The number is staggering but the logic is mundane. Animoca Brands co-founder Yat Siu told Forbes this week that he expects 50 to 100 billion AI agents to be operating within the next few years, and virtually none of them will have a bank account. The reason is not technological but procedural: banks require identity verification tied to a human being. An AI agent—whether it is booking compute, paying for an API, or settling a microtransaction—cannot pass a KYC check. Siu’s conclusion is that these agents will have no choice but to use crypto wallets and stablecoins as their financial backbone, making onchain rails the default settlement layer for machine-to-machine commerce by sheer force of exclusion.
The KYC Wall and the Agent Identity Gap
Siu’s argument rests on a structural incompatibility between the traditional financial system and autonomous software. Banks are legally required to identify their customers under anti-money laundering and counter-terrorist financing regulations. A customer is a natural person or a registered legal entity. An AI agent is neither. Even if an agent acts on behalf of a human, the bank’s relationship is with the human, not the software. The agent cannot open its own account, hold a balance, or sign a transaction. This is not a temporary oversight; it is a foundational design choice of the regulated banking system. Siu’s forecast implies that no regulatory fix is coming fast enough to accommodate the scale of agent deployment he anticipates.
Why Stablecoins Fit the Agent Use Case
If agents cannot access bank accounts, they need a bearer asset that settles digitally, programmatically, and without human intervention. Stablecoins—particularly USDC and USDT on public blockchains—fit this description. An agent can generate a wallet, receive stablecoins, and spend them via smart contracts or protocols like x402 without ever touching a bank. The transaction cost, settlement time, and counterparty risk are all functions of the blockchain, not a banking intermediary. Siu’s framing reframes stablecoins not as a speculative tool or a dollar substitute for humans in high-inflation markets, but as the native currency of autonomous economic actors. This is a significant shift in the stablecoin narrative, moving it from retail remittances to the plumbing of an agent-driven economy.
The Scale Question: 50 to 100 Billion Agents
Siu’s estimate of 50 to 100 billion agents deserves scrutiny. For context, there are roughly 8 billion humans on Earth. His number implies an order-of-magnitude more agents than people, each potentially transacting multiple times per day. If even a fraction of those agents use stablecoins for low-value, high-frequency payments—paying for inference, data access, or API calls—the aggregate transaction volume would dwarf current stablecoin settlement figures. Visa’s stablecoin pilot, which we covered earlier this week, hit a $7 billion annualized run rate. Siu’s vision suggests a future where that number is a rounding error. The infrastructure to handle that load—wallets, key management, gas abstraction, and payment protocols—is still under construction, but the direction of travel is clear.
What This Means for the Stablecoin Stack
If Siu is even directionally correct, the stablecoin market must evolve from serving human users to serving software. That means wallets designed for agent custody, not human UX. It means payment protocols that assume the payer is a machine, not a person clicking a button. It means identity systems—like World’s AgentKit, which links an agent to a verified human via World ID—that bridge the gap between autonomous action and human accountability. The pieces are being assembled in real time. Mastercard’s agent payment network, Circle’s Machine Payments Protocol, and Coinbase’s x402 integration all point toward a world where stablecoins are the settlement layer for a non-human economy. Siu’s prediction gives that trend a number: 50 to 100 billion reasons to build it.