Will AI Agents Use Stablecoins or Virtual Cards?

Stablecoin-Linked Cards Are in Hyper Growth

The episode opened with major news: Visa is expanding its partnership with Bridge into over 100 countries by end of year, powering stablecoin-linked cards at a scale that would have been unthinkable even 12 months ago. Sheffield was blunt about the demand picture.

“Stablecoin-linked cards are in hyper growth mode. It's hard to keep up with all of the companies coming to us that want to issue stablecoin-linked cards.”

— Cuy Sheffield, Head of Crypto at Visa

Taylor noted the infrastructure achievement hiding behind the headline. Historically, issuing cards in 100 different markets meant securing bank partnerships country by country. Having a single API that handles all of them through stablecoin rails is a genuine inflection point — one that flips fintech from a developed-market activity into a truly global one.

Crossmint's Alfonso Gómez-Jordana confirmed the trend from the merchant side: roughly half of Crossmint's customers now have a stablecoin-linked card product. The business model is attractive — acquiring users via stablecoin rails, then monetizing through card interchange. It's the same revenue engine that built companies like Chime, now plugged into crypto-native distribution.

AI-Exclusive Virtual Cards: Crossmint's Bet

The episode's headline story was Crossmint's launch of AI agent-exclusive virtual cards, powered by Visa Intelligent Commerce. The mechanism: a consumer saves a regular credit card on file. When their AI agent needs to make a purchase, it requests a virtual card number with a specific spend limit. The agent uses that virtual card, the consumer keeps earning their credit card points, and the agent operates within defined guardrails.

Gómez-Jordana explained why they went this route over stablecoins, despite having built extensive stablecoin integrations:

“We started serving the agent tech finance ecosystem a year ago. At the beginning we started thinking it would just be stablecoins for everything. Then we hit the market with it, had people, and realized that was not what people were looking for.”

— Alfonso Gómez-Jordana, Founder of Crossmint

The demand signal was clear: agent developers don't want to onboard new payment complexity. They have a credit card. They want their agent to use it. Virtual cards make that possible without forcing anyone into a new ecosystem.

The use cases are immediate and practical. An agent that's vibe coding a project needs to buy a domain name, call a Twilio API, or pay for cloud compute. Each time it hits a credit card form, it needs a payment method. A virtual card with a spend limit solves that without the agent leaving the development environment — what Sheffield calls “command line commerce.”

Where Stablecoins Still Win

If virtual cards are the pragmatic short-term answer, stablecoins remain the infrastructure for use cases that cards weren't designed for. The panel identified a clear dividing line: existing merchants versus new ones.

For established merchants — SaaS companies, cloud providers, domain registrars — cards already work. The agent finds a card form, enters the virtual card number, and the transaction flows through existing rails. Nothing needs to change on the merchant side.

But vibe coding is creating an entirely new class of merchant. Someone builds a microservice from their phone, wants to charge for API access, and has no Stripe account, no business entity, and no way to pass traditional payment onboarding. For these merchants, stablecoin acceptance via protocols like x402 is the fastest path to getting paid.

“These people are writing these services from the toilet, by coding with Cursor, and you cannot bother with the onboarding process. It's much easier to spin up an x402 facilitator and just start taking payments.”

— Alfonso Gómez-Jordana, Founder of Crossmint

Taylor drew a parallel to Stripe's early days. The Collisons pitched investors on a customer that didn't exist yet — the developer as merchant. Many investors rejected it because the market wasn't visible. Today's equivalent is the vibe coder who builds a service in an afternoon and needs to monetize it before they even have a company.

The “Agents Need Bank Accounts” Myth

Sheffield pushed back hard on a narrative that's become common in crypto circles: that AI agents can't get bank accounts, so they must use stablecoins.

“Few things drive me crazier than when I hear the argument that AI agents aren't going to be able to get bank accounts, so they have to use stablecoins. Every agent is created or controlled by some human, and that human probably has a bank account, probably has a card. Why can't that human delegate their credential to that agent?”

— Cuy Sheffield, Head of Crypto at Visa

This reframing matters. The dominant narrative has positioned stablecoins as the only solution for agent payments because agents aren't legal entities. Sheffield's counterargument: agents act on behalf of humans, and those humans already have payment credentials. The real question isn't “how does an agent open a bank account?” but “how does a human securely delegate spending authority to software?”

Virtual cards answer that delegation question cleanly. The human retains control (their card, their credit limit, their points), while the agent gets a scoped credential it can use within defined limits.

The Agentic Co-Founder vs. the Agentic Shopping Assistant

Sheffield made a distinction that reframes the entire agentic commerce conversation. Six months ago, the pitch was conversational commerce — asking ChatGPT to help you find and buy shoes. That's incremental. What's happening now with vibe coding is fundamentally different.

“I think it's a lot more interesting to have an agentic co-founder than an agentic shopping assistant. We're seeing a whole new class of entrepreneurs that start vibe coding, that might have never written code before, and now they're creating projects — but they run into all of these roadblocks and bottlenecks.”

— Cuy Sheffield, Head of Crypto at Visa

The value proposition of agentic commerce has shifted from saving time to creating value. A vibe coder building a product needs their agent to buy domains, provision APIs, purchase hosting, and procure data — all without breaking out of the coding flow. Every time the agent hits a payment wall, the creative momentum stops.

This is why command line commerce matters so much. The goal is to keep the agent inside its working environment, able to acquire whatever resources it needs without the human having to context-switch into a browser to enter card details.

Stablecoins for Enterprise Treasury: Stable Sea's Approach

While much of the agentic commerce discussion centers on consumer-grade agents, Stable Sea is exploring AI agents for enterprise treasury management. Taddeo described a fundamentally different use case from the vibe coding world.

“We're experimenting with an AI agent that serves as your de facto chief of staff to the CFO — managing your treasury across two or three subsidiaries where you have spare idle cash in Argentine peso, South African rand, and US dollars, converting into and out of stablecoins and parking in money market funds.”

— Tanner Taddeo, CEO of Stable Sea

Stable Sea serves real-economy enterprises — companies that have been around 50 to 150 years, with deep supply chains and subsidiaries across the globe. Their core pain points are pre-funding times and trapped cash in emerging markets. An AI treasury agent that can autonomously move funds across jurisdictions using stablecoin rails — converting idle local currency into yield-bearing instruments and back again — is a natural evolution of what these companies already do manually.

This matters because it shows agentic commerce isn't just a consumer play. The enterprise use case for AI agents using stablecoin infrastructure is arguably larger in dollar terms, even if it's less visible than the vibe coding narrative.

Morgan Stanley's National Trust Charter: What It Signals

The episode also covered Morgan Stanley applying for a de novo National Trust charter — a new entity called Morgan Stanley Digital Trust — to own custody, settlement, and fiduciary plumbing for blockchain finance under US bank supervision. They also plan to introduce Bitcoin, Ether, and Solana trading via E*Trade in the first half of 2026.

Sheffield offered a useful framework for understanding why large financial institutions are now applying for new charters rather than just outsourcing to crypto-native companies:

“When the products were really crypto trading, the default instinct was to move fast and outsource. Now the demand is tokenizing securities, using tokenized assets as collateral, on-chain lending — things that are the core guts of the product they're in the business of offering. That changes the build vs. buy calculus entirely.”

— Cuy Sheffield, Head of Crypto at Visa

Taddeo added an important observation: the regulatory arbitrage that crypto-native companies currently enjoy will close over the next five to fifteen years as banks build the licensing infrastructure to bring these services in-house.

“The one signal that is absolutely clear is that on-chain financial services are here to stay, and all the banks are posturing for the right regulatory status so they can crush the arbitrage in the next three to seven years.”

— Tanner Taddeo, CEO of Stable Sea

The Death of “Crypto” as a Risk Category

One of the episode's sharpest insights came from the panel's discussion of how financial institutions categorize risk. Sheffield described a pattern that many in the industry will recognize: during previous crypto hype cycles, institutions created a blanket “crypto” risk category and marked it as restricted. Now that their own customers are adopting stablecoins, those binary categorizations are breaking down.

“When we said we don't do crypto, what does that mean? Are we going to off-board some of our largest customers because they now use a blockchain database? Meme coin trading and tokenized deposits are very, very different activities that both could arguably be bucketed under crypto.”

— Cuy Sheffield, Head of Crypto at Visa

Taylor put it simply: saying you use crypto technology today is like saying you use the internet. The risk-appetite frameworks written for 2017 and 2021 are no longer fit for purpose. Financial institutions are being forced to add nuance — distinguishing between the stablecoin use case, tokenized securities, DeFi lending, and speculative trading, rather than treating them all as one monolithic category.

Key Takeaways: What This Means for the Industry

Episode 73 crystallized a market that's moving faster than most people realize. The key conclusions:

• Virtual cards and stablecoins will coexist. Neither is replacing the other. Cards win where merchants already accept them and consumers want to keep their existing credit card benefits. Stablecoins win where new merchant types emerge (vibe-coded services, AI inference, micropayments) and where card onboarding is too heavy.
• The consumer entry point is cards; the agent can handle complexity. Humans fund with credit cards. Agents can convert to stablecoins, switch rails, and manage the underlying complexity — but only if they have the right tools like virtual cards and wallet infrastructure.
• Vibe coding is the catalyst. The shift from agentic shopping assistant to agentic co-founder is creating real payment demand. Every commerce moment where an agent has to leave its workflow to pay for something is friction that the market is racing to eliminate.
• Enterprise treasury is a massive, quieter opportunity. AI agents managing cross-border stablecoin treasury for multinational enterprises is a use case with potentially larger volume than consumer agentic commerce, even if it gets less attention.
• Banks are building, not just buying. Morgan Stanley, Kraken, Revolut, and others applying for charters signal that on-chain financial services are moving from experimental to core infrastructure. The “partner now, build later” strategy is shifting to “build now.”
• “Crypto” as a binary risk category is dead. Financial institutions need nuanced frameworks that distinguish between stablecoin payments, tokenized securities, DeFi, and speculative trading. The blanket restrictions from 2017 and 2021 are breaking down as their own customers adopt the technology.

As Sheffield summarized, the zero-sum framing of cards versus stablecoins misses the point entirely. There will be trillions of dollars of volume going to both. The real question is who builds the infrastructure fastest — and which merchants, both old and new, adopt it first.