What Is Restaking? How Reused Stakes Buy Extra Yield and Extra Risk

Restaking is one of those ideas that sounds like free money until you look closely at what the second layer of yield is actually paying for. The pitch is simple enough: your ether is already staked, already earning, already helping to secure Ethereum. Why not let that same stake underwrite a handful of smaller services and collect a second income on top?

You can, and the second income is real. So is the second set of ways to lose money. Restaking turns one deposit into collateral for several jobs at once, and every job brings its own rules about what counts as a mistake and how much that mistake costs.

This guide explains how restaking works in practice, who the players are, where the extra yield comes from, and what changed when slashing finally switched on.

The problem restaking was built to solve

Every proof-of-stake network needs validators, and validators need something valuable at risk. That is what makes an attack expensive: to rewrite history or approve fraudulent transactions, an attacker has to put up more capital than the attack is worth.

Assembling that security from scratch is brutally expensive for a new protocol. You need a token, independent operators, enough stake to make an attack irrational, and years of track record. Most projects fail at one of those steps.

Ethereum already solved the problem at massive scale. Restaking takes that pool and rents it out, so a smaller service can inherit a slice of Ethereum's economic security instead of building its own.

The concept is closely associated with Sreeram Kannan, a University of Washington professor who founded Eigen Labs in 2021 to commercialise it. EigenLayer is the protocol that put the idea into production on Ethereum, and for better or worse, it defines how most people understand the category today.

Three roles, one deposit

Restaking splits into three jobs, and it helps to keep them separate.

Restakers supply the capital. They either point an existing Ethereum validator's withdrawal credentials at an EigenLayer contract, or they deposit a liquid staking token such as stETH into a strategy contract.

Operators do the work. They run the software for each service they support, keep it online, and accept the consequences when it misbehaves. Restakers delegate their stake to an operator rather than running infrastructure themselves.

Services, usually called Actively Validated Services or AVSs, consume the security. A data availability layer, an oracle network, a bridge validator set, a price feed, a zero-knowledge prover — each needs a group of operators doing verifiable work, with something at stake if they cheat.

The flow runs in one direction. Capital moves from restaker to operator, and the operator chooses which services to join. That matters: a restaker picks an operator and then inherits whatever services that operator later opts into, so collateral can end up backstopping work you never evaluated.

Native restaking versus liquid restaking tokens

There are two practical routes in, and they carry different risk.

Native restaking means committing ether you already validate, or staking fresh ether directly through EigenLayer. You keep control and you keep the full yield, but you need at least 32 ETH per validator, you take on the operational burden, and your capital is locked while it is committed.

Liquid restaking is the popular route. You deposit a liquid staking token or ether into a provider such as ether.fi, Renzo or Kelp, and receive a liquid restaking token in return. That token represents your restaked position and can often be used elsewhere in DeFi, which is precisely why it became popular: people did not want their capital sitting idle while it earned a second yield.

The trade-off is a longer chain of dependencies. With a liquid restaking token you are trusting the underlying staking protocol, EigenLayer itself, the operator your provider delegates to, the provider's contracts, and any protocol where you deploy the token. In 2026 several of those layers produced expensive lessons: exploit-driven losses and peg scares around restaked assets showed how fast a problem at one provider spreads through lending markets.

Where the extra yield actually comes from

Restaking rewards arrive from three places, and the mix matters more than the headline rate.

Service fees are the honest source. An AVS pays for the security it consumes, usually in ether or in its own token. If a service has real users, those fees are real revenue.

Token emissions are the promotional source. Protocols hand out their own tokens to attract stake early, on the theory that security begets usage. These rewards are only worth what the token is worth when you sell, and selling pressure from farmers tends to work against that.

Points programmes are the speculative source. You earn non-transferable points that may convert into a future allocation. That is a bet, not a yield.

For most of the category's short history, emissions and points have dominated. Base Ethereum staking yields roughly 3%, and restaking has typically added a percentage point or two on top, often paid in tokens rather than ether. That is a modest premium for taking on extra slashing exposure — a point worth sitting with before you chase a double-digit advertised rate.

Slashing: when the risk became real

For roughly two years, restaking had a credibility problem. Operators could register for services and collect rewards, but the mechanism that was supposed to confiscate their stake for misbehaviour had not shipped. Every claim about "slashable security" rested on reputation.

That changed on 17 April 2025, when EigenLayer activated slashing on mainnet under a protocol upgrade known as ELIP-002. Two design pieces came with it.

Operator Sets let each service define its own group of operators, with specific registration rules, hardware expectations, task assignments and slashing conditions. An operator has to actively accept those terms rather than drifting in by default.

Unique Stake Allocation is the accounting model underneath. An operator divides its delegated stake into slices and assigns each slice to one operator set. Only that service can slash that slice. If a data availability layer holds a fifth of an operator's stake and a bridge holds another fifth, a failure in one cannot reach into the other.

That compartmentalisation does not make restaking safe. It caps the blast radius of a single bad service, which was the most-cited systemic worry. It does not stop an honest operator with buggy service software from losing money, and it does not stop a service writing slashing conditions so broad that ordinary downtime becomes punishable.

Withdrawals add another wrinkle. When you queue an exit, your shares can still be slashed while they wait, and slashed shares go through a resolution delay before they clear. The money in the interface is not always the money you receive.

What the EIGEN token adds

Restaked ether handles faults that can be proven on-chain: a double signature, an invalid attestation, a block that contradicts consensus. Some failures cannot be proven that way. An oracle publishing bad data, or a service censoring requests, may be obviously wrong to any reasonable observer without being provable by a smart contract.

EIGEN exists for that grey zone, an idea Eigen Labs calls intersubjective slashing. Holders of EIGEN can be penalised for faults that a community agrees on but code cannot verify, and the ultimate backstop is a fork: if a majority of EIGEN stakers collude, honest participants can fork the token and strip the attackers of its value. It became transferable on 1 October 2024 and is separate from the ether you restake. It is also a governance token with an unlock schedule, which means its price behaves like a young altcoin rather than a bond.

How big the market got, and how quickly it cooled

EigenLayer's deposits crossed $20 billion during 2024, a period when points farming made restaking one of the most crowded trades in DeFi. The number told you more about incentives than about demand for security.

The unwind was equally instructive. DefiLlama put EigenCloud, the umbrella that now houses EigenLayer, at roughly $6.1 billion in total value locked in October 2026, still around 94% of the restaking category but a fraction of the peak. Reporting through 2026 noted that slashing had been live for over a year without a major event hitting a large service — reassuring, but also a reminder that the mechanism had not yet been stress-tested by a real claim.

Restaking beyond Ethereum

Ethereum is not the only place the idea took hold. Babylon built a market for using bitcoin to secure proof-of-stake chains, letting holders put idle BTC to work without bridging it into a wrapped asset, and on Solana restaking arrived through providers that let staked SOL be redeployed to additional services.

These variants differ in mechanics, guarantees and maturity, and their standards are far less settled than Ethereum's. The underlying idea is the same: capital already committed somewhere is asked to do a second job.

Questions worth asking before you restake

Start with the service, not the yield. Which AVSs would your stake actually secure, and what are their slashing conditions in writing? If a provider answers vaguely, that is the answer.

Then check the operator: uptime history, how many services it runs, whether it publishes a slashing policy of its own, and whether it has been slashed before. If you use a liquid restaking token, ask who decides which services the underlying stake supports, what happens if the token trades below redemption value, and how long an exit really takes.

Finally, size the position as risk capital. Restaking pays a small premium for a tail risk that has not yet materialised at scale. If you would not hold through a slashing headline and a temporary depeg, the extra point of yield is not compensation worth having.

If you are still working out the base layer, it helps to understand how yield farming rewards are actually funded before layering a second source of yield on top, and how decentralised exchanges price and route the tokens those rewards are paid in.

What would make restaking matter

The category has solved its technical credibility problem. Slashing exists, operator sets exist, and the isolation model answers the design worry that dominated 2024.

The economic problem is unsolved. Restaking is only durable if services pay fees that exceed the risk they impose, and if those fees reach the people whose capital is on the line. Much of the current yield is still subsidised, and the first serious slashing event has yet to be tested.

That is not a reason to dismiss the idea. Renting security is a sensible primitive. It is a reason to treat restaking as infrastructure in an early, uneven phase rather than a yield product. The second income is real. So is the second bill, and it arrives less predictably.