a permanent-sounding summary can miss the disclosed conditions.
Drawer T: selecting directors is a distinct power
The Long-Term Benefit Trust's Class T role concerns election and removal of Class T directors, who under the proposed arrangement generally constitute a majority of the authorized board. The filing distinguishes that authority from ordinary general shareholder voting and describes Class T as lacking economic rights.
This means a headline about general shareholder control does not fully describe director-selection authority. A party can dominate matters submitted to stockholders generally while a different institution selects a specified group of directors. The resulting governance map needs both relationships.
The board itself then has responsibilities for the company's business and oversight. Selecting directors is not identical to personally making every operational decision. A trust's power to influence board composition should not be described as a direct vote on each model release unless the actual documents create such a procedure.
For the rights inventory, the reader records appointment and removal, the relevant director category, and the conditions affecting that category. Evaluation of whether the arrangement serves public benefit is a further governance argument. First, the mechanical question must be clear: who can select which decision-makers?
Drawer E: employee voice has a defined scope
Class E is proposed for employees other than co-founders and has limited voting functions without economic rights. The filing identifies specified governance matters rather than an ordinary vote on every shareholder proposal. A statement that employees receive shares would conceal the distinction if readers assumed those shares necessarily carried a tradable financial interest.
The terms also describe issuance and redemption conditions tied to employment and other specified events. The existence of an employee governance class therefore should not be treated as an indefinite, freely transferable vote held on the same terms as public Class A. The scope and lifecycle are part of the right.
Conversely, limited does not mean meaningless. A class-specific approval can matter on the decisions to which it applies. The relevant analysis asks whether the decision is covered, which voting threshold applies, and which other approvals are also required. A small set of protected matters can have significance that an aggregate vote count misses.
The employee's financial compensation and this governance participation remain separate questions. Options, restricted units, ordinary shares, and Class E should not be merged into one balance labeled employee equity. A record that preserves their different functions makes both compensation and governance easier to understand.
The inventory table keeps the questions separate
| Proposed class | Economic role in the filing | Governance role to inspect |
|---|---|---|
| A | Economic common-stock rights | One vote per share on general matters, within the full structure |
| N | Economics generally aligned with A, subject to stated distinctions | Generally nonvoting; conversion and required-law exceptions matter |
| F | No economic rights | Fixed share of general voting power through the founder entity |
| T | No economic rights | Specified board-selection and related governance functions |
| E | No economic rights | Limited employee governance matters and lifecycle conditions |
The table summarizes roles, not every provision. It should be read with the source sections on capital stock, governance, and the agreements. Its value lies in preventing a number from one column from being silently used in another calculation.
A hypothetical holder has three different percentages
Imagine one thousand economic Class A shares and one thousand economic Class N shares, with no other economic instruments in the simplified example. A holder owns one hundred Class A shares. The holder has five percent of the two thousand economic shares and ten percent of the Class A shares.
Now add a Class F voting mechanism modeled solely for illustration on the stated fifty-point-one-percent general-vote proportion. If ordinary Class A votes total one thousand, the Class F votes must equal one thousand multiplied by fifty-point-one divided by forty-nine-point-nine, approximately one thousand four-point-zero-zero-eight. Total general voting power is about two thousand four-point-zero-zero-eight.
The holder's one hundred votes are therefore four-point-nine-nine percent of the general voting total. Five percent economic ownership, ten percent of ordinary Class A, and four-point-nine-nine percent of general votes are all correct within the model. None describes the holder's share of Class T director-selection authority.
The arithmetic does not estimate Anthropic's capitalization. It demonstrates why an ownership report must name its denominator. A precise percentage without that name is incomplete, especially when the legal structure deliberately separates kinds of rights.
New issuance can dilute one interest without changing another
Suppose the hypothetical company issues another thousand Class A shares. Economic shares rise from two thousand to three thousand. The original holder's one hundred shares now represent about three-point-three-three percent of that simplified economic total. The holder has not sold anything, but the denominator has grown.
Under the stipulated Class F mechanism, the fixed general-vote percentage is maintained rather than divided by the new ordinary-share count. This illustrates a structural asymmetry. Public economic interests and individual Class A voting percentages can dilute while the specified controlling proportion remains fixed under its terms.
This does not establish whether the issuance benefits or harms economic value. New capital can fund useful activity, and dilution of percentage ownership is not automatically destruction of value per share. The economic judgment requires the proceeds, price, and use of resources. The rights calculation simply identifies what changed mechanically.
The analyst should therefore report percentage dilution, per-share economics, and governance effects separately. Combining them into the single word dilution makes it difficult to tell whether the concern is a smaller ownership fraction, a lower value per share, or reduced influence over decisions.
Conversion can change votes without diluting economics
Return to the example with one thousand Class A and one thousand Class N economic shares. Assume, solely for this exercise, that all the Class N shares are transferred in a way that satisfies the stipulated conversion conditions and become Class A. No new economic shares are issued. The total remains two thousand, and the holder of one hundred original Class A shares still owns five percent of the simplified economic interest.
Ordinary Class A votes, however, rise from one thousand to two thousand. Under the example's fixed-percentage Class F mechanism, Class F votes also increase to maintain fifty-point-one percent of the relevant general total. The original holder's one hundred votes now represent two-point-four-nine-five percent of that total, compared with four-point-nine-nine percent before conversion. The economic percentage is unchanged while the percentage of general voting power falls by half.
This is not a claim that such a transfer will occur, that all real transfers have identical conditions, or that the example reconstructs the issuer's final capitalization. It is a controlled demonstration of why conversion deserves its own column. Calling every reduction in voting percentage economic dilution would misstate this case. Saying nothing changed because economic shares remained constant would miss the governance change.
The timing matters too. Before a qualifying transfer, the analyst should retain the existing rights. After it, the rights register should identify the converted shares and the resulting decision-specific denominator. A rumored sale is not evidence that conversion has completed. Use the effective date when updating the register; an announcement describing a proposed change and a record establishing completed conversion identify different states, even if they concern the same holders. The update needs the event and the provision that connects the event to the new rights.
This exercise suggests a useful reporting format with two reconciliations. The economic reconciliation explains issuances, repurchases, cancellations, and other changes affecting the ownership base. The voting reconciliation explains class conversion, special voting formulas, and changes relevant to the decision under discussion. The two can share events without becoming the same calculation.
A valuation model also needs to resist importing the vote percentage as an economic share. A fifty-point-one-percent general voting instrument with no economic rights does not by itself receive fifty-point-one percent of a hypothetical dividend. Conversely, an economically substantial nonvoting class can participate in the relevant distribution without controlling a ballot. A reader can assess whether that separation is desirable only after describing it accurately.
The practical advantage of the specimen cabinet is that it keeps these changes legible. The analyst can say what moved, why it moved, and which consequence follows. That is more durable than attaching one ownership percentage to a person and reusing it for every economic and governance claim that later appears in a story.
A board majority and a shareholder majority are different claims
The proposed structure requires the reader to distinguish voting on general shareholder matters from selecting a majority of directors through Class T. These powers can coexist without one being a mathematical contradiction of the other. They operate through different decision channels.
A hypothetical acquisition might require board action and shareholder approval under its governing conditions. A director election might occur through a specified class. A change to the trust could require several approvals. The reader should follow the particular decision rather than ask only who controls the company as though one percentage settled every channel.
This is also a limit on simplistic forecasts. A founder-controlled general vote does not establish that every board proposal succeeds. A trust-selected board majority does not establish what the directors will decide in every future tradeoff. Rights create authority and incentives; actual decisions depend on people, procedures, evidence, and the applicable duties.
The practical map begins with a decision type and traces each required approval. It can then identify whose consent is decisive, whose participation is limited, and where a proposed change could be blocked. That is more useful than a pie chart that mixes economic ownership and governance into one circle.
Dividends and liquidation belong in the economic drawer
An economic common share can participate in distributions under the relevant rights and decisions. That does not promise a dividend on a particular timetable. The company's dividend policy, financial condition, restrictions, and board decisions matter. A reader should not infer an income stream simply because a security is called stock.
Liquidation rights answer a different question: what may holders receive after the relevant obligations and priorities are addressed? Economic ownership is a residual claim within a capital structure, not ownership of an identified server or a right to withdraw a proportional part of current cash on demand.
The proposed non-economic classes should not be added to the economic denominator as if each could claim the same distribution. Conversely, excluding them from economic per-share arithmetic should not erase their governance significance. Each calculation needs the classes relevant to the right being measured.
The SEC's stock overview provides useful general vocabulary, but issuer-specific rights come from the filing and governing instruments. General descriptions of common stock cannot override a deliberately unusual structure. Investor.gov, Stocks
A special right can depend on another institution's decision
Some rights are exercised by an entity rather than by a natural person acting alone. The founder entity and the trust each have their own arrangements. A reader should distinguish the right held by the entity from the internal procedure through which the entity decides how to exercise it.
This matters for succession and disagreement. If several people participate in an entity's decision, their relationships and thresholds can affect its actions. A summary saying one founder controls a right may be inaccurate if the documents assign it to a collective entity with specified procedures.
The same caution applies to independence. Describing trustees as financially disinterested identifies one intended feature; it does not answer every question about selection, information, expertise, or accountability. Those questions belong in the institutional assessment after the formal rights have been mapped.
Anthropic's earlier public explanation of the trust provides historical context for the design. The current proposed instruments in the S-1 remain the appropriate source for the IPO structure where the details have developed. A 2023 explanation should not be used to overwrite a more specific 2026 provision. Anthropic, The Long-Term Benefit Trust
Sunset provisions turn a static picture into a sequence
A right can end after a specified event, period, or change in participation. The reader should identify the triggering conditions and the resulting structure. A current majority may not be perpetual, while a transition may not occur merely because one holder sells an economic share.
The inventory therefore has a before-and-after page. It names which class remains, which votes change, which directors are selected through which route, and which employee functions continue. The purpose is not to forecast the trigger date without evidence. It is to understand what the document says would happen if the conditions occur.
Redemption also needs its stated meaning. A non-economic governance share redeemed for no consideration has a different function from an ordinary financial security bought back for cash. Using one generic share-buyback label could mislead readers about both company spending and the change in authority.
The model should preserve uncertainty about future personnel and events. A clear transition rule can be described without predicting how the institution will behave years later. The right is a mechanism with conditions, not a complete forecast of governance quality.
The holder's final evidence request
Before accepting a percentage claim, the reader asks for the security class, date, denominator, conversion assumptions, and decision type. Before accepting a control claim, the reader asks which approvals and director-selection routes are included. Before accepting an employee-ownership claim, the reader asks whether it concerns economic compensation or limited governance stock.
These requests are practical because they can change an interpretation. A corrected denominator can transform a headline without changing a single certificate. A disclosed sunset can qualify a permanent-sounding claim. A conversion condition can explain why rights differ before and after a transfer.
The proposed Anthropic offering makes these distinctions unusually visible. Its structure should be read as a set of particular rights rather than as one large ownership percentage. Once the drawers are separated, the reader can evaluate the economic claim, the voting opportunity, and the institutions that select decision-makers without making one stand in for all the others.