Part 2 worked through the on-chain machinery for linking two legs. This part turns to the leg that decides whether any of it matters in practice: the cash leg. Atomicity is a property of the ledger, not of tokenisation in general, and that distinction is the first thing any operator-level reader needs to internalise. Once the cash leg is honest, the capital and operational consequences stack up cleanly, and the unit economics of an atomic DvP product can be priced with the same arithmetic you would use for any other infrastructure investment.

The cash-leg problem

The hard problem in atomic delivery-versus-payment (DvP) is the cash leg. Same-ledger atomicity works when the cash is tokenised on the same ledger. Cross-ledger atomicity works when both ledgers are participating in a common settlement protocol. Neither works when the cash leg sits in a non-tokenised currency at a correspondent bank, settled through SWIFT messages and end-of-day netting, which is still most of the world.

The atomic property is, in this sense, a feature of the ledger architecture, not of tokenisation in general. You cannot tokenise the asset, leave the cash in fiat, and call the result atomic DvP. What you have is asset-on-chain and a payment instruction off-chain, sequenced as best you can. That is a real product, and it has real settlement-cycle improvements over the legacy alternative, but it is not atomic and the rulebook should not pretend it is. The BIS CPMI tokenisation taxonomy is careful with this distinction and an institutional rulebook drafted to it will be too.

This is why the cash-leg solution defines the architecture. Three production options in 2026, each covered fully in later chapters.

Tokenised deposits on the same ledger as the asset, the Project Ensemble and Canton Network pattern. Single-name credit risk to the issuing bank persists; the cash leg moves at the speed of the asset. See Chapter VII.

Regulated stablecoins under a defined perimeter, where the stablecoin is on the same chain as the asset and the chain is treated as a settlement venue. Legal characterisation depends on jurisdiction (see Stablecoin types).

Wholesale central bank digital currency (wCBDC), where the central bank issues a settlement token for use among supervised institutions, and asset ledgers commit against the wCBDC ledger. This is the Project Agorá direction. Chapter VIII covers it.

The first two are live in production with named participants and real volume. The third is in advanced pilots in multiple jurisdictions and is the architecture most regulator-led work converges toward. Each comes with a different distribution of credit, operational, and political risk; none is dominant on all three.

Capital and operational implications

Three things change on the bank's books with atomic DvP.

Settled positions release from the credit-risk capital line earlier. A trade that previously sat in pending status for a settlement cycle, generating counterparty exposure that consumed risk-weighted assets, can be released the instant the chain commits. The exact saving depends on trade mix and netting agreement structure, and quoting a specific basis-point figure across the industry would be unhelpful, but the direction is clear and capital teams have started to surface it internally. The basel sco60 cryptoasset standard sets the prudential framework against which the capital release has to be argued, with Group 1a treatment available for tokenised traditional assets that meet the classification conditions (Basel SCO60).

Collateral that moves atomically can be repurposed faster. The intraday repo product set, what BNY Mellon (BNY), JPMorgan, and the Eurex repo platform are building, depends precisely on this property. If you can DvP a tokenised Japanese government bond (JGB) against tokenised cash and then repo it back atomically, the capital cost of holding the bond is materially lower than under an overnight settlement cycle. JPMorgan's tokenised collateral disclosures on Onyx (now Kinexys) are the only operational write-up with named counterparties and timestamps, and the recurring product theme is intraday velocity rather than headline trade size.

Fail-to-deliver risk drops sharply. A fail in a regulated market is not just an operational annoyance; depending on jurisdiction it triggers buy-in procedures, reputational damage, and regulatory reporting. Atomic settlement removes the failure mode by design. The operational cost of running the chain replaces it, but it is a different cost with a different shape, and one the bank controls. Operationally, a fail rate measured in basis points becomes a chain reorg rate measured in basis points of basis points, and the supervisor's expectations about how the residual is monitored shift accordingly.

Unit economics

The unit economics are not zero-sum. Atomic DvP costs more in liquidity (Model 1 gross settlement) and demands more operational sophistication (running the chain, integrating to systems of record, reconciling continuously) than legacy Model 2 or Model 3 arrangements. The savings come from capital release, faster collateral velocity, and lower fail-rates. Whether the maths works in any product depends on volume, the funding cost of gross legs, and the fail-rate baseline.

For high-volume, high-value flows, the maths works. Repo, large-ticket bond settlement, intraday liquidity products, and triparty collateral are the natural fits, and most named production examples cluster there. For low-volume products, particularly retail-adjacent securities with thin order books, the funding cost of gross settlement often outweighs the capital saving, and a hybrid Model 2 design with on-chain securities and off-chain net cash settlement can be the better answer even though it is not atomic.

The decision is product-level, not platform-level. A single tokenisation programme can run atomic DvP for the high-value institutional flows and a hybrid model for the rest, on the same chain, with the same legal documentation. That flexibility is part of why the permissioned-ledger architecture survives: the chain is one design choice, but the settlement workflow on top of it is another, and the second one tracks the economics of the specific product line rather than the dogma of the platform.

Part 4 turns to the question of why this cycle is the one in which atomic DvP graduates from pilot to production, and clears out the four or five common confusions that still derail tokenisation conversations between TradFi and Web3 readers.