21 November 2026 · 10 min read
Tokenized deposits: the fast path is easy, the hard parts move
A practitioner’s map of what really changes in settlement, and what still breaks at 02:00.

The industry keeps describing tokenized deposits with one word: speed. Faster settlement, 24/7 availability, fewer intermediaries, less trapped liquidity.
Speed is real. It is also the least interesting part.
What matters is where the old friction goes when you wrap a bank liability in a token and connect it across currencies. Some steps genuinely disappear. Others reappear as “just an API call” until the first exception hits, the first sanction alert blocks a transfer, or the first Monday morning arrives with a weekend’s worth of unsettled edge cases.
The trigger for this piece is the Hong Kong and Swift tests connecting tokenized bank deposits across currencies, building on earlier work involving digital central bank money. Good. We need more real experiments. But if you are a European bank thinking “how would we ship this safely?”, you need a map that is operational, not promotional.
Here is that map, from the standpoint of what changes in treasury and payments plumbing: prefunding, cut-off times, nostro and liquidity work, and the part everyone underestimates, exceptions.
What tokenized deposits actually change in the payment chain
Tokenized deposits are still bank money. That matters. A stablecoin tries to be cash-like without being a bank deposit. A tokenized deposit is a deposit, issued by a bank, represented as a token, moved on some shared infrastructure.
So the promise is not magic. The promise is a shorter, more programmable path between two banks’ ledgers, possibly across currencies, with fewer batch boundaries.
In practical terms, three changes are meaningful.
- Prefunding pressure can drop. If you can atomically deliver deposit tokens versus whatever the counter-asset is, you do not need to park as much liquidity “just in case the window closes”. That is one of the core reasons banks keep chasing stablecoin-style flows.
- Cut-off times become less dominant, but do not vanish. The system can run 24/7. Your people and your controls usually do not. The cut-off shifts from “payment system closes” to “risk window closes”. It becomes internal rather than external.
- Message-to-settlement latency compresses. When the rail is closer to finality, the time between “I sent it” and “it is done” can shrink from hours to minutes or seconds. That reduces daylight exposure. It also changes how you detect and respond to problems, because there is less time to intervene.
- Finality definition. At what point is the transfer irrevocable from a legal and operational standpoint? “On-chain” is not a legal answer. You need “final when X ledger posts” and “reversal requires Y process”.
- Atomicity across legs. If cross-currency is involved, you need payment-versus-payment (PvP) behavior, not “we’ll do the other leg next”. The whole prefunding game comes back the moment one leg can fail after the other succeeds.
- Credit exposure recognition. If you settle against another bank’s tokenized deposits, you still have exposure to that bank. The exposure may be shorter. It is still there, and it has to land in limits, monitoring, and escalation.
- You invent internal cut-offs. Treasury will still set windows where certain corridors are allowed, certain sizes are allowed, and certain counterparties are allowed. The policy just moves from “because the system closes” to “because we close our risk window”.
- You need tiered service levels. Not every payment should be real-time at 03:00. You will end up with categories: always-on low value, always-on for pre-approved counterparties, restricted for new beneficiaries, restricted for high-risk geographies, and so on.
- You create a new kind of incident. The rail works, but your controls or staffing do not. The classic example is a sanctions or fraud hold that triggers outside business hours. If you cannot resolve it quickly, you can end up with liquidity stuck in an intermediate state that the customer experiences as “the bank lost my money”.
- From batch sweeps to continuous positioning. When settlement becomes more continuous, your liquidity position moves more continuously. That sounds nice until you need monitoring, alerting, and decision rights that can act just as continuously.
- New fragmentation. If tokenized deposits exist on multiple networks or schemes, you can accidentally recreate the nostro problem inside the token world. You end up with “token balances” that need to be funded and rebalanced across venues.
- Collateral and central bank access questions do not go away. European banks live inside a dense set of rules about what counts as high quality liquidity, what is eligible collateral, and what can be mobilized quickly. A deposit token might be a deposit, but its operational mobilization path is new. Treasury will ask, correctly, what happens in stress.
- Compliance holds. Screening, name matches, sanctions, AML patterns. These do not get easier because the asset is tokenized. They often get harder because the pace is higher and the customer expectation is “instant”.
- Investigation and recall flows. Wrong beneficiary, wrong amount, duplicated transfer, disputed authorization. Token rails often have weaker built-in recall semantics than traditional correspondent paths. If your product promise is “instant and final”, your recovery playbook must be brutally honest.
- Operational mismatches. One side posts, the other rejects. One side’s reference data differs. One system rounds differently. One side’s token contract version changes. The failure mode is not “payment rejected”, it is “two ledgers disagree”. Those are expensive days.
- Partial outages. The network is fine, the custody component is fine, but the API gateway is rate-limiting. Or your fraud system is down and you have to decide whether to keep settling without it. Every new component you add becomes a new way to be half-broken.
- Reconciliation risk becomes state-consensus risk. Instead of reconciling message logs to ledger postings, you reconcile token movements to internal books, and you reconcile smart contract state to off-chain accounting. The break types change. The need does not.
- Liquidity risk becomes timing risk. Faster settlement reduces some exposures, but it increases sensitivity to sudden flows. If outflows can happen instantly at any hour, buffers must be designed differently.
- Third-party and concentration risk migrate. If a token scheme relies on a small set of vendors, nodes, or infrastructure providers, you concentrate your critical path. The operational impact of one outage can be larger than in the old correspondent web.
- Model risk moves into rules. Programmability means rule-based transfers, conditional payments, automated limits. That is great until an edge case triggers the wrong rule at scale. Then it is not a coding bug. It is a balance-sheet event.
- What is the moment of final settlement, and who declares it?
- What is the supported reversal and recall process, and what is explicitly impossible?
- Who is liable for each failure class: sender bank, receiver bank, scheme, infrastructure provider?
- A single queue for holds, rejects, and investigations with ownership, timers, and escalation.
- Deterministic state machines for “posted”, “held”, “reversed”, “stuck”, “needs manual approval”.
- Metrics that matter: time-in-exception, exception rate by corridor, unresolved aging.
- Real-time position visibility, not end-of-day reports.
- Automated limit checks that can degrade gracefully when data is stale.
- Clear decision rights for pausing a corridor when flows become abnormal.
- Assume partial failures: one component down, others up.
- Run game days: sanctions hold during peak volume, vendor outage, corrupted reference data.
- Prove you can reconcile after downtime without inventing money.
- Do not sell “instant” if your compliance review is staffed for office hours.
- Do not sell “final” if customers will expect reversibility.
- Tell the truth about disputes and error recovery, up front.
- Which prefunding pools are we trying to shrink, and what replaces that buffer?
- Which cut-offs are we removing, and which internal risk windows will we impose instead?
- Which exceptions will rise, and who will be awake, authorized, and equipped to resolve them?
The danger is assuming these improvements are linear. They are not. Every time you remove one buffer, you increase the importance of controls elsewhere.
Prefunding: yes, you can free liquidity, but only if you trust finality
Prefunding is a symptom of mistrust. You prefund because you do not fully trust timing, finality, or reversibility in the chain you are using.
Tokenized deposits can reduce that mistrust, but only when three things are crisp.
This is where a lot of pilots look good and then stall. The pilot assumes the participants are friendly, limits are loose, and exception volume is near zero. Production is where limits tighten, exception volume becomes non-trivial, and people start asking who eats the loss when something lands in the wrong state.
Cut-off times do not disappear, they relocate into risk and staffing
Everyone wants 24/7 settlement until they remember what 24/7 actually costs.
Cut-offs exist for two reasons: operational capacity and risk containment. When you push the rail to always-on, you remove the external constraint. You do not remove the need to contain risk.
So what happens?
This is the same theme I wrote about in When Microsoft Goes Dark, What Keeps Your Organisation Moving?. Availability is not one system being up. It is the whole chain of people, tooling, and escalation staying coherent under stress.
Nostro and liquidity operations: less idle cash, more intraday precision
Nostro management is where the headline benefit is supposed to show up: fewer funded accounts sitting around, less trapped cash, less manual sweeping.
Tokenized deposits can help. They also change the shape of the liquidity job.
The operational KPI changes too. In a batch world, you chase end-of-day breaks. In a near real-time world, you chase time-in-exception and time-to-recovery.
Exception handling is where “stablecoin-style speed” goes to die
The industry likes straight-through processing diagrams. Real payments are mostly exceptions with good marketing.
Exception handling is not one thing. It is a family of failures, each with a different owner and a different clock.
I have built trading systems where milliseconds mattered, and the surprising lesson was not about speed. It was about cleanup. The system that wins is the one that can explain what happened, reverse what must be reversed, and reconcile the rest without heroics when the week gets messy.
If you want tokenized deposits to work in production, design exceptions first, not last. If your pilot deck does not have an exception section that is longer than the happy path, you are still in storytelling mode.
“Same risk, new API”: the operational risks that simply reappear
Tokenization changes the interface, not the physics. These risks do not disappear. They show up with new names.
This is why I keep coming back to the same discipline across domains: the system is only worth building when it changes something real for people. For tokenized deposits, “real” is not a prettier settlement diagram. It is fewer trapped funds, fewer failed payments, fewer days lost to investigations, and fewer customers stuck between two institutions pointing at each other.
What a European bank needs to ship this safely
Europe is not short on good engineers. It is short on tolerance for unclear accountability in regulated flows. If you want to deploy tokenized deposits under EU reality, build it like a safety-critical product. Here is the checklist I would insist on before broadening beyond a controlled corridor.
1) Define finality, reversals, and liability in writing
2) Build an exception operations layer, not a “support inbox”
3) Treat treasury integration as product scope, not a downstream task
4) Make resilience an explicit design constraint
5) Align customer promise with the hardest edge cases
If you are serious about this, also read Tokenization that survives the weekend: the treasury KPIs that matter. The weekend is where the marketing ends and the system reveals what it really is.
My position: tokenized deposits are a treasury product, not a blockchain project
I like the direction of the Hong Kong and Swift work on connecting tokenized deposits across currencies because it forces the conversation toward bank money and interoperability, not speculative wrappers.
But if you lead a European bank’s effort here, do not start with a chain. Start with three questions:
If you cannot answer those cleanly, you are not building faster payments. You are building faster confusion.
And if you can answer them, tokenized deposits stop being a buzzword and become what they should be: a more direct way to move bank liabilities that reduces dead time for customers and frees liquidity for the real economy.
Newsletter
Working notes, straight to your inbox.
Occasional, no-noise notes from inside the work: what shipped, what broke, and what I would do differently.