EigenLayer Mechanics and the AVS Ecosystem

The Three Roles That Make Restaking Work

Imagine 100,000 Ethereum validators each pointing their withdrawal credentials at a single programmable contract — and that contract's rules rippling into the security guarantees of a data-availability layer, an oracle network, and a cross-chain bridge simultaneously. That is EigenLayer in a sentence. But the mechanics underneath that sentence are what determine whether the yield you receive is compensation for genuine risk or simply a number that will hurt you later.

Part 1 of this series introduced restaking as a mental model. This article opens the hood. EigenLayer has three distinct participant classes — restakers, operators, and AVSs — and they interact in a triangle, not a hierarchy. Understanding where each role begins and ends is the prerequisite for evaluating any position in the restaking ecosystem.

Restakers supply the collateral. They point an Ethereum validator's withdrawal credentials at an EigenPod contract (native restaking) or deposit a liquid staking token such as stETH into an EigenLayer strategy contract (LST restaking). Either way, their ETH becomes the economic backstop for services further down the chain.

Operators are the actors who run the software. They register on-chain, set commission rates, and opt into specific services. A restaker cannot interact with those services directly — they must delegate their stake to an operator they trust. One restaker, one operator at a time per deposit.

AVSs — Actively Validated Services — are the consumers. Each AVS defines what operators must do, what constitutes misbehavior, and how many operators it needs. It pays fees in return for access to the pooled cryptoeconomic weight that flows from restakers through operators.

The flow is directional: restaker delegates to operator → operator opts into AVS → AVS inherits the economic weight of every delegating restaker. If the operator violates the AVS's conditions, the delegating restakers absorb the loss. Note the asymmetry: the restaker chooses an operator, but the operator chooses which AVSs to join. After delegation, restakers cannot veto new AVS opt-ins.

This design lets a single unit of ETH simultaneously back multiple AVSs — a feature called rehypothecation that multiplies yield but also multiplies the number of ways that ETH can be slashed.

Native Restaking vs LST Restaking

How collateral enters EigenLayer determines its risk profile from day one.

Native restaking requires running or controlling a validator — 32 ETH minimum under the pre-consolidation rules. The validator's withdrawal credentials are pointed at an EigenPod contract. The underlying ETH stays on the Beacon Chain; the EigenPod acts as a pass-through that lets consensus-layer rewards flow normally while subordinating the stake to EigenLayer's slashing logic on top of standard Beacon Chain penalties.

LST restaking is accessible without a full validator. Users deposit stETH, rETH, mETH, or other supported tokens into EigenLayer strategy contracts. Because LSTs are already derivatives of staked ETH, LST restakers carry layered risk from three separate systems simultaneously:

1. The LST issuer's smart-contract risk
2. The underlying Beacon Chain slashing risk
3. The EigenLayer slashing risk

As of late 2024, native ETH accounted for roughly 84% of restaked collateral; LSTs made up the remaining 16%, primarily stETH. Both paths still require delegating to an operator before any AVS economic weight — and therefore any restaking yield — becomes active. Depositing into a strategy contract without delegating earns nothing.

Operators, Operator Sets, and Unique Stake Allocation

Operators register on-chain with metadata and commission rates. Restakers choose which operator to delegate to, and that choice is exclusive per deposit — stake cannot be split across multiple operators from one position.

How operators plug into AVSs changed fundamentally on April 17, 2025, when EigenLayer activated on-chain slashing. Before that date, operators opted into AVSs through social commitment and reputational stakes alone. There was no cryptoeconomic punishment for misbehavior. The entire security guarantee was theater backed by nothing enforceable.

After April 2025, EigenLayer introduced Operator Sets: structured groupings that each AVS uses to organize the operators it trusts. An AVS defines registration requirements, hardware specifications, and slashing conditions per operator set.

Critically, this launch introduced Unique Stake Allocation. An operator now designates specific amounts of delegated stake exclusively to one operator set. That portion can only be slashed by that AVS. The allocation cannot be claimed by a different AVS's slashing event — which partially isolates but does not eliminate cross-AVS exposure.

At the time slashing launched, EigenLayer had over 2,000 registered operators. By March 2026, active operator count was approximately 1,900. Operator concentration remains a live concern: a small number of large professional operators hold the majority of delegated stake. A configuration error or exploit at one large operator affects a disproportionate share of the protocol's overall security budget.

The AVS Ecosystem: What Gets Built on Pooled Security

The first AVS mainnet launch occurred on April 11, 2024, alongside the EigenLayer Stage 3 general mainnet. Six initial services launched:

• AltLayer MACH — fast finality for Optimism rollups
• Eoracle — an Ethereum-native oracle network built on restaked validators
• Lagrange State Committees — ZK light client infrastructure for optimistic rollup bridging
• Brevis ZK Coprocessor — co-processing for ZK computation
• Witness Chain's DePIN Coordination Layer — coordination infrastructure for decentralized physical infrastructure
• Xterio MACH — fast finality for gaming rollups

EigenDA is EigenLayer's own data-availability AVS and by far the most widely adopted. EigenDA V2 launched on mainnet on July 30, 2025, delivering 100 MB/s sustained throughput, approximately 60x latency reduction to ~5 seconds average, and a peak of 124 MB/s in testing. It secures over $2 billion in L2 customer assets.

The AVS taxonomy spans five practical categories:

1. Data availability — EigenDA
2. Oracle networks — eoracle (see oracle networks for the broader price-feed problem)
3. Fast finality and rollup acceleration — AltLayer MACH, Xterio MACH
4. Cross-chain messaging and bridges — Hyperlane
5. ZK infrastructure and co-processors — Lagrange State Committees, Brevis

From June 2025, Eigen Labs repositioned the protocol as EigenCloud — a developer platform backed by a fresh $70 million purchase from a16z (bringing total a16z commitment to over $170 million). EigenCloud unifies EigenDA, EigenVerify (dispute resolution), and EigenCompute (off-chain execution verification) under one roof and extends the verifiability model to AI inference, gaming, and Web2 applications. Verifiable AI workloads launched on EigenCloud mainnet in late 2025; EigenCompute entered mainnet alpha in January 2026.

By the time slashing launched in April 2025, the ecosystem counted 190+ AVSs in development with 40 live on mainnet.

Slashing: How Enforcement Actually Works

The gap between EigenLayer's marketing and its actual cryptoeconomic guarantees was substantial until recently. Before April 17, 2025, there was no on-chain slashing. Operators chose AVSs and delegators chose operators, but there was no mechanism that could forcibly reduce a restaker's collateral. The entire security guarantee was reputational.

Slashing went live on mainnet on April 17, 2025, making EigenLayer's security model what the protocol called "feature-complete."

The mechanism works as follows:

1. Each AVS defines slashing conditions attached to specific Operator Sets.
2. An operator is enrolled in an Operator Set with a defined Unique Stake Allocation.
3. If the operator violates the AVS's conditions, the AVS can trigger a slash against that allocated stake.
4. The slash fires against the restakers who delegated to that operator.

Two important limitations constrain the current slashing model. First, operators do not automatically opt into slashing conditions when an AVS adopts them. They must actively accept new slashing terms, which means the security model strengthens incrementally rather than uniformly. Second, slashed stake is burned, not redistributed. Redistribution — where slashed funds compensate harmed parties rather than being destroyed — was announced as forthcoming but was not live as of mid-2026.

Slashing also introduces a composability concern that goes beyond any single AVS. When the same ETH backs multiple services simultaneously, a failure in one service's slashing logic can cascade in ways that are difficult to model in advance. Part 4 of this series explores exactly that dynamic through the composability risk lens and the Kelp contagion event.

The EIGEN Token

EIGEN is a separate asset from restaked ETH, and it serves a different purpose.

Where restaked ETH backs tasks with objective, on-chain-verifiable outcomes — "did this block hash match?" — EIGEN is designed for intersubjective tasks: situations where correctness cannot be verified by a smart contract alone but can be judged by the community when a dispute arises. The dispute-resolution mechanism of last resort is forking the EIGEN token itself, which creates an economic deterrent against misbehavior on services that require this type of validation.

EIGEN launched with a stakedrop in 2024. Transferability and exchange trading began on October 1, 2024, with the token debuting at a fully diluted valuation of approximately $6.5 billion. Total initial supply was approximately 1.67 billion EIGEN. As of May 2026, circulating market cap was roughly $145 million — a significant decline from the FDV at launch. Governance over protocol parameters and future upgrades flows through EIGEN stakers.

Risks to Understand Before You Participate

Operator concentration risk. A small number of large professional operators hold the majority of delegated stake. A configuration error, exploit, or poor AVS selection at one large operator affects everyone who delegated to that operator — far beyond what a distributed operator set would imply.

AVS opt-in and slashing risk. Operators choose their AVS exposure after you delegate. Each additional opt-in introduces a new slashing surface for every restaker who delegated to that operator — even if the restaker never evaluated that specific AVS. You cannot veto new opt-ins retroactively.

Smart-contract risk. EigenLayer's strategy contracts, EigenPod contracts, and each AVS's slashing-condition contracts are separate attack surfaces. The codebase is large and actively evolving. Audits reduce but do not eliminate this risk.

Rehypothecation risk. The same ETH earning EigenLayer restaking rewards also earns Beacon Chain consensus rewards — and if an LST is used, the LST protocol's yield on top. Stacking three yield layers means a failure at any layer can cascade. Part 4 of this series covers the Kelp contagion as a concrete case study of how this plays out.

Liquidity and exit risk. Unstaking from EigenLayer is multi-step: withdraw delegation, unstake from the strategy, complete any remaining Beacon Chain exit queue. In stress scenarios, these queues extend unpredictably. BTC restaking via Babylon follows a structurally similar multi-step logic — both designs share the exit-queue problem.

Key Takeaways

• EigenLayer's three-role triangle — restakers, operators, AVSs — is non-hierarchical. Restakers delegate to operators; operators opt into AVSs; AVSs inherit the economic weight. Risk flows back down the same chain.
• Native restaking (EigenPod) and LST restaking (strategy contracts) are mechanically different and carry different layered risks. LST restakers face three stacked risk systems, not one.
• On-chain slashing did not exist before April 17, 2025. Every security argument made about EigenLayer before that date was reputational, not cryptoeconomic.
• Unique Stake Allocation partially isolates cross-AVS slashing exposure, but operator concentration and post-delegation opt-in rights remain unresolved structural concerns.
• EigenDA V2 and EigenCloud mark the protocol's expansion from a restaking primitive toward a broader verifiable compute platform — the AVS surface area is growing, not shrinking.
• Before entering any liquid restaking token position, map the operator set, the AVS exposure, and the exit path. Part 3 covers how the major LRTs — weETH, ezETH, and rsETH — make those choices for you.