The cross-bank transferability problem
A tokenised deposit at Bank A is, by construction, a claim on Bank A. Sending the token to a customer of Bank B does not, in itself, change the issuing bank. Without further mechanics the recipient now holds a claim on Bank A while sitting at Bank B, which is not what they wanted.
Three solutions are in production. The simplest is correspondent-style: Bank B credits a customer's deposit account in equivalent value when it receives a Bank A token, with inter-bank netting and a separate settlement asset (typically central bank reserves) handling reconciliation off-chain. The chain is the transfer instruction; the settlement is the legacy correspondent mechanism. This is how most cross-bank flows actually settle in 2026. It is also why the BIS Garratt-Shin paper on the singleness of money frames the problem at the system level: if Bank A's token does not consistently exchange one-for-one with Bank B's token, the unit of account fragments, and the policy work is to keep the inter-bank rail tight enough that the fragmentation never shows up in the price.
The second is the consortium-ledger approach, where participating banks have agreed in advance to extinguish and reissue tokens at the receiving bank as a single logical operation. The Bank A token is burned; a Bank B token is minted in equivalent value; the inter-bank settlement happens against a netting cycle or, in more sophisticated implementations, atomically against a settlement asset at a settlement bank. Partior operates this model across DBS, JPMorgan, and Standard Chartered for cross-currency PvP.
The third is the tiered ledger, where the inter-bank leg settles atomically against wholesale CBDC (wCBDC) at the central bank's layer. This is the Project Ensemble direction in hong kong, coordinated by HKMA and described in the HKMA Phase 2 release. Project Agorá at the BIS Innovation Hub is generalising the same architecture across borders, with the wholesale-CBDC settlement asset at each participating central bank acting as the connective tissue.
The takeaway: cross-bank transferability is not a chain feature. It is an inter-bank arrangement layered on top of the chain. The chain is the rails; the settlement of the inter-bank leg is the contract.
Reconciliation discipline
The bank's general ledger remains the system of record. The chain entry must reconcile with it continuously, and if they diverge the books govern.
If a chain entry shows a transfer that the general ledger does not, the bank cannot simply update the general ledger to match. The general ledger is the legal record of the bank's deposit liabilities for accounting, prudential reporting, and regulatory purposes. The auditor reads it, the supervisor reads it, the central bank's reserve calculation flows from it. A divergence has to be investigated and the chain rolled back or reconciled by manual journal entry as appropriate.
This is why no major bank has put deposit tokens onto a permissionless chain it does not operate. The reconciliation surface area is unbounded if the bank cannot freeze the chain. A permissioned chain the bank or consortium operates can be paused or rolled back under defined emergency procedures; a permissionless chain cannot. Every production tokenised deposit system has a permissioned holder set, with KYC enforced at the chain level rather than only at the boundary. Chapter V on Permissioned blockchains covers the trade-offs in detail; here it is enough to note that the choice is forced by the reconciliation requirement, not by an aesthetic preference.
A useful operational frame. Treat the chain as one of two ledgers in a typical reconciliation pair, alongside the bank's core. The chain emits transfer events; the core books deposit movements. A reconciliation engine matches them within a defined window. Breaks are routed to a control desk for investigation. Staffing and tooling for that desk is where most of the operational risk capital around a tokenised deposit programme actually sits, and underestimating the headcount has been a recurring lesson in the early production cohorts.
Capital and accounting treatment
A tokenised deposit is a deposit. It sits on the same balance sheet line, gets the same Liquidity Coverage Ratio (LCR) and Net Stable Funding Ratio (NSFR) liquidity treatment, and gets the same prudential capital treatment as a non-tokenised deposit. The token is not a separate asset class for capital purposes.
For the depositor, the tokenised deposit sits in cash and cash equivalents, subject to the existing investment policy on counterparty exposure to the issuing bank. No separate policy line is needed; the existing deposit policy applies, with the chain treated as a transfer rail. This is part of why corporate treasurers have moved into tokenised deposit usage faster than into payment stablecoin usage in jurisdictions where both are available; the policy work is already done.
Custody for clients is custody, with the same basel sco60 cryptoasset standard considerations as custody of any tokenised traditional asset. Group 1a treatment is available where the standard's classification conditions are met, which for a tokenised deposit issued by a regulated bank under a designated system is typically straightforward. Operational risk capital against the chain itself is the additional consideration, sized to chain complexity and ops maturity.
This was not true in 2021. Before SCO60, banks faced the question of whether holding a tokenised deposit on the balance sheet would attract Group 2 capital treatment by analogy to crypto. With SCO60 settled, a bank's own tokenised deposit is a deposit, full stop. The BIS Annual Economic Report 2024 treats this as a settled question, and the live debate has moved on to how Group 1a applies to multi-bank consortium tokens and to settlement-coin constructions where the issuer is not a bank.
The accounting and capital story is, in short, the part of a tokenised deposit programme that needs the least rebuilding. The interesting and unfinished work is upstream of the balance sheet, in the chain design and the inter-bank leg.