What Is Restaking?

Restaking lets ETH that is already securing Ethereum be pledged, at the same time, as collateral for other protocols — data availability layers, oracle networks, bridges — in exchange for extra rewards and extra ways to lose it. Pioneered by EigenLayer, it became one of the fastest-growing sectors in Ethereum's history and one of its most debated, because it deliberately stacks new risks on top of the safest yield in crypto.

Staking first: the security restaking reuses

To understand restaking you need the baseline it builds on. Ethereum is a proof-of-stake network: validators lock up ETH as collateral, propose and attest to blocks, and earn protocol rewards for doing it honestly. Running your own validator requires a deposit of 32 ETH, per ethereum.org. If a validator misbehaves — signing conflicting blocks, for example — the protocol destroys a portion of its stake. That penalty is called slashing, and it is the economic teeth behind Ethereum's security: attacking the chain means burning your own capital. (Since the Pectra upgrade activated in May 2025, a single validator can also consolidate up to 2,048 ETH via EIP-7251, but 32 ETH remains the minimum.)

Because most holders don't have 32 ETH or the appetite to run infrastructure, liquid staking emerged. You deposit any amount of ETH into a protocol such as Lido and receive stETH, a token representing your staked position; Rocket Pool's rETH works similarly. The two differ in accounting — stETH rebases, meaning your token balance grows daily as rewards accrue, while rETH keeps a fixed balance but appreciates against ETH via an exchange rate — but both let staked capital stay liquid and usable across DeFi.

By 2023, this created an enormous pool of yield-bearing, liquid, staked ETH. Restaking is the answer to a follow-up question: if that capital is already posted as collateral for Ethereum, could the same collateral back other things too?

The staking backdrop in 2026

The pool restaking draws from has kept growing. Roughly 41 million ETH — about a third of the entire supply — is staked as of 2026, with Lido alone accounting for around 8.7 million ETH, per Datawallet's staking statistics. Demand accelerated through 2026: the validator entry queue swelled to over 3.5 million ETH by late May, driven by yield-distributing spot ETH ETFs, corporate treasury staking, and post-Pectra validator consolidation. The base staking rate this all earns is modest — around 2.6% annually, per ethereum.org — which is precisely why a product promising "the same capital, plus extra yield" found such a ready audience.

What restaking adds: opt-in slashing for extra duties

Before restaking, any new decentralized service — an oracle network, a data availability layer, a bridge — faced a cold-start problem. To be credibly secure it needed its own validator set, which meant issuing its own token, convincing people to stake it, and hoping the token's value grew large enough that attacking the network became expensive. Bootstrapping that trust is slow, dilutive, and frequently fails.

EigenLayer's insight was that Ethereum's staked capital could be rented. A staker opts in to EigenLayer's smart contracts and accepts additional slashing conditions on their stake: alongside Ethereum's rules, they now also promise to perform duties for external services, and can be penalized under those services' rules if they fail. In return, the services pay for the security they receive. The same ETH ends up doing double (or tenfold) duty — which is precisely why both the rewards and the risks compound.

There are two ways in. Native restaking is for validator operators: they point their validator's withdrawal credentials at an EigenLayer contract (an EigenPod), putting their actual staked ETH under the protocol's slashing authority. LST restaking is for everyone else: deposit a liquid staking token like stETH into EigenLayer's contracts and it becomes restaked collateral. You can watch how much capital sits in this sector on our restaking dashboard, and inspect EigenLayer's own protocol page for its TVL trend.

Inside EigenLayer: restakers, operators, and AVSs

EigenLayer is a marketplace with three roles. Understanding who does what makes the risk conversation much clearer.

Restakers (delegators)

Restakers supply the capital: natively restaked ETH or deposited LSTs. Most restakers do not run any infrastructure. Instead they delegate their stake to an operator, sharing in that operator's rewards — and, crucially, in any slashing it incurs. Delegation is a trust decision, not a passive index bet.

Operators

Operators are the professional node runners. They register with EigenLayer, accept delegations, and choose which services to validate for. Each service defines its own tasks (store this data, attest to this price, verify this proof) and its own penalty conditions. Since the mainnet slashing release, operators can also carve out "unique stake" — portions of their allocated stake that only a single AVS can slash — which limits how far one service's failure can reach, per EigenLayer's slashing documentation.

AVSs (Actively Validated Services)

AVSs are the customers: external protocols that rent pooled security instead of bootstrapping it. An AVS gets, from day one, a large operator set backed by billions in restaked collateral — security that would otherwise take years to accumulate. When EigenLayer's slashing went live on mainnet on April 17, 2025, the protocol secured over $7 billion in restaked assets across 39 AVSs, per CoinDesk. That launch mattered: before it, restaking's penalties were largely promissory; after it, misbehaving operators could actually lose stake.

What AVSs actually do

"Actively Validated Service" is deliberately generic. The categories that emerged first:

• Data availability (DA). EigenDA, the first AVS on EigenLayer, is a data availability layer: rollups post compressed transaction data to it so the data is provably retrievable, at lower cost than posting everything to Ethereum itself. Operators attest that they hold the data and can be slashed for failing to serve it. If you follow Layer 2s, DA is the piece of the rollup stack these networks most often outsource.
• Oracle networks. Price feeds and off-chain data attestation, where restaked collateral backs the honesty of reported values.
• Decentralized sequencers. Services aiming to decentralize the ordering of L2 transactions, currently a centralized single point in most rollups.
• Bridges and interoperability layers. Cross-chain message verification backed by slashable stake rather than a trusted multisig.
• Coprocessors and keeper networks. Off-chain computation (proofs, automation, AI inference verification) whose correctness is bonded by restaked capital.

The common thread: each service defines objectively checkable duties, and failure to perform them is punishable by taking the operator's (and its delegators') stake.

Slashing grows teeth: redistribution and the EigenCloud pivot

Two 2025 developments reshaped what restaking actually is. The first came three months after slashing went live: redistributable slashing (ELIP-006, July 2025) lets an AVS route slashed funds to a designated recipient instead of only burning them, per the EigenCloud announcement. That sounds like a technicality; it is not. Burn-only slashing can punish but never compensate — redistribution turns an operator's stake into an insurance pool that can make a service's users whole when commitments are broken, enabling AVS designs like lending guarantees and insurance that need a payout path, not just a punishment. The flip side: a compromised or buggy AVS slashing condition is no longer merely destructive but potentially extractive, which is why redistribution arrived alongside stricter opt-in controls for operators.

The second was strategic. In June 2025, Eigen Labs rebranded the platform as EigenCloud, repositioning restaking as the trust layer of a broader "verifiable cloud" — EigenDA for data availability, EigenCompute for off-chain computation, EigenAI for AI inference, and EigenVerify for dispute resolution — with a16z buying an additional $70 million of EIGEN to back the pivot, per Coin Bureau's 2026 review. The subtext matters for anyone reading the sector: pure security-renting to crypto-native AVSs had not yet produced yields that justify the risk stack, so the leading protocol widened its market toward verifiable computation generally. Restaking in 2026 is less "extra yield on ETH" and more "the bond underneath a cloud platform" — a quieter, more infrastructural business than the 2024 narrative promised.

AVS economics: what does rented security actually cost?

Restaking's long-term viability comes down to an unglamorous question: can AVSs afford the security they rent? The arithmetic is unforgiving. Restakers already earn Ethereum's base staking rate — around 2.6% — for free, so an AVS must pay a spread above that large enough to compensate for its slashing conditions, its contract risk, and the operator's infrastructure costs. Multiply a meaningful spread by billions in restaked collateral and the implied security bill quickly exceeds what a young protocol earns in total revenue. During the points era this gap was papered over with expectations; once slashing went live and points stopped, the gap became visible — one reason sector TVL deflated so hard through 2025–2026.

The counter-argument is that security should be sized to the value it protects, not to the collateral available. A data availability layer securing modest rollup traffic needs enough slashable stake to make cheating irrational — not $15 billion of it. On that view, the sector's contraction toward a smaller base of genuinely paid stake is the market finding the right price for security rather than a failure. It also explains the EigenCloud pivot: verifiable compute and AI inference are attempts to find customers whose willingness to pay for cryptoeconomic guarantees is measured in real budgets rather than token emissions. When you evaluate any AVS or LRT yield today, the two questions from our DeFi guide apply verbatim: who pays this, and would they keep paying if the incentives stopped?

Liquid restaking tokens: eETH, ezETH, pufETH

Restaking directly through EigenLayer means choosing operators, tracking AVSs, and managing withdrawals yourself. Liquid restaking protocols abstract all of that away — and mint a token against the position, exactly as Lido did for plain staking.

• ether.fi (eETH). Deposit ETH; the protocol stakes it, restakes it via EigenLayer, and issues eETH, a rebasing token that accrues staking plus restaking rewards. A wrapped, non-rebasing version (weETH) is widely used as DeFi collateral. See ether.fi's documentation, and track deposits on the ether.fi protocol page.
• Renzo (ezETH). A reward-bearing (non-rebasing) LRT: ezETH's redemption value against ETH rises as rewards accrue. Renzo manages operator delegation and AVS strategy on the depositor's behalf.
• Puffer (pufETH). A native restaking protocol that lowers the barrier for independent node operators, using "validator tickets" — prepaid, per-day licenses to run a validator — and anti-slashing signing tooling, per Puffer's docs. Depositors hold pufETH against the pooled position.

The result is a second derivative of staked ETH. It is worth being precise about what each layer adds:

Beyond EigenLayer: Symbiotic and the shared-security market

EigenLayer created the category, but restaking is now a market with competing designs. The most notable is Symbiotic, which positions itself as a permissionless, modular shared-security layer: networks that rent security through it can configure their own collateral assets, operator selection, slashing logic, and dispute "resolvers," rather than accepting one protocol's fixed design. Unlike EigenLayer's ETH-centric model, Symbiotic is multi-asset by design — collateral can be a wide range of tokens held in configurable vaults, per Symbiotic's introduction. Karak pushed the multi-asset idea furthest, accepting LP tokens, stablecoins, and wrapped bitcoin as restakable collateral, per Protofire's 2026 protocol comparison. The broader lesson is that "shared security" has become a product category: multiple protocols now compete to match capital that wants yield with services that want security.

The market verdict so far is lopsided. EigenLayer retains roughly 90% of Ethereum restaking despite its drawdown; Symbiotic holds around $344 million and Karak has fluctuated in the low hundreds of millions, per DefiLlama and Protofire. Shared security exhibits strong network effects — operators, AVSs, and capital all want to be where the others already are — which is why the challengers compete on flexibility rather than scale.

The idea has also escaped Ethereum entirely. Babylon applies the same logic to Bitcoin: BTC holders lock coins using Bitcoin-native scripts — no bridge, no wrapping — and that stake secures external proof-of-stake networks, with misbehavior punished by burning the locked bitcoin, per Protofire's comparison. Bitcoin's enormous, largely idle capital base makes it a natural security reservoir, and Babylon's growth alongside EigenLayer and Symbiotic confirms the broader thesis: "sell security from an established asset to young networks" is now a design pattern, not a single protocol's product.

When comparing these ecosystems, be careful with headline TVL: restaked capital is often counted at several layers at once (the LST, the restaking protocol, the LRT, and the DeFi pools the LRT sits in). Our guide on reading DeFi TVL covers this double-counting problem in detail.

The risk stack

Restaking's pitch is "extra yield on the same capital." The honest version is "extra risk on the same capital, hopefully compensated." The risks are layered, and they interact.

1. AVS slashing risk

Every AVS you (or your chosen LRT) opt into is a new set of conditions under which stake can be destroyed. A bug in an AVS's slashing logic, an operator misconfiguration, or genuinely malicious behavior can all trigger penalties. Diversification across AVSs multiplies reward sources — and penalty surfaces.

2. Smart contract layering

A pufETH or eETH position routes through at least three protocol layers: the staking layer, EigenLayer's contracts, and the LRT protocol's contracts — before it ever touches a DeFi lending market. Each layer is code that can contain bugs, and a failure at any layer impairs everything above it. Risk here is multiplicative, not additive.

3. Operator risk

Delegators inherit their operator's performance. An operator that gets slashed loses its delegators' stake too; an operator that opts into aggressive, poorly audited AVSs drags its delegators along. LRT holders sit one step further back: they inherit the LRT team's choice of operators and AVSs, usually with no direct say.

4. LRT depeg and liquidity risk

LRTs trade on secondary markets, and their price can detach from the value of the underlying stake — especially because exits are slow (staking withdrawal queues plus restaking unbonding periods), which weakens the arbitrage that normally holds pegs. The canonical case study is Renzo's ezETH on April 24, 2024: after a controversial token announcement, holders rushed to sell into thin DEX pools and ezETH briefly traded around $688 on Uniswap — a discount of tens of percent to its backing — triggering cascading liquidations of leveraged restaking positions on lending platforms before recovering, per Cointelegraph. Depth matters enormously here; our token liquidity guide explains why a token with shallow pools can gap violently on ordinary sell flow.

5. Systemic leverage

The ezETH episode also illustrated the systemic concern: LRTs are widely used as collateral to borrow ETH, which is then re-deposited to mint more LRT — a loop that builds leverage on top of the restaking stack. Critics, including voices in the Ethereum research community, have warned that if restaking grows large enough, a correlated slashing event or major LRT failure could transmit stress back into the core staking layer that secures Ethereum itself. That is the sense in which restaking is sometimes described as rehypothecation of Ethereum's security budget.

When the stack fails: the KelpDAO exploit (April 2026)

Every risk above stopped being hypothetical on April 18, 2026, when KelpDAO — a liquid restaking protocol issuing rsETH against EigenLayer positions — suffered the largest DeFi exploit of the year. Attackers minted roughly 116,500 rsETH, worth about $292 million and around 18% of the token's supply, per CoinDesk. Notably, nothing in the restaking machinery itself was broken: the failure was in KelpDAO's LayerZero-powered cross-chain bridge, configured with a single data-verification network as its only verifier. Attackers — attributed by LayerZero, Mandiant, and CrowdStrike to North Korea's TraderTraitor group — compromised the RPC nodes feeding that verifier and injected fabricated cross-chain messages, minting rsETH that no deposit backed, per Halborn's post-mortem.

The contagion dwarfed the theft. The attackers posted stolen rsETH as collateral and borrowed real assets against it, so lending markets — not KelpDAO depositors alone — were left holding the bag. Aave, SparkLend, and Fluid froze their rsETH markets to stop bad debt accumulating, and unnerved depositors pulled roughly $8.45 billion from Aave and more than $13 billion from DeFi overall within 48 hours. Read against the risk stack above, the episode checks nearly every box: contract layering (the weakest link was a bridge most holders never thought about), operator-style trust in the LRT team's infrastructure choices, depeg dynamics as frozen rsETH became unpriceable, and systemic leverage transmitting one protocol's failure across the ecosystem. It is the single best case study for why LRT due diligence must extend to every layer a token routes through.

The points era: how restaking grew so fast

No account of restaking is complete without the speculative engine that inflated it. Through late 2023 and 2024, EigenLayer and every major LRT protocol ran points programs: non-transferable balances tracking deposits over time, universally understood as claims on future token airdrops. Depositors farmed EigenLayer points and LRT points simultaneously; secondary markets even priced and traded the points themselves.

The payoff arrived in 2024. EigenLayer's foundation announced the EIGEN token with 15% of supply reserved for a multi-season "stakedrop"; season one allocated roughly 5% of supply based on a March 15, 2024 snapshot, with claims opening May 10, 2024. The launch was controversial — tokens were initially non-transferable and many jurisdictions were geoblocked, per CoinDesk. Renzo's REZ airdrop the same spring was poorly received and became the proximate trigger for the ezETH depeg described above.

The lesson for reading dashboards today: capital that arrives to farm an airdrop leaves when the airdrop ships. Deposit growth during a points meta is not the same signal as deposit growth driven by AVSs actually paying for security. When you evaluate restaking's size on our restaking map, ask which regime the flows belong to — our momentum methodology explains how we try to separate persistent trends from incentive-driven spikes.

Restaking in 2026: after the points era

That reckoning has now happened, and the numbers tell the story. EigenLayer's TVL sits around $5.1 billion as of August 2026, per its DefiLlama page — down from roughly $20 billion at the 2024 peak and above $15 billion as recently as early 2026, per Coin Bureau. The drawdown compresses three forces into one line: airdrop capital exiting after the stakedrops shipped, live slashing forcing an honest repricing of risk that points-era deposits had ignored, and the 2026 market correction marking down the ETH the deposits are denominated in. The LRT layer sorted into clear winners and losers — ether.fi's staking vault holds about $3.3 billion while Renzo, whose ezETH once anchored the number-two spot, has fallen below $100 million, per DefiLlama.

A shrinking headline number is not the same as a failing sector, though. What remains is closer to restaking's original thesis: capital that is knowingly pricing slashing risk, AVSs that pay in revenue rather than expectations, and infrastructure — above all EigenDA — with genuine rollup customers. The honest summary of 2026 is that restaking got smaller and more real at the same time.

The whole arc in one timeline

Sources for each row appear in the sections above and the reading list below.

How to evaluate a restaking position in 2026

If you are considering an LRT or a direct restaking position, the questions worth asking have sharpened since the points era:

1. Which AVSs, exactly? "Restaked" is not one risk. Ask which services the position is opted into, what their slashing conditions are, and whether stake is isolated per-AVS as unique stake or shared across many.
2. Where does the yield come from now? Post-points, demand real numbers: what are AVSs actually paying, in what token, and how much of the advertised APY is still incentive emissions? Our DeFi guide's "who pays this yield" test applies with full force.
3. What is the exit path? Trace the full withdrawal route — LRT redemption queue, restaking unbonding, staking exit — and the secondary-market depth you would rely on if you needed out faster. Thin pools were the ezETH lesson.
4. What infrastructure does the token route through? The KelpDAO exploit came through a bridge configuration, not the restaking contracts. Multi-chain LRTs inherit the security of every bridge and verifier network they touch.
5. How does the protocol handle a slashing or exploit event? Look for insurance funds, redistribution arrangements, and the governance process that decides who eats a loss — before you need to know.

How to follow restaking on ETHBubbles

The restaking dashboard visualizes the sector's protocols as bubbles sized by TVL, so you can see at a glance how EigenLayer, ether.fi, Renzo, Puffer, and peers compare — and which are gaining or losing capital. From there, drill into individual protocol pages (for example EigenLayer or ether.fi) for trend lines, or zoom out to the live bubble map to see restaking in the context of the whole ecosystem. Cross-reference with how to read TVL before drawing conclusions from the raw numbers.

Related guides

• Ethereum DeFi Explained — the lending, DEX, and staking landscape that LRTs plug into.
• How to Read DeFi TVL — why restaking TVL is especially prone to double-counting.
• Token Liquidity & Slippage Explained — the mechanics behind LRT depegs like ezETH's.
• Ethereum L2s Explained — rollups, data availability, and sequencers: the main AVS customers.

Frequently Asked Questions

Sources & further reading

• ethereum.org — Staking
• CoinDesk — EigenLayer Adds Key 'Slashing' Feature (April 17, 2025)
• EigenLayer blog — Intro to Slashing on EigenLayer
• Symbiotic — Introducing Symbiotic: Permissionless Restaking
• ether.fi docs — Restaking
• Puffer docs — Validator Tickets
• Cointelegraph — Renzo's ezETH depegs to $688 (April 2024)
• CoinDesk — EigenLayer Opens Claims for EIGEN Airdrop (May 2024)
• EigenCloud blog — Redistribution Coming to EigenLayer (ELIP-006)
• Coin Bureau — EigenLayer Review 2026: Restaking, EigenCloud, EIGEN (TVL trajectory, EigenCloud pivot)
• DefiLlama — EigenLayer (TVL figures, August 2026)
• DefiLlama — Symbiotic and DefiLlama — Renzo (competitor and LRT TVLs)
• Protofire — EigenLayer vs Symbiotic vs Babylon (2026) (protocol comparison, Karak collateral model)
• CoinDesk — Kelp DAO Exploited for $292 Million (April 2026)
• Halborn — The Kelp DAO Hack Explained (April 2026) (bridge exploit mechanics)
• Datawallet — Ethereum Staking Statistics (2026) (staked ETH totals, entry queue, Lido share)

Disclaimer: Restaking protocols involve experimental cryptography and complex smart contracts. They are subject to extreme slashing risks. This is not financial advice.