Ethereum is positioning its base layer to coordinate autonomous agents, a move that puts machine, to machine commerce on a direct path to on-chain settlement in the coming year.
This month, the Ethereum Foundation formed a dedicated dAI Team with a remit to advance agent identity, trust, and payments, including support for ERC-8004, a draft standard for agent credentials and verification that would anchor identity and attestations at the protocol level.
The initiative frames Ethereum as a settlement and coordination layer for agent economies, with censorship resistance and open access as core design goals, while community drafts around ERC-8004 outline how on-chain identity and trust could allow automated systems to negotiate, post bonds, and execute escrow without custodial intermediaries.
The near-term deliverable is research and standards progress that can be adopted by wallets, middleware, and dApps in 2026, creating a shared trust substrate for agentic applications.
Token flows already reflect an AI tilt in crypto markets.
AI-focused tokens such as Bittensor, Fetch.ai (ASI), Internet Computer, and Render have maintained on-chain activity and relative price stability through Q3, outpacing broader altcoins during the recent market drawdown.
Koinly’s market roundups point to continued demand for decentralized compute, inference, and agent frameworks, while ecosystem reports show ICP’s push for native app hosting and Render’s GPU marketplace drawing steady usage from AI workloads.
Per Token Metrics, DeFi total value locked has rebounded from roughly $72 billion in early 2025 toward the $100 billion area, with new AI-native DeFi rails such as Blackhole DEX on Avalanche, Sahara AI, and Moby AI contributing to volumes and fee generation that persisted through volatility. Token Metrics places this in a broader rotation toward automated liquidity and agent execution that can operate across chains via messaging and omnichain abstractions.
The payments stack is converging on agent use cases at the protocol boundary. Google introduced the Agents to Payments, or AP2, protocol in September to let software agents request and confirm consumer payments through standardized flows, a building block for machine-to-machine billing and subscription patterns that can interface with crypto settlement rails.
According to Google Cloud, AP2 is designed around explicit user consent, verifiable agent identities, and reversible transactions for compliance, and early pilots include Ethereum and ICP integrations via third-party connectors that bridge fiat accounts with on-chain transfers.
As these pilots mature, wallets may treat agents as first-class actors, with ERC-8004 style attestations allowing policies that cap spend per timeframe, restrict counterparties, or require human co-sign for high value thresholds.
Forward models now tie plumbing upgrades to measurable network demand.
Token Metrics’ September scenario work projects AI smart agents reaching 15 to 20 percent of DeFi transaction volume by late Q4, which, if sustained and amplified by Ethereum’s dAI roadmap, places AI-integrated protocols in the $200 to 300 billion TVL range by end-2026.
The same analysis frames a feed-through into base-layer utilization, with gas usage for agent identity and execution contracts rising 30 to 40 percent quarter over quarter in 2026 once standards like ERC-8004 see broad adoption across custody, consumer wallets, and DAO middleware.
In practice, this means governance, treasury rebalancing, fee routing, and cross-chain liquidity management could be executed by software agents that operate with risk limits, insurance, and verifiable credentials on chain.
Security outcomes are another lever in the adoption curve. Academic and industry research on adaptive, AI-assisted contracts points to a sharp drop in successful exploits when contracts can detect anomalies, tune parameters, and quarantine suspicious flows in near real time.
Early models show reductions of up to 70 percent in successful attacks for systems that pair rule-based controls with learned heuristics, relative to static parameter schemes. This result depends on transparent update policies and monitorable on-chain behavior to avoid creating opaque control surfaces, a point that dovetails with supervisory attention on smart contract auditability and incident reporting.
The macro context is turning from concept to pilot.
Regulatory agendas in the United States and Europe include workstreams on automated financial agents, transparency for adaptive contracts, and disclosures around model risk.
DLA Piper’s September brief and other legal trackers describe a path where agent identities, usage policies, and exception handling must be legible to regulators and counterparties, a requirement that aligns with Ethereum’s identity and attestation thrust rather than contradicts it.
Recent enforcement themes place emphasis on control effectiveness, not technology bans, which supports a runway for compliant agent operations as standards mature.
Hiring data remains supportive, with Recruitblock recording a 22 percent year-over-year increase in 2025 for roles at the AI and blockchain intersection, spanning protocol engineers, data infrastructure, and applied cryptography, a pipeline that matters if agent frameworks are to reach production scale across consumer and enterprise touchpoints.
Cross-market, the machine economy lens is not confined to a single stack. Avalanche hosts AI-governed liquidity via Blackhole DEX, Ethereum focuses on identity and settlement, NEAR and ICP court on-chain app hosting and low-latency inference, and Render supplies GPU resources for training and model serving.
Koinly’s and Token Metrics coverage place these in complementary roles rather than direct substitutes, with a thesis that demand for decentralized inference and marketplace coordination expands as agents become default actors in payments, fulfillment, and protocol operations.
If ICP’s growth model for native AI hosting holds, on-chain inference cycles could cut latency by half by 2026, which would make agent interactivity viable for user-facing applications like intent routers, real-time hedging, and supply-chain or IoT settlement.
| Protocol | Primary AI function | On-chain volume or TVL, Sept 2025 | Forward focus |
|---|---|---|---|
| Ethereum | Agent identity and settlement, ERC-8004, dAI Team | $38B+ | Trust and coordination layer for agents |
| Bittensor, TAO | Decentralized training and inference markets | $1.4B est. | Open AI compute exchange |
| Fetch.ai, FET | Autonomous economic agents, dApp infrastructure | $640M est. | Machine-to-machine coordination |
| Render, RNDR | Decentralized GPU and inference | ~$985M | Compute backbone for on-chain AI |
| Internet Computer, ICP | Native on-chain AI app hosting | $800M+ | Lower latency for agentic dApps |
| Blackhole DEX, Avalanche | AI-governed AMM and liquidity | $193M | Permissionless agent trading |
The scenarios fall into three buckets.
A base case has Ethereum consolidating the identity and trust layer, as at least a quarter of new dApps adopt agent automation by 2026, converging governance, treasury, fees, and payments into programmable policies anchored in attestations.
A bull path turns on a fuller machine economy where agents handle bilateral negotiation and fulfillment across consumer and enterprise contexts, with DeFi TVL moving beyond $300 billion and decentralized AI API marketplaces reaching critical mass for long-tail services.
A bear case centers on regulatory licensing of agents and ongoing centralization of compute and model access, which would cap open participation and bottleneck innovation to a small number of well-funded teams.
DLA Piper’s overview and policy trackers point to transparency and control standards as the fulcrum, not outright prohibitions, yet compute centralization remains a known constraint.
Investors and builders shift from token narratives to measurable adoption triggers.
On the standards side, ERC-8004 is a core watch item, since wallets and custody providers will need to implement attestation checks, recovery flows, and policy enforcement for agents to operate safely in consumer contexts.
On the payments side, AP2 pilots, if extended into crypto rails at scale, would provide the first repeatable pattern for subscriptions, usage billing, and fulfillment between non-human actors, and would pressure bridges and account abstraction stacks to expose fine-grained limits and approvals.
On the security side, field evidence that adaptive controls reduce realized loss would unlock more autonomous governance, especially for parameter tuning in volatile markets. Each of these tracks has public milestones that can be monitored without relying on price charts alone.
The open question is goes beyond whether agents will transact; it is where the settlement and trust checks occur.
If identity, attestations, and policies live on the chain, the machine economy will default to public ledgers, and DeFi will become the operating system for non-human economic activity. If those checks remain in closed platforms, crypto’s role will collapse to bridges and payout rails.
With Ethereum’s dAI mandate, the AP2 pathway for agent payments, and a measurable shift in developer hiring toward AI x crypto roles, the center of gravity is moving toward verifiable, on-chain coordination that treats agents as first-class participants in markets.