On Saturday, Sept. 5, DBS Bank and Citi used tokenized deposits and a new ledger built by the Society for Worldwide Interbank Financial Telecommunication, or SWIFT, to complete a U.S. dollar transfer from Singapore to New York in minutes. The significance was not only the speed. It happened on a weekend, and it used blockchain-based coordination between banks.
In the article, Simon Taylor argues that banks are trying to make account balances move around the clock through tokenized deposits without ripping out the aging core systems that still run much of global banking. For institutions that generate billions of dollars in payment revenue, this is not a side project. It is a defensive move in a market they cannot afford to lose.
SWIFT was not the main bottleneck
Taylor writes that SWIFT has often seemed oddly named, because the network itself was never the only reason cross-border transfers felt slow. SWIFT sends messages. It does not move funds. The actual transfer of value depends on what the banks on each side can do with those messages once they receive them.
These efforts are still in the pilot stage. Even so, the article says it is a mistake to dismiss them as little more than a database swap. A single bank can modernize its own internal records. Getting Citi, HSBC, and DBS to confirm obligations to one another across systems, countries, and time zones is a different problem. That is the kind of coordination challenge blockchain systems are built to handle.
Corporate clients, in Taylor’s view, are not looking to abandon banks altogether. Treasury teams value the credit lines, FX services, and human support their banking partners provide when something goes wrong. What they want is simple: the ability to move money on weekends as well.
Why weekend transfers have been so hard
A bank deposit is, at its core, a database entry that represents a liability owed by the bank to the customer. Sending money to another bank means sending instructions and updating records across more than one institution. Those updates do not happen all at once.
Most book-entry operations are still processed during weekday business hours, and each country runs on its own timetable. That is one reason settlement typically stops over the weekend.
Taylor compares the setup to email. A message can arrive instantly, but the person who needs to read and act on it may only do so during office hours. In this analogy, the giant mainframe systems inside banks are the people checking the inbox.
If the sending and receiving banks do not have a direct relationship, correspondent banks get involved. That is the core of correspondent banking: one bank maintains accounts with another so payments can be completed on behalf of clients. Large correspondent banks also provide liquidity, FX conversion, and compliance screening.
Even in a simplified U.S. dollar payment chain where the sending and receiving banks share the same correspondent bank, each stage has cutoff times. Miss a deadline by a minute and the payment can roll into the next batch. Miss the window on Friday and the transfer may not move again until Monday.
Time zones make the process harder. One bank’s 4 p.m. may be another bank’s 4 a.m. Add multiple currencies, several local payment systems, more than one correspondent bank, and each institution’s own risk controls, liquidity rules, and processing windows, and the setup becomes far more complex.
That is why SWIFT messages can arrive right away while the actual transfer still waits in line. A single bank may be running 10, 20, or even 30 separate core banking systems across different countries.
Why banks are not simply replacing core systems
Any realistic solution has to work with those legacy systems. Taylor writes that no one is going to make a large mainframe in Jakarta process transactions on a Sunday. The common suggestion, just upgrade the database, ignores how tangled these environments have become.
He describes a conversation with the chief architect of a large global bank. That bank operated in more than 60 countries, each with its own mainframe stack. Many of the engineers who wrote the underlying code had already retired. Over decades, the systems had been adapted to local regulatory requirements and connected to thousands of surrounding services. The bank, according to Taylor, could not even map every dependency across the environment with confidence.
Replacing a core banking system is an enormous task, and business cannot stop while it happens. With more than 60 independent systems in play, the more practical strategy is to reduce dependence on old mainframes over time, so new features do not require a global rebuild every time.
That leads to a different model. Instead of rebuilding the core, banks add a new road around it.
What tokenized deposits are doing
Tokenized deposits are onchain representations of bank deposits. They are still bank liabilities. What changes is the way those balances can be moved: outside traditional operating hours and under programmable payment logic.
The article says these services are not replacing existing core systems. They run alongside them, which some in the industry call a “side core.” The technical setup can vary, but the basic idea is straightforward. A bank segregates customer funds, creates the corresponding tokenized balance, and lets clients move that balance through a new service while ensuring records stay consistent and the same money cannot be spent twice.
Taylor points to JPMorgan’s Kinexys, Citi’s token services, and HSBC’s tokenized deposit offering as examples. Within the supported markets and currencies, corporate clients can move funds around the clock.
What the client actually gets: payments on Saturday
The article frames the benefit in operational terms. A supplier may not ship goods until payment is received. If money cannot arrive over the weekend, the buyer either prepays and ties up capital, borrows to cover the gap, or waits. Delivery slips. Treasury teams then have to plan funding around cutoff times for each currency and jurisdiction.
Ant Group’s international Alipay business is presented as one of the deepest real-world deployments so far. Its treasury platform, Whale, is connected to tokenized deposit services from HSBC, Standard Chartered, DBS, JPMorgan, and at least five other banks. Through that setup, Ant can move U.S. dollars, euros, Singapore dollars, Hong Kong dollars, and offshore renminbi between Ant entities at those banks on any day and at any time.
That arrangement required integrations with nine banks, which meant nine separate system integrations. Ant had the engineering resources to do that. Most corporate treasury teams do not.
Even Ant, the article notes, can currently move funds only between its own entities within the same bank. If an Ant entity at HSBC wants to pay a supplier banking with DBS on a Saturday, HSBC’s side-core system has to interoperate with DBS’s side-core system. That raises the central question: whose database should be the source of truth?
The harder problem is coordination across banks
A single bank could build its side core on PostgreSQL or another traditional database, Taylor writes. Tokenization is not the only path to weekend settlement. Banks could build extra payment services on conventional infrastructure if they chose to.
The difficult part starts when those services need to talk to other banks. Once Citi, HSBC, and DBS each offer a 24/7 internal transfer layer, customers will want to pay across them. The banks then need a shared way to confirm which payments are prefunded, which transactions are complete, and what each institution owes the other.
If Citi hosted the ledger, HSBC would have to trust Citi’s database administrators. If SWIFT ran a conventional centralized ledger, every participant would have to trust SWIFT. In Taylor’s telling, that would effectively turn SWIFT into a bank, and arguably the most important one in the world.
The value of a SWIFT ledger, then, is that it offers a shared record all banks can trust without placing control in a single institution’s hands. That is where blockchain is useful. It does not move the money itself. It keeps records aligned among parties that either do not fully trust one another or do not want their shared data housed under a single commercial owner.
Even for those comfortable with centralized systems, blockchain features still matter. Each bank holds its own keys. Every instruction must be signed. No one can fake Citi’s payment commitment. Payments can also be handled as a state machine, moving through stages such as initiated, prefunded, and completed. Smart contracts lock in the rules so states change only when the required conditions are met, and each bank can verify those conditions on its own.
In that model, there is one global ledger record visible to everyone, and no one can quietly alter it later. The SWIFT ledger supplies the shared record and the rules for how payment states change. Key signatures, state transitions, and multi-party consensus are the features doing the heavy lifting here.
How a Saturday payment works in practice
Taylor gives a simple example. You hold an account at Bank A. Your supplier banks with Bank B. It is Saturday.
You send a $100 payment instruction to Bank A. Bank A checks your balance and authorizes the payment. It then sends a signed message to the shared ledger stating that Bank A owes Bank B $100 for the supplier payment. The ledger verifies the conditions and records the obligation. Bank B sees the update in real time and credits the supplier’s account.
The supplier gets usable funds within minutes instead of waiting until Monday.
No asset is moving across chains here. No token is transferred from Bank A to Bank B. Your deposit stays recorded at Bank A. The supplier’s new deposit is recorded at Bank B. What moves across the ledger on Saturday is a signed IOU between the two banks. Taylor says the tokenized deposit transaction previously completed by HSBC and Standard Chartered worked on the same logic: records of obligations stay within each bank’s own systems, while SWIFT matches the transaction and handles net settlement.
Final settlement still happens on Monday
Bank B credits the supplier before funds are actually delivered from Bank A. That means Bank B either keeps prefunded liquidity ready for possible weekend flows, which ties up capital, or extends short-term credit to Bank A and takes on the related risk.
Taylor notes that this is not radically different from how the current SWIFT system already works. The difference is that banks have not had a practical way to reflect that flow to clients faster, even if they wanted to.
On Monday, the banks use existing payment rails to complete final settlement. If Bank A owes Bank B $100 while Bank B owes Bank A $80, they only need to settle the $20 difference. That is net settlement, which banks strongly prefer because it is more efficient than moving the gross amount back and forth.
The ledger, in other words, records performance commitments and checks conditions, but it does not supply the money itself. The risk does not disappear. It is documented, signed, and made easier to net. Taylor describes this as an upgraded form of correspondent banking with better record-keeping, not as a fundamentally different category.
Banks are using this to answer stablecoins
SWIFT is not the only group building this kind of infrastructure. Taylor says The Clearing House, jointly owned by JPMorgan, Bank of America, Citi, and Wells Fargo, plans to launch a similar network in the first half of 2027. Some banks refer to it as a “bridge.”
With these systems, a bank’s own 24/7 payment service can connect to other institutions and create a cleaner, programmable, end-to-end workflow. The article presents that as part of banks’ response to pressure from stablecoins.
Large corporate clients are already comfortable with banking relationships, not because banks are the most innovative organizations or have the best interfaces, but because they are reliable and can solve complicated problems. For a multinational company, a banking partner may provide credit, cross-border cash management, FX pricing, and a real person to call when something breaks. The client may not want to leave. It may simply want better service.
The revenue at stake is large. Citi’s Treasury and Trade Solutions business generated $21.3 billion in revenue in 2025, about a quarter of Citi’s total revenue, according to the article. Incremental revenue in that segment was $1.6 billion, contributing more than one-third of the group’s revenue growth. JPMorgan’s comparable unit generated $19.3 billion in 2025, about one-tenth of managed revenue, up $1.2 billion from the prior year. That increase accounted for roughly one-quarter of the group’s revenue growth.
Fintech firms such as Airwallex and Nium, along with global platforms like Airbnb and Spotify, are identified as a customer segment that values innovation. They generate large cross-market payment volumes, care deeply about API access and pricing, and do not accept an answer that amounts to “come back on Monday.”
That demand has created room for companies such as Augustus. Taylor says Augustus is building a 24/7 correspondent banking platform that extends dollar and euro accounts into Latin America, Southeast Asia, and Africa. Augustus also says it is building a native stablecoin core banking system, which the article presents as one competitive route.
Taylor’s view is that better technology and lower unit costs could pull some clients away from incumbent banks and toward new transaction banks or fintech platforms such as Airwallex. The banks’ response is to make the existing bank account more useful while keeping the credit, FX, and service layers clients still need. The speed with which banks are moving reflects how seriously they take that threat.
Stablecoins still matter outside the banking system
The article does not frame the market as a binary choice. In finance, Taylor writes, firms are often partners and competitors at the same time. A single customer can use both models.
Stablecoins still offer something distinct. If the recipient can accept tokens, the payer can send money to an individual or business outside the banking network. That matters most where banking coverage is thin or expensive. But using stablecoins does not replace the full service set companies obtain from banks.
The key difference is whose liability the customer ends up holding. Taylor expects global platforms to use both. Operating cash may stay with banks, and bank borrowing may continue, while stablecoins are used to pay overseas suppliers. Those suppliers can then choose whether to hold the stablecoin or convert it into bank deposits.
The next fight is over payment flow
Taylor closes with a broad direction of travel: bank deposits are moving toward 24/7 availability, first inside individual banks and then across networks of connected banks. Money will be able to move between bank deposits and stablecoins across institutions, currencies, and networks.
In that setup, tokenization gives money constant availability and programmability. Stablecoins create an open payment loop. SWIFT, in his telling, is finally starting to live up to its name. What remains contested is who captures the payment flow.
