Who Actually Holds the Shares Behind a Tokenized Stock, and What They Owe You

Tokenized equities are usually described in terms of the token: how it trades, where it trades, what it costs to move. That is the part a reader can see onchain. The part that determines whether the token is worth anything is off the chain entirely, in a custody chain that ends with an actual share sitting in an actual account at an actual institution. This piece walks that chain from the token backwards, because the questions worth asking about a tokenized stock are almost all custody questions wearing other clothes.

The chain of custody, from the token backwards

Start at the token and work down. You hold a token in a wallet. The token was issued by an issuer, which is a legal entity somewhere with a regulator, or without one. The issuer says the token is backed one for one by a share. That share is not held by the issuer directly in most structures. It is held by a custodian, a regulated entity whose business is holding client assets, on behalf of the issuer. The custodian, in turn, usually does not hold a paper certificate either. It holds a position in an account at a central securities depository, the market infrastructure that holds the definitive record of who owns what in a given jurisdiction. The company whose shares these are keeps its own register, maintained by a transfer agent, and in most cases that register shows the depository's nominee as the holder of record for the entire float.

So there are typically four or five layers between you and the register: wallet, issuer, custodian, depository, transfer agent. Each layer is a promise from one party to another. The token is a promise from the issuer to you. The custody agreement is a promise from the custodian to the issuer. The depository position is a promise inside the market's plumbing. None of these promises are the share itself. The share, in the legal sense that a court would recognise, lives at the bottom.

What each layer actually owes you

This is where the structures diverge, and where the reading matters. In the strongest arrangements, the shares are held by the custodian in a segregated account, in the name of the issuer or a bankruptcy-remote vehicle, for the benefit of token holders, and the token confers a direct or beneficial claim on those shares. The document to look for is the one that says what a token holder is entitled to demand and from whom. If it says you may redeem the token for the underlying share or its cash value, you have a claim. If it says the token is a note, a contract for difference, a certificate tracking the price, or a synthetic exposure, you have a claim against the issuer's balance sheet and nothing more.

That distinction is not academic. A claim on segregated shares survives the issuer failing, at least in theory, because segregated client assets are not supposed to be available to the issuer's creditors. A claim against the issuer's balance sheet does not survive the issuer failing. It joins the queue. The tokens can look identical onchain: same ticker, same decimals, same one-to-one accounting, same attestation page. The difference is entirely in the paperwork, and the paperwork is what gets litigated.

Omnibus versus segregated, and what is traded away

Custodians hold client assets in one of two broad ways. In a segregated account, the issuer's shares sit apart from every other client's, identifiable as a distinct pool. In an omnibus account, the custodian holds one large position covering many clients and maintains its own internal books showing who owns what fraction of it. Omnibus is cheaper, faster to operate, and much more common, because the alternative means opening and reconciling separate accounts for every client relationship.

The tradeoff is reconstruction risk. If a custodian using omnibus accounting fails, or its records are wrong, establishing which shares belonged to which client is an accounting exercise conducted after the fact, sometimes by an administrator, sometimes over years. Segregation buys certainty of identification at the cost of operational overhead. Neither is fraud. Both are ordinary market practice. But a reader deciding how much comfort to take from the phrase held by a regulated custodian should know which one is meant, because the phrase covers both.

Where the chain breaks

There are four ordinary failure points, none of them exotic. The first is the issuer being a different entity from the one implied by the marketing, often incorporated in a jurisdiction with a light company registry, so the entity that owes you the share is not the recognisable brand on the website. The second is the attestation covering the wrong thing: a report confirming that a certain number of shares existed at a moment in time is not a report confirming that the shares are unencumbered, that they have not been lent, or that the count matches the tokens outstanding right now.

The third is rehypothecation, meaning the shares are lent out or pledged as collateral elsewhere while the tokens circulate. This is legal in many arrangements and disclosed in the custody agreement, usually in a clause about the custodian's or issuer's right to use client assets. It converts your claim on a specific share into a claim on a promise to return one. The fourth is the gap between the onchain record and the register. The transfer agent's book is the legal record of ownership. The token contract is a separate ledger that a court has no obligation to recognise unless a legal structure explicitly links the two. Most structures link them by contract rather than by law.

What to read, and in what order

For any tokenized equity, the useful reading order is: the terms document that says what a token entitles the holder to, the named issuing entity and its jurisdiction, the named custodian, whether the account is segregated or omnibus, whether assets may be lent or pledged, the redemption mechanism including who may use it and at what minimum size, and only then the attestation. The attestation is the last item, not the first, because it verifies a number inside a structure. If the structure gives you a claim against a thinly capitalised entity in a jurisdiction with no securities regulator, a clean attestation confirms that the entity accurately counted things it does not owe you.

One further test is worth applying: can retail redeem, or only authorised participants? Many structures allow redemption only for a small set of institutional counterparties above a minimum size. That is a normal design, and it is how creation and redemption keeps the token near the share price. But it means the individual holder's remedy in a stress scenario is to sell into whatever market exists, not to redeem. If the market is a single pool on a single venue, the practical value of the backing depends on someone else choosing to arbitrage it.

Why it matters more as volumes grow

None of this is a reason to avoid tokenized equities. It is a reason to know which of several very different products you are holding, because they trade under names that do not distinguish them. Tokenization moves the settlement layer onchain. It does not move the custody layer, the register, or the legal definition of ownership onchain, and current structures mostly bridge that gap with contracts rather than with statute. That gap is where the interesting policy work is happening, in transfer agent modernisation and in custody rulemaking, and it is the gap a reader should keep an eye on. The token is the easy part. The share underneath it is the whole product.