Tech giants like Google and Amazon are in the business of predicting where society is headed, and in recent months, both companies have started making moves to corner the development of AI agents—automated assistants authorized to zip around the internet, completing complex tasks on behalf of their human overlords and other machines.
The push to develop a formidable AI agent economy is still far from complete. But when robots are eventually let loose en masse to transact efficiently with both the existing economy and each other, experts predict their productivity and output will rival that of humans.
The main question looming over the development of the AI agent economy is what infrastructure will best facilitate this explosion. Increasingly, top minds in Silicon Valley and crypto are coalescing around a single answer to that lucrative question: Ethereum.
Ethereum’s core developers have recently arrived at the conclusion that the network is uniquely well positioned to become the foundational layer of the AI agent economy, given its ability to provide three key ingredients the ecosystem currently lacks: payment rails, identity verification, and trust.
The team is confident that within a handful of years, Ethereum will not just be foundational to the AI agent economy, but also that AI agents will become the core user base of the network.
“For us, it’s very important. It’s a strategic area,” Davide Crapis, an Ethereum core developer focused on AI, told Decrypt this week.
Crapis said that within three to five years, he believes the majority of traffic on Ethereum will be coming from machines.
ERC-8004
Earlier this month, Crapis debuted ERC-8004: a proposed interface for Ethereum that would standardize how AI agents discover each other on the network and establish trust sufficient to engage in economic interactions.
The proposal fixes what Crapis sees as the major flaws in existing ecosystems for agent-to-agent interactions. In April, Google unveiled the Agent2Agent protocol, which it promised would allow AI agents to seamlessly collaborate and “drive unprecedented levels of efficiency and innovation.”
But the framework has its shortcomings. For one thing, it doesn’t currently enable payments—an essential ingredient for a genuinely autonomous robot economy. Two, it doesn’t give agents the means to identify and trust each other out on the open internet. That means, in practice, that the protocol can only be used effectively to facilitate the interaction of agents within a single organization, on tasks that don’t involve financial transactions.
By its nature, Ethereum can easily fill these fundamental gaps, Crapis said. The payments issue is instantly solved by on-chain transactions, which AI agents already are capable of completing. As for identity and trust: that’s Ethereum’s bread and butter. NFTs, for instance, provide a secure means of establishing a unique digital identity. ERC-8004 provides a simple framework for how AI agents would go about validating each other’s identities on-chain.
And if Ethereum were to provide that framework to undergird the AI agent economy, it’s not like the blockchain network would be going up against the likes of Google. On the contrary, the Silicon Valley behemoth is actually backing Crapis’ Ethereum proposal. Jordan Ellis, one of the core Google employees behind its Agent2Agent protocol, is a co-author of ERC-8004.
“This, for me, is a signal that it’s not too early,” Crapis said of the collaboration. “In the sense that even in the traditional AI space, people are looking into agent-to-agent payments, and agent-to-agent identity.”
Powerful stakeholders in the burgeoning AI agent economy want to see the ecosystem as universally standardized as possible, to increase its potential reach and ease of navigation. These companies may not be crypto maximalists, necessarily; but if blockchain networks solve their problems far more easily than other approaches, what’s the downside?
The perfect use case?
Time and again over the last decade, crypto projects have struggled to reach mass adoption, in large part because they’ve failed at convincing mainstream consumers that the pain of navigating complicated blockchain networks is worth the gain of financial incentives or privacy benefits.
But in the looming age of the robot economy, crypto’s marketing woes may become far less of a liability. Crapis, who is now back at the Ethereum Foundation after a few years working on AI-related projects, is adamant that when the AI agent economy booms, robots will unemotionally choose the most efficient terrain on which to complete transactions—and that this best market will unquestionably be Ethereum.
“Our challenge has been making [Ethereum] more UX-friendly for humans to use, trying to shift their behavior,” Crapis said. “But if the user is an agent or a machine, then it’s fairly easy. Robots don’t have any problems remembering their private keys.”
The traditional economy was built for humans, and designed to verify human activity. (What’s your mother’s maiden name?) Ethereum, on the other hand, almost seems like it was built for robots, years before they possessed the capability to roam the internet on their own. That long-perceived liability—the network’s convoluted user experience—may now finally reveal itself as a boon in the era of the agent-dominated internet.
Even among other blockchains, the Ethereum team feels the network’s signature multi-layer structure is uniquely well poised to absorb the massive amount of AI agent traffic likely to arise in coming years.
The base Ethereum blockchain will provide foundational security and stability to handle the deluge and to verify particularly high-stakes transactions, they say, while an ever-customizable, expandable, cheap, and speedy legion of layer-2 networks will be able to handle the likely massive quantity of everyday, smaller-scale settlements.
0/ Autonomous agents are about to become Ethereum’s biggest power users.
Guest thread from @kleffew94 and @MurrLincoln on how a long-forgotten HTTP status code, ‘402 Payment Required’ could unlock a new frontier for Ethereum: agentic commerce. 🧵
— Ethereum (@ethereum) August 13, 2025
Other blockchains will have an immensely difficult time carrying the weight of the entire AI agent economy on their shoulders, Crapis said.
“Solana, in its current design, cannot sustain the machine economy,” he said, giving the example of Ethereum’s rival network. “They have no idea how much activity can come on-chain, once these machines start using it.”
The software developer predicts that once the AI agent economy arrives in full force, it will redefine the function of Ethereum, just as decentralized finance (DeFi) did back in 2020.
Getting the Ethereum developer ecosystem to agree on a standard for agent-to-agent encounters is the first crucial step in preparing for that day. Crapis said he intends on tweaking ERC-8004 over the next few months, as he gets feedback from community members.
But the standard will then be finalized in short order, to prepare for the arrival of an army of intelligent, crypto-wielding robots.
“I cannot predict when this takeoff will happen,” Crapis said, “but I feel that we have some urgency to build for it.”
