a16z: Why Do AI Agents Need Stablecoins for B2B Payments?
TechFlow Selected TechFlow Selected
a16z: Why Do AI Agents Need Stablecoins for B2B Payments?
Stablecoins work from day one—even without negotiated merchant agreements or complex B2B terms.
Navigating Web3 tides with focused insightsBy Sam Broner, a16z crypto
Translated by AididiaoJP, Foresight News
As a tourist walking through a bazaar, you’ll witness this scene: people bustling about, scrutinizing goods, comparing items, sampling products, haggling with each vendor, and finally paying. It looks like a one-off transaction—each interaction a small negotiation—with trust mediated by cash in hand or value exchanged via credit card.
But that’s not how most business gets done in the bazaar. Look closer: most people are locals who head straight to their favorite merchants. A restaurant owner visits his friends—the butcher, the fishmonger, and the farmer. A tailor seeks out the mechanic, the weaver, and the artisan. Both sides extend and receive credit.
When we talk about how intelligent agents will pay, we default to thinking like tourists.
But intelligent agents will act like locals. The very traits that distinguish intelligent agents from humans—unlimited replicability, flexible resource allocation, and zero marginal cost of creation—mean that only a few intelligent agents can dominate a niche market. Even as building intelligent agents becomes easier, relationships, partnerships, and trust remain essential for delivering winning experiences. Dominant intelligent agents don’t need the tourist’s payment rails. They need supplier relationships, working capital, and credit.
What would that look like? As intelligent agents consolidate into business-like platforms, agent payments must shift from retail payment rails to pre-negotiated B2B terms and credit—a gap current rails cannot fully fill. This is the opportunity window for the next-generation payment rail—such as stablecoins—if entrepreneurs build robust solutions tailored to next-gen use cases: intelligent agents, streaming payments, and high-volume, low-value global commerce.
This article explores that thesis in three parts: how intelligent agents differ from humans—and how those differences shape winning payment strategies; why current approaches fall short; and what the next-generation payment rail must build to win.
How Intelligent Agents Differ from Humans
To understand intelligent agents and payments, we must ask two questions: Will intelligent agents behave more like people—or like businesses? And will they play the long game—or the short game?
Intelligent agents will behave more like businesses, cultivating long-term relationships with suppliers and partners. Intelligent agents will be lightly customized instances built atop a larger enterprise structure—like a flawless tour guide at a well-connected travel agency, or a franchisee adapting its operations manual to local tastes without renegotiating its entire supply chain.
Why Intelligent Agents Behave Like Businesses
First, the best experiences are carefully designed. I don’t want an agent that dawdles at checkout, comparing prices and haggling with suppliers on the spot. I want an agent that has already done that work—an agent that knows which suppliers are reliable, has pre-negotiated pricing, and can settle instantly. That’s a business relationship—not a tourist-style transaction.
In fact, human agents already exist: travel agencies, literary agents, talent agents, watch dealers, real estate brokers, and more. These agents cultivate critical, multi-round relationships—with publishers, studios, watch distributors, or mortgage lenders—customizing each transaction atop that foundation.
Second, intelligent agents are infinitely replicable—but scale advantages (and the benefits they confer) are not. Top-performing intelligent agents will leverage the cost and revenue advantages of scale: cheaper compute, better supplier pricing, deeper integrations, and more predictable components. Scale begets scale: a travel agency booking one million flights per year secures far better airline terms than one booking ten.
We’ve already seen this. Only ChatGPT has the channel access to negotiate partnerships with Shopify, Amazon, Expedia, and others. Smaller startups are stuck scraping websites or reverse-engineering APIs—while paying retail-tier fee structures.
That’s why intelligent agents consolidate—or at least why most are built atop larger platforms. Building intelligent agents is easy, but economics favor a small number of winners per vertical—each backed by deep supplier relationships and healthy margins to reinvest in superior experiences. And vertically specialized agents with deep supplier ties can partner with user-facing agents to deliver the best of both worlds.
Two Payment Relationships
If intelligent agents behave like businesses, then two distinct payment relationships must be designed: user → agent, and agent / agent platform / agent “guide” → supplier.
Users pay the agent—via subscription, per-task fees, credit lines, or delegated account access. The agent pays suppliers under negotiated B2B terms: volume pricing, net-30 invoicing, or sub-agency arrangements. By analogy to today’s corporate spend, agents may occasionally use retail rails to pay suppliers—but even then, it represents only a small fraction of total spend.
This is, in fact, how credit cards work today: the issuing bank maintains a retail relationship with consumers—bearing risk, crafting custom rewards, and extending credit. The acquiring bank maintains a commercial relationship with merchants—including negotiated terms, bulk transfers, and complex working-capital conversations.
Agents and Credit Cards: A McKinsey-Perfect Fit
As many have noted, credit cards are actually quite reasonable payment products for agent use cases. They’re widely accepted; payments between $20 and $1,000 are considered appropriate; and they come with built-in arbitration, cancellation, and digital functionality.
Credit cards also offer monthly statements—the key moment for consumers to understand what they’ve paid. As intelligent agents replace iPad-wielding kids as the primary source of surprise charges, that concept will certainly evolve.
But two problems remain: first, credit card technology is ill-suited for agents at a technical level. Second, its pricing model pushes the credit card industry into a classic innovator’s dilemma.
Credit Card Tech Is Hard to Upgrade
Nearly all credit card technology presumes human involvement: an approver, a UI layer, and a traditional payment type (one-time, subscription). Stripe Link, Visa 3D Secure, and dozens of other virtualization products—those letting you save cards on websites for future purchases or register them for recurring monthly subscriptions—now work well. But it took over 15 years to get there.
Agent adoption is happening too fast for thousands of PSPs, POS systems, merchants, and client endpoints to slowly upgrade their interfaces, programmability, and fraud-detection systems for this new payment flow.
Credit Cards Fail at Both High- and Low-Cost Purchases
Imagine an agent streaming payments to a compute provider—or making micropayments for API access. Neither works on credit card rails. First, Visa doesn’t support payments below one cent. Second, its economic model expects fixed fees around $0.30 per transaction. Visa could technically build streaming or micropayment capabilities—but convincing stakeholders to accept lower per-transaction revenue would be far harder.
More problematically, credit cards are trapped in the innovator’s dilemma. Though their user relationships and payment requirements resemble agent payments, agent payments often fall outside the $20–$1,000 range. Worse, many early use cases involve paying for API calls—fees that are hard to refund or easily resold (i.e., subject to fraud).
Even beyond credit cards, traditional rails will retain a role going forward.
Traditional Payments Still Have a Role
As intelligent agents consolidate into business-like platforms, most high-volume spend will shift to pre-negotiated B2B terms: invoicing, net-30 payment, discounts, and credit lines. In that world, the “payment rail” can be anything—often just a mundane, asynchronous settlement over traditional rails. Fees amortize across larger transactions; working capital negotiates directly between enterprises.
But agents won’t live solely in that world. Agents are emerging now—and they’re operating precisely where traditional payments fall short: establishing first-time relationships, cross-border checkout, simplifying complex reconciliation, enabling novel agent-supplier models, facilitating instant payments to reduce lending costs, and enabling microloans.
In these scenarios, stablecoins are the superior payment choice—and critically, it’s far easier to build next-generation functionality on programmable money than on legacy infrastructure. New relationships built using stablecoins become old relationships that continue using stablecoins. Over time, as full stablecoin payment platforms launch, stablecoins—which are already cheaper, faster, and more global—will likely capture a larger share of the payment mix.
Opportunity for New Payment Technologies
To anticipate what comes next, we should focus on technologies best suited to rapidly growing use cases.
Stablecoins—1:1 backed by highly liquid assets, faster, cheaper, and globally accessible—are a new platform meeting the needs of underserved commercial categories like international payments and streaming payments. Key stablecoins are programmable: core features like arbitration, monthly (or hourly) billing, credit, escrow, and conditional payments can be flexibly extended to support many new use cases. Unlike bank or card payments, stablecoin payments integrate seamlessly into APIs, databases, and agent checkout flows—greatly simplifying reconciliation, approval, and onboarding processes—a massive advantage for founders racing to build agent-based businesses.
At a practical level, stablecoins solve credit cards’ unit-economics problems at the extremes. No $0.30 minimum fee makes micropayments viable. No fees eroding margins on large transfers mean an agent streaming $0.001/sec to a compute provider and a manufacturer settling a $50,000 supplier invoice can use the same rail. This flexibility is critical when engineers and founders choose their next build platform.
Building More Stablecoin Infrastructure
The most common objection to stablecoin adoption is expensive on- and off-ramps. That’s true for uninformed “tourists”—but disappears when users are guided by a “guide,” i.e., an agent. Guides can help tourists convert currencies and execute precisely needed transactions—saving on fees.
Add billing and arbitration features to our stablecoin-guided system, and we’re close to the required architecture.
Imagine shopping at a department store like Dell. You browse multiple brands, select items, and check out once at a unified register. The store handles the complex split-payment logistics across brands. Agents need the same pattern: a unified cart view spanning multiple suppliers, with one-click batch approval. Users see “Your agent wants to book a flight, reserve a hotel, and rent a car”—not three separate checkouts. The agent platform manages supplier relationships; the user manages intent. Users can approve, review, or dispute transactions.
Credit cards handle arbitration well—but new rails need that layer too. Arbitration is easiest when goods are high-value or easily returnable: flights cancellable within a 24-hour window, subscriptions not yet started, luxury goods with healthy margins—suppliers can absorb reversals. But early agent use cases often involve low-margin digital goods—compute power, API calls—or food delivery.
Summary
Intelligent agents won’t pay like tourists. They’ll pay like locals—through relationships, credit, and repeat transactions. That means the real volume of payments will flow via pre-negotiated B2B terms—not credit card swipes. Frankly, pre-negotiated B2B terms don’t require a new payment rail. The settlement layer can be anything—wire transfer, ACH, or plain-vanilla batch processing. For established relationships, traditional payments work well.
But we’re at a fork in the road. Agents are happening now. Founders are building—and they need payment tools that work today—not after multi-year credit card tech stack upgrades. Credit cards aren’t ready: too expensive for micropayments, too challenging for reconciliation, burdened by technical debt, and requiring manual intervention for fraud decisions. Stablecoins are ready. They’re programmable, global, easy to reconcile with digital services, and integrate effortlessly into APIs and agent checkout flows. They work from day one—even without negotiated merchant agreements or complex B2B terms.
That’s the opportunity window. Founders building agents today will seek tools that work well today. Payments are sticky. Ultimately, new relationships built on stablecoins become old relationships that stay on stablecoins. Over the next few years, the ecosystem will mature, on-ramp friction will vanish, and infrastructure gaps—billing, arbitration, credit, batch approvals, interoperability—will be filled by a wave of startups building on stronger foundations.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
a16z
