What is restaking? EigenLayer in plain English
EigenLayer let validators re-pledge their 32 ETH bond to secure additional services. TVL hit $20bn in 2024 before correcting. The promise, the slashing surface, and what AVS economics actually look like.
On 14 June 2023, EigenLayer opened its first mainnet deposits and introduced a word that the staking economy had been missing: restaking. Within twelve months, EigenLayer’s TVL had crossed $20bn, briefly making it the second-largest DeFi protocol on any chain. The deposit cap was raised, then removed entirely; the EIGEN token launched in October 2024; the first wave of “actively validated services” — AVSs in the EigenLayer vocabulary — went live with real slashing in early 2025. TVL then corrected as yields normalised and as a wave of restaking-adjacent protocols (Symbiotic, Karak, Babylon for BTC) entered the same competitive space. By August 2026 EigenLayer’s restaked assets had fallen to about $5.10bn, down from $22.06bn a year earlier, as the sector’s economics came under pressure and its single largest customer began to unwind its restaking entirely.
Three years in, restaking is past the hype curve and into the part where the economic model actually has to work — and, on the evidence of 2026, is struggling to. This piece is a plain-English walk-through of what restaking is, what an AVS is, where the slashing surface widens, and why the answer to “is the yield worth it?” is more interesting than the marketing makes it look.
September 2026 Update: Restaking has contracted sharply. On 8 September 2026 the sector held about $10.02bn in TVL but generated only around $100,000 in weekly fees — a far smaller fee base than liquid staking’s. EigenLayer’s own restaked assets had fallen to roughly $5.10bn by August 2026, down from $22.06bn a year earlier. The category’s largest player, ether.fi, removed restaking from its weETH token in August 2026 and said less than 1% of its assets would still be restaked by the end of the third quarter of 2026; it plans to remove its remaining EigenPod withdrawal credentials by year-end 2026, cutting its last technical link to EigenLayer. We have updated the figures throughout this piece to reflect these changes; treat any developments after early September 2026 as unconfirmed until we update again.
The basic restaking insight
The Ethereum validator set is the largest pool of bonded cryptoeconomic security ever assembled. As of 2026, roughly 34 million ETH — about $80-100bn depending on price — is locked up backing consensus. That capital is doing one job: making it expensive to attack Ethereum. EigenLayer’s founding observation was that this capital could be re-pledged to back other services, layering additional slashing conditions on top of the existing one. If you trust the validator already to do consensus, the argument goes, you can trust them to run an oracle, a data-availability committee, or a bridge committee, in exchange for an additional fee stream.
The mechanism is precise. A validator points their withdrawal credentials at an EigenLayer contract instead of a regular address. They then opt in to one or more AVSs, each of which can publish its own slashing conditions. If the validator misbehaves on an AVS, the AVS can call slash on the EigenLayer contract, which burns or seizes a portion of the underlying ETH bond. Crucially, this slashing is in addition to — not instead of — Ethereum’s own slashing for consensus faults. The validator is now backing two protocols with the same bond, and is exposed to both slashing regimes simultaneously.
Native restaking vs LST restaking
EigenLayer accepts two flavours of deposit. Native restaking means the user runs a real validator with withdrawal credentials pointed at the EigenPod contract; the bond is the literal 32 ETH on the consensus layer. LST restaking means depositing a liquid staking token — stETH, rETH, cbETH, sfrxETH, mETH, LsETH, and a handful of others — into a corresponding EigenLayer strategy contract. The two are economically similar but operationally different.
| Type | Minimum | Withdrawal time | Slashing reach | Counterparty layers |
|---|---|---|---|---|
| Native (EigenPod) | 32 ETH | 7 days EL queue + consensus exit queue | Direct on ETH bond | EigenLayer contract |
| LST (stETH/rETH/etc) | Any | 7 days EL queue + LST exit | Indirect via LST | EigenLayer + LST protocol |
| Liquid restaking token (LRT) | Any | Secondary market or 7+ days | Indirect via LST and LRT | EigenLayer + LST + LRT issuer |
The third row — liquid restaking tokens, or LRTs — is where the action moved in 2024 and 2025, and where it has since unwound fastest. Protocols like ether.fi (weETH), Renzo (ezETH), Kelp DAO (rsETH), and Puffer (pufETH) accept ETH or LSTs from users, deposit into EigenLayer on their behalf, and issue a tradable receipt token that earns both base staking yield and restaking rewards. At peak, the LRT category held more than half of all EigenLayer deposits. That model has since reversed sharply: in August 2026 ether.fi — the largest issuer — removed restaking from weETH, expecting less than 1% of its assets to remain restaked by the end of the third quarter of 2026, and it plans to strip out its remaining EigenPod withdrawal credentials by year-end. The convenience was always real; so was the counterparty stack, three layers deep, in which a user holding ezETH is implicitly trusting Renzo’s smart contracts, Lido or another LST issuer (if Renzo deposits via LST), and EigenLayer’s slashing logic.
What an AVS actually is
An actively validated service is, in plain language, any off-chain or cross-chain protocol that wants to rent cryptoeconomic security rather than build its own validator set. The first AVSs were straightforward: EigenDA, EigenLayer’s own data-availability layer, was the launch AVS and remains one of the largest. Lagrange, AltLayer, eOracle, Witness Chain, Brevis, and Hyperlane have all built AVSs with substantial restaker participation. The use cases span data availability, oracle attestation, fast finality for L2s, cross-chain message verification, and shared sequencing.
Economically, an AVS does three things. It registers a slashing contract with EigenLayer that defines what counts as misbehaviour. It publishes a rewards schedule, paid in its own token, ETH, or stablecoins. And it specifies which operators can opt in and what their quorum requirements are. Operators — distinct from passive restakers — actually run the AVS software, sign messages, and bear the day-to-day risk of slashing. Restakers delegate to operators, much as ETH-holders delegate to validators in proof-of-stake.
Where the slashing surface widens
The most important and least-discussed property of restaking is that it changes the slashing surface from a one-dimensional thing into a multi-dimensional one. Pre-EigenLayer, a validator could be slashed only for two narrowly defined consensus offences: double signing and surround voting. Post-EigenLayer, an operator opted into N AVSs is exposed to N additional slashing conditions, each defined by the AVS team. These conditions can include failing to sign a quorum message, signing a conflicting message, going offline beyond a threshold, or — and this is the part that worries researchers — anything else the AVS team puts in the contract.
The risk is composability. If a single operator opts into ten AVSs and each AVS has a 5% slashing penalty for misbehaviour, the operator’s maximum loss in a worst-case correlated failure approaches the entire bond. EigenLayer has implemented a “slashing magnitude” cap and a “veto committee” governance check that can override unjustified slashing in the first few years, but the long-term equilibrium requires AVS teams to publish well-specified, auditable slashing conditions — and requires operators to actually read them. The EigenLayer documentation is improving, but the AVS-by-AVS slashing detail is still inconsistent across teams.
The yield: where it actually comes from
Marketing dashboards for LRTs typically advertise a headline APR somewhere between 4% and 8%. The first 2.6% is base ETH staking yield, which any LST also earns. The remainder is restaking rewards, and that number deserves scrutiny. Restaking rewards come from three sources:
- AVS protocol payments: Direct fees an AVS pays for security, usually denominated in ETH or stablecoins. These are the “real” yield. Still tiny — most AVSs are bootstrapping and paying primarily in their own tokens.
- AVS token rewards: Newly minted AVS tokens distributed to restakers. Headline-grabbing during launches; their value depends entirely on the AVS token’s market.
- EIGEN rewards: EigenLayer’s own token, distributed through programmable incentive schedules to restakers and operators.
The honest reading of 2025 was that most restaking yield in dollar terms came from category 2 and 3 — token incentives — not from category 1, protocol payments. That is normal for a bootstrapping ecosystem, but it means the long-run sustainability of headline APRs depends on AVSs eventually generating enough revenue to pay restakers in cash terms. By 2026 the thinness of that economic base was undeniable. The sector’s roughly $10.02bn in TVL was throwing off only about $100,000 in weekly fees as of 8 September 2026 — a fraction of what liquid staking earns on a comparable asset base. And the five largest liquid restaking protocols outside ether.fi — Renzo, Kelp, Swell, Puffer Finance, and Bedrock — earned a combined $953,350 in gross profit in Q2 2026, down from $2.18m just three quarters earlier. The bull case was always that data-availability, fast finality, and shared-sequencing services are real businesses that will eventually charge real fees; the bear case was that most AVSs would not find product-market fit and that token incentives were subsidising users into a thin economic base. On the 2026 numbers, the bear case is winning.
The Symbiotic, Karak, and Babylon alternatives
EigenLayer is no longer alone, though it created and long dominated the category. Symbiotic, backed by Paradigm and operationally close to the Lido ecosystem, launched in 2024 with a more permissionless model: any ERC-20 can be used as collateral, and the slashing logic is more modular. Karak takes a similar approach with broader asset support. Babylon is the BTC-native equivalent, letting Bitcoin holders restake without leaving the Bitcoin chain through a clever timelock-based design. Each has carved out a niche. By August 2026, though, EigenLayer’s own restaked assets had fallen to roughly $5.10bn — about half of the sector’s roughly $10bn total — as the category shrank overall rather than simply redistributing between venues.
The original concern about restaking — that it would create a single point of failure where the entire DeFi stack rested on EigenLayer’s slashing logic — has been moderated less by competition than by contraction: as restakers and issuers such as ether.fi have withdrawn, the amount of value depending on any one protocol’s slashing logic has fallen with them. Track the relative shares on our market dashboard.
The systemic concern: cross-AVS correlation
Vitalik Buterin’s “Don’t overload Ethereum’s consensus” essay from May 2023 remains the canonical warning. The concern is not that restaking is bad in itself; it is that if a large fraction of validators are restaked into a small number of AVSs, and one of those AVSs has a contentious failure mode, the resulting slashing could destabilise Ethereum’s own consensus. The mitigation is layered: capped slashing magnitudes, distinct social slashing for “AVS bugs” vs “consensus faults,” and — perhaps most importantly — keeping the restaking opt-in fully voluntary so the Ethereum protocol itself stays neutral.
EigenLayer has been responsive to these concerns. The “fork” recovery mechanism, in which Ethereum’s social consensus could theoretically reject a malicious AVS slashing, is explicit in the protocol design. The veto committee and the gradual unlock of slashing magnitudes give the network time to react. None of these mitigations are foolproof — but with restaking’s share of Ethereum’s validator set having fallen sharply through 2026, the acute version of this systemic risk has receded for now, even as the design questions remain relevant for whatever the next cycle looks like.
Practical advice for an actual user
If you are considering restaking — directly, through an LRT, or by allocating a validator — three questions matter more than the headline APR. First, which AVSs are you opted into, and have you read their slashing conditions? An LRT with auto-allocated AVSs is functionally a black box; the better LRT issuers publish their AVS allocations and update them as the set changes. Second, what fraction of the advertised yield is in protocol payments versus token incentives? The two have very different durability, and the 2026 fee data suggests the durable part is small. Third, what is the withdrawal path under stress? EigenLayer’s seven-day exit queue plus the underlying LST’s queue can mean two weeks or more to get to spot ETH in a bad scenario.
For most retail users the right answer in late 2026 is more cautious than it was a year ago: if you restake at all, keep it a small allocation to a well-audited LRT, treated as a higher-risk position than vanilla staking and sized as you would size any shrinking DeFi niche. Note too that the largest issuer, ether.fi, has been winding restaking out of weETH, so a “restaking” LRT may increasingly be a plain liquid-staking token in disguise — read what your token is actually exposed to. For larger or more sophisticated allocators, native restaking through your own EigenPod gives the cleanest visibility into AVS exposure, but requires running a real validator. Run a few scenarios in our staking calculator, and watch the AVS launch calendar on the events page.
The honest summary
Restaking is a real and useful primitive. It lets new protocols bootstrap security without building their own validator set, and it lets ETH-holders earn additional yield in exchange for additional risk. It is also a new shape of risk — wider slashing surface, deeper counterparty stack, more correlated failure modes — that the staking economy did not have to think about before EigenLayer. But 2026 has been a reality check: TVL roughly halved from its 2025 peak, fees stayed tiny, and the category’s biggest customer walked away. Treat the marketing APR as a starting point, not an answer, and read the AVS list — and whether there even is one — before you delegate.