AI Agents Will Need Money: Why Stablecoins Could Become The Native Payment Layer For AI

Ken Nizam is the co-founder of AsiaTokenFund Group, ATF Capital and 1M Technology, a Web3 and Fintech venture studio and accelerator.

gettyFor decades, software has been remarkably good at processing information but remarkably bad at moving money. An application can search the internet, analyze millions of documents, generate software and make recommendations in seconds. But when it needs to pay for an API call, purchase computing resources or compensate another software service, it usually has to stop and wait for a human, a credit card or a traditional billing system.

AI agents are changing that equation. As AI moves from answering questions to independently taking actions, agents will increasingly need to become economic actors. They will book services, purchase data, consume computing power, negotiate with vendors and potentially transact with other agents.

That raises a fundamental question technology leaders should start addressing now: If AI agents can act independently, how will they pay?

The first generation of AI assistants primarily generated information. The next generation will increasingly execute tasks.

Imagine telling an AI agent, “Find the best flight to Tokyo, book it within my budget, arrange transportation and update my calendar.” The agent may need to interact with several services, each involving a transaction.

Now take this further. Imagine a software development agent that needs additional computing power for 15 minutes. Instead of relying solely on a monthly cloud subscription, it could purchase computing resources dynamically. Or a research agent could pay a data provider a few cents for access to a proprietary dataset.

These transactions are fundamentally different from traditional e-commerce. They may be extremely frequent, low value and machine-to-machine.​​

The existing payment system was designed around humans. A person buys something, enters payment credentials and receives a receipt. The merchant processes the transaction and absorbs the associated costs.

AI agents operate differently. An agent might make hundreds or thousands of transactions during a single task. Many could be worth only a few cents or fractions of a cent.

Traditional card transactions involve authorization, network fees, merchant fees, fraud controls and settlement processes that can make little economic sense for a $0.02 API request.

Stablecoins offer a different model. A dollar denominated stablecoin can move on blockchain infrastructure 24/7, with programmable transaction rules and potentially low transaction costs. This makes stablecoins particularly interesting for machine-to-machine payments.

The significance isn’t that stablecoins are another form of cryptocurrency. It is that they could become programmable digital dollars for software.​

In 2026, Amazon Web Services introduced a preview of Amazon Bedrock AgentCore Payments, developed with Coinbase and Stripe. The service allows AI agents to autonomously access and pay for APIs, MCP servers, web content and other agents.

An agent can receive an HTTP 402 “Payment Required” response, authenticate its wallet, make a stablecoin payment and continue its task. Developers can also impose spending limits and monitor transactions. Payment is no longer necessarily a separate checkout experience.​

Coinbase and AWS are pushing this concept further. In June 2026, they announced an integration allowing publishers and API providers using AWS CloudFront and WAF to accept payments from AI agents through x402.

The model is simple: An agent requests content, receives a payment request, makes the payment and receives the content.

The web is effectively gaining a mechanism for saying, “If your AI agent wants this information, it can pay for it.” That could fundamentally change how the internet monetizes information.​

Traditional financial institutions are moving in the same direction. In June 2026, Visa announced new infrastructure combining AI, stablecoins and tokenized payments. Its initiatives include agent verification, transaction controls and infrastructure designed for AI-initiated commerce.

Visa also reported that its stablecoin settlement activity had reached an annualized run rate of approximately $7 billion.

Cards may remain important for consumer purchases. Bank accounts and traditional payment networks will remain essential. Stablecoins may become particularly useful for global settlement and machine-scale micropayments.

The winning infrastructure may abstract all of this away.​

Giving an AI agent a wallet is relatively easy. The difficult question is, “How much authority should that wallet have?”

We should not give an AI agent unrestricted access to a corporate bank account any more than we would give an employee an unlimited company credit card. Agentic payments will therefore require programmable financial controls.

A company might allow an AI procurement agent to spend up to $500 per transaction and $10,000 per day, only with approved vendors. A software development agent might be allowed to purchase computing resources but prohibited from transferring money to an unknown wallet. A research agent might spend $100 on information services without asking for approval, while anything above that threshold requires human authorization.

This is where blockchain becomes particularly interesting. Smart contracts and programmable wallets can enforce rules at the infrastructure level. Circle, for example, is developing programmable agent wallets with spending limits, service caps, allowlists and time-bounded sessions.​

If this model scales, the economic structure of the internet could change. Today, many APIs are sold through subscriptions. A company might pay $99 per month regardless of whether it makes 10 requests or 100,000.

Agentic commerce enables a different model. An API could charge $0.002 per request. A data provider could charge $0.01 for a specific dataset. An AI model could charge based on inference consumed. A computing provider could charge by the second.

Instead of humans subscribing to software, software could purchase capabilities dynamically from other software. This creates the possibility of a genuinely machine-native economy.

The model could eventually extend beyond AI. IoT devices could pay for network access. Autonomous vehicles could pay for charging. Robots could purchase maintenance services. Enterprise software could negotiate and settle transactions without human intervention.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?