the most interesting piece of plastic on the planet

I have spent a meaningful chunk of my career around payments, and for an embarrassingly long time, I thought I understood how credit cards made money.

Everybody knows the dinner-party version of the answer. Interests. Fees. Rewards funded by merchants (somehow). The usual words, in roughly the usual order.

Enter, a new job, where I found myself tasked with building a pipeline for processing credit card transactions at scale. Curiosity took over, and I ended up understanding the entire CC industry at a much more granular level than I ever had before.

I'll skip the storytelling here, but in essence, I had badly misread what business I was actually looking at.

From the outside, the system looks insultingly simple. A customer swipes (or more lately, taps) a card. A bank approves it. A merchant gets paid. It feels like a thin layer of software and banking glue wrapped around a purchase.

But the swipe is effectively a toll booth, which leads your money through a chain of institutions, each with their own claim on the transaction and each with a very different set of financial incentives. In this blog, I'll try to encapsulate what I've learned reading about them over the past few weeks. Also, for folks who use a metal credit card and took the title too seriously, fret not, the mechanism works exactly the same.

Layers

Credit cards are built on a layered business model, with at least four interlocking components, each nested within the next.

The outermost layer is payments: a simple, convenient way for merchants to get paid, with a small fee skimmed off each transaction for the service. But that fee supports an economic system far more complex than the payment itself. Lending, distribution, loyalty, and customer acquisition all sit beneath it, drawing value from the same swipe. That is why the system has remained structurally intact despite two decades of regulatory pressure, merchant lawsuits, and fintech disruption: pull on any one thread, and the others tighten around it.

If you think about a single transaction, the customer experiences it as a purchase, and the merchant experiences it as a payment, but underneath that moment sits a chain of intermediaries, each taking a different kind of economic exposure and each earning a different kind of return. Some make money immediately through fees, some over time through revolving balances, some through distribution, and some through the data, loyalty, and customer behavior the system generates. To see why the structure holds together so well, it helps to follow one transaction from end to end and map who touches it, who gets paid, and why.

Anatomy of a $100 Swipe

Suppose you walk into a Nike store and tap your Visa card for a $100 pair of running shoes.

From your perspective, the experience is over in two seconds. A beep, a green checkmark, a receipt. From the system's perspective, a chain of messages has just fired across at least six distinct counterparties, each with a contractual claim on some fraction of that hundred dollars.

The Merchant is where the transaction begins. Nike does not interact with Visa or your bank directly. It has a relationship with a payment processor (like Adyen, Stripe, or Fiserv), which connects it to the broader card network. After a settlement period of one to two business days, Nike receives the $100 minus a deduction called the merchant discount rate - typically 1.5% to 3.5% for a U.S. card-present transaction. On our $100 purchase, Nike might receive around $97.80.

The Processor is the merchant's technical gateway. It routes the authorization request, handles settlement messaging, and ensures the data meets the network's formatting requirements. Indispensable but low-margin - on our $100 swipe, the processor keeps roughly $0.25 to $0.30.

The Acquirer (sometimes the same entity as the processor, sometimes not) is the financial institution on the merchant's side. It underwrites the merchant's risk - fraud, chargebacks, business failure. In modern stacks the processor and acquirer are often bundled (Stripe acts as both), but the function is distinct: the acquirer holds contractual responsibility for the merchant's behavior on the network.

The Network - Visa, in this case - sits at the center. It does not issue cards, hold deposits, or lend money. It operates the messaging rails connecting acquirer to issuer, sets the rules, and charges both sides for the privilege. Visa's cut comes from assessment fees - roughly 0.13% to 0.15% of transaction value - plus a growing menu of charges for authorization, clearing, and settlement. On our $100, Visa collects about $0.13 to $0.15. Small per transaction, but Visa processed over $14 trillion in payments volume in fiscal year 2025 across 258 billion transactions. At that scale, thirteen basis points is a $40-billion-a-year business.

The Issuer is the bank on your side - the one that gave you the card and extended you a line of credit. If you never pay your bill, the issuer eats the loss. For bearing that risk (and funding the interest-free grace period), the issuer receives the largest slice of the merchant discount: the interchange fee. On a standard Visa consumer credit transaction in the U.S., interchange runs 1.50% to 2.30% depending on card tier and merchant category. On our $100 purchase, the issuer collects $1.80 to $2.10.

The Customer - you - pays nothing at the point of sale. The card frequently pays you, in cashback, points, or miles. More on how this is possible shortly.

Where the $2.20 goes

That merchant discount - the $2.20 that Nike does not receive - looks like a single fee. It is actually three fees wearing a trenchcoat.

Interchange (~$1.80-$2.10) flows from the acquirer to the issuer. The rate depends on a thicket of variables: card tier (Visa Infinite Preferred carries higher interchange than Visa Classic), merchant category (supermarkets qualify for lower rates than restaurants), transaction type (card-present vs. card-not-present), and data quality. Visa's published rate schedule runs thirty-plus pages. The issuer is being compensated for fraud liability, credit losses, the grace period, and the cost of running a consumer banking relationship.

Network assessment fees (~$0.13-$0.15) flow to Visa or Mastercard. Historically small, these have been growing faster than payment volumes. The networks have layered on digital commerce fees, integrity risk fees, and transaction processing excellence fees - charges merchants rarely see itemized because they arrive bundled into the total cost of processing.

Processor/acquirer markup (~$0.25-$0.30) is the only component the merchant can negotiate. A large retailer processing millions of transactions might pay a few basis points above interchange. A small shop on a flat-rate processor like Square pays significantly more - a convenience premium for not having to understand the rate table.

When a politician or a merchant group complains about "swipe fees," they are almost always talking about interchange, because it is the biggest component and it is set by the networks rather than negotiated by the market. But interchange is also what funds the credit system - fraud prevention, credit losses, the grace period, and the rewards that make customers want to use the card. You cannot meaningfully cut interchange without altering the incentive structure that supports the rest of the ecosystem.

Which raises the question at the center of credit card economics: if the issuer is paying you 1.5% cashback on a transaction where it earns 1.8% interchange, how does anyone make money?

The answer is that the swipe itself is not where the profit is. The swipe is the acquisition mechanism.

The Revolver Subsidy

The credit card industry has a secret hiding in plain sight: the payment function and the lending function are two different businesses sharing the same piece of plastic, and they have radically different economics.

The industry divides cardholders into transactors (pay in full every month) and revolvers (carry a balance). The split matters more than almost anything else in the business.

Transactors generate interchange revenue but almost no interest income. After subtracting rewards, fraud losses, the cost of funding the grace period, and operating expenses, a transactor is often marginally profitable - and for premium rewards cards, frequently unprofitable. McKinsey estimated the profit per account at roughly $25 for transactors.

Revolvers carry a balance at APRs that, as of early 2026, average north of 20%. Federal Reserve research shows that heavy revolvers - roughly 20% of all accounts - pay more than $60 per month in interest and account for over 70% of all interest income in the system. McKinsey pegged the profit per revolver account at around $240 - nearly ten times the transactor figure. Revolvers make up about 60% of accounts but generate 85% to 90% of issuers' revenues net of rewards.

This disparity reframes everything. The 2% cashback card that a financially disciplined transactor uses to earn free money is economically viable because somewhere in the same portfolio, a revolver is paying 24% APR on a balance they are struggling to pay down. Rewards are not a thank-you for using the card. They are customer acquisition cost for a lending business, laundered through the payment system.

The credit card system is not a payment system that happens to lend. It is a lending system that uses payments as a distribution mechanism. Once you see the distinction, a great deal of otherwise puzzling behavior - the arms race in rewards, the tolerance for unprofitable transactors, the aggressive solicitation of new cardholders - becomes legible as ordinary business logic applied to a lending portfolio.

The Mint in the Sky

If the revolver subsidy is the hidden engine of credit card profitability, airline loyalty programs are the hidden engine of the hidden engine.

When a bank issues a co-branded airline credit card - a Delta SkyMiles card from Amex, a United Explorer card from Chase - the cardholder earns miles per dollar spent. Those miles do not materialize from nothing. The bank purchases them from the airline at a negotiated rate, typically 1.5 to 2.5 cents per mile. When a cardholder earns one mile per dollar, the bank pays the airline roughly two cents, then recovers that cost through interchange and (for revolvers) interest income.

The airline creates those miles at effectively zero marginal cost. A mile is not a claim on fuel or airframe time. It is an entry in a database. The airline can create as many as it wants, sell them at a markup, and control the redemption cost by adjusting award availability, imposing dynamic pricing, or inflating the currency - raising the number of miles required for a flight. Airlines are functioning as central banks for their own currencies, controlling supply, setting prices, and managing the redemption cost of the goods you can buy with them.

The financial results are remarkable. Delta's co-branded Amex relationship generated $7.4 billion in revenue in 2024 - equivalent to the airline's entire operating profit. American Airlines received $6.1 billion from card partnerships. United collected $2.9 billion from Chase. For the major U.S. carriers, loyalty revenue from bank partnerships is now the single largest source of profit.

During the pandemic, when airlines could not fill seats, the loyalty programs kept generating cash - banks kept buying miles because consumers kept spending on co-branded cards even when they could not travel. When airlines needed emergency capital, they used their loyalty programs as collateral. United raised $6.8 billion backed by MileagePlus. American secured $10 billion backed by AAdvantage. Delta borrowed $9 billion against SkyMiles. The implied valuations were, in several cases, higher than the market capitalization of the airlines themselves. AAdvantage was valued at roughly $25.5 billion when American's equity cap was around $6 billion. The market was effectively saying that the planes, crews, gates, and routes were worth less than zero - all the enterprise value resided in the loyalty currency business.

(Today, 57% of all frequent-flyer miles issued in the U.S. come from credit card spending rather than actual flying.)

The Relationship Behind the Swipe

A premium rewards card issued to a transactor who pays in full and uses every perk - lounge access, travel credits, free checked bags - can easily lose money on a standalone basis. So why does the bank issue it?

Because the card is not the product. The relationship is the product.

Banks like Chase, BofA, and Citi evaluate card profitability in the context of the total customer relationship. A premium card brings in an affluent customer who might otherwise bank elsewhere - someone the bank wants for checking, savings, mortgages, brokerage, and eventually wealth management. Chase can offer the Sapphire Reserve at a loss because it is a beacon for high-income customers who bring deposits and mortgage originations. The card is marketing spend. The profit center is the relationship.

This also explains the billions spent on co-brand partnerships. The bank is not buying miles. It is buying access to a self-selected population of high-spending consumers who may eventually want a mortgage or a private banking relationship.

The whole system is a customer-acquisition flywheel dressed up as a payment method.

Why the System Is So Hard to Disrupt

A credit card is a bundle of at least four businesses: payments processing, consumer lending, loyalty currency issuance, and customer acquisition for a banking relationship. Each layer depends on the others in a way that makes the bundle extraordinarily difficult to take apart.

The regulatory approach attacks interchange directly. The 2011 Durbin Amendment capped debit card interchange at roughly $0.21 + 0.05% of the transaction value for banks with over $10 billion in assets. (The Fed had originally proposed 12 cents; the final rule landed at 21 cents - nearly three times the average per-transaction cost of 7.7 cents at the time.) The theory: lower merchant fees would flow through to lower consumer prices. What actually happened: issuers steered customers toward credit cards (higher, unregulated interchange), eliminated free checking, and cut debit rewards. The Richmond Fed found 98% of merchants raised prices or kept them the same.

The Credit Card Competition Act - reintroduced in 2026 with bipartisan support and a White House endorsement - aims to extend routing mandates to credit cards, requiring large issuers to enable processing on at least two unaffiliated networks. The bill has been attached as a rider to everything from the GENIUS Act (a stablecoin bill) to the NDAA, and has not yet passed. But the structural tension - merchants wanting lower fees, banks protecting the lending ecosystem, networks guarding fee revenue - is permanent.

The fintech approach unbundles the payment from the credit. Buy Now Pay Later (Affirm, Klarna, Afterpay) moves short-term lending to different rails, funded by the merchant rather than through interchange. BNPL genuinely unbundles one piece of the credit card, but does not replicate the others: loyalty programs, the ongoing credit line, bank relationship, fraud protection, universal acceptance. Complement, not substitute.

The crypto approach replaces the payment rails entirely. This solves a real problem (cross-border settlement cost) but ignores the reason the network exists: it is not just a messaging system; it is a trust framework - guaranteeing payment, managing fraud, running chargebacks, and scaffolding unsecured lending at scale. The payment rail is the least valuable part of the credit card stack. Replacing only the rail is solving the wrong problem.

The merchant approach builds alternative networks. Walmart Pay, MCX (now defunct), and various ACH-based checkout options have tried to route around the cards. The challenge is the cold-start problem every payment network faces. Credit cards solved this decades ago - first with BankAmericard's mass-mailing campaigns in the 1960s, then with the Visa and Mastercard cooperative networks - and the installed base of over a billion active cards creates a moat no startup has crossed.

The deepest reason the system resists disruption is that every participant's incentive is aligned just enough to make defection unattractive. The merchant pays interchange but gets guaranteed payment, fraud protection, and higher-spending customers. The issuer funds rewards but gets a lending portfolio. The network charges fees but provides the trust framework. The consumer pays nothing but provides spending data and, if they revolve, interest income. Every attempt to disrupt attacks one layer while ignoring the others. But the layers are load-bearing. Remove one and the structure does not simplify - it collapses.

This is why a product that has existed in recognizable form since the Diners Club card of 1950 continues to process tens of trillions of dollars annually, shrugging off each generation of would-be disruptors with the quiet confidence of a system that knows exactly what it is optimizing for.

The next time you tap your card for a pair of shoes, remember that the beep you hear is the sound of a lending business acquiring a customer, a network collecting a toll, a loyalty program minting currency, and a bank evaluating whether your spending habits make you a good candidate for a mortgage. The payment is the least interesting thing happening.

This is the real reason credit cards have survived every wave of disruption: they are not a payment product that happens to do other things. They are a financial operating system that uses payments as an interface. The plastic (or metal) is just the part you can touch.