Walmart’s Card Fee Fight: The AI-Powered Way Out

A class action that covers roughly 20 million U.S. merchants sounds like a show of force—until the biggest merchants start arguing they’re better off alone.

That’s the tension behind Walmart’s latest move in the Visa/Mastercard interchange (swipe fee) litigation: Walmart is pressing a federal judge to split or redefine the plaintiff class so large national merchants can pursue remedies that actually fit their scale. Walmart’s filing calls the proposed settlement “nakedly inequitable” for large merchants and says the relief on offer—especially tweaks to “honor all cards”—doesn’t meaningfully change the economics.

From the standpoint of our AI in Payments & Fintech Infrastructure series, this isn’t just courtroom drama. It’s a loud signal that the largest retailers are trying to take control of their payments stack, reduce reliance on card rails, and demand more direct negotiating power. And if you’re running payments for a large enterprise, Walmart is basically describing your 2026 roadmap—whether you admit it yet or not.

Why Walmart wants out of the “one-size-fits-all” merchant class

Walmart’s core point is simple: the merchant “plaintiff class” is too broad to represent conflicting economic realities. A small business may want different settlement terms than a national retailer processing billions in card volume.

Walmart argues that the proposed settlement’s main practical change—loosening Visa/Mastercard’s “honor all cards” rule so a merchant can decline certain network-branded card products—doesn’t help a merchant that can’t realistically tell customers “no.” When your brand promise is low friction and high acceptance, refusing card types becomes a theoretical option, not an operational strategy.

The underlying conflict: acceptance is non-negotiable for large retailers

Large merchants live under a hard constraint: customers arrive with a preferred payment method (and a tolerance for checkout friction that’s close to zero). If you’re a top national retailer, you don’t win by educating customers about why their premium card product isn’t welcome.

That’s why Walmart says the real fix is issuer-level competition—the ability to negotiate interchange rates directly with issuing banks in exchange for volume. Put bluntly: Walmart doesn’t want permission to reject card products; it wants a market where issuers compete for its acceptance.

What this tells the market (even if the settlement changes)

Even if the judge denies the objection, the strategic message stands:

• Mega-merchants are done waiting for incremental rule tweaks.
• They want rate negotiation, smarter routing, and alternative rails that move money without credit card interchange.
• Payments is being treated less like a vendor relationship and more like core infrastructure.

That shift is the real story.

Card fees aren’t just a cost line—they’re an infrastructure constraint

Interchange is often discussed like a tax. For payments leaders, it’s more like a constraint that shapes product decisions:

• Which tender types you promote
• How you design loyalty (points vs. cash-back vs. closed-loop)
• Whether you build your own wallet
• How aggressive you get with pay-by-bank
• How much risk you’re willing to hold when you push customers off cards

And December is the perfect time to notice this. Holiday peak load is when payments teams feel every basis point—because volume spikes amplify everything: fees, fraud, chargebacks, and outage risk.

The “honor all cards” change is a distraction for enterprises

For enterprise merchants, the question isn’t “Can we reject some card types?” It’s:

• Can we shift 5–15% of volume to lower-cost rails without hurting conversion?
• Can we reduce fraud losses while pushing customers to irrevocable payments?
• Can we do it without turning checkout into a science project?

This is where AI-driven payments infrastructure becomes practical rather than buzzworthy.

The real alternative: route payments like a network engineer, not a finance team

Here’s what most companies get wrong: they treat “payment method” as a front-end choice and “processing” as a back-end commodity.

The better approach is to treat payments as a routing problem under constraints—cost, authorization rate, fraud, latency, dispute exposure, and customer preference.

AI isn’t magic here. It’s math plus feedback loops.

What AI-driven transaction routing actually does

At enterprise scale, “smart routing” means choosing the best path per transaction (or per customer segment) in real time:

• Card routing optimization: dynamically selecting acquirer paths, using network tokens, applying retries with rules that minimize issuer irritation
• Tender steering: deciding when to present pay-by-bank, wallet, card, BNPL, or store account—based on customer propensity and margin
• Risk-based authentication: stepping up only when risk warrants it, rather than blanketing everyone with friction
• Approval rate uplift: detecting false declines patterns and tuning retries and descriptors

A useful way to phrase it internally:

Every basis point of approval rate is revenue. Every basis point of fee reduction is margin. AI can target both at once—if your data is wired correctly.

The enterprise playbook: reduce card dependence without breaking checkout

If your leadership team hears “reduce interchange,” their first fear is conversion loss. That fear is justified—unless you take a measured approach.

A practical migration plan looks like this:

1. Segment customers (high-trust repeat buyers vs. first-timers; high basket vs. low)
2. Offer alternatives selectively (pay-by-bank prompts to customers most likely to adopt)
3. Tie incentives to margin (discounts funded by interchange savings, not blunt coupons)
4. Control risk (bank payment confirmation, fraud scoring, velocity checks)
5. Measure relentlessly (A/B tests across authorization, cart abandonment, and support contacts)

AI matters because it makes steps 1–5 fast enough to run continuously.

Pay-by-bank and instant payments: the hardest part isn’t the rail

Walmart has already explored pay-by-bank and instant payments. That direction makes sense: account-to-account payments can reduce card fees and lower disputes.

But the rail isn’t the hard part. The hard part is everything around it:

• Identity and account validation
• Fraud and social engineering prevention
• Refunds and exception handling
• Customer support tooling
• Reconciliation and settlement visibility
• Regulatory/compliance requirements

Where AI helps most in pay-by-bank

AI is most valuable in the “messy middle”—where operational complexity kills adoption.

1) Fraud detection tuned for irrevocable payments Card fraud tooling doesn’t transfer perfectly to pay-by-bank because dispute mechanics differ. AI models can score risk using:

• behavioral signals (typing cadence, device posture, session anomalies)
• identity consistency (address/phone/email graph signals)
• historical repayment/return behavior

2) Exception prediction and automated resolution A hidden cost of new rails is exceptions: misapplied payments, returns, partial shipments. AI can classify exceptions and route them to the right queue, reducing support cost.

3) Reconciliation automation If finance has to manually match payouts to orders, your “fee savings” evaporate. AI-assisted matching and anomaly detection turns reconciliation into a near-real-time process.

If you’re serious about moving volume off cards, this is non-negotiable.

What merchants should do in 2026 (even if you’re not Walmart)

Most retailers won’t file objections in federal court. You can still borrow the strategy: treat payments like infrastructure and build negotiating power.

A concrete 90-day plan for payments leaders

If you want results you can defend in a QBR, start here:

1. Quantify your “effective interchange rate”

   • Break it down by channel (in-store, ecommerce), basket size, and card product mix.

2. Map your routing and approval stack

   • Identify where decisions are static (one acquirer, one retry pattern, one fraud threshold).

3. Run two experiments (not ten)

   • Example A: smart retries + issuer-aware logic to lift approvals.
   • Example B: pay-by-bank prompt for a high-trust cohort with a margin-funded incentive.

4. Build an AI-ready data layer

   • Unify authorization outcomes, fraud signals, chargebacks, and customer IDs. AI projects fail when teams can’t agree on what happened in the transaction.

5. Negotiate with evidence

   • Better routing and richer data give you leverage with processors, acquirers, and even issuers.

“People also ask” style answers (because your CFO will)

Does pay-by-bank always cost less than cards? Not automatically. The rail fee can be lower, but the total cost depends on fraud, exceptions, and customer support. AI helps keep that total cost predictable.

Will steering customers away from cards hurt conversion? It hurts conversion when it’s blunt. It works when it’s targeted—offered to the right customers at the right moment with the right incentive.

Is this only for massive retailers? No. The difference is scale and budget. But the pattern—smart routing, better fraud controls, automated reconciliation—applies to any merchant doing meaningful volume.

What Walmart’s stance signals for AI in payments infrastructure

Walmart’s objection isn’t just a legal complaint. It’s an infrastructure statement: large merchants want direct control over the economics of acceptance.

Card networks and issuers have built a system where merchants compete on customer experience while paying largely non-negotiable pricing. When the biggest merchant in the U.S. argues that “honor all cards” relief is “useless,” it’s telling the market that incremental policy tweaks won’t meet enterprise needs.

AI-driven payments infrastructure is how merchants close the gap without setting checkout on fire: optimize transaction routing, reduce false declines, manage fraud on new rails, and automate reconciliation so fee savings show up as real margin.

If you’re planning your 2026 payments roadmap, here’s the forward-looking question that matters: What would it take for your business to treat cards as one option among several—rather than the default rail you can’t negotiate?