The thesis
When a large bank tells an infrastructure vendor "stablecoins today, not tokenised deposits on a third-party network," it is not expressing a technology preference. It is expressing a position about liability location and supervisory control. A tokenised deposit on someone else's network means the bank's own liability settles on infrastructure it does not control and cannot fully see, which is a non-starter for a treasury and a hard conversation with a supervisor. A regulated stablecoin is someone else's liability: the bank can custody it, orchestrate it, and learn the mechanics of always-on settlement without putting its own deposit base on a third party's ledger. Decode the sentence that way and the demand signal stops being confusing.
This is the first piece in a demand-signals series: reading what the buy side is asking for, not just tracking what the sell side ships. The position here is built from practitioner conversations at GSIB-scale digital-asset and treasury seats, read against what those same institutions have actually shipped. Treat it as our interpretation, not a disclosed strategy.
The setup
The supply side of tokenisation publishes constantly, so it is easy to track. The demand side mostly speaks in procurement conversations, and the sentences that escape are compressed to the point of being cryptic. "Stablecoins over tokenised deposits from a third-party network" is a representative example: taken at face value it sounds like banks prefer the instrument with the weaker bank-money pedigree, which would be strange, since a tokenised deposit is structurally the instrument a bank should love. It stays strange only as long as you read it as a ranking of instruments. It resolves once you read it as a ranking of exposures.
Separate two questions that compressed sentence runs together. Whose liability is the instrument? And whose rail does it settle on? A tokenised deposit is the bank's own liability; a stablecoin is an issuer's liability backed by segregated reserves. A rail can be operated by the bank itself, by a consortium the bank co-owns and governs, or by a third party the bank merely connects to. Banks are not ranking stablecoins above deposit tokens in the abstract. They are ranking the combinations.
The argument
Own liability on a third-party rail is the combination that fails
For a bank treasury, a deposit is not just a product; it is the balance sheet. Letting that liability circulate on a ledger the bank does not operate raises questions a treasurer cannot answer comfortably: who can see the flows, who controls the validator set or the operator's change pipeline, what happens to finality in the operator's insolvency, and how the supervisor's operational-resilience and outsourcing expectations map onto a record of the bank's own deposits held outside the bank. None of those questions is about blockchains. They are the same questions that would arise if a bank proposed running its core deposit ledger on an unaffiliated company's mainframe, and the supervisory conversation lands the same way.
The shipped record matches the read. JPMorgan built its own rail (Kinexys) and issues JPMD on networks of its choosing; the DBS and Kinexys interoperability framework is two banks negotiating transfers with conversion at each bank's own perimeter, precisely so that neither bank's liability lives natively on infrastructure the other controls. Where banks do accept shared rails for deposit-like instruments, ownership does the work the bilateral perimeter otherwise would: Partior is co-owned by its founding banks, and Japan's consortium deposit-token programmes (Progmat, DCJPY) run inside a governance and regulatory perimeter the member banks collectively control. The pattern is consistent: banks join rails they govern; they do not float deposits on rails they merely use.
Someone else's liability is a learning vehicle, not an exposure
A regulated stablecoin inverts the exposure. The issuer carries the redemption obligation and the reserve requirements; the bank's role is custody, distribution, settlement services, or treasury usage, all of which let it touch always-on rails and build operational muscle without its own deposit base at stake. The post-GENIUS Act wrapper makes that posture easier to defend internally and to a supervisor, because the instrument now sits inside a recognisable prudential perimeter (the structural convergence argued in stablecoins are bank money in disguise). And the banks moving first are monetising the regime without taking issuer risk: JPMorgan's JLTXX, a registered government money-market fund designed for GENIUS-eligible stablecoin reserves, is a GSIB selling picks and shovels to the stablecoin economy while its own deposit tokens stay on rails it controls.
So "stablecoins today" does not mean banks believe in stablecoins more than in tokenised deposits. It means stablecoins are the combination available today at acceptable exposure: third-party liability, any rail, learning at low balance-sheet cost. Tokenised deposits remain the endgame for the banks themselves, on rails they own or govern; what they are declining is the specific combination of their liability on your network.
Implications for practitioners
For an infrastructure or payments provider selling into banks, the sequencing follows directly. Lead with what banks can adopt without balance-sheet exposure: stablecoin custody, orchestration, reserve management, and settlement integration, which is where the near-term budget is. Pitch tokenised-deposit infrastructure only with a control story a treasurer can repeat to a supervisor: bank-governed deployments, consortium ownership, or conversion-at-the-perimeter interoperability of the DBS-Kinexys shape, rather than a neutral network asking banks to park deposits on it. And watch the GENIUS-reserve products (JLTXX first) as the tell for how banks will monetise the stablecoin economy without issuing; reserve plumbing is a real wedge for third parties. The pattern to internalise: sell banks exposure-free learning first, control second, and never their own liability on your ledger.
Where this could be wrong
Three honest outs. First, the signal base is conversations and shipped behaviour, not disclosed strategy; a bank's procurement sentence can reflect one quarter's compliance bottleneck rather than a structural position. Second, interoperability frameworks could mature into supervised industry utilities the way correspondent messaging did, and if a deposit-token interchange emerges with bank-controlled governance and regulator blessing, the third-party-rail objection fades on a five-year horizon. Third, regulatory friction could flip the ranking from the other side: if stablecoin regimes tighten (interest prohibitions, reserve constraints, extraterritorial reach) faster than deposit-token frameworks clarify, banks may leapfrog the learning phase and push consortium deposit tokens harder, in which case "stablecoins today" turns out to have been a two-year window, not a posture.