How restaking works, and the risk you take on with it

Restaking asks a simple question: can the ETH already staked on Ethereum do a second job without unstaking it? EigenLayer built the plumbing to make the answer yes, and roughly $12 billion in deposits followed. The catch is that the same ETH now backs promises to more than one system at once.

What does restaking actually do

Normally, staked ETH secures one thing: the Ethereum base layer, through a validator running consensus software. Restaking lets that same validator, or its staked ETH, opt into securing other services too.

• You deposit ETH or an LST (like stETH) into a restaking protocol such as EigenLayer.
• You (or an operator you delegate to) opt into securing one or more "actively validated services" (AVSs) - things like oracles, bridges, data availability layers, or new rollups.
• Each AVS sets its own slashing conditions. If the operator misbehaves or goes offline in a way the AVS defines, part of the restaked ETH gets slashed.
• In exchange, the AVS pays a yield on top of normal staking rewards, usually in its own token plus a cut of protocol fees.

The pitch is capital efficiency. Instead of every new service bootstrapping its own token and validator set from scratch, it rents security from ETH that's already locked up and battle-tested.

What does the yield actually look like

Say you're running 32 ETH as a validator earning roughly 3-4% from Ethereum staking. You restake through EigenLayer and delegate to an operator securing two AVSs: an oracle network and a new data availability layer.

• Base ETH staking yield: ~3.5% APY, paid in ETH.
• AVS 1 (oracle) reward: ~1-2% APY, often paid in the AVS's own token.
• AVS 2 (DA layer) reward: ~1-3% APY, also token-denominated.
• Total stacked yield: could land anywhere from 5% to 9%+ nominal, but most of the upside is in tokens with unclear liquidity and price risk, not ETH.

The nominal APY looks better than plain staking. Whether it's actually better depends on whether those AVS tokens hold value, and on what you risk to earn them.

What's the actual risk

The core tradeoff is that one set of collateral now backs multiple sets of slashing rules, and those rules aren't uniform or, in most cases, fully battle-tested.

• Compounded slashing risk. A bug or malicious action in an AVS you secure can get your restaked ETH slashed, even if your Ethereum validator behaved perfectly.
• Operator risk. If you delegate to an operator instead of running your own node, their mistakes become your losses. You're trusting their key management and uptime across every AVS they support.
• Smart contract risk, stacked. EigenLayer's contracts, the AVS's contracts, and any LST contract underneath all need to work correctly. More layers, more surface area.
• Correlated risk. Restaking concentrates a lot of Ethereum's staked capital into a small number of middleware protocols and operators. A bad bug there doesn't stay contained to one AVS.
• Underpriced early yield. Early AVS rewards are often subsidized with token emissions to bootstrap adoption. That yield can shrink fast once emissions taper or token prices drop.

EigenLayer's own docs are explicit that slashing conditions are set by each AVS individually, which means the risk profile changes every time you opt into a new service. There's no single "restaking risk number" - it's additive and specific to what you've delegated into.

What to check before restaking

Don't treat the advertised APY as the full picture. Check these first.

• Which AVSs is your ETH or LST actually securing, and what are their specific slashing conditions?
• Is the operator you're delegating to running multiple AVSs, and what's their track record on uptime?
• How much of the yield is paid in ETH versus in an AVS token you'd need to sell to realize value?
• Has the AVS been audited, and for how long has it run in production versus testnet?
• What's your actual unbond or withdrawal time if you want out?

Restaking isn't free yield. It's ETH doing more jobs at once, which means more ways for one of those jobs to go wrong and touch your principal.