Interchange Arbitrage Fuels Embedded Finance
Banking-as-a-Service: The $1T Market to Build the Twilio of Embedded Finance
This arbitrage is what turned small sponsor banks from back office utilities into the economic engine of embedded finance. Marqeta and BaaS platforms sit between the fintech and the bank, then route card programs through banks under the $10B asset threshold so the program can earn small issuer debit interchange that is exempt from the standard cap. The spread is then divided among the fintech, the platform, the bank, and the network, with the fintech usually gaining leverage as volume scales.
-
In practice, the sponsor bank supplies the regulated shell, BIN sponsorship, and ledger relationship, while Marqeta or a BaaS platform supplies the APIs that let a customer issue cards, set spend limits, approve or decline transactions, and launch quickly without building a bank stack from scratch.
-
The money flow is simple. A merchant pays interchange on each debit or card transaction, then the network, issuer, platform, and fintech split it. Internal unit economics work shows banks can fall from roughly 20 to 30 basis points to 2 to 3 basis points at scale, because exempt sponsor banks can still earn attractive economics even after giving up most of the spread.
-
This is why scale shifts bargaining power upward. Interviews and market data show the fintech owns the end customer and usually takes the biggest share, while Marqeta like issuer processors and full stack BaaS firms face take rate compression as their largest customers renew and demand more interchange. Marqeta already showed this pattern as revenue concentrated in Square and take rates declined with scale.
Going forward, the winning platforms will be the ones that do more than pass through sponsor bank access. As interchange spreads get competed away, value will move toward the provider that controls onboarding, compliance, fraud, and product speed, while any change to the small issuer exemption would put even more pressure on pure interchange arbitrage models.