Banks stopped picking winners because they own every horse

HSBC published a piece this month that would have been unthinkable from a global bank 3 years ago. Digital currencies: an institutional payments toolkit declares the which-token-wins debate over, sorts CBDCs, tokenised deposits and stablecoins into separate lanes by use case, and positions the bank as the guide across all 3. The framing is genuinely useful, and I want to give it credit before I start pulling on the parts that deserve a closer look.

The best paragraph in the piece is not about technology at all. It describes a company with weekend payout obligations, prefunding an account on Friday against forecasted needs, where transferring too little fails the payment and transferring too much strands the liquidity. Tokenised deposits dissolve that trade-off through just-in-time funding from an operational account, at any hour. That is the clearest treasurer-level articulation of the value proposition I have seen from any bank, and it is the right level to argue at. Nobody adopts an instrument because of its consensus mechanism.

The toolkit consensus is real

The piece's core claim matches what the production evidence has been saying for a while. Different instruments are winning different jobs, and the winner-take-all framing died quietly somewhere in 2025. Tokenised deposits carry the wholesale flows where counterparties are known and the regulatory perimeter matters. Stablecoins carry the flows that cross perimeters. CBDCs, where they are moving at all, are moving in wholesale cross-border pilots rather than in anyone's wallet. I have made a version of this argument myself in the rail economics chapter, so on the consensus, no quarrel.

The numbers HSBC brings are worth recording too. mBridge has processed more than 4,000 transactions worth roughly USD 55.5 billion cumulatively, against USD 22 million in its 2022 pilot, per the article's own citations. The bank's Tokenised Deposit Service now spans Hong Kong, Luxembourg, Singapore, the UK and the US across 6 currencies, with AED added in June following the UAE launch. The Ant International partnership moved cross-border tokenised deposit transfers into production on ISO 20022 standards in 2025. These are operational facts from a bank running real flows, and they are more informative than most of what circulates in this category.

Diving deeper into lane assignments

A bank sorting instruments into lanes will, quite naturally, draw the lanes where they suit a bank. The piece assigns tokenised deposits to wholesale transfers and stablecoins to retail payments, where individuals sign up to an app and transact with merchants. That split is cleaner in a bank's interest than it is in the data.

The best available measurement of where stablecoin payments actually happen points the other way. Artemis's survey-based work, the methodology most industry estimates trace back to, put B2B stablecoin payments at roughly USD 226 billion in 2025, with about 60% of measured volume originating in Asia. Corporate treasurers moving supplier payments through stablecoin corridors are not a retail phenomenon, and they are precisely the flows a tokenised-deposit franchise would rather carry itself. From my perspective, the retail lane is where stablecoins threaten a bank's deposit funding least, and that makes the lane assignment worth reading as positioning rather than as a neutral map. The instruments compete at the boundary, and the boundary is being drawn by the incumbent with the most to lose from where it lands.

The hedge itself is not a flaw. It is worth understanding why HSBC can afford to make it so comfortably. The bank now runs a tokenised deposit service, holds 1 of only 2 Hong Kong stablecoin issuer licences, advises central banks on CBDC design, and sits inside mBridge, Project Agorá and Project Ensemble at the same time. A G-SIB (global systemically important bank) stops needing to pick winners once it owns every horse in the race, and that is a genuinely rational strategy. It also means the race card is worth reading with the ownership column in view.

What the piece says quietly

3 details reward a careful read.

• The Canton pilot was a simulation, and the piece does not explain why. HSBC's tokenised-deposit exercise on the Canton Network is presented as a successful pilot paving the way for interoperability. Reporting at the time linked the simulation-only scope to the Basel framework's punitive capital treatment of exposures on permissionless-style chains, a constraint no bank would advertise in its own material. The pilot is real progress. The distance between simulation and live issuance is regulatory, not technical, and that distance is the actual news.
• mBridge and Agorá appear as parallel initiatives, which is diplomatically graceful. One is backed by central banks in Asia and the Middle East, the other is a BIS-convened project whose live phase has run without the New York Fed's participation in real-value testing. Treating them as complementary interoperability efforts elides that they are, in part, competing answers to whose rails carry cross-border wholesale value. A bank operating in both camps has every reason to blur that line. An operator planning corridor infrastructure does not.
• The HKD stablecoin's scope is telling. HSBC's coin, launching in H2 2026 under its HKMA licence, is framed around peer-to-peer payments and subscribing to tokenised investments. That is the retail-and-distribution lane, sitting alongside Anchorpoint's HKDAP in a 2-licence market. The wholesale HKD lane stays with tokenised deposits and EnsembleTX, which is exactly where the lane logic above predicts a bank would keep it.

The part worth keeping

The interoperability section is the one I would underline rather than discount. Fragmentation across bank-specific and chain-specific silos is the binding constraint on institutional scale, the piece says, and that matches both the official-sector analysis and my own view that depth is the one liquidity that cannot be engineered per-issuer. It has to pool. A bank saying this plainly, while building multi-chain plumbing to act on it, is more useful than another interoperability white paper.

Lewis Sun closes with the observation that the best solution might not require the most sophisticated technology, and that is the sentence I would keep. The toolkit era ends the religious wars over instruments, which is progress. The fights that remain are about lane boundaries, and lane boundaries get set by regulation and incentives rather than by technology. Read every bank's map of digital money with that in mind, this one included. It is a good map, drawn by a cartographer who owns several of the territories.