Costs and pricing

How the money moves

Merchants pay for most consumer card programs, through interchange. The program keeps what's left after the network, the processor, the bank, the card maker, the identity vendors and the wallet take their share.

The bank's split of interchange is the biggest single negotiation. One banking-as-a-service vendor's example gives the fintech 70% and splits the rest evenly between the vendor and the bank; The Bancorp's filing suggests something closer to 6-8 basis points of volume for the bank. On debit the bank also earns the spread on deposits and may share some of it.

How companies of different sizes use it

I measure size here by active cards rather than headcount, because a program's costs and choices follow the cards in use, and a small fintech can run a million cards while a large retailer runs none. Each label names the company that usually sits in that band. Nobody publishes card programs by size, so the bands and costs below are my rough estimates, built from the costs above and public filings. Spend per card matters too: a payroll card that workers drain on payday behaves like a program one band smaller.

A startup (under about 10,000 active cards) is usually an early fintech or a software platform adding cards to its product. It issues stock or virtual cards through an embedded issuing platform such as Stripe Issuing, Adyen, Lithic, Highnote, Increase or Column, whose bank holds the BIN and whose compliance program covers the startup. There are no startup credits; past the free sandbox it pays per card, and the platform keeps a share of interchange. A founder or the head of product owns it. My rough estimate is $40,000-650,000 a year in program costs, mostly netted out of interchange by the platform rather than invoiced. It moves on when the interchange it gives up grows bigger than what its own bank and processor would cost, or when it needs something the platform won't do, such as its own ledger, credit or a second bank.

A scale-up (about 10,000 to 250,000 active cards) is a growth-stage fintech or vertical software company, and the payroll platform sits here in its first years. It runs its own program on a sponsor bank such as The Bancorp, Pathward, Sutton or Coastal, with a processor such as Marqeta, Galileo, Lithic or Highnote (or a managed program where the processor brings the bank), plus custom plastic, identity checks and fraud tools. A head of card program owns it, with the compliance officer the bank requires and finance running the daily tie-out. My rough estimate is $300,000 to $15 million a year in program costs, before people. This is where the bank's Durbin status decides the business case: by my rough model of the payroll platform, a program that comfortably covers its variable costs at an exempt bank barely covers them at a capped one, before paying for people and compliance. Its biggest decision is the bank contract, with its minimums, interchange split and exit terms. It moves on when depending on one bank starts to feel dangerous, as it did for programs frozen by the 2024 orders, or when the bank nears the Durbin threshold.

A mid-market program (about 250,000 to 2 million active cards) usually belongs to a later-stage or listed fintech. It adds another sponsor bank before it needs one, keeps a ledger both banks can read every day, and sweeps deposits across small banks so none crosses the Durbin threshold. It negotiates processor tiers, sometimes runs a processor migration, and toward the top of the band earns network incentives large enough to offset much of its network fees. Mercury's move to Choice and Column after Evolve's order is the kind of project this size has to be ready for. A payments or bank-partnerships lead owns it, with a chief compliance officer and separate fraud, disputes and treasury teams. My rough estimate is $5-60 million a year in program costs. It moves on when processor fees grow big enough that building its own processor pays, and its volume gives it real leverage over its banks.

An enterprise program (about 2 million to 20 million or more active cards) belongs to a large neobank or payments app: Chime, Dave, Cash App. Chime now runs its own processor and ledger, while Dave still runs on a sponsor bank, Coastal, which also lends to it. Member counts overstate card scale, since many members barely use the card. The CFO or COO owns the bank relationships and a platform engineering group owns processing. My rough estimate is $50-500 million a year in processing, network, bank, card and fraud costs; no filing I read breaks them out. The next step is a bank charter, which Nubank, Mercury and Revolut have been conditionally approved for, with the catch that a chartered bank above the Durbin threshold gets capped like any other.

A Fortune 500 brand (revenue above about $7.5 billion) is different in kind: it doesn't run a card program, it rents one. A big issuing bank holds the BIN, the balance sheet, underwriting and servicing, and the brand owns the customer and the rewards, sometimes through a fintech arm: Walmart left Capital One, and its cards now come from Synchrony through Walmart's OnePay. I haven't found good data on who owns this inside the brand; I'd guess finance and the loyalty team, while the bank's partnership team runs the program. The money runs the other way here: the bank pays the brand, and American Express paid Delta about $8 billion in 2025. There's no next size up. The contracts are long and exclusive, and when one ends the brand re-tenders it and the new issuer buys the card book. A big company's own corporate cards are in B2B payments: spend management.

What changes as a program grows (costs are my rough estimate):