A bank paying in a token it issued itself has not settled anything

Settlement is not a synonym for “fast”

Settlement is one of the few words in finance with a precise meaning, and it is not “the money arrived quickly.” It means the obligation is discharged, finally and irrevocably, in an asset the receiver does not have to trust anyone to honour. In a card transaction that asset has for decades been central bank money. The issuing bank’s balance at the Federal Reserve goes down, the acquiring bank’s goes up, and afterwards nobody owes anybody anything. That last clause is the whole product.

SoFi Bank and Mastercard now propose to run SoFi’s entire card programme on a different asset: SoFiUSD, a stablecoin the bank issues itself. The published detail is a headline and a line of summary, so what follows is an argument about the structure, not about an implementation nobody outside the two firms has seen. The structure is enough to argue about, because it has one feature you cannot design away.

You cannot discharge a debt with your own IOU

When the paying party settles in its own liability, the receiving party has not been paid. It has been issued a promise with a redemption process behind it. Banks do this all day — a deposit is exactly that — but a deposit is not how interbank obligations are extinguished, and for good reason. The acquirer holding SoFiUSD at the end of the day is a creditor of SoFi Bank until it redeems. The moment of finality has not been abolished; it has been moved one step later, into the redemption queue, with a token sitting in front of it.

So the central bank does not leave the chain. It is still there, at the end, when somebody wants actual dollars. What changes is that a new intermediate leg has been inserted, and the exposure during that leg belongs to whoever is holding the token. Whether that exposure is trivial or serious depends entirely on things a press release does not tell you: what the reserves are, whether they sit apart from the bank’s general estate, whether token holders rank ahead of ordinary creditors, and who eats the loss if redemption is gated on a Sunday afternoon.

Card settlement is the least forgiving place to find out. An acquiring bank cannot decide to be paid tomorrow instead. Merchants are funded on a schedule. A redemption problem does not stay a redemption problem; within hours it is a merchant funding problem, and merchant funding problems are how payments incidents become newspaper stories.

Four per cent changes the arithmetic

There is a reason this idea is arriving now rather than in 2021. The Federal Open Market Committee lifted its target range to 3.75 to 4 per cent last week. Balances that used to sit overnight at negligible opportunity cost are now worth real money, and the reserves behind a stablecoin earn the money market rate for whoever holds them. Any design that lengthens the period during which settlement value sits inside your own reserve pool rather than someone else’s is worth something, and the number is not small across a full card programme.

I am not saying that is the stated reason; I have not seen the statement. I am saying the interest rate arithmetic is not neutral, and readers should price the enthusiasm accordingly. Float has always been a payments business model. Tokenising it does not change what it is.

The strongest case for doing it anyway

Here is the honest version of the other side. Deposits are already private liabilities, correspondent balances already carry intraday credit exposure, and a Mastercard settlement obligation is already a claim on somebody. Purity about central bank money is a bit rich when the existing plumbing is full of private credit. Cards run every hour of every day; settlement does not. The weekend gap the current system tolerates is a real mismatch, and continuous settlement closes it. And if the reserves are Treasury bills and central bank deposits held outside the bank’s general balance sheet, the acquirer’s claim may be better collateralised than the correspondent balance it replaces.

All of that could be true. Every clause of it turns on documents. Reserve composition, segregation, redemption rights, insolvency ranking, and the network rule that says what happens when redemption fails at two in the afternoon on a business day. The well-built version and the badly-built version look identical from the outside until the first bad day, which is precisely why the documents, not the announcement, are the story.

The advantage of doing this inside a bank

There is one genuinely good thing here. This is a chartered bank, not an offshore issuer, which means a supervisor can simply require the answers as a condition rather than asking politely afterwards. The Office of the Comptroller of the Currency did something adjacent this week, conditionally approving three national trust charters for firms in this part of the market. Conditional is the operative word: with novel plumbing, the conditions are where all the substance lives, and they are public.

The same discipline should apply here, and it should apply before migration rather than after. There is also a question nobody has answered in public and someone should: whether a holder of a bank-issued settlement token has a deposit claim or something weaker. That is not a technicality. It is the difference between two entirely different products wearing the same name.