Unfiltered: the cost of free payments

In his latest Unfiltered column for The Paypers, Dwayne Gefferie, Payments Strategist, looks at the Eighth Circuit ruling against the Fed’s debit interchange cap and India’s postponed UPI merchant fee, and argues that when a payment rail is priced below its running cost, someone else ends up paying for it.

In October 2026, a US appeals court and India's payments authority worked on the same problem from opposite ends. I think both are about to learn the same lesson.

On 26 October, the US Court of Appeals for the Eighth Circuit upheld a ruling that the Federal Reserve's debit interchange cap (Regulation II) is unlawful. The court's view is that the cap is too generous, because the Fed included costs the Durbin Amendment doesn't allow. TD Cowen estimates that a cap built the court's way could fall below five cents for the largest banks. Today it's 21 cents plus 5 basis points, with another cent for fraud prevention.

India went the other way. NPCI was due to start charging merchants for UPI, the largest real-time payment system in the world, on 15 October, after almost six years at zero. After pushback from traders, it postponed the start until after Diwali, citing the festive season. The fee is still coming.

My thesis is simple. You can price a payment rail below what it costs to run, but you can't make that cost disappear. It moves to someone else, usually someone who wasn't part of the conversation. The US is about to push debit pricing towards marginal cost, and India has just shown what six years at zero actually costs.

What the court actually decided

On the law, Judge David Stras has a point. The Durbin Amendment says the cap should reflect the incremental cost of authorising, clearing and settling each transaction. The Fed went further and also counted fixed processing costs, network fees, fraud-monitoring costs, and fraud losses. The court called out the flat 21-cent cap, applied to every transaction whatever it cost, as the clearest example of the Fed going beyond what Congress wrote.

That's a fair reading of the statute. As a way to price a payment system, it's the wrong yardstick.

The economics of a payment rail

Anyone who has looked at an issuing or processing P&L (profit and loss statement) knows how this works. Most of the cost of a debit programme sits in things that don't change when one more card is used: the processing platform, fraud models, dispute operations, compliance, and customer service. The cost of one extra authorisation is close to nothing. In my view, that's the nature of a scale business, and it's why nobody in payments chooses to price at marginal cost. Telecom operators don't charge you the cost of carrying your next call either.

The amounts at stake are significant. According to the Fed's latest data, there were 100.7 billion debit and prepaid transactions in 2023, generating USD 34.12 billion in interchange. Covered transactions averaged 24 cents on single-message networks and 22 cents on dual-message, and fraud losses ran at 17.6 basis points of value. If a new cap takes out most of that revenue, the fixed costs it was paying for stay exactly where they are.

Durbin: we've seen this before

The US has run this test once already. When the original Durbin cap took effect in October 2011, merchants got cheaper debit acceptance. Two Federal Reserve economists later looked at what happened on the issuer side.

Banks covered by the cap became 35.2 percentage points less likely to offer free checking. Without Durbin, they estimated, 65.2% of the non-interest checking accounts offered by those banks would have been free. In reality, 30% were. Monthly fees on those accounts rose by about 20%.

I don't see that as bad behaviour by the banks. A basic current account is funded by the revenue it generates, and debit interchange was a large part of that revenue. Cut it, and the bank reprices the account, so the cost moves from the merchant to the account holder. A cap at less than a quarter of today's level is the same mechanism with a bigger number.

There's a second effect I'd watch, which I'll flag as conviction because I can't document it yet. If debit stops covering its own costs at the largest banks, the easiest fix is to steer customers towards credit, where interchange is several times higher and there's no Fed cap. Merchants could win the debit case and still pay more per transaction once the card mix shifts.

UPI: zero was never the price

UPI is the best argument anyone has for free payments. It processed 24.07 billion transactions in September 2026, worth about USD 309 billion, without charging merchants anything.

But zero was a subsidy, not a price. The government compensated the industry through an incentive scheme that went from about USD 380 million in FY24 to USD 110 million in FY25, with roughly USD 210 million budgeted for FY27. NPCI, citing industry estimates, puts the annual cost of running UPI at around USD 2.1 billion. The subsidy covered a fraction of that, and the banks and payment apps absorbed the rest.

The government's own FAQ on the new fee, published by the Department of Financial Services, is open about this. It describes the incentives as short-term bridge funding, not a permanent arrangement. Once the fee starts, merchants will pay 0.4% on UPI payments above INR 2,000 (about USD 21), capped at INR 300 (USD 3) for payments of INR 75,000 (USD 790) and up. Railways, telecoms, insurance, fuel, and utilities pay a flat INR 5 (about five cents) instead. Small merchants receiving less than INR 1 lakh (USD 1,050) a month through UPI QR stay exempt, person-to-person payments stay free, and merchants can't pass the fee on to customers.

India isn't giving up on cheap payments. It's acknowledging that someone has to pay to keep the rail running.

Where the cost goes instead

When a regulator pushes one price down, the cost usually shows up somewhere else.

Europe is the clearest example. When the EU capped consumer debit interchange at 0.2% and credit at 0.3% in 2015, merchants' acceptance costs did fall. But the European Commission's own review in 2020 found that scheme fees and acquiring margins rose afterwards, taking back part of the saving.

Canada is going through a smaller version now. On 24 October 2026, Visa extended its lower small-business interchange rates to merchants with up to CAD 750,000 in annual Visa sales, from CAD 300,000. In the same release that welcomed the cut, the Canadian Federation of Independent Business warned of new per-transaction network fees. Mastercard's took effect this month, and Visa plans one from April 2027.

Wero is more transparent about this than most. The Dutch market parties behind it have committed to keep scheme pricing broadly at iDEAL levels until 31 December 2028, with purchase protection in place by 1 January 2028. Purchase protection costs money, and so does running a pan-European rail. My view, again flagged as conviction, is that what Wero charges from 2029 will be the first real indication of what European A2A costs.

What this means for the industry

To be clear, this isn't a defence of today's interchange levels. They may well be too high. Merchants see cost-based pricing as fairness and regulators see it as competition, and both have a point.

But any rule that pushes a payment towards marginal cost has to answer one question: who covers the fixed cost? If the rule doesn't answer it, the market will, and so far the answer has been consistent. Account holders pay through higher fees, taxpayers pay through subsidies until the budget runs out, or merchants pay through a new line on their statement under a different name. None of that reduces the cost. It just changes who carries it.

For issuers, the question now is how to fund a debit programme if the cap drops to single-digit cents. For merchants and acquirers, the useful thing to track is which line on the statement goes up next.

The Eighth Circuit has sent the Fed back to write a cap that matches the statute. Stifel's Brian Gardner expects bank groups to petition the Supreme Court, and a pending case in the Sixth Circuit [VM17] could produce a split between courts. Whatever happens there, the cost of running the US debit system won't shrink because a court says it should.

India has already learned that a free payment rail still has a price. Someone else is just paying it.

About the author