In short
A tokenized deposit is a commercial bank deposit recorded on a distributed ledger: it stays a claim on the issuing bank, remains inside the regulated perimeter, and carries that bank's credit. A payment stablecoin is a separate instrument backed by a reserve its issuer holds. The choice is therefore not about technology, since both settle in seconds on a ledger, but about who the holder is exposed to and who they can reach. Deposits keep the relationship and the supervision; stablecoins reach counterparties the banking network does not serve. Most institutions with a serious cross-border business end up running both, which makes the abstraction layer between them the part that actually has to be engineered.
The distinction that carries the decision
Both instruments do the same visible thing: they move value on a programmable ledger far faster than the correspondent chain they replace. Because the visible behaviour is identical, they get discussed as competing implementations of one idea, and the decision gets framed as a technology preference. It is not. They differ in what the holder owns.
A tokenized deposit is the bank's liability. The holder's claim is against a supervised institution with capital requirements, deposit protection where it applies, and a regulator who will be involved if something goes wrong. A stablecoin is a claim against an issuer's reserve, and its behaviour under stress is determined entirely by what that reserve holds and how quickly the issuer will honour redemption.
That difference is invisible on a good day and it is the only thing that matters on a bad one. Any comparison that leads with settlement speed or transaction cost has skipped the question.
Where each instrument is strongest
Tokenized deposits are the stronger instrument inside a known set of counterparties. Treasury movement between entities of the same group, settlement between banks that already have a relationship, and corporate payments where both sides bank somewhere supervised. These gain programmability without introducing a new credit exposure or a new regulatory question. The money never leaves the perimeter, which is precisely why banks are building shared networks for it instead of ceding the ground.
Stablecoins are the stronger instrument where the counterparty is not reachable through that perimeter at acceptable cost or speed. Paying a supplier in a corridor where correspondent banking has withdrawn, settling with a crypto-native counterparty, or moving value at a weekend. These are cases where a tokenized deposit has nothing to settle against, because the other side has no account in the network.
The unhelpful framing is that one will win. They are addressing different halves of the same problem, and an institution that needs both halves solved will hold both instruments and needs a system that treats that as normal rather than as an exception.
Interoperability is the unsolved part
Tokenized deposits face a coordination problem that stablecoins do not. A stablecoin is one issuer's instrument and is fungible with itself everywhere; a tokenized deposit is a claim on a specific bank, and a claim on one bank is not interchangeable with a claim on another. Making them interoperable means either a shared consortium instrument or a clearing layer that converts between them. And both concentrate the risk that bilateral settlement had distributed.
This is why shared network initiatives matter more than any individual bank's pilot. A tokenized deposit that only settles between customers of the same bank is an internal ledger with extra steps. The value appears when it settles across institutions, and that requires agreement on standards, legal structure and who operates the middle. Which is slow, commercially sensitive work that no amount of engineering shortens.
The planning consequence for anyone building now is to avoid designing around a specific network's assumptions before that network exists. Build the ledger and the abstraction so that the settlement venue is replaceable, because the venue that wins is not yet decided.
The engineering obligation is identical
Whichever instrument is chosen, the same requirement appears: the on-chain position and the institution's ledger must never disagree, and proving that must not require a person. The ledger is authoritative for what is owed; the chain is authoritative for what has settled; every movement on one has a counterpart on the other, and unmatched items are exceptions raised immediately rather than discovered at month end.
The mistake that produces the worst outcomes is treating the chain itself as the ledger. Chains have reorganisations, failed transactions, pending states and fee-driven ordering. They are an excellent record of what settled and a poor record of what was intended, and a system that conflates the two will eventually report a balance it cannot defend.
Redemption deserves the same rigour under both models. A tokenized deposit redeems into the bank's own account structure and a stablecoin into a reserve, but in both cases the question is what happens when redemption volume is one-directional for a week. And in both cases the answer is decided by architecture rather than by intent.
How to decide
Start from the counterparties. List the parties you actually need to pay and be paid by over the next two years, and mark which are reachable inside a regulated network. If nearly all are, tokenized deposits solve the problem with less regulatory surface. If a material share are not, no amount of deposit tokenization reaches them.
Then ask what you are permitted to hold. An institution that cannot take reserve exposure to a third-party issuer has made the decision already; one whose product requires settlement at a weekend has made a different one.
The answer that survives contact with a roadmap is usually both, sequenced. Build the ledger and the abstraction first, add one settlement venue, prove reconciliation and redemption are boring, then add the second. Institutions that started from the chain and worked backwards to the ledger have consistently arrived at the same place more slowly.