Walmart, Kroger, and Ahold Delhaize are routing payments around card networks. Trump backs the bill that would force the rest. And AI agents will optimize for the cheapest route automatically.

The three largest grocery operators in the United States are all building the same thing: a way for customers to pay without touching a card network.

Walmart is live with Fiserv. Ahold Delhaize expanded its program this week. Kroger is signaling interest. Three companies, three separate efforts, one shared objective. Route money from customer bank accounts directly to merchant bank accounts. Skip Visa. Skip Mastercard. Skip the interchange fee entirely.

This matters because grocery is the highest-frequency retail category in America. People buy groceries every week, sometimes more. Every single transaction that leaves card rails is interchange revenue that the networks permanently lose. Not temporarily. Not until a contract renegotiation. Permanently.

US merchants paid $148.5 billion in card processing fees in 2024. Current estimates put 2026 on pace for $172 billion.

Think about that number for a second. $172 billion. That is not revenue. That is the cost of accepting payment.

Visa processed $14 trillion across 258 billion transactions in fiscal year 2025, generating $40 billion in net revenue. Mastercard handled $9.2 trillion. Together, they sit atop virtually every card transaction in America, collecting a toll on each one.

The average US interchange rate is 1.80 percent. That is the highest in the world. In-person transactions run about 1.70 percent. Online transactions hit 1.90 percent.

Now look at grocery economics. The industry runs on one to three percent net margins. When interchange alone eats 1.8 percent of the transaction value, it consumes roughly half the profit on every card payment. Half.

That is why grocery is the front line. No other retail category combines this level of purchase frequency, this kind of margin pressure, and this much transaction volume. A furniture retailer can absorb interchange because margins are 40 percent. A grocer cannot.

Walmart went first. Its partnership with Fiserv's NOW Network is live for online purchases, using RTP and FedNow for bank-to-bank settlement. No card network in the middle. The project remains on track for broader rollout, and analysts at Javelin Strategy have called it a potential transformation of real-time account-to-account payments.

Then this week, Ahold Delhaize expanded aggressively. Stop & Shop, Giant Food, and The GIANT Company, all Ahold brands, moved pay-by-bank from pilot to fully available on e-commerce. Tens of thousands of customers had already registered during the pilot phase. Now it is open to everyone.

Kroger has signaled interest but has not announced a partner yet.

The mechanics are straightforward. A customer links their bank account. When they check out, the payment flows through RTP or FedNow instant rails directly to the merchant's bank. The merchant pays a flat fee or a small percentage, far below card interchange rates. No Visa. No Mastercard. No issuing bank takes a cut.

Here is the thing. We have seen this playbook before.

In 2011, the Durbin amendment capped debit interchange from 44 cents to roughly 24 cents per transaction. That cost banks between $6.6 billion and $8 billion per year in lost revenue. Banks screamed. Lobbyists mobilized. But the cap held, and merchants saved billions.

Pay-by-bank does not cap interchange. It eliminates it. That is a fundamentally different proposition, and it is why the card networks should be more worried about grocery self-checkout than about any bill in Congress.

If you want to see where the US is heading, look across the Atlantic.

TrueLayer has hit 20 million users. It adds one million per month. The company has processed over $10 billion in payments and handles 47 percent of all UK pay-by-bank transactions. We covered TrueLayer's model in depth in our open banking review, and the growth trajectory has only steepened since.

Pay by bank is now the fastest-growing payment method in Europe. Not crypto. Not BNPL. Bank transfers.

eBay invested in TrueLayer in February 2026 and integrated pay-by-bank into its UK checkout. When a marketplace of that size commits, smaller merchants notice.

Plaid is running more than half of European payment volume through virtual accounts, building a parallel payment infrastructure that does not depend on card rails at all.

The UK is roughly two years ahead of the US on pay-by-bank adoption. The pattern is clear: once the rails exist and consumers enroll, it scales fast. FedNow is entering year three with increasing users and volumes. The rails are there. Walmart and Ahold Delhaize are proving that the enrolment model works. The rest is momentum.

The market is moving on its own. Washington wants to push it faster.

The Credit Card Competition Act was reintroduced in January with something it never had before: a presidential endorsement. Trump posted on Truth Social calling on Congress to "stop the out of control Swipe Fee ripoff." That matters. This bill has been floating around Capitol Hill for years with bipartisan support, but never enough momentum. A sitting president telling Congress to pass it changes the calculus.

The CCCA would require banks with $100 billion or more in assets to enable a second network on every credit card. Merchants choose which network to route through. The estimated savings: $17 billion per year.

The Senate Banking Committee held expedited hearings on January 15. House markup followed on January 22. A floor vote is likely before the mid-2026 summer recess. Sponsors are seeking larger legislative packages to attach it to, which is the fastest path to passage.

And the CCCA is not the only front. Merchant groups are challenging the Visa/Mastercard interchange settlement, with two damages trials scheduled: April in New York, September in Chicago. Australia's Reserve Bank announced interchange caps and a surcharging ban. The global regulatory direction is obvious.

The Durbin amendment cut debit interchange and cost banks billions. The CCCA does the same thing for credit. If it passes, card networks face real competition on every credit card transaction for the first time. Ever.

This is the part most people are missing. And we think it matters more than any regulation.

When AI agents start shopping for groceries, and they will, Walmart, Shopify, and Amazon are all building agent-ready commerce interfaces, those agents will optimize for the cheapest payment route. Not the most convenient. Not the one with the best rewards program. Not the one the consumer has used for 15 years out of habit. The cheapest.

FIS is already developing real-time least-cost routing at the individual SKU level. Adyen, Stripe, and Worldpay offer dynamic transaction routing as a standard feature. AWS built a Cognitive Payments Director specifically for real-time route optimization. Industry reports cite 20 to 80 percent operational savings from intelligent routing.

An AI agent has no brand loyalty to Visa. None. It has a cost function. Pay-by-bank wins that function every single time interchange is the alternative.

Think about what that means at scale. A consumer might choose to pay with their Visa card because they like the rewards, or because it is the default in their Apple Wallet, or because they have always done it that way. An AI agent making 50 grocery orders a week for a household does none of that. It reads the cost of each payment rail and picks the cheapest one. Every time.

This is the threat that card networks cannot lobby against. They cannot regulate it away. They cannot settle it in court. It is algorithmic. Once an agent can choose the cheapest rail, it will. The convergence of identity and payment infrastructure makes this even more inevitable, because agents will handle authentication and payment selection as a single optimized decision.

We wrote about Fiserv's positioning in agentic commerce in our deep dive on Visa, Mastercard, and Fiserv's agentic strategies. The companies building payment infrastructure know this shift is coming. The question is whether the card networks can adapt fast enough.

They are not standing still. Give them that.

Visa is moving up the stack from transaction processing to data intelligence. Value-added services now include fraud detection, tokenization, analytics, and Visa Intelligent Commerce. The pitch: even if transactions route around our rails, merchants still need us for fraud prevention and data.

Mastercard is running the same play. Agent Pay. Agentic Tokens. Positioning as the trust layer rather than just the payment rail.

The bet is reasonable. Merchants do need fraud prevention. They need identity verification. They need dispute resolution. If the networks can become the intelligence layer sitting on top of any payment rail, they survive the loss of interchange revenue by replacing it with services revenue.

But there is a risk in that bet. Pay-by-bank rails are not standing still either. The Paysecure and Yaspa partnership is building fraud layers directly into bank payment rails. Plaid already handles identity. TrueLayer does verification. If the new rails develop their own fraud and identity infrastructure, the value-added services argument gets a lot weaker.

The networks are betting they can climb the stack faster than the new rails can build up. That is a race, not a certainty.

Three things to watch.

Grocery adoption rates. If Walmart's pay-by-bank hits meaningful transaction volume, Kroger and the rest will follow within months. Grocery is the domino. When it falls, it falls fast, because every grocer faces the same margin arithmetic.

The CCCA vote. If the Credit Card Competition Act passes before summer recess, card networks face mandated competition on credit transactions for the first time. Combined with pay-by-bank momentum, that is a one-two punch that fundamentally changes the economics of card acceptance.

Agentic commerce routing. The moment AI agents start choosing payment methods based on cost, interchange models face pressure from a direction the networks cannot control. That moment is closer than most people think.

The grocery aisle has always been where American spending habits play out in real time. Now it is where the future of payments is being decided. Will card networks adapt fast enough to survive a world where every transaction gets routed by an algorithm, not a habit?