IBM Is Selling the Bridge to Tokenized Deposits Bank Money
IBM has connected its Digital Asset Haven platform to Swift’s blockchain-based shared ledger, in beta, and extended Haven to on-premises deployment on its own mainframes. The company announced both moves related to tokenized deposits on September 24.
It reads as routine enterprise plumbing. The context is not. It lands days before Sibos convenes in Miami. It follows Swift’s declaration, in July, that its shared ledger is ready for initial use. And it arrives mid-race in the payments industry’s hardest problem: moving bank-issued tokens between institutions that do not hold each other’s money.
The prize is real. Tokenized deposits let banks move customer funds around the clock, with final settlement catching up through existing systems. The question is who supplies the plumbing that lets a regulated bank touch that world without rebuilding its operations.
What the Integration Actually Does
Swift’s shared ledger is an orchestration layer, not a payment system. Each participating bank issues tokenized deposits on its own ledger. The shared layer records and validates the payment commitments between banks, so client funds can move overnight and on weekends. Final settlement still happens the old way: through RTGS systems and correspondent banking.
That architecture creates an integration problem, and the problem is IBM’s pitch. Banks instruct payments today in ISO 20022 messages, the standard format behind Swift’s network. Haven’s new ISO 20022 Messaging Adapter lets institutions keep using those messages to command tokenized deposit transactions on the shared ledger. The alternative is blockchain-specific workflows: new code, new audit surface, new operational risk. The adapter, IBM says, means banks “build on existing payment message formats and operational processes.”
One caveat belongs beside the announcement. The adapter is a beta. The Swift ledger itself, by multiple independent accounts, is a minimum viable product — live for initial use, with first transactions targeted later in 2026. IBM is offering a bridge to a destination still under construction.
Why the Timing Matters
Swift unveiled the shared ledger at Sibos in Frankfurt in 2025. It designed the platform with feedback from more than 40 financial institutions and activated it in nine months, on July 9, 2026. Seventeen banks from six continents are preparing to pilot live transactions: ANZ, BNP Paribas, BNY, Citi, DBS, First Abu Dhabi Bank, FirstRand, HSBC, Itaú Unibanco, Lloyds, Mashreq, MUFG, OCBC, Standard Chartered, UBS, UOB and Wells Fargo.
Note the verb. The banks are preparing; none has announced a live transaction. Reuters framed Swift’s launch as an effort to compete with the emerging stablecoin industry. Blockstories, citing people familiar with the work, described the platform as an MVP with scalability still some way off.
The clock matters too. Sibos returns at the end of September in Miami, and Swift will want visible progress on stage.
The BIS’s Project Agorá completed its prototype in May with seven central banks and more than 40 institutions. Agorá differs on the critical dimension: it places tokenized central-bank reserves on the shared platform, letting payment chains settle atomically. Swift’s design leaves final settlement off-ledger. Two philosophies of tokenized money are competing for the same banks, and IBM’s connector is a bet that Swift’s wins — or at least survives.
The On-Prem Half: Custody on the Mainframe
The second announcement extends Haven to on-premises beta deployment on IBM Z and LinuxONE, inside the client’s own data centre, with no public cloud dependency. The solution layer and the key management layer stay entirely within the client’s environment. Keys sit in IBM Crypto Express hardware security modules. Key ceremonies follow formal, auditable processes that produce documentation banks can hand to regulators, with cold storage through IBM’s Offline Signing Orchestrator. IBM’s blog adds that clients evaluate the stack within their own OpenShift environments.
Then there is the number. The release claims “configurations achieving industry-leading 99.999999% availability” — eight nines. The footnote does real work. IBM derived the figure from its own measurements and projections, on a specified stack: LinuxONE Rockhopper 5 machines, z/VM single-system-image clusters, GDPS 4.6 disaster recovery, Red Hat OpenShift 4.14 and a MongoDB test workload. Application-induced outages are excluded.
Treat it as a mainframe marketing benchmark, not a bank’s lived experience with digital asset operations.
The positioning is still legible. Cloud-native custody and wallet infrastructure vendors — Fireblocks, Metaco, Dfns — serve institutions comfortable renting the stack. IBM is courting the banks whose core operations already live in the glass house, and whose risk committees will not approve either.
Where It Fits in the Market
IBM is not competing with the tokenized-deposit issuers; it wants to sell them plumbing. And the issuers are busy. JPMorgan’s Kinexys has run JPM Coin, a tokenized deposit, since 2019, and reports more than $3 trillion in cumulative volume. Its JPMD token went natively to Base in November 2025 and to Canton Network in early 2026.
Citi Token Services runs continuous 24/7 transfers between its own branches. HSBC’s Tokenised Deposit Service operates in five markets and currencies, and HSBC says it connected the service to Swift’s ledger.
Consortium architecture is arriving too. On June 5, The Clearing House announced a bank-led on-chain money initiative with JPMorgan, Bank of America, Citi and Wells Fargo. Reporting targets a shared tokenized deposit network by mid-2027, linked to the RTP and CHIPS rails. In Europe and Asia, Partior, Fnality, Japan’s Progmat and the BIS’s Agorá attack the same interbank problem from different directions.
IBM’s differentiator is not blockchain. Its Hyperledger heritage runs deep, but the levers here are older: an unmatched Z and LinuxONE install base in bank data centres, decades of operational trust, and a security story regulators already understand. Digital Asset Haven launched in October 2025 with wallet-infrastructure partner Dfns, months after the GENIUS Act gave U.S. banks stablecoin clarity. The product is one year old. The distribution channel is sixty.
What the Release Doesn’t Answer
No client is named anywhere. “Participating IBM clients have successfully tested tokenized deposits on the Swift shared ledger,” the release says — without saying who, at what scale, or with what result. None of Swift’s 17 named pilot banks has publicly credited Haven. There are no volumes, no pricing, and no general-availability dates for either beta.
The deeper gap is conceptual. A deposit token from Bank A remains a claim on Bank A’s balance sheet; it is not automatically money at Bank B. Every architecture in this market — Swift’s ledger, Agorá, The Clearing House — is an attempt to solve that settlement gap, and none has yet demonstrated the answer in production. IBM’s connector makes a bank’s arrival on Swift’s network easier. It does not make the network’s core problem easier.
What This Means for Bank Technology Buyers
The due-diligence questions are straightforward. Which of Swift’s 17 pilots, if any, run through Haven? Does the on-prem beta integrate with existing key-management estates, or parallel them? What breaks if Swift’s ledger design changes as it scales beyond MVP — and who maintains the adapter? And how does the ISO 20022 path compare with the integration work a bank could do itself, given that the adapter is, so far, a beta to a pilot?
The strategic read is simpler. Tokenized deposits have crossed from innovation theatre into infrastructure procurement. A bank that buys now is buying the option to participate, not a production capability. IBM’s pitch — keep your messages, keep your keys, keep your mainframe — is precisely calibrated for the slow-moving, audit-driven institutions that will be last in and largest. Whether that conservatism becomes IBM’s advantage, or just the reason the deal closes after the winners are decided, is the question the release cannot answer.
Editor’s Note
This article draws on an IBM press release dated September 24, 2026, and on independent reporting. Company-reported information includes all Haven capability descriptions, the eight-nines availability claim, the unnamed-client testing claim, and the J.P. Morgan Payments statistic on payments modernization. Swift’s ledger timeline, pilot-bank list and architecture come from Swift’s July 9, 2026 press release and reporting by Reuters, Ledger Insights, CoinDesk, ClearingPost and Blockstories; the MVP characterization and Sibos Miami timing reflect that reporting. The Clearing House initiative is documented in its June 5, 2026 announcement and Forbes reporting.
Kinexys figures come from JPMorgan’s own disclosures via The Block and GARP. Digital Asset Haven’s October 2025 launch details come from IBM’s newsroom and coverage by The Paypers, Banking Exchange and SiliconANGLE.
TechRecast could not independently verify any named IBM client using the Swift adapter, transaction volumes, pricing, or the availability figure beyond IBM’s stated measurement conditions.