AI Agents May Soon Dominate Ethereum Transactions — Here’s Why
Coinbase developers Kevin Leffew and Lincoln Murr predict that artificial intelligence (AI) agents will become Ethereum’s most vital users, leveraging the blockchain’s decentralized infrastructure for autonomous transactions.
Speaking at a recent industry conference, they identified the HTTP 402 “Payment Required” standard and Ethereum Improvement Proposal (EIP) 3009 as key enablers. These upgrades allow AI agents to pay with stablecoins without human intervention, which has the potential to change decentralized applications.
The HTTP 402 allows programmatic payments for digital services, such as API calls, cloud storage, or computational resources.
EIP-3009 improves this with secure, programmable ERC-20 token transfer receipts, allowing AI agents to sign and settle transactions in real time. Leffew compared the process to a vending machine, where payments unlock services in real time.
Diagram of V1 protocol sequencing (Source: Github)
Ethereum’s trustless settlement layer supports these atomic transactions and programmable policies, which positions the blockchain as a foundation for AI-driven commerce with the capacity to process billions of dollars in transaction volume.
Developers Test AI Payments, Boosting Ethereum’s Growth
Early adopter developers are already testing AI-driven payments on Ethereum. Hyperbolic Labs is using HTTP 402 to allow its large language models to pay for computational resources on Ethereum, creating autonomous economic loops.
Prodia Labs is applying similar mechanisms for media generation, with AI agents paying for rendering jobs in real time, streamlining content creation workflows.
Leffew and Murr outlined potential use cases, including self-driving taxis covering operational expenses or AI models generating on-demand content. These rely on Ethereum’s intermediary-free transaction system.
Etherscan data shows Ethereum processed around 1.74 to 1.88 million transactions daily in August 2025, with transfers of stablecoins like USDC and USDT comprising around 40% of activity.
This strong transaction base suggests Ethereum can support the microtransactions AI agents require, from API calls to data processing.
Analysts predict AI-driven transactions could add $500 million to Ethereum’s annual transaction fees by 2030, assuming high adoption. This projection is based on Ethereum’s Q1 2025 validator revenue of $1.45 billion, majorly from DeFi and stablecoin activity.
While speculative, the projection is a testament to the anticipation for AI’s role in improving Ethereum’s economy, especially as developers refine payment standards for mainstream adoption.
Will Rules and Scale Limit Ethereum’s AI Agent Dreams?
While the future is promising, there are issues to overcome. Ethereum’s mainnet has 15–20 transactions per second (TPS), which could face scalability challenges under potential AI-driven microtransactions.
Layer-2 solutions like Optimism and Arbitrum are addressing this, collectively processing over 10 million transactions in July 2025. Arbitrum alone settled 46 million monthly transactions, and Optimism’s total transactions numbered 661.9 million as of August 2025, per Etherscan. These platforms are essential to settling AI-driven demand.
Regulatory uncertainty poses another obstacle. Autonomous AI agents conducting financial transactions face scrutiny in jurisdictions with strict digital asset laws.
The U.S. GENIUS Act clarifies stablecoin regulations but does not fully address AI-driven systems; this could potentially affect adoption.
Coinbase developers are working with ecosystem partners to standardize HTTP 402 and EIP-3009 for improved interoperability. “Ethereum’s trustless payment system is the backbone for AI economies,” Leffew said.
