Agentic commerce won't kill cards, but it will open a gap

Cards will dominate agentic commerce for existing merchants. Stablecoins will power the ones that don’t exist yet.

The wrong fight

A few weeks ago, a piece from Citrini Research arguing that stablecoins would disintermediate Visa and Mastercard sent card stocks down sharply. Crypto Twitter cheered. The thesis felt clean: AI agents will optimize every transaction, interchange is a tax, stablecoins route around it. I spend my days in crypto, and I wanted it to be right, but most of it is wrong. Not because stablecoins don’t matter. But because the real opportunity has nothing to do with replacing cards. It is about the merchants that will struggle to accept them.

Cards win most of this

The Citrini thesis rests on an assumption: that AI agents, freed from human habit, will optimize away interchange fees. But cards don’t just move money. They extend unsecured credit, pre-authorize uncertain transactions, and guarantee fraud protection with chargeback rights. Stablecoins can move money. They cannot yet do the rest.

Say your agent books a hotel room and it’s nothing like the listing. With a card, you dispute the charge. With a stablecoin, the money is gone.

Eighty-two percent of Americans carry a rewards card. Eighteen billion cards are in circulation globally. For most types of transactions, consumers are not going to voluntarily give up purchase protection and points for an irreversible payment that offers neither. Fraud detection compounds the advantage: card networks run models across billions of transactions in real time. Stablecoins lack a comparable network-level fraud layer today.

The counterarguments get more specific, but the pattern repeats.

Micropayments are often cited as cards’ weakness. But card networks have adapted to ill-fitting transactions before. Visa has processed over 2 billion transit fares by aggregating taps into daily settlements. The card industry has never conceded a transaction category. It has always invented new products to capture each one.

Then there is the identity objection: “Agents can’t hold cards.” But an agent is just a new device. Your phone, your watch, and your laptop each carry a separate token pointed at the same card. Same technology as Apple Pay. Your phone never passed KYC; it just carries your token. So does an agent. Visa has issued over 16 billion tokens. Agents will get these tokens too. Visa’s Intelligent Commerce framework is in pilot. Mastercard’s Agent Pay is live for all US cardholders. The Agentic Commerce Protocol, built by Stripe and OpenAI, already has Etsy live and over a million Shopify merchants set to come online.

For the merchants and consumers that already exist, cards will likely dominate agentic commerce. The stablecoin opportunity lies elsewhere: the merchants that do not exist yet.

The merchants that don’t exist yet

Every platform shift creates a wave of merchants that the existing payment system cannot serve. When eBay created a marketplace of people selling to their peers, those sellers could not get merchant accounts. PayPal served them and grew to a million users, handling 40% of eBay auction payments by 2000. Shopify went from 42,000 merchants to 5.5 million in 13 years. And as Alex Rampell and James da Costa have pointed out, Stripe was founded before many of its eventual customers even existed.

The pattern is consistent: the winners serve the merchants that incumbents cannot yet justify underwriting.

The AI wave will likely create these merchants faster than any previous platform shift. Thirty-six million new developers joined GitHub in the last year alone. In YC’s Winter 2025 batch, a quarter of companies had codebases that were 95% or more AI-generated. On Bolt.new, one of the most popular AI coding platforms, 67% of 5 million users are not developers. Millions of people who could not write production code two years ago are shipping software now. Each one is a new buyer of developer infrastructure, purchased through a command line, not a sales call.

These same developers are also new sellers. A vibe coder needs services as part of their workflow: data endpoints, testing infrastructure, deployment tools. Anyone with the same tools can build those services and sell them right back. The same force is creating buyers and sellers simultaneously.

Imagine a vibe coder builds a tool that cleanly presents financial data for public companies. The project might take four hours of work with AI coding tools. No website, no terms of service, no legal entity. Another developer’s agent calls it 40,000 times in a week at a tenth of a cent per call, generating $40 in revenue with no human ever visiting a checkout page.

I see vibe coders build tools like this every week. The first question is always, “How do I get paid?” For most of them, the answer right now is that they cannot.

Existing payment processors will find it difficult to onboard these merchants. Not because the technology is lacking, but because when a processor says “yes” to a merchant, it takes on that merchant’s risk. If the merchant commits fraud or racks up chargebacks, the processor is on the hook. Processors reject applicants they cannot underwrite. A tool with no website, no entity, and no track record is extremely difficult to underwrite.

The system is working as designed. It just was not designed for this.

Processors could adapt. They have before, creating new risk tiers for payment facilitators and marketplace platforms. But it took 16 years from PayPal’s launch to the first industry underwriting guidelines for the payment facilitator model it pioneered (aggregating merchants under a single account and absorbing their risk). These merchants need to get paid now.

For these merchants, accepting stablecoins is the equivalent of a street vendor only taking cash. Not because cash is better, but because merchants with their profile have historically struggled to get approved for card acceptance.

For that gap, stablecoins are the only option that works today, despite wallet UX that remains rough and compliance frameworks that are still forming. Protocols like x402 already embed stablecoin payments directly into HTTP requests, requiring no merchant account, no processor, no onboarding, no chargeback liability. None of that requires anyone to agree that stablecoins are better than cards. It requires only that incumbent processors have yet to adapt.

These merchants will not be choosing stablecoins over cards. They will be choosing stablecoins over nothing.

What gets created

Every wave of new merchants has eventually been absorbed by the traditional payments system. The same will probably happen here, on some timeline. Still, the merchants come first, and the underwriting catches up later. In the gap between those two moments, stablecoins are the infrastructure.

Cards serve every merchant a processor can underwrite. Stablecoins serve every merchant a processor cannot. The next wave of commerce will be built in that gap.

---

You’re receiving this newsletter because you signed up for it on our websites, at an event, or elsewhere (you can opt out anytime using the ‘unsubscribe’ link below). This newsletter is provided for informational purposes only, and should NOT be relied upon as legal, business, investment, or tax advice. This newsletter may link to other websites or other information obtained from third-party sources — a16z has not independently verified nor makes any representations about the current or enduring accuracy of such information. Furthermore, the content is not directed at nor intended for use by any investors or prospective investors in any a16z funds. Please see a16z.com/disclosures for additional important details, including link to list of investments.