The $200 Billion Interchange Settlement Isn't Settled
A settlement is supposed to be the point where an argument stops. The proposed $200 billion interchange settlement between Visa, Mastercard and the merchants who sued them has turned into the next round of the argument instead. Walmart and roughly 1,000 other merchants have filed to have the deal rejected, Payments Dive reports. The class the settlement would cover is about 12 million businesses, so by count the objectors are a small slice of it. They include one of the best-known retailers in the world, though, and their objection goes straight at the numbers.
My working days are spent in payment orchestration, so interchange isn’t an abstract policy subject for me. It’s one of the biggest lines in most card-payments cost models, and the one the people building those models can do the least about. That’s why I read this story less as legal news and more as a test of an assumption I expect a lot of finance teams to make over the next few months: that the settlement is a rate change they can put into next year’s budget. I don’t think it is, and the reasons say something about how interchange gets priced in the first place.
What the deal does, and what the merchants object to
The mechanics first, as Payments Dive reported them. The settlement trims credit interchange by 0.1% for five years. It caps interchange on standard consumer cards at 1.25% for eight years. The objecting merchants set that against a weighted average of 2.35%, which is what merchants actually paid in 2024, and against swipe fees that have risen 80% since 2020, to $198 billion last year. For comparison, the EU caps card fees at 0.3%. The final-approval hearing is set for November 16, 2026.
Interchange, for anyone who doesn’t live in it, is the fee that goes to the bank that issued the customer’s card. The card networks set it through their rate schedules, and the merchant pays it on every card transaction through its acquirer, bundled with everything else it pays to accept cards.
It isn’t approved yet
Until November 16, this is a proposal. A court still has to decide whether to approve it, and it will make that decision with a formal objection from a large, well-known group of merchants in front of it. I’m not going to predict the outcome, and I’d be wary of anyone who says they know it.
For a cost model, the treatment of an unapproved settlement is simple: it goes in as a scenario. If the finance team wants to see what the numbers look like with the cap in place, that’s a reasonable thing to model, as long as it sits next to the current-rate version and nobody quietly deletes the current-rate version because the headline sounded final.
The relief comes with an expiry date
Even if it’s approved, both of the settlement’s main terms are time-limited. The credit trim runs for five years. The cap on standard consumer cards runs for eight. The reporting doesn’t describe what replaces either one when it ends, and I wouldn’t assume anything kinder follows. I’d model the end of each period as the point where the concession stops.
Five and eight years sound far away until you look at what gets modelled over that kind of horizon. Customer lifetime value on a subscription business, the payback period on a new market, the business case for a multi-year processor contract: all of these routinely stretch past year five. A model that treats the cap as the new permanent rate will look fine in its first few years and be wrong for the rest of its life.
The cap covers a category, not your blend
This is the part I’d spend the most time on, because it’s where I expect the most expensive mistakes.
Let’s imagine a situation. A finance team at a mid-sized online business is building next year’s budget, and the news reaches them as something like “card fees capped at 1.25%.” Someone opens the cost model, finds the line that assumes something close to the 2.35% average, and changes it to 1.25%. The budget suddenly looks much better. It’s also wrong on several counts at once.
The cap applies to one category, standard consumer cards. The 2.35% is a weighted average across everything merchants paid. Interchange isn’t one rate; it varies by card product and by transaction type, which is exactly why an average and a category cap aren’t the same kind of number. How much of the distance between 1.25% and 2.35% a particular merchant would ever see depends on how much of its volume falls inside the capped category. The headline can’t tell you that. Your own transaction data can.
The 0.1% trim on credit interchange is the easier term to size. Next to a 2.35% average, it’s a small adjustment for most cost models, and I wouldn’t let it change any decision on its own.
The 80% figure is the one I expect to see quoted most, and it’s the one I’d use least in a model. Swipe fees rising 80% since 2020, to $198 billion, is a total. A total grows with card volume as well as with rates, so on its own it tells you how much money is flowing through card fees. It doesn’t tell you how far the price of an individual transaction moved. It’s a fair number to put in front of a court. For modelling, the 2.35% is far more useful, because it’s a rate.
What I’d do before touching the model is pull the last twelve months of card transactions, split them by the card categories your processor reports, and find out what share of volume would fall under the capped category. That share, applied to the cap, is the number worth modelling, and only in the scenario tab until November 16.
Interchange is priced by leverage
The EU comparison is the most telling detail in the story. A card payment in the EU and a card payment in the US involve the same basic parties: a cardholder, an issuing bank, a network, an acquirer, a merchant. In the EU, card fees are capped at 0.3%. In the US, merchants paid a weighted average of 2.35% in 2024. I wouldn’t lean on the precise ratio, since fee comparisons across two regimes rarely measure exactly the same thing. Even allowing a generous margin for that, the gap is far too large for me to believe differences in the cost of running a card transaction explain it.
What explains it, in my view, is who sets the price. In the EU, a regulator did. In the US, the price is being contested through litigation, and the settlement terms are what the parties to the case could agree on. A 0.1% trim, a 1.25% cap, five years, eight years: these are negotiated outcomes. As I read them, they reflect what each side thought it could win or lose in court, which is a different question from what a card transaction costs to process.
I want to be clear about where this argument stops. I’m not saying the EU’s 0.3% is the correct price and the US number is wrong. A regulated cap is a leverage outcome too, just with a regulator holding the pen instead of a courtroom. I’m also not taking a position on whether the settlement is too small; that’s the court’s call, and the objecting merchants have an obvious interest in their side of it. My point is narrower and more practical. For anyone modelling payment costs, interchange behaves like a price set by whoever holds leverage at the moment, and it tends to move in steps when that leverage shifts, through a ruling, a regulation or a change in the networks’ schedules, rather than drifting with underlying costs. Model it that way, as a variable with discrete jumps and dates attached.
What orchestration can and can’t do about interchange
For clarity about my own side of this: an orchestration layer, which is what we build at Corefy, sits above processors and routes payments between them. It doesn’t change the interchange on a given card. That fee follows the card and the transaction and is set by the network, so for the same card and the same transaction it doesn’t meaningfully change based on which processor you send the payment through. If someone pitches routing to you as a way to cut interchange on credit cards, ask them to show you the mechanism.
What routing does affect is the layer around interchange. Processors charge their own margin on top of it, and those margins differ. More importantly, routing affects whether a payment is approved at all, and a declined payment costs the whole sale. I’ve argued before that approval rate on its own can mislead, and that the more honest measure is approved revenue you kept, per attempt. Interchange is one of the things that measure subtracts, and a settlement that changes it on some cards, for a fixed number of years, doesn’t change the case for having more than one processor. If you run a single-market business on one processor that covers your customers well, this settlement calls for a quick review of one line in your cost model and not much more.
One more data point on card economics
In my previous piece I said I expect most agent payments to move over stablecoin rails rather than card rails, and that I believe those rails will eventually take over from the card networks more broadly. This fight doesn’t prove that, and I wouldn’t point to it as a sign that it’s close. Merchants accept cards because customers pay with them, and nothing in a settlement changes that.
What the fight does show is that the price of card acceptance is openly contested by the businesses that pay it, and that a proposed deal hasn’t ended the contest. That’s the pressure any alternative rail is ultimately answering, and the settlement, approved or not, leaves the question of who sets the price where it was.
For anyone running payment operations between now and November 16, the practical list is short. Keep the settlement out of the base case. Find out how much of your volume falls under the capped category before assuming any benefit. Put the five- and eight-year end dates into the model if the deal is approved. And watch the hearing, because whatever the court leaves standing is the number that belongs in the model.