AI Agents Can't Open Bank Accounts — Why Crypto Is Becoming the Default Infrastructure for Machine Finance

The next billion users of crypto might not be human. On March 9, 2026, Coinbase CEO Brian Armstrong posted a thesis that is reshaping how both Wall Street and Silicon Valley think about blockchain: AI agents cannot open bank accounts, but they can own crypto wallets — and that single fact could redirect trillions of dollars in economic activity onto decentralized rails.

Within days, Binance founder Changpeng Zhao amplified the argument with a blunter claim: AI agents will eventually make one million times as many payments as humans, and they will use crypto. Bitwise CIO Matt Hougan called agentic finance "a big emerging catalyst," predicting that most internet transactions will ultimately settle on-chain.

This is not a theoretical debate. The infrastructure is already live, the transaction volumes are real, and the biggest names in fintech are racing to capture a market that barely existed twelve months ago.

The Core Argument: Why Banks Cannot Serve Machines

Traditional banking rests on a foundational assumption — the customer is a person. FinCEN's Customer Identification Program requires institutions to verify the identity of any individual opening an account. That means a name, a date of birth, an address, and a government-issued ID.

An AI agent has none of these. It is software — a set of weights, a prompt, a runtime. It cannot satisfy Know Your Customer requirements because there is no customer to know.

This is not a temporary gap waiting for regulation to catch up. The entire architecture of identity-based banking is structurally incompatible with autonomous software that creates, executes, and terminates in milliseconds. A single orchestration layer might spin up thousands of sub-agents for a complex task, each needing to send or receive payments. No bank can onboard thousands of ephemeral entities in real time.

Crypto wallets face no such constraint. A wallet is generated from a private key — no identity check, no approval process, no waiting period. An AI agent can create a wallet, fund it, transact, and dissolve it within the span of a single API call.

The x402 Protocol: HTTP Meets Money

The most tangible evidence that machine finance is already here is Coinbase's x402 protocol, which has processed over 50 million transactions since its launch. The protocol's name comes from HTTP status code 402 — "Payment Required" — a response code that was defined in the original HTTP specification but never implemented because no native internet payment layer existed.

x402 embeds stablecoin payments directly into HTTP requests. When an AI agent hits a paywall or an API with a usage fee, it pays in USDC within the same interaction and continues its task. No redirect to a checkout page. No card number entry. No human approval.

The protocol is backed by Cloudflare, Circle, AWS, and Stripe — a coalition that represents a significant portion of global internet infrastructure. This is not a crypto-native experiment; it is a bet by mainstream tech companies that machine-to-machine payments will become a core internet primitive.

Coinbase's Agentic Wallets, launched on February 11, 2026, build on this foundation. They provide AI agents with autonomous spending, earning, and trading capabilities while maintaining programmable guardrails — session spending caps, transaction size controls, and gasless trading on the Base Layer 2 network. Developers can deploy an agent wallet via command-line interface in under two minutes.

The Economics That Make It Inevitable

The argument for crypto as machine payment infrastructure is not ideological — it is economic.

Credit card networks charge minimum fees of roughly 30 cents per transaction. For a human buying a $50 item, that is negligible. For an AI agent making 10,000 micro-transactions per hour — paying fractions of a cent for API calls, compute cycles, data feeds, and model inference — those minimums make card rails economically impossible.

Crypto settlements happen in under 500 milliseconds at costs below $0.001 on modern Layer 2 networks. This makes sub-cent, high-frequency machine-to-machine payments not just viable but cheap.

The numbers tell the story of where this is heading:

• Gartner estimates AI "machine customers" could influence or control up to $30 trillion in annual purchases by 2030.
• McKinsey projects agentic commerce could generate $3 to $5 trillion globally by 2030.
• The AI agents market is projected to grow from $7.84 billion in 2025 to $52.62 billion by 2030 — a 46.3% compound annual growth rate.
• The stablecoin market is forecast to expand 10x to $3 trillion by 2030, driven significantly by AI agent adoption.

TradFi Fights Back — But On Whose Terms?

The traditional financial system is not standing still. Visa launched its Trusted Agent Protocol in October 2025, and Mastercard completed Europe's first live AI-agent bank payment inside Santander's regulated infrastructure in March 2026. Both approaches layer cryptographic verification onto existing card rails.

But there is a fundamental mismatch. Card networks were designed for human-initiated, merchant-terminated transactions with dispute resolution, chargebacks, and fraud detection built around human behavior patterns. AI agent transactions look nothing like this — they are high-frequency, low-value, programmatic, and often between two machines with no human in the loop.

Stripe's response is telling. Rather than forcing AI payments through card rails, the company reintroduced USDC payments on Solana, Ethereum, and Polygon — explicitly for machine-to-machine commerce. When the largest payment processor in Silicon Valley chooses stablecoins over its own card infrastructure for a specific use case, the signal is hard to ignore.

The Emerging Machine Finance Stack

A full infrastructure stack for AI agent finance is crystallizing rapidly:

Wallet Layer: Coinbase Agentic Wallets (EVM + Solana), MoonPay's Ledger-secured AI agents, and open-source alternatives provide identity-free wallet creation for autonomous software.

Payment Protocol: x402 enables native HTTP payments. AI agents pay for resources the same way browsers request web pages — as part of the protocol itself.

Settlement Layer: Base (Coinbase's L2), Solana, and Ethereum provide sub-second finality at minimal cost. The Ethereum Foundation's new dAI division is specifically positioning Ethereum as the preferred settlement layer for machine-driven commerce.

Verification Infrastructure: EigenCloud's partnership with Google Cloud provides verifiable infrastructure supporting AI agent transactions, ensuring that autonomous payments maintain cryptographic proof of execution.

Currency: USDC and other regulated stablecoins serve as the unit of account — stable enough for commerce, programmable enough for machines, and liquid enough for instant settlement.

The Skeptic's Case

Not everyone is convinced the future arrives this cleanly. Bloomberg reported in March 2026 that "stablecoin firms bet big on AI agent payments that barely exist," highlighting the gap between infrastructure investment and actual transaction volume.

The skeptics raise valid concerns. Most AI agents today operate within closed ecosystems where internal accounting suffices. The number of truly autonomous agents making real economic decisions with real money remains small. Regulatory frameworks for machine-initiated financial transactions do not exist, and when they arrive, they could impose identity requirements on wallets that currently operate without them.

There is also the question of liability. When an AI agent makes a bad trade or pays for a service that does not deliver, who bears the loss? Traditional finance has centuries of contract law and dispute resolution built around human accountability. Machine finance has none of this.

Fortune's analysis noted that while the case for AI and crypto is strong, "you have to squint" to see it in current transaction data. The infrastructure is ahead of the demand — a familiar pattern in technology, but not a guarantee of success.

What Happens Next

The most likely near-term outcome is a split. Regulated commerce — consumer purchases, payroll, mortgages — stays on traditional rails with AI augmentation via Visa and Mastercard protocols. Machine-to-machine payments — agents hiring agents, paying per API call, buying compute on demand — migrate to stablecoins because the economics demand it.

The critical variable is speed of AI agent deployment. If autonomous agents remain a niche developer tool, crypto's machine finance thesis stays theoretical. But if the current trajectory holds — agent-related crypto projects surged from 5% to 36% of all crypto AI deals between H2 2023 and H1 2025 — then the infrastructure being built today will look prescient.

As Armstrong put it: AI agents are not the future users of crypto. They are the current users of crypto. The 50 million transactions already processed through x402 are not a proof of concept. They are the opening chapter of what happens when software gets a wallet.

BlockEden.xyz provides high-performance RPC endpoints and API infrastructure across multiple blockchain networks including Ethereum, Solana, Base, and Sui — the same settlement layers that AI agent wallets depend on for fast, reliable transactions. Explore our API marketplace to build the infrastructure your autonomous agents need.