SoFi Bank began settling its own Mastercard card volume in a stablecoin it issues itself on 22 September 2026. The disclaimer at the foot of the announcement is the part worth reading twice: SoFiUSD is not a deposit and is not insured by the FDIC.
That combination is new. A nationally chartered US bank, regulated by the OCC, is now clearing debit and credit card transactions in a bearer token that sits outside deposit insurance, against a programme SoFi expects to exceed $25 billion in annualised volume. It is easy to collapse "issued by an FDIC-insured bank" into "an FDIC-insured token", and plenty of coverage has. Those are different instruments with different failure modes, and if you are building treasury or reconciliation logic, the difference is the whole story.
What Is SoFiUSD and What Went Live on Mastercard in September 2026?
SoFiUSD, ticker SOFID, is a dollar-denominated token issued by SoFi Bank, N.A. It launched on Ethereum in December 2025, built on BitGo's stablecoin-as-a-service infrastructure for minting and custody, and expanded to Solana during 2026 as the settlement layer for SoFi's enterprise banking product.
What changed on 22 September is scope. This is not a pilot on a side product: SoFi is migrating settlement for its entire Mastercard debit and credit card programme, with issuers, acquirers and merchants settling through SoFi Bank.
The operational claim is round-the-clock settlement, intraday, weekends and holidays, with merchants withdrawing to cash at no cost through SoFi's Big Business Banking platform. Anyone who has waited for a Monday batch to clear a Friday night of card volume knows why that matters. Same thesis as Mastercard's earlier move to settle cards on Solana, now with a bank supplying the token.
Is SoFiUSD FDIC Insured?
No. SoFi's own language is unambiguous: the token "is not a deposit, is not insured by the FDIC or SIPC", is not bank guaranteed and may lose value.
What holders get instead is a redemption promise. SoFiUSD is redeemable 1:1 for US dollars, backed by reserves consisting primarily of cash. SoFi has signalled that converting the token into tokenised deposits, which would earn interest and carry insurance, is a planned feature. Planned is not shipped. Today it is a bearer claim on the issuer, not a deposit liability.
This matters mechanically because of where the reserves sit: as cash at SoFi Bank, the same entity that issues the token. The token's credit risk and the bank's credit risk are one risk, and the instrument carrying it has been carved out of the insurance protecting the bank's other dollar liabilities.
SoFiUSD vs USDC vs a Tokenised Deposit: Which Claim Do You Hold?
This is the comparison the announcement does not make, and the one that belongs in your risk register.
| SoFiUSD | USDC | Tokenised deposit | |
|---|---|---|---|
| Issuer | SoFi Bank, N.A. (OCC) | Circle (non-bank) | A bank |
| Reserves | Primarily cash at the issuing bank | Short-dated Treasuries and cash, in segregated funds away from the issuer | The bank's own balance sheet |
| Legal character | Bearer token, not a deposit | Bearer token, not a deposit | Deposit liability |
| FDIC cover | None | None | Yes, to applicable limits |
| Pays interest | No | No | Typically yes |
| Who can hold it | Anyone with a wallet | Anyone with a wallet | Usually permissioned account holders |
The trade-off is not "bank good, crypto bad". USDC's reserves are deliberately remote from Circle, short-dated government paper in segregated vehicles, so an issuer insolvency and the reserve pool are separate problems. SoFiUSD's reserves are cash at the issuer, which is excellent for redemption speed and unhelpful for insolvency remoteness. A tokenised deposit closes the insurance gap and gives up the permissionless transferability that makes a token useful as a settlement asset.
Pick the property you need. If you want an asset that keeps working when the issuer does not, reserve remoteness beats a bank charter. If you hold for hours rather than weeks against a regulated US balance sheet, the bank-issued version wins.
How Does SoFiUSD Card Settlement Work on Solana?
The flow is mint-and-burn, and the reserves never leave the bank. An enterprise deposits dollars with SoFi Bank. SoFi mints an equivalent quantity of SoFiUSD against those reserves. The tokens move on Solana to settle. To exit, the counterparty redeems and SoFi burns the tokens, returning dollars.
Solana carries the settlement leg rather than Ethereum for reasons that are pure arithmetic. SoFi cites transaction costs near $0.00025, finality from sub-second to seconds, and thousands of transactions per second at the base layer. Ethereum needs roughly 13 minutes for full economic finality and fees run into dollars at peak demand. Ben Reynolds, who runs SoFi's big business banking, cited cost, speed and throughput. For a card programme pushing $25 billion a year, a dollar of gas per transfer is not a rounding error, it is the business case inverted.
Two things follow for integrators. Finality assumptions are chain-specific, so a confirmation policy written for the Ethereum deployment will be wrong on Solana. And because the token exists on both chains, address validation and metadata must be chain-aware: verify decimals and contract addresses per deployment rather than assuming they match. That assumption quietly corrupts amounts, the same family of bug as minor-unit mismatches in ISO 4217.
Which Blockchains Does Mastercard Settle Stablecoins On in 2026?
Mastercard's settlement framework now spans eight chains: Arbitrum, Base, Canton, Ethereum, Polygon, Solana, Tempo and XRP Ledger. SoFiUSD uses two of them. Integration with Mastercard's Multi-Token Network is described as planned rather than done.
Visa has taken a narrower path, running production settlement on Solana and using Circle's Arc for institutional validation, while separately expanding its stablecoin-linked card work with Bridge towards more than 100 countries by year end. Reported figures put Visa's annualised stablecoin settlement volume near $20 billion, the same order of magnitude as SoFi's projection for one bank's card book.
Mastercard is building a multi-chain menu and letting issuers bring their own token; Visa is concentrating on fewer chains and third-party stablecoins. Chain count is not a feature though, it is a maintenance surface: eight chains means eight sets of finality rules, reorg behaviour and fee dynamics to keep current. I would rather integrate two well-understood chains than a menu of eight.
Is a Bank-Issued Stablecoin Actually Safer? My Take
Not yet, and the marketing is running ahead of the legal structure.
My prediction: the bearer-token form is transitional. Within about eighteen months the serious bank products will be tokenised deposits, insured and interest-bearing, with the permissionless version kept only where it must interoperate with public-chain counterparties. SoFi has effectively conceded this by putting deposit-token conversion on its roadmap. The bearer design survives because it can move to any wallet on Solana today and a deposit token cannot.
The risk I would watch is concentration. A token whose reserves are cash at its own issuer has no insolvency remoteness and no insurance, and it settles a card programme whose merchants did not choose it. That is defensible at $25 billion of annualised flow with one bank in the loop. It becomes a supervisory question if a dozen banks copy the structure and card settlement depends on uninsured intraday balances at individual institutions. Circle's Arc poses the same question from the other direction: purpose-built chain, non-bank issuer.
What This Means for Payment Engineers
Concrete steps if SoFiUSD or a peer token lands in your settlement path:
1. Classify the instrument correctly in your ledger. This is not a cash-equivalent deposit balance. If your treasury model has one bucket for "money at a bank", the token does not belong in it. Bearer tokens need their own account class and credit exposure. 2. Write the redemption path before you need it. Know who can redeem, over what hours, at what minimum size, and how long burn-to-dollars takes in practice. A 1:1 promise with an unclear operational route is a liquidity risk, not a guarantee. 3. Make confirmation policy per-chain, not per-token. Same ticker, different finality. Store the chain identifier alongside every balance and movement. 4. Cap intraday exposure explicitly. Round-the-clock settlement lets you hold the token for minutes instead of days. The safest position in an uninsured bearer instrument is a small one held briefly. 5. Read the disclaimer, not the headline. "Issued by an FDIC-insured bank" is a statement about the issuer. Whether the token is insured is a separate question with, currently, a different answer.
The shift itself is real. Card settlement ran on multi-day batch cycles for decades because no alternative asset moved at weekends. One exists now, and a regulated US bank has put its whole card programme behind it. The instrument design will change; the direction will not.
Written by Tom Wang, a payments engineer working on cross-border and stablecoin infrastructure.