Thirty-second in a series on jobs whose pay system is stranger than the salary. Recent entries covered the title insurer, who collects a one-time premium on a risk that almost never comes true, and the pharmacy benefit manager, who sits between your prescription and your pharmacy and keeps a spread nobody can see. The card network and the issuing bank belong right beside them, because the swipe fee is the most-collected toll in American commerce and the least-named. It rides on trillions of dollars a year, it is split three ways, and not one of the three parties appears on your receipt. Every figure below is cited to the Federal Reserve, a company's SEC filing, or the Government Accountability Office, and where a number could not be verified at a primary source I say so rather than estimating.
Look at your last card receipt. It shows the store, the date, the total, and the last four digits of your card. It does not show the three parties who each took a cut of that total before the store saw a dime.
That is not an accident. It is the design.
Every time a card is swiped, tapped, or typed, a small slice of the sale is peeled off and divided among three toll booths: the bank that issued your card, the bank that serves the merchant, and the network, Visa or Mastercard, that carries the transaction between them. The Government Accountability Office describes the mechanics without any euphemism. When you pay by card, "the merchant does not receive the full purchase amount because a certain portion of the sale is deducted to compensate the merchant's bank, the bank that issued the card, and the card network that processes the transaction."
The merchant pays the toll. You fund it, through prices set high enough to absorb it. And the receipt names none of the collectors. This article is about who they are and how the money actually reaches them, because the biggest slice of that toll does not go where most people assume, and even a federal law written to shrink it ended up holding it in place.
The anatomy of a swipe, and who keeps which piece
Start with the piece of jargon that hides the largest toll: interchange.
Interchange is the fee the merchant's side pays to your bank on every transaction. Here is the part that surprises people. Visa and Mastercard set the interchange rate, publish the schedules, and enforce them, but they do not keep the interchange. It passes straight through them to the bank that issued your card. Mastercard spells this out in its own annual report to the SEC: in the settlement of a purchase, "the issuer pays the acquirer an amount equal to the value of the transaction, minus the interchange fee," and "the acquirer pays the amount of the purchase, net of a discount (referred to as the merchant discount rate), to the merchant." Interchange fees, the filing says plainly, "are collected from acquirers and paid to issuers."
Read that chain slowly, because it is the whole trick.
- The merchant sells you something and agrees to receive the price minus a cut, the merchant discount rate.
- The merchant's bank, the acquirer, keeps part of that cut and forwards the rest as interchange.
- The interchange lands at your bank, the issuer.
- Visa or Mastercard routes the whole thing and charges its own separate fees for doing so.
So the toll is not one fee. It is a stack. And the biggest layer of the stack, interchange, is set by a network that never books it as revenue and collected by a bank whose name is on your card but nowhere on the receipt. The party that writes the rule does not pocket the money, and the party that pockets the money did not write the rule. That separation is exactly what makes the toll so hard to see and so hard to fight.
The bank's toll: interchange, and the cost it dwarfs
Now put a number on what the issuing bank collects and what it actually costs the bank to do the work.
In 2023, Americans ran 100.7 billion debit and prepaid card transactions worth $4,658.42 billion, roughly $4.66 trillion, through this machine. That is the river the toll rides on. On that flow, issuers collected $34.12 billion in debit and prepaid interchange fees in a single year.
Here is the gap that makes it a toll rather than a service charge. The Federal Reserve, which by law must study this, measured what it costs a large regulated bank to authorize, clear, and settle one debit transaction. In 2023 that cost, the network's own term is authorization, clearing, and settlement cost, came to about 4.1 cents per transaction, excluding fraud losses. The average interchange fee those same covered banks collected on a debit swipe was about 23 cents in 2024.
Sit with those two numbers. Four cents to do the job. Twenty-three cents collected for doing it. The fee is not proportional to the work. It is a toll on the transaction, sized to what the traffic will bear, and the difference between the cost and the fee is the whole reason a payments business is a good business.
There is a legal standard that is supposed to keep interchange tethered to cost, and its wording is almost poignant next to those figures. Under the Federal Reserve's Regulation II, which implements the part of the 2010 financial-reform law known as the Durbin Amendment, debit interchange for large issuers must be "reasonable and proportional to the costs incurred by the issuer." Reasonable and proportional. Four cents of cost, 23 cents of fee, and the law calls the fee proportional to the cost because a regulator wrote a formula that says so.
The cap that became a floor
That formula is the strangest part of the whole structure, because it was written to hold the toll down and instead it drew a line that issuers hug.
Regulation II covers only large banks, those with $10 billion or more in assets, and it caps what they may charge. The cap is a base fee of 21 cents, plus 0.05 percent of the transaction value, plus a 1-cent fraud-prevention adjustment for banks that qualify. A covered bank on an average-sized debit purchase is allowed to collect roughly 24 cents. It collects about 23. In other words, the big banks the rule was meant to restrain now cluster right against the ceiling the rule set. A cap that was supposed to be an upper limit became the number everyone aims for. The 21-cent base is not a leash. It is a target.
You can see the cap working, and you can see its hole, by splitting issuers into the two groups the law creates.
| Debit interchange, 2024 average | Per transaction | Share of value |
|---|---|---|
| Covered issuers (banks with $10B+ in assets) | $0.23 | 0.47% |
| Exempt issuers (small banks, some prepaid, government programs) | $0.51 | 1.21% |
| All transactions | $0.34 | 0.73% |
The covered banks, the ones the cap binds, collect about 23 cents. The exempt banks, small institutions and certain prepaid and government-administered programs that Regulation II leaves alone, collect about 51 cents, more than double, on the same act. The law reached the big issuers and left everyone else to charge what they like.
Now look at the totals, because they hold a small surprise. Of the $34.12 billion in debit interchange collected in 2023, covered issuers took $14.02 billion and exempt issuers took $20.10 billion. The banks the government did not cap collected more interchange in total than the banks it did. Whatever else the Durbin Amendment accomplished, it produced a two-tier toll where the exempt tier out-collects the regulated one.
I want to be precise about a few things I could not confirm at a primary source. A Federal Reserve proposal reported in October 2023 to lower the 21-cent base cap, widely cited as a cut toward roughly 14.4 cents, did not appear on the Fed's average-interchange page I reviewed, so I am not asserting its status or its number. The specific statutory chain often quoted for the Durbin Amendment, its section number in the 2010 law, the code section it added, and its October 2011 effective date, was not restated on the Regulation II overview page I used, so I name the law and the verified $10 billion threshold but not those particulars.
The network's toll: the smaller slice with the fatter margin
If interchange passes through the networks and lands at the banks, then how does Visa keep about 50 cents of profit out of every dollar it books? Because the network charges its own fees, on top of interchange, for carrying the traffic and connecting the parties, and those fees are where the network lives.
Visa's revenue is not interchange. Look at how it is built. In its fiscal year ended September 30, 2025, Visa reported $40.0 billion in net revenue, assembled from data processing revenue of $19,993 million, international transaction revenue of $14,166 million, service revenue of about $17,539 million, and other revenue of $4,053 million, less $15,751 million of incentives paid back to its bank clients. Not one of those lines is interchange. The network's toll is the processing charge and the assessment, the fee for running the rails, and the interchange it sets for the banks never touches its own income statement.
And the margin on running those rails is extraordinary. On $40.0 billion of net revenue in fiscal 2025, Visa earned net income of $20,058 million, a net margin of about 50.1 percent. Roughly half of every dollar of net revenue drops to profit. The two prior years were nearly as rich: net income of $19,743 million in fiscal 2024 and $17,273 million in fiscal 2023, on net revenue of $35,926 million and $32,653 million.
Mastercard runs the same shape on its own calendar. In the year ended December 31, 2025, it reported total net revenue of $32,791 million, split into payment network revenue of $19,476 million and value-added services and solutions of $13,315 million, up from $28,167 million in 2024 and $25,098 million in 2023. Its operating income was $18,897 million, an operating margin of 57.6 percent, and its net income was $14,968 million, a net margin of about 45.6 percent.
These are company-specific figures, Visa's and Mastercard's own reported numbers, not industry universals, and I present them as such. But the point they make is structural. The bank collects the largest slice, interchange, and does ordinary bank things with it. The network collects a smaller slice, the processing and assessment fee, and converts nearly half of it into profit. The toll booth that keeps the richest margin is the one whose name you do at least half-recognize, and even it is charging you for something you never chose: the privilege of the merchant being connected to your bank.
The merchants who fought the toll, and mostly lost
A toll this large, collected this invisibly, has been in court for twenty years. The story of that litigation is the clearest proof that the swipe fee behaves like a private tax rather than a competitive price, because the people who pay it have spent two decades trying to lower it and have very little to show.
The central case carries a name as heavy as the subject: In re Payment Card Interchange Fee and Merchant Discount Antitrust Litigation, the multidistrict litigation known as MDL 1720 in the Eastern District of New York. On the damages side, the class of merchants and the networks reached an amended settlement dated September 17, 2018, and the district court granted final approval on December 13, 2019. The defendants had already made about $1.1 billion in what the filings call takedown payments, and an additional $900 million payment was required. Those are the verified mechanics from Visa's own filing. The widely reported headline value of that settlement, often cited as roughly $5.54 billion net after merchants opted out of an original figure near $6.24 billion, I could not pin to the filing text, so I flag those specific numbers as secondary and do not assert them.
Money is one thing. Changing the rules is another, and that is where the merchants have hit a wall. A separate part of the case sought not damages but injunctive relief, an actual change to the network rules that would lower or constrain swipe fees going forward. Plaintiffs filed for preliminary approval of such a settlement on March 26, 2024. The court denied the motion on June 25, 2024. The deal that was supposed to bend the fee itself was rejected. Its widely reported terms, a figure near $29.79 billion in estimated merchant savings and interchange-rate reductions of at least 0.04 percentage points for a few years, are secondary and unconfirmed at the filing, so I report only what the filing confirms: it was filed, and it was denied.
Even the government's own cap is not settled. Regulation II, the 21-cent rule itself, is under active legal challenge, and the courts have split. A federal district court in North Dakota ruled that the Federal Reserve exceeded its authority and vacated the debit interchange standard, finding the Fed had improperly folded fraud losses, network fees, and other fixed costs into a cap that the statute meant to tie more narrowly to cost. A federal district court in Kentucky ruled the opposite, that the Fed acted within its discretion. So the one number the government did impose on this toll, 21 cents, is itself in dispute, with two courts reaching opposite conclusions about whether it is even lawful.
The networks carry the running cost of all this as a line on the balance sheet. Mastercard reported an accrued liability of $637 million for the U.S. interchange litigation as of December 31, 2025, up from $559 million a year earlier. That is what it costs to keep the toll booth open against two decades of legal siege: a few hundred million in reserves against a toll that rides on trillions. The arithmetic tells you who is winning.
What a salaried reader should take from this
The biggest toll on the economy is the one nobody itemizes. The swipe fee is split three ways, among your bank, the merchant's bank, and the network, and it appears on none of your receipts. The GAO says flatly that the merchant does not receive the full purchase amount. You pay it anyway, folded into the price of everything, the same way the title insurer's premium is folded into your closing and the pharmacy benefit manager's spread is folded into your co-pay. When a cost is deliberately unlabeled, that concealment is doing work. The first move in understanding any price is to ask which parties took a cut before the seller saw the money.
A cap is only a cap if it sits above what the market would charge. Below that, it is a floor. Regulation II set 21 cents as a ceiling for big banks, and big banks now collect about 23 cents, right against it. A limit placed low enough that everyone clusters at it does not restrain the toll, it standardizes it. The same logic runs through this whole series: a number set by a rule, whether it is a COLA formula or an interchange cap, pulls behavior toward the rule and away from what a competitive market would have produced, and the direction it pulls is not always the one the rule intended.
Follow the money to the party that keeps the fattest margin, not the biggest number. The bank collects the largest dollar slice, interchange, but the network converts nearly half of its smaller slice into profit. Visa's roughly 50 percent net margin and Mastercard's roughly 46 percent are the signature of a toll booth, not a service business, because no ordinary service keeps half of revenue as profit when real competition is allowed to work. When you see a margin that fat and that stable, stop asking how good the company is and start asking who is legally or structurally prevented from undercutting it.
When the payers spend twenty years in court and the reserves stay small, the toll is winning. Merchants have litigated the swipe fee since the mid-2000s, won damages measured in billions, and still could not get a court to approve a change to the rules that set the fee. A few hundred million in litigation reserves against a fee that rides on $4.66 trillion of debit purchases alone is not a threat to the system, it is a cost of doing business. A toll that survives that much organized opposition is not a price. It is an institution, and the institution is the point.
Related reading
- How title insurers are paid: another one-time toll baked invisibly into a transaction, on a risk that almost never materializes.
- How pharmacy benefit managers are paid: an unnamed intermediary that keeps a spread between what you pay and what your pharmacy receives.
- How credit rating agencies are paid: a gatekeeper whose blessing is required and whose fee is paid by the party being rated.
- How car dealers are paid: the toll layered into a purchase everyone assumes is a simple sale.
- How mortgage bond sellers are paid: the same asset tolled at several stations, each a different paycheck, none of them the making.
Fact-check notes and sources
The debit interchange caps, averages, totals, and per-transaction cost come from the Federal Reserve. The network revenue, margin, and litigation figures come from Visa's and Mastercard's own SEC filings and are that company's figures, not industry universals. The three-party split comes from the Government Accountability Office. Where a figure is secondary or could not be verified at a primary source, it is flagged in the text and here.
- The Regulation II debit interchange cap (a base fee of $0.21, plus 0.05 percent of the transaction value, plus a $0.01 fraud-prevention adjustment for eligible issuers) and the 2024 average debit interchange fees by issuer type (covered issuers $0.23, or 0.47 percent of value; exempt issuers $0.51, or 1.21 percent; all transactions $0.34, or 0.73 percent) are from the Federal Reserve, Regulation II Average Debit Card Interchange Fee data, a primary source. A reported October 2023 Fed proposal to lower the 21-cent base cap (widely cited toward roughly 14.4 cents) did not appear on that page and its status is not asserted here.
- The statutory standard that debit interchange be "reasonable and proportional to the costs incurred by the issuer," and the exemptions for small issuers, government-administered programs, and reloadable general-use prepaid cards are from the Federal Reserve, Regulation II overview (Durbin Amendment implementation), a primary source. The $10 billion asset threshold that defines a covered issuer (issuers with assets under $10 billion are exempt) is stated on the Federal Reserve's average debit card interchange fee page, not the overview page; both are primary Federal Reserve sources. The specific statutory citation chain sometimes attached to the Durbin Amendment (its section in the 2010 law, the code section it added, and the October 2011 effective date) was not restated on either page and is not asserted.
- The 2023 flow and toll totals (100.7 billion debit and prepaid transactions worth $4,658.42 billion; $34.12 billion in total debit and prepaid interchange collected, of which $14.02 billion by covered issuers and $20.10 billion by exempt issuers) and the 2023 authorization, clearing, and settlement cost of about $0.041 per transaction for covered issuers, excluding fraud losses, versus the roughly $0.23 average covered interchange fee, are from the Federal Reserve, 2023 Interchange Fee Revenue report, a primary source. The 4.1-cent cost figure is a 2023 measurement and the 23-cent fee is the 2024 covered-issuer average; the Fed presents them as the cost-versus-fee comparison for covered issuers.
- The settlement mechanics of a card transaction ("the issuer pays the acquirer an amount equal to the value of the transaction, minus the interchange fee," the acquirer pays the merchant "net of a discount (referred to as the merchant discount)," and interchange fees "are collected from acquirers and paid to issuers"), the Mastercard net revenue split (total net revenue of $32,791 million in FY2025, comprising payment network revenue of $19,476 million and value-added services and solutions of $13,315 million; $28,167 million in FY2024 and $25,098 million in FY2023), the Mastercard operating and net figures (operating income $18,897 million, operating margin 57.6 percent, net income $14,968 million), and the accrued U.S. interchange-litigation liability ($637 million at December 31, 2025; $559 million at December 31, 2024) are from the Mastercard Incorporated FY2025 Form 10-K, via SEC EDGAR. These are Mastercard's own figures; the roughly 45.6 percent net margin is my arithmetic on $14,968 million over $32,791 million.
- Visa's net revenue ($40,000 million in FY2025, $35,926 million in FY2024, $32,653 million in FY2023, fiscal years ended September 30), its revenue components (data processing revenue $19,993 million, international transaction revenue $14,166 million, service revenue about $17,539 million, other revenue $4,053 million, less client incentives of $15,751 million, netting to $40,000 million in FY2025), its net income ($20,058 million in FY2025, $19,743 million in FY2024, $17,273 million in FY2023), and the MDL 1720 litigation facts (the damages-class amended settlement dated September 17, 2018, with final district-court approval on December 13, 2019, about $1.1 billion in prior takedown payments and an additional $900 million payment required; the injunctive-relief class's motion for preliminary approval filed March 26, 2024 and denied June 25, 2024; and the split rulings on Regulation II, with a North Dakota district court vacating the debit interchange standard and a Kentucky district court upholding it) are from the Visa Inc. FY2025 Form 10-K, via SEC EDGAR. These are Visa's own figures; the roughly 50.1 percent net margin is my arithmetic on $20,058 million over $40,000 million. The often-cited headline value of the 2018 damages settlement (about $5.54 billion net, reduced from about $6.24 billion) and the reported terms of the rejected 2024 injunctive settlement (about $29.79 billion in estimated savings, at least a 0.04-percentage-point rate reduction) are secondary and were not confirmed in the filing text, so they are not asserted.
- The three-party split of the swipe fee ("the merchant does not receive the full purchase amount because a certain portion of the sale is deducted to compensate the merchant's bank, the bank that issued the card, and the card network that processes the transaction") is from GAO-10-45, Credit Cards: Rising Interchange Fees Have Increased Costs for Merchants, but Options for Reducing Fees Pose Challenges, a primary source.
- Not asserted, and why: any all-in annual total for U.S. card swipe fees across credit and debit and all networks (the commonly cited figure of roughly $170 billion or more for 2023 comes from the Nilson Report as relayed by merchant coalitions, and could not be confirmed at a free primary source; the Fed's verified $34.12 billion is debit and prepaid interchange only and excludes credit interchange and network assessments); the combined Visa-and-Mastercard share of U.S. card volume (the "duopoly" framing, often cited above 80 percent when American Express and Discover are excluded), which was not confirmed at a primary source; and the specific dollar terms of both the 2018 and 2024 settlements beyond the dates and payment amounts the Visa filing confirms.
This post is informational and journalistic, not career, legal, or financial advice, and nothing here is a recommendation to buy or sell any security. It describes Federal Reserve data, a Government Accountability Office report, and the SEC filings of Visa Inc. and Mastercard Incorporated. Interchange rules, fee schedules, litigation status, and company figures change year to year, and several figures here are as of the fiscal years or reporting periods noted or are flagged as unverified, so verify current data before relying on any of them. Mentions of specific companies, agencies, banks, and litigation are drawn from the public record and are nominative fair use, and no affiliation is implied.
