Imagine a future where your AI shopping assistant is so good at finding deals, it starts looking at how you pay for things. And it decides your credit card's 3% fee is a terrible deal. That's the core of a speculative scenario laid out by Citrini Research in a recent note, framed as a pure thought exercise about how AI could upend the entire economics of the card payment world.

The central idea is pretty simple. If AI agents are constantly running transactions for us, they'd eventually zero in on the 2-3% interchange fee charged by card networks as a major, unnecessary cost. "Once agents controlled the transaction, they went looking for bigger paperclips," the analysts wrote. The logical move for these cost-cutting bots? Shift payments to stablecoins settled on super-cheap networks like Solana or Ethereum Layer 2s, where transaction costs are, as the note puts it, "measured in fractions of a penny."

This isn't happening today, but it's a scenario that would put immense pressure on giants like Mastercard (MA) and Visa (V). Interestingly, the note points out that Mastercard has been building out its own stablecoin infrastructure—a move that might look very smart if this future comes to pass.

To make the risk concrete, Citrini constructed a hypothetical earnings headline for Mastercard in Q1 2027. In this imagined world, purchase volume growth has slowed to just 3.4% year-over-year, down from 5.9%, with company management blaming "agent-led price optimization." The result? Mastercard's stock drops 9% the next day. Visa, perceived as having a slightly stronger position in stablecoins, only partially recovers in the scenario.

The pain wouldn't stop at the networks. Citrini argues card-focused banks would face a compounded risk. In their model, companies like American Express (AXP), Synchrony Financial (SYF), Capital One (COF), and Discover Financial Services each fall more than 10%. They'd be hit by a double whammy: the AI bypass gutting their interchange revenue, and potential white-collar job losses (perhaps also driven by AI) shrinking their customer base. The analysts summed up the vulnerability starkly: "Their moats were made of friction. And friction was going to zero."

It's crucial to remember this is just a model, a stress test of an idea. The authors were clear: "The sole intent of this piece is modeling a scenario that's been relatively underexplored." But sometimes, the most useful thought experiments are the ones that point out where the cracks might form in a seemingly solid business model. For the card industry, built on a foundation of small percentage fees, the idea of an AI relentlessly optimizing those fees away is a fascinating—and potentially terrifying—glimpse of a possible future.