Interchange Economics

Eric Glyman offers one of the clearest plain-English breakdowns of card economics available. In card businesses there are two basic ways to make money: a transaction-based model, interchange, and a lending model, interest on revolving balances. Interchange is the fee that moves every time a card is swiped.¹

Who keeps what on a swipe

| Party | Role | Net take, approximate | |---|---|---| | Merchant | Sells the good | The lion's share | | Processor, or acquirer | Accepts and routes the payment | Roughly 0.1 to 0.5 percent net, though the gross headline rate is closer to 2.9 percent plus 40 cents; most of that is passed through | | Merchant bank | Holds and deposits the funds | Roughly 10 cents | | Network | The card brand | Roughly 0.1 to 0.4 percent | | Issuer plus issuer-processor | The bank whose brand is on the card | Most of the interchange |

The headline rate a merchant sees, something like 2.9 percent plus 40 cents, is gross: the processor collects a lot but keeps little, since it pays the networks and the other parties in the chain.¹

Why the issuer keeps the most

Because the issuer takes the risk and the cost. The issuer commits to paying the merchant even if the customer does not pay the issuer back, and if the customer later defaults, that loss sits with the issuer. The issuer also bears the operational cost of standing up and running the card program. This is the structural reason the issuer occupies the most lucrative seat in the chain, and it is why some fintechs choose to become issuers themselves rather than simply route payments through one.¹

Historical note

These rates used to be far higher. In early-1900s department stores, a bank would set up shop inside the store itself, and interchange could run 5 to 6 percent. The compression to today's sub-1 percent issuer take is a story of scale, networks, and competition, tracing back through the history of American credit from A.P. Giannini's early California banking, through Bank of America and the creation of BankAmericard, and eventually to Visa.¹

The co-brand variant

The co-brand or partnership business, a retailer-branded card or a university card, is interchange applied to a store's existing customer loyalty: convert a small fraction of loyal customers to a branded card and earn interchange when they spend both with the retailer and everywhere else. These are powered by large issuers, among them Synchrony, Capital One, American Express, Barclays, and MBNA, and in aggregate represent tens of billions of dollars in business.¹ Building a modern creator or small-business co-brand card was judged, by at least one operator who considered it, to be a decades-long build rather than a quick win.

Interchange as one leg, not the whole stool

Interchange, from card spend, is still Ramp's largest single revenue line, but the contribution-profit mix behind it has flipped. By Glyman's later account, card was more than 90 percent of contribution profit a few years ago, and lines two through five, bill pay and software, treasury, procurement, and travel, become the majority by year end. Bill pay and software are predominantly a float and foreign-exchange business rather than an interchange business, while treasury is predominantly a deposit and yield business. The interchange primitive that founds many of these companies becomes the wedge rather than the destination, with the platform increasingly monetizing through float, yield, and outcome-based pricing rather than transaction fees alone.

The agent-scale stress test

Interchange economics come under real strain at machine scale, and the argument belongs to Sam Broner of a16z crypto, who made it in an essay called "Tourists in the Bazaar." Cards carry a fixed floor per transaction plus a percentage on top. Broner works from a floor of roughly 30 cents, a little under the 40-cent fixed component in the merchant-facing headline above, and the combination implies an effective sweet spot of somewhere between 20 and 1,000 dollars. Below that range the fixed floor dwarfs the purchase, so an agent buying a two-cent API call cannot pay a thirty-cent fee, fifteen times the price of the thing it bought. Above that range the percentage becomes a large tax that a party moving tens of thousands of dollars will route around. Agents making thousands of small, high-frequency payments live precisely in the tails that cards were never priced for, which is Broner's structural argument for why stablecoin rails, rather than card rails, are likely to win the net-new payment flows generated by autonomous agents.