💵 Agentic Commerce 💵
It’s hard to talk about technology in 2026 without talking about AI agents.
They're now heralded as the future of how humans interact with software and the internet.
But this raises an important question. ‼️
What happens when these agents need to move money?
Not just assist with tasks, but negotiate prices, purchase services, and settle payments with other agents.
This is where the concept of agentic commerce enters the conversation.
——————————
💰 The $5 Trillion Opportunity
Many researchers believe AI agents could soon mediate a massive share of online economic activity.
According to a report by McKinsey, AI agents could handle $3–5 trillion in global commerce by 2030. Stripe CEO Patrick Collison has also suggested that agents could eventually make orders of magnitude more payments than humans.
But not all agentic commerce looks the same.
According to Delphi Digital, there are two main models emerging.
1️⃣ Version 1: Agent-Assisted Commerce
Here, AI agents help humans complete purchases — booking flights, ordering food, or shopping online. Humans still make the decisions; the agent simply executes them.
2️⃣ Version 2: Agent-to-Agent Commerce
This is where things get more interesting. In this model, agents transact directly with each other without human intervention — paying for compute, purchasing data, or hiring specialised agents to complete tasks.
Some refer to this second model as the machine economy.
——————————
🔍 Where Crypto Fits In
The infrastructure powering these two models may look very different.
For Version 1, traditional payment rails still make the most sense. Consumers want fraud protection, chargebacks, and dispute mechanisms — features that credit cards already provide.
That’s why companies like OpenAI, Stripe, and Google are building agentic payment infrastructure compatible with existing card networks. Systems like the Agentic Commerce Protocol (ACP) and Google’s AP2 allow agents to transact using credit cards with predefined spending limits and merchant restrictions.
But Version 2 agentic commerce is a different story.
Agent-to-agent transactions will likely involve high-frequency micropayments — things like paying a fraction of a cent for API access, compute, or data.
Traditional payment networks simply can’t handle that. ❌
Most charge ~$0.30 per transaction, making sub-cent payments economically impossible.
Crypto solves this problem.
Stablecoins and blockchains allow near-instant settlement and micropayments down to fractions of a cent. They also provide a decentralized trust layer where agents can identify each other, coordinate, and transact without intermediaries.
Several crypto protocols are already building the infrastructure for this machine economy.
➡️ ERC-8004 introduces a registry system allowing AI agents to verify the identity and reputation of their counterparties.
➡️ x402, pioneered by Coinbase, embeds stablecoin payments directly into API calls — allowing agents to pay for services automatically.
➡️ Virtuals Protocol’s ACP enables agents to discover services, negotiate terms, and execute agreements onchain, complete with escrow systems and evaluator agents to verify completed work. As of early 2026, the platform has reported over 23,500 active wallets and nearly $480 million in agent-generated economic output.
——————————
❓ The Big Question
While there are sure to be growing pains, it’s clear that Agentic commerce is inevitable.
If AI agents truly become independent economic actors, the internet may soon have machines transacting with machines at a scale humans never could.
And if that future materializes, the real battle won’t just be about AI. ⚔️
It will be about which financial rails power the machine economy.
The real question is whether version 2 agentic commerce can grow as large, if not larger than, version 1 agentic commerce. 🤔
The answer to that will determine whether crypto or traditional card networks capture the most value from the rise of agentic commerce.
