Payment Networks: Toll-Booth Economics Without Credit Risk

Ask ten investors what a credit card stock is and nine will name a company that lends money. The tenth will name a company that never lends a dollar. Both are right, and that gap is the whole subject of this post.

When you swipe a card, two very different businesses get paid. One extends credit, funds the purchase for weeks or months, and bears the loss when the cardholder doesn't repay. The other carries the message that says "approve this" from the merchant's bank to the cardholder's bank and back, takes a sliver of the transaction, and goes home. The first is a lender. The second is a toll booth. Most screens file them under one heading anyway.

I'd argue that's the single most expensive classification shortcut in the financials sector. The industry structure underneath a Visa and the industry structure underneath a mid-tier card lender have almost nothing in common: different capital needs, different cycle exposure, different ways to lose money. If you can't say which side of the split a company sits on, you can't say what its earnings are really made of.

The Toll Booth and the Lender

Start with how a card payment actually moves. The merchant's bank (the acquirer) sends the transaction through a network. The network routes it to the cardholder's bank (the issuer), which decides whether to approve. Later, the network handles clearing and settlement between the two banks. The network owns the rails. It does not own the loan.

That distinction decides who carries which risk. If the cardholder defaults, the issuer eats the charge-off, not the network. If purchase volume rises, the network earns more without committing a cent of balance-sheet capital. The economics resemble a toll road: the operator collects per vehicle, indifferent to whether the driver is rich or broke, as long as the traffic flows. We made the same structural point about pipelines in our look at midstream energy pipeline moats, where fee-based volume beat commodity exposure. Same logic, completely different sector.

Scale shows up in the filings. Visa's fiscal 2024 Form 10-K (year ended September 30, 2024) reports net revenue of about $35.9 billion, up 10% from the prior year, on roughly 233.8 billion processed transactions. Net income was about $19.7 billion. Think about that ratio for a second: more than half of every revenue dollar reaching the bottom line, in a business with no loan book to reserve against. A typical bank can't get within shouting distance of it.

A crude division of revenue by transactions gives around 15 cents per processed transaction. Treat that as a rough guide only, since some network revenue (data processing, services, international fees) isn't tied one-for-one to a processed transaction. But the order of magnitude tells you what the business is: tiny fees, enormous count.

Where does the money come from, then? The network charges fees on the volume it carries and on related services: authorization, clearing, currency conversion on cross-border purchases, and data products sold back to banks and merchants. A separate fee, interchange, flows from the merchant's bank to the issuer and compensates the lender for credit risk and rewards. It is set within the network's rules, but the issuer collects it, not the network. People mix those two up constantly, and the mix-up is why they credit Visa with fees it never keeps.

Three Competitive Roles in the Card Business

Grouping companies by what they actually do, rather than by the logo on the plastic, makes the competitive landscape much easier to read.

• Pure network operators: Visa and Mastercard. They set the rules, run the authorization and settlement plumbing, and earn fees tied to payment volume and transaction counts. They don't issue cards to consumers and don't take credit losses.
• Issuer-only players: JPMorgan Chase, Citigroup, Bank of America, and private-label specialists such as Synchrony. They put their name on the card, fund the receivables, earn interest and fees, and carry the charge-offs. Most of them ride on someone else's network.
• Hybrid issuer-networks: American Express and Discover, each of which runs its own network and lends to its own cardholders. Capital One has moved toward this group as well, following its acquisition of Discover. Hybrids capture both revenue streams, and both risk streams.

The hybrids are the interesting middle case, and the one I'm least comfortable generalizing about. A closed-loop network that also underwrites the credit gets richer data and tighter control over the merchant and cardholder relationship. It also owns the downside when a recession hits. Whether that trade is worth it depends on the cycle you're in. Nobody gets both the toll and the insulation.

A short note on what each group is really paid for. Networks are paid for coordination: making sure a card issued in one place works in another. Issuers are paid for risk: pricing and funding a revolving loan to millions of people whose finances change every month. Those are different skills and they reward different temperaments. One compounds quietly, the other has to be underwritten with discipline every year.

Why the Toll Booth Is Hard to Rebuild

A network is a two-sided market. Merchants accept a card because cardholders carry it, and cardholders carry it because merchants accept it. Break the loop for a new entrant and you need to win both sides at once. That's the textbook case for what we covered in network effects as a moat source, and payments may be its cleanest real-world version.

Add scale. Fraud models improve with transaction volume. Issuers and acquirers have spent decades integrating with the dominant networks. Switching costs aren't dramatic for any one party, but coordinating a switch across thousands of banks and millions of merchants is another matter entirely. And so the incumbents hold.

Not that the position is untouchable. Regulators have long scrutinized interchange fees, and merchants have litigated them for years; a rule change or settlement that trims fees or loosens routing rules would hit revenue directly. Real-time account-to-account payment systems are another potential bypass of the card rails, though adoption in the US has been slower than the evangelists predicted. I don't have a confident view on how fast either pressure builds. The honest read is that both are real and neither has broken the model yet.

There is also a secular side to the story that deserves a plain statement. For decades, card and digital payments have taken share from cash and checks, and that shift is a genuine secular tailwind for volume. But a long-running trend can hide the cyclical wobble inside it. Volume growth in a recession year is slower, not negative for long, and investors who confuse the trend with the cycle tend to overpay at the top.

Metrics That Separate the Two Businesses

Different business, different dashboard. Using the wrong one is how analysts end up confused about what a company's growth means.

• For networks: payments volume growth (Visa reported 8% growth in constant-dollar payments volume for the twelve months ended September 30, 2024), processed transaction counts, cross-border volume, and net revenue per transaction or take rate.
• For issuers: loan growth, net interest margin, net charge-off rate, delinquency trends, and the reserve build or release running through the provision line.
• For hybrids: all of the above, plus how much of the volume on their network comes from their own cards versus third-party issuers.

Notice what's missing from the network list. No charge-off rate, no provision for credit losses, no funding costs. Their earnings track consumer spending, not consumer solvency. Spending is a secular tailwind as cash and checks keep losing share to cards and wallets; it also has a cyclical headwind, because nominal volume slows when consumers pull back and cross-border travel is the first thing to wobble. Issuer earnings carry the same spending sensitivity and then add a second, nastier layer: credit losses that tend to arrive exactly when volume weakens.

Risks the Toll Booth Doesn't Escape, and Where Issuers Fit

No credit risk doesn't mean no risk. Networks are exposed to regulation of interchange and routing, to litigation, and to disruption of the payment rails themselves. Their growth depends on consumer spending and cross-border flows, which means a deep recession bites revenue even though no loan goes bad. There's also concentration: two dominant networks means each is a large, visible target for antitrust attention.

Issuers have the opposite problem. They're not short on moat in every case, but their returns are hostage to the credit cycle. A card book underwritten generously in a hot economy can look brilliant for two years and dreadful in the third. The reserves analysts watch are estimates, and estimates move. For a broader view of how bank balance sheets absorb these swings, see our overview of US banks industry structure.

One more thing worth separating out: settlement risk. Networks that guarantee settlement between banks take on a form of counterparty exposure in stress, which is a different animal from consumer credit. If you want the mechanics of how counterparty exposure works in contracts more generally, our explainer on what a derivative is covers the logic. For networks it's a tail risk, not a driver.

Choosing a Spot in the Value Chain

Issuers are not the villains of this piece. A well-run lender with a sticky customer base, low funding costs, and conservative underwriting can earn good returns through a cycle, and some do. The point is that those returns come from credit judgment, which is an operating skill, while a network's returns come from position, which is a structural feature. You can underwrite a skill poorly. You can't really mis-underwrite a position, though you can pay too much for it.

Don't ask whether the card sector is attractive. Ask which part of the value chain you want exposure to, and what you're paid for the risk you take there.

Networks tend to be priced as the high-quality compounders they are, which means the debate is usually about what's already in the price, not about the business. Issuers trade on the credit cycle, and the sensible habit is to look at them through a full cycle rather than at peak earnings. Hybrids deserve a separate lens because you're buying two businesses with different cycle sensitivities under one ticker.

None of this is a recommendation on any security. It's a way to read the filings. Read the segment disclosures, find the revenue line that doesn't depend on a loan, and ask how much of the earnings it supplies.

Key Takeaways

• Payment networks route transactions and collect fees; card issuers fund loans and bear charge-offs. Treating them as one 'credit card' category hides the real industry structure.
• Visa's fiscal 2024 10-K shows the network model at scale: about $35.9 billion of net revenue and $19.7 billion of net income, with no loan book behind it.
• Use the right metrics: payments volume and take rate for networks, net charge-offs and net interest margin for issuers, both for hybrids.
• Networks avoid credit risk but not regulatory, litigation, or spending-cycle risk.