SoFi and Mastercard Just Complicated My Stablecoin Argument
Last week I wrote that I’m a big believer stablecoin rails are going to take over from the card networks. Two days before that piece went up, SoFi had started settling its entire card program in a stablecoin, live on Mastercard’s network, and the piece didn’t account for it. On the surface that looks like a point in my favour, and in one sense it is. It’s the most concrete evidence I’ve seen that stablecoin settlement works at real scale inside a regulated bank. It’s also evidence against the way I framed the argument, and I think that part is more useful to look at than the part that flatters me.
What SoFi and Mastercard launched
Here is what the reporting says. On September 22, SoFi Bank began settling transactions from its debit and credit card programs in SoFiUSD, its own dollar-backed stablecoin, on Mastercard’s network. SoFi is moving its entire card program to this kind of settlement, and The Block reports the program is expected to process more than $25 billion in annualized volume. SoFi says it is the first bank to go live with stablecoin settlement across Mastercard’s network.
SoFiUSD is issued by SoFi Bank, which is nationally chartered and regulated by the OCC. According to The Block, it is redeemable one for one in US dollars, backed by reserves consisting primarily of cash, and available on Ethereum and Solana. SoFi launched it in December 2025, and PYMNTS notes that at launch SoFi described it as running on a public, permissionless blockchain. It became available in SoFi’s banking app in May, and SoFi and Mastercard agreed on plans for this settlement work in March. SoFi’s CEO, Anthony Noto, put the pace this way: “In six months, SoFi and Mastercard took stablecoin settlement from an idea to a live product that materially improves how money moves for businesses.”
The merchant side matters most for what follows. Merchants don’t need to hold SoFiUSD or change their existing payment systems. Through SoFi’s Big Business Banking platform, a merchant can receive settlement funds instantly in a SoFi Bank account and withdraw them as cash around the clock at no cost.
Two more details round it out. Mastercard’s own stablecoin settlement infrastructure spans eight blockchains (Arbitrum, Base, Canton, Ethereum, Polygon, Solana, Tempo and the XRP Ledger), and in June it added USDC, PYUSD and RLUSD among the stablecoins it supports. And both companies say they’re exploring further uses for SoFiUSD on the network, including cross-border payments and remittances, while SoFi is separately discussing stablecoin-based settlement with large US merchants.
What I argued
In Most Agent Payments Won’t Run on Cards, the section heading was “stablecoin rails, not card networks.” The argument under it was that a card payment is organised around a cardholder, that agents paying other agents often have no cardholder in the picture, and that a stablecoin transfer from one wallet to another suits that shape of spending better. I also said, more broadly, that I believe stablecoin rails will take over from the card networks.
I did hedge it. I wrote that the card networks aren’t going anywhere soon, that my view was about direction rather than next year, and that it was an opinion rather than something I’d present as settled. I still think those hedges were right. But the hedges were about timing. The framing itself was a competition between two rails, where one wins and the other loses ground. “Take over from” means instead of. The SoFi launch doesn’t fit that picture cleanly, and I’d rather adjust the view than explain the evidence away.
Where the launch fits my view, and where it doesn’t
A card transaction is several things bundled together, and it’s easy to talk about “the card network” as if it were one. There’s the credential in the customer’s pocket and the fact that almost every merchant accepts it. There’s authorisation, the yes or no at the moment of payment. There’s the rulebook around disputes and chargebacks. There’s the brand that makes a customer trust an unfamiliar checkout. And underneath all of that there’s clearing and settlement, the part where money moves between institutions after the fact.
What SoFi changed is the settlement part. Everything above it, as far as the reporting describes, stayed where it was. A SoFi cardholder still pays with a Mastercard. The merchant still accepts a Mastercard, doesn’t touch a stablecoin, and doesn’t change its systems. The dollars show up in a bank account. From the point of view of the two people standing at the counter, nothing about the transaction looks different. What looks different is what happens behind it: settlement moving onto a stablecoin on public blockchains, instantly and around the clock, rather than on the usual banking-day schedule.
That is stablecoin rails winning something real. It’s just not the card network losing something. If anything, Mastercard comes out of this looking more capable, with the network brand, the acceptance and the merchant relationship all intact and a faster settlement layer added underneath. Mastercard’s own language points the same way. Its Global Head of Digital Commercialization, Sherri Haymond, called the launch “another step toward giving businesses more choice in how money moves.” PYMNTS also reports that Mastercard’s CFO, Ling Hai, said earlier this month that the company is embedding stablecoin orchestration and white-label wallet capabilities into its payments stack.
That last phrase caught my attention for professional reasons. Orchestration is the thing I work in: a layer that sits above several ways of moving money and decides which one a given payment goes through. A card network describing itself as embedding stablecoin orchestration into its own stack is a network positioning itself as that layer, the one that chooses between rails, rather than as one of the rails being chosen. I don’t know how far that plan goes beyond the phrase, and I wouldn’t read a strategy into one reported sentence. But it’s the opposite of how an incumbent behaves when it expects to be bypassed. It’s how an incumbent behaves when it intends to absorb the new thing.
So here is where I’ve landed, for now. I think the “cards versus stablecoins” framing was too binary. For the payments people make at merchants, the more likely shift, based on this launch, is that stablecoins move into the settlement layer first, where the end user never sees them, while the card network keeps the parts people do see. On that path, stablecoin rails can take over a large share of how money moves without taking over from the card networks in the sense I originally meant.
Where I don’t think the launch changes much is the case the original article was built on. SoFi’s program is a person with a card, paying a merchant. My argument about agents was about payments with no cardholder at all: software paying other software, in small amounts, many times a day, with no checkout and nobody deciding in the moment. Nothing in this launch speaks to that. A stablecoin under a card still assumes there’s a card, and the card still assumes there’s a cardholder. For agent-to-agent spending, I still expect most of it to move from wallet to wallet without a card network involved. What I’d now say is that there are probably two paths running in parallel, one where stablecoins sit underneath the card, and one where the card isn’t part of the transaction to begin with, and that last week I wrote as if there were only the second.
What I could still be wrong about
There are ways this could go that would push my view back in either direction.
One bank is one bank. SoFi issues its own stablecoin and runs its own business banking platform, so it controls both ends of the arrangement in a way most issuers don’t. The instant settlement a merchant gets is settlement into a SoFi Bank account, through a SoFi platform. That may say as much about SoFi’s structure as about what stablecoin settlement will look like when other banks follow, if they do. Mastercard’s broader infrastructure, with eight chains and several third-party stablecoins, suggests it expects more than one participant, but a launch with one bank tells me the model works for that bank and not yet much more.
The detail that could pull things back toward my original framing is SoFi discussing stablecoin settlement arrangements directly with large US merchants. The reporting doesn’t say whether those arrangements would run through Mastercard’s network or around it. If a bank with its own regulated stablecoin starts settling with large merchants directly, the network’s position in that relationship becomes a real question, and that would be a much more direct test of “instead of” than anything in this launch. I’m not going to guess which way it goes. It’s the part of the story I’d watch.
It’s also worth being clear about what the reporting doesn’t cover. Neither source says anything about the fees merchants pay on these card transactions. As far as I can tell, a change in what settlement is denominated in doesn’t touch interchange, which follows the card and the transaction and is set through the network’s schedules. I wouldn’t assume a merchant’s card costs move because the settlement layer changed, and the fight over the interchange settlement is a reminder that those costs are set by whoever holds leverage over them, which has little to do with how efficiently the money moves underneath.
Does a public chain underneath help merchant reconciliation?
In the earlier article I made a second argument, that a public chain changes what it takes to know what was spent, because the record sits somewhere anyone can read it rather than inside one company’s systems. It’s fair to ask whether that carries over here, with a card network’s settlement now running on public blockchains.
My answer is mostly no, at least not by itself. The visibility argument depended on the payer’s wallet being on-chain and readable by whoever operates it. Here, the merchant doesn’t hold SoFiUSD and doesn’t touch the chain. It receives dollars in a bank account. The reporting doesn’t describe which transfers in this arrangement appear on-chain, or whether merchants are given any way to map an on-chain settlement back to their own transactions. Without that mapping, a public record doesn’t help a merchant much. As I wrote before, a public record is not the same as a usable one, and that applies even more when the merchant never sees the record at all.
What does change for a merchant is timing. Reconciliation in card payments is largely the work of matching what was authorised against what settled, and when. The gap between those two events, and the batch schedules that sit inside it, is a good part of what makes the matching tedious. Settlement that arrives instantly and can be withdrawn at any hour shrinks that gap. It doesn’t remove the matching, because the merchant still has to tie each settled amount back to the sales that produced it, but it should make the timing side of that work simpler. That’s a smaller benefit than the one I described for agents on public chains. It’s still a real one, and it’s the kind that tends to matter more to a finance team than a headline about blockchains.
If I rewrote last week’s section heading today, I wouldn’t call it “stablecoin rails, not card networks.” I’d say stablecoin rails are coming for settlement first, underneath the card, and for agent spending directly, without it. The first half of that is now live at a US bank. The second half is still the bet I’m making, and I’m less sure than I was that the two halves will end up looking like the same shift.