Darius Baruo
Feb 19, 2026 19:01
Andreessen Horowitz partner Sam Broner makes the case for stablecoins over credit cards in the emerging $15.88 trillion B2B payments market for AI agents.
Andreessen Horowitz is betting that credit cards won’t cut it for the AI agent economy. In a new analysis published February 19, a16z crypto partner Sam Broner argues that stablecoins—not traditional payment rails—will capture the emerging market for autonomous AI agents making business purchases.
The thesis hinges on a counterintuitive observation: AI agents won’t shop like consumers. They’ll operate like businesses, with pre-negotiated vendor relationships, volume pricing, and credit lines. That distinction matters enormously for which payment infrastructure wins.
The Tourist vs. Local Problem
Broner uses a bazaar metaphor to frame the opportunity. Tourists haggle at every stall, paying retail prices with cash or cards. Locals—the restaurant owner buying from the butcher, the tailor visiting the mechanic—move on credit and established terms.
“Dominant agents don’t need tourists’ payment rails,” Broner writes. “They need vendor relationships, working capital, and credit.”
This isn’t theoretical. ChatGPT already has negotiated partnerships with Shopify, Amazon, and Expedia that smaller AI startups can’t access. Scale begets scale. A travel agent booking a million flights annually gets better airline terms than one booking ten.
Why Cards Fall Short
Credit cards work reasonably well for purchases between $20 and $1,000. But AI agents will frequently operate outside that sweet spot—streaming micropayments to compute providers or settling large vendor invoices.
Visa doesn’t support sub-cent payments. The 30-cent fixed fee makes micropayments economically impossible. And card technology assumes humans in the loop for approvals and fraud detection—assumptions that break down when agents need to transact autonomously at machine speed.
“Agents adoption is happening too quickly for the thousands of PSPs, POSs, merchants, and client endpoints to slowly upgrade,” Broner argues.
The Stablecoin Opportunity
The timing matters. The global B2B payments market hit $11.69 trillion in 2024 and is projected to reach $15.88 trillion by 2030. That growth coincides with rapid AI agent development—and whoever captures the payment layer for autonomous commerce could lock in significant market share.
Stablecoins offer specific advantages for this use case: no minimum fees killing micropayments, no interchange eating margins on large transfers, and programmable features that can be built without waiting for legacy infrastructure upgrades.
“An agent streaming $0.001/second to a compute provider and a manufacturer settling a $50,000 vendor invoice can use the same rail,” Broner notes.
The on-ramp friction that plagues stablecoin adoption today becomes less relevant when AI agents handle the complexity. The agent becomes the tour guide, facilitating currency exchange and transactions while the user simply approves batched purchases.
What Needs Building
Stablecoins still lack key infrastructure that cards have refined over decades: arbitration mechanisms, statement generation, credit facilities, and batch approval systems. Broner acknowledges these gaps but argues they’re easier to solve on programmable money than legacy rails.
The Capital One-Brex acquisition announced February 14 for $5.15 billion signals traditional finance recognizes the B2B payments opportunity. But the question is whether incumbents can move fast enough.
“Payments are sticky,” Broner concludes. “New relationships built on stablecoins will become old relationships still built on stablecoins.”
For entrepreneurs building AI agents today, the choice of payment infrastructure could determine which ecosystem captures the next wave of autonomous commerce—and the window for that decision is narrowing.
Image source: Shutterstock
