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

Interchange fees are one of the few line items that grow automatically when your sales grow. No hiring plan, no new stores, no additional product lines—just more volume flowing through the same card rails, with the same pricing power concentrated in the same places.

That’s why Walmart’s latest move—pushing a federal judge to split the merchant class in the Visa/Mastercard swipe-fee litigation—matters far beyond this one case. Walmart is effectively saying: “Stop treating every merchant like we have the same bargaining position and the same needs.” I think they’re right.

For leaders building or buying payments infrastructure in 2026, the bigger lesson isn’t legal strategy. It’s this: the card model is structurally misaligned with how modern commerce wants to optimize cost, control, and risk. AI-driven payment systems are the most practical path to fixing that—because they can make routing, authorization, and settlement decisions in real time based on economics and risk, not just brand rules.

Why Walmart wants out of the class (and why you should care)

Answer first: Walmart wants a settlement that changes the economics of card acceptance, not cosmetic rule tweaks.

In its court filing opposing the proposed Visa/Mastercard settlement, Walmart argues the class lawyers don’t represent large national merchants’ interests. The retailer specifically takes aim at relief that might help small businesses but does little for enterprises that can’t realistically refuse major payment types.

Two parts of Walmart’s argument should grab any CFO, payments head, or infrastructure architect:

1. “Honor all cards” changes aren’t enough. The settlement reportedly loosens rules so a merchant could decline some Visa/Mastercard-branded cards. Walmart calls that “useless” because a large merchant generally can’t tell millions of shoppers “no” at checkout without taking a reputational hit.
2. Walmart wants issuer-level negotiation. The company says real competition requires being able to bargain interchange rates directly with issuing banks—effectively pushing for a market where acceptance is a negotiated relationship, not a take-it-or-leave-it schedule.

This matters because interchange isn’t just “a fee.” It’s a design choice baked into the card system: issuers set rates, networks enforce rules, and merchants mostly comply because consumer preference is non-negotiable.

If you’re a mid-market merchant, platform, marketplace, or SaaS provider, you might not have Walmart’s scale—but you still have the same problem: your payment costs rise with success, and your control is limited.

The real issue: cards aren’t optimized for merchant economics

Answer first: Card rails optimize for ubiquity and issuer incentives, not for merchant cost efficiency.

The standard card stack was built to do three things exceptionally well:

• Create global acceptance rules
• Protect brand trust with standardized dispute processes
• Incentivize issuance through interchange

Those are legitimate goals. But they lead to predictable outcomes:

• Limited price competition at the point of sale. Even when merchants can steer, many can’t push hard without losing conversion.
• One-size-fits-most pricing. Interchange schedules bundle risk, rewards, and category assumptions into tables that don’t reflect your specific customer mix this week.
• Slow feedback loops. Merchants discover cost issues after the fact—at reconciliation—when the only move left is reporting, not optimization.

Walmart’s stance implicitly challenges that model: if you can’t decline issuers, and you can’t negotiate rates, then “choice” is theoretical.

Here’s the thing about payment economics: small differences compound brutally at scale. When volume spikes (hello, holiday season), you feel it in fees immediately. If your stack can’t respond dynamically, you’re stuck.

Where AI fits: from static fees to intelligent payment routing

Answer first: AI doesn’t “remove interchange,” but it can materially reduce effective payment cost by optimizing routing, tender mix, and risk decisions in real time.

In the “AI in Payments & Fintech Infrastructure” series, we keep coming back to a practical idea: payments are a decision system. Which rail? Which authentication method? Which fraud control? Which fallback? Most organizations still treat these as hard-coded rules.

AI changes that by making the decision layer adaptive.

1) AI-driven routing: choose the cheapest successful path

A modern payments orchestration layer can route transactions across methods and providers. AI improves this by forecasting outcomes per route:

• Approval probability by issuer, BIN, channel, and cart profile
• Expected fraud and chargeback risk
• Expected cost (including network fees, processor pricing, and operational overhead)

Instead of “send everything to Processor A,” you can choose: the route with the best expected margin—and keep learning.

Snippet-worthy truth: The cheapest payment method is the one that actually gets approved. AI helps you stop optimizing for sticker price and start optimizing for expected profit per authorization attempt.

2) Tender mix optimization: make alternatives actually usable

Walmart has publicly explored pay-by-bank and instant payments. The hard part isn’t launching an alternative. It’s getting customers to use it.

AI helps you target the offer, not blast it:

• Identify segments likely to adopt pay-by-bank (based on behavior, not demographics)
• Personalize incentives to the minimum needed to switch
• Decide when not to steer (e.g., high urgency baskets where any friction risks abandonment)

If you’re running a marketplace or subscription business, AI can also optimize:

• When to suggest ACH vs card
• When to request account updates
• When to move customers to lower-cost rails during renewal cycles

3) Risk and dispute automation: lower the “hidden” card tax

Interchange is visible. The hidden cost stack is often larger than teams admit:

• Manual review labor
• False declines (lost revenue)
• Chargebacks and representment ops
• Customer support contacts tied to payment failures

AI in fraud detection and dispute workflows reduces cost by:

• Scoring risk with more context (device, session patterns, historical behavior)
• Automating evidence collection for chargeback responses
• Predicting which disputes to fight vs accept (based on win rate and internal cost)

This is where infrastructure pays off fast: fewer exceptions means fewer humans in the loop.

“Honor all cards” vs “honor all outcomes”: a better way to frame it

Answer first: Merchants shouldn’t be forced to accept identical economics for every transaction; they should be able to optimize for outcomes—approval, cost, fraud, and customer experience.

Walmart calls partial relief “useless” because a large merchant can’t realistically refuse popular card types. That’s the right diagnosis. But the fix isn’t only legal; it’s architectural.

What merchants actually want is an operating model where:

• They can offer multiple rails without harming conversion
• They can route intelligently without breaking compliance
• They can measure unit economics per payment choice in near real time

In practice, that means building (or buying) these layers:

1. Payment orchestration (multi-PSP, multi-rail, fallback logic)
2. Data foundation (normalized auth, cost, dispute, and settlement data)
3. AI decisioning (routing, risk, retries, and steering)
4. Experimentation (A/B testing steering, incentives, and checkout UX)

Most companies get this wrong by starting with “add another payment method.” The better approach is to start with the decision layer and measurement—then add rails.

What changes if big merchants get their way?

Answer first: If large merchants can negotiate more directly (or create credible alternatives), the market will shift toward dynamic pricing and programmable payments.

If the court forces a rethinking of who the class represents—or if a future settlement grants more meaningful flexibility—expect second-order effects:

• More pressure for bank-linked payments (especially for high-ticket retail and recurring)
• More investment in instant payment acceptance where available
• More “payments as product” thinking inside retailers and marketplaces

But there’s a catch: negotiating interchange directly is complex. You’d need issuer relationships, compliance guardrails, and a way to operationalize different rates without breaking checkout.

That’s why I’m bullish on AI-enabled infrastructure. It scales negotiation outcomes into execution. A negotiated rate that you can’t enforce at the transaction level is just a slide deck.

Practical steps: a 30-day plan to reduce effective card costs

Answer first: You don’t need Walmart-scale to act; you need clean data, orchestration, and a test-and-learn cadence.

If you’re responsible for payments cost and performance, here’s a pragmatic month-one plan:

1. Build a true “cost per successful order” metric
   • Include: interchange, network fees, processor fees, chargeback losses, manual review, and retry overhead.
2. Instrument issuer/BIN-level visibility
   • Even basic BIN insights reveal where approval rates or costs are out of line.
3. Add intelligent retries before adding new methods
   • AI-informed retry timing and routing often lifts approvals without changing UX.
4. Pilot pay-by-bank with targeted segments
   • Start with internal employees, loyalty members, or high-frequency buyers.
   • Incent with a capped budget and measure net margin impact.
5. Automate dispute triage
   • Classify disputes by likely win rate and internal handling cost.
   • Reduce the workload first; then improve win rates.

The goal isn’t to “replace cards” overnight. The goal is to stop being price-takers on every transaction you process.

The bigger signal for 2026: payments are becoming programmable

Walmart’s filing is a reminder that merchant frustration with card economics isn’t fading—it’s escalating. The most credible response isn’t another fee dashboard. It’s infrastructure that can act.

In this AI in Payments & Fintech Infrastructure series, we’ve been tracking a clear shift: payments are moving from static rails to programmable systems. When your payment stack can decide—based on real-time economics and risk—you gain the flexibility merchants have wanted for years.

If you’re planning your 2026 roadmap, here’s the question worth sitting with: when fees change, fraud patterns shift, or an issuer starts declining more often, does your system adapt automatically—or do you find out two weeks later in a report?