Restaking

Restaking lets ETH that is already staked on Ethereum — or a liquid staking token (LST) that wraps a native staking position — be reused to secure additional services on top of the base chain. On EigenLayer these services are called AVSs (Actively Validated Services). Stakers delegate their stake to operators, operators opt into a set of AVSs, and in return for the extra yield they accept the extra slashing conditions those AVSs impose.

The core tradeoff is capital efficiency versus concentrated risk: the same stake earns yield from several sources at once, but it also becomes exposed to the failure of every service it secures.

How restaking works

• Stakers deposit ETH or an LST and delegate it to an operator.
• Operators run the software for the AVSs they choose to serve and allocate a portion of their delegated stake to each.
• AVSs consume this pooled economic security and define the conditions under which stake can be slashed.

A liquid restaking token (LRT) wraps a restaked position into a transferable token, the same way an LST wraps a native staking position — adding yet another contract layer between the holder and the underlying ETH.

Security surfaces

Restaking is not a single risk but three distinct failure surfaces sharing one product:

• Cryptoeconomic — slashing, and the redistribution variant (ELIP-006) that lets slashed funds be routed to a recipient address rather than burned, turning a compromised slashing key into a potential theft. Unique Stake and Operator Sets localize slashing so one bad AVS cannot trigger systemic loss.
• Market structure — an LST or LRT can depeg not because it is unbacked but because it cannot be redeemed fast enough; duration risk in the withdrawal queue widens the discount, and leverage loops (often powered by flash loans) turn that discount into a liquidation cascade.
• Smart-contract and mechanism — the withdrawal queue is a shared resource that can be griefed, and the accounting around native-stake verification, share conversions, and reentrancy is where real audit findings live.

The surfaces are distinct, but a single trigger — a slashing event, for example — can propagate across all three, so the blast radius is not.