Restake DeFi Protocols

Restaking is the practice of pledging capital that already secures one network as slashable collateral for additional jobs, and collecting a second stream of rewards for carrying the extra risk. This page is for people who already hold an LST, run validators, or have a liquid restaking token sitting as collateral in a money market, and who want to know what they signed rather than what the landing page promised.

The thing most people get wrong is the direction of the trade. Restaking is not a bonus yield stacked on staking yield; it is a payment for underwriting, and since EigenLayer shipped slashing to mainnet in April 2025, the underwriting is enforceable in code rather than aspirational in a whitepaper. The second thing people get wrong is where the money comes from. Actively validated services pay for security out of their own budgets, and those budgets are overwhelmingly denominated in the service's own token, which means a large fraction of the advertised restaking APR is emissions with a pretty wrapper. The fee revenue, the part paid in ETH or stablecoins by customers who would have paid for the service anyway, is small enough that most dashboards prefer not to break it out.

There are three production designs in restaking and they do not share an architecture. Treating them as one category is how people end up with risk they cannot describe.

EigenLayer takes deposits through two doors. The first is the StrategyManager: you approve an LST, call depositIntoStrategy, and receive non-transferable shares in a per-asset strategy contract. There is no token and no secondary market, which is the entire reason liquid restaking tokens exist. The base layer underneath this is still ordinary liquid staking, and if the mechanics of the receipt token and its exchange rate are unfamiliar, that is the prerequisite rather than this page.

The second door is native restaking. You deploy an EigenPod, a per-user contract, point a validator's withdrawal credentials at the pod address, then submit a Merkle proof of the validator's credentials against a beacon block root. That proof is only possible in the EVM because EIP-4788 exposes the parent beacon block root to contracts, a Dencun change that gets very little credit for how much of this design it enables. Afterwards you run periodic checkpoints, which are further proofs reconciling the pod's accounting with the validator's actual balance. Skip the checkpoints and your rewards simply do not exist as far as the protocol is concerned.

Either way, shares do nothing until you delegate them. delegateTo assigns your entire share balance to a single operator, all or nothing, one operator per staker address. Splitting across operators means splitting across addresses, with the gas and bookkeeping that implies.

What changed in April 2025 is what the operator can do with that delegation. Under ELIP-002, an operator allocates a magnitude of each strategy to a specific operator set belonging to a specific AVS. Only the allocated magnitude is slashable by that AVS, not the whole deposit. When a slash lands, the AVS calls in with a proportion, the corresponding shares burn, and the loss is spread pro rata across every delegator of that operator. You get no vote, no appeal and no notice beyond what you can read on chain. A later upgrade added redistribution, so slashed stake can be routed to a recipient instead of the zero address. That makes restaking usable as collateral for lending and insurance-style services. It also means slashing can now be profitable for whoever configured the slasher, which is a different risk shape entirely from burning.

Symbiotic takes the opposite stance: it ships modules and refuses to have opinions. A vault holds collateral, which can be any ERC-20 the collateral contract accepts, including LSTs, wrapped BTC, sUSDe and project tokens. A delegator module decides how that collateral is allocated across networks and operators. A slasher module decides whether slashing is instant or vetoed, and a resolver, typically a committee or another contract, gets a veto window to reject a slash before it executes. Withdrawals settle at vault epoch boundaries, and the epoch length is a vault parameter, not a protocol constant. The Symbiotic documentation is explicit that the risk configuration is the curator's job, which is honest and also means reading one vault tells you nothing about the next.

Babylon does Bitcoin, and it does it without bridging, wrapping or a smart contract, because Bitcoin has none of those. Stake sits in a timelocked script the staker controls, co-signed by a covenant committee. Finality providers vote using extractable one-time signatures: sign two conflicting votes at the same height and the private key falls out of the arithmetic, at which point anyone can assemble the slashing transaction that spends the staked output. It is a cryptographic trap rather than a contract call, and it is the most elegant piece of engineering in the whole category. Babylon Genesis, the chain that consumes this security, launched in April 2025 with the BABY token.

On Solana, Jito Restaking provides vault and node consensus network programs, with vault receipt tokens as the liquid wrapper. The detail worth noting is that the first meaningful NCN was Jito's own TipRouter, which replaced a centralised tip distribution process with an operator-voted one. That is a real workload with real value at stake, and it is also the protocol eating its own cooking, which is what most AVS usage still looks like across every chain in this sector.

Almost everyone ends up in one of three positions. The differences that matter are who selects the risk, how long the exit takes, and how many contracts sit between you and the collateral.

The LRT column is where retail sits, and it is the column with a real incident attached. The direct column is slower, duller and gives you the only risk control that has ever mattered in this sector: the ability to see exactly which operator sets your stake is exposed to, on chain, without asking anyone.

What follows is the direct path, because it is the one you can verify at every step. The LRT path is two clicks and a signature, and its entire difficulty is the research you do before you click.

1. Work out what you are buying. Pull the operator's current allocations and list the AVSs behind them. For each one, find out whether it has non-affiliated paying customers or whether its only user is the team that built it. This is the single highest-value hour in the process, and it is the one nobody spends. Done when: you can state, in one sentence per AVS, what service it sells and who buys it. Cost: an hour, no gas.

2. Pick your collateral path. LSTs go through the StrategyManager and require no infrastructure. Native restaking requires a validator, which means 32 ETH minimum, or a consolidated 0x02 validator holding up to 2,048 ETH since the Pectra upgrade activated on 7 May 2025. Native pays the full consensus and execution reward with no LST fee; it also puts you on the hook for proof maintenance forever. Done when: you have chosen and can explain why. Cost: nothing yet.

3. Deposit. For the LST path, approve the token to the StrategyManager, then call depositIntoStrategy. You should see your share balance on the strategy contract update in the same block. These shares do not transfer, do not appear as a token in your wallet and cannot be sold. For the native path, deploy an EigenPod, set the validator's withdrawal credentials to the pod address, wait for the credentials to be visible in a beacon block root, then submit the verification proof. Done when: the pod reports restaked shares equal to the validator's effective balance. Cost: one or two transactions for LSTs; several hundred thousand gas per validator for proofs, which is lunch money at 3 gwei and an actual decision at 40.

4. Delegate. Call delegateTo with the operator's address. Check three things first: their total delegated stake, because concentration is the sector's quiet problem; their commission split; and whether they have ever been slashed anywhere, on any network. Done when: the DelegationManager shows your address delegated and your shares counted in the operator's total. Cost: around 100,000 gas.

5. Wait for the allocation delay. New allocations by your operator do not become slashable immediately. The protocol enforces a configuration delay first, and that window is the only advance warning you will ever receive that your slashable surface is about to grow. Check the current value in the docs rather than trusting a blog post, including this one, because the parameter has been changed before. Done when: you have a monitoring method, even if it is a calendar reminder to look at the operator's allocation page fortnightly. Cost: attention.

6. Claim rewards. Distributions land as Merkle roots posted by the RewardsCoordinator on a periodic cadence. You call the claim; nothing arrives automatically. Unclaimed rewards sit there indefinitely, which is fine, but they are not compounding and they are not in your wallet. Done when: the reward token balance appears at your address. Cost: one transaction per claim, which is why claiming tiny balances weekly is a way of paying gas to feel productive.

7. Exit. Undelegate or queue a withdrawal, serve the escrow period, then complete the withdrawal. Fourteen days is the current escrow after the slashing release, up from seven previously. Native restakers then exit the validator, join the beacon chain exit queue, and wait for the sweep. During the unstaking rush of summer 2025 that queue ran for weeks, which surprised a number of people who had budgeted for days. validatorqueue.com shows the current depth. Done when: ETH is in your wallet. Cost: two transactions and a fortnight, minimum.

On custody: the address that holds strategy shares and signs delegation changes is not a trading wallet, and it will be signing infrequent, high-consequence transactions for as long as the position exists. A hardware signer is the appropriate tool, and it is worth confirming your device firmware can display the delegation and withdrawal calldata in a readable form rather than as a hex blob you approve on faith.

The loop, priced by the wrong oracle. On 24 April 2024, ezETH traded far below its redemption value after a points programme ended and holders rushed a shallow pool. Positions that had borrowed against ezETH on lending markets were liquidated at market price, not redemption price, and the losses ran into tens of millions of dollars. How to recognise it: your collateral is a receipt token whose redemption is queued or closed, and the lending market prices it from an AMM rather than from the protocol's exchange rate. What to do: check the oracle type before you borrow. If the position can be liquidated faster than the token can be redeemed, the leverage is not on the yield, it is on the peg.

Points treated as yield. The advertised return for two years was points, multipliers and boosts, with the value of a point set by a token that did not exist. EIGEN transfers opened on 1 October 2024 and the LRT tokens followed; deposits left as soon as the reason to be there stopped accruing. How to recognise it: the APR in the interface has no denominator. What to do: work out the fee-based yield separately from the emissions-based yield and decide whether you would hold the position at fee yield alone.

Risk your operator chose and you inherited. Your delegation follows the operator into every operator set they allocate to. They do not need your consent, and their incentive is to maximise fee-earning allocations. How to recognise it: new allocation events in the AllocationManager for your operator. What to do: monitor allocations, and undelegate during the configuration delay if the new exposure is not one you want. This is precisely the sort of thing a structured scoring approach catches, and our notes on weighting contract risk and concentration apply directly to operator selection.

Double-counted TVL. The same ETH appears in Lido's TVL, in EigenLayer's TVL, in the LRT's TVL and again in the lending market it was deposited into. Aggregators are reasonably clear about this in their methodology; the tweets citing them are not. How to recognise it: someone quoting a restaking figure that implies more ETH than exists in the staking set. What to do: use DefiLlama's restaking category with double-counting filters switched on and treat headline sums as marketing arithmetic.

Native restaking left on autopilot. Pod accounting only updates when someone proves it. Validators that get partially withdrawn, consolidated or exited without a corresponding checkpoint leave pods reporting stale balances, and the reconciliation gets more painful the longer it is postponed. How to recognise it: pod share balance diverging from the validator's real balance. What to do: run checkpoints on a schedule, and re-read the pod documentation after any consensus-layer upgrade that touches withdrawal credentials.

Assuming the absence of slashes means the absence of slashing. Publicly documented punitive slashes on restaking AVSs remain rare, and Ethereum's own base-layer slashing history amounts to a few hundred validators since the Beacon Chain launched, nearly all of it operator error rather than attack, as beaconcha.in's slashing records show. That is a statement about how young the mechanism is, not about how safe it is. The redistribution upgrade in particular created a category of slasher that has something to gain from firing.

Deposits made before April 2025. The pre-slashing architecture had operators opt in to AVSs without any enforceable penalty, which is why the early numbers looked so good: there was no downside to register. Legacy stake is not retroactively slashable, but the delegation relationship persists, and an operator allocating that stake to an operator set today makes it slashable today. If you deposited in 2024 and have not looked since, you are running a different position from the one you opened.

Below roughly 5 ETH. Run the arithmetic before the enthusiasm. If the incremental fee yield from restaking is half a percentage point, three ETH earns you 0.015 ETH a year. A handful of reward claims and a couple of withdrawal transactions at 30 gwei will eat a meaningful share of that, and native restaking with its proof overhead is simply not viable at that size. Small positions belong in an LRT, where the gas is socialised, or nowhere.

Bitcoin. Babylon's design has no bridge and no wrapped asset, so the custody risk profile is genuinely different: your BTC never leaves a script you can spend from after the timelock. The counterweight is the covenant committee and the fact that slashing is a key-extraction trap rather than a governed contract call. Read the Babylon documentation on unbonding times before assuming BTC restaking behaves like ETH restaking with a different ticker.

Solana. Jito's restaking programs support slashing, but most node consensus networks have not implemented punitive conditions in production. The risk today is mostly smart contract risk and vault receipt liquidity rather than slashing, which makes the current yields look better than the eventual steady state. See the Jito developer docs for the vault and NCN interfaces.

United States tax and securities treatment. The SEC's Division of Corporation Finance issued staff statements in 2025 covering protocol staking and then liquid staking receipt tokens, taking the view that those activities did not involve securities offerings under the facts described. Restaking, AVS-paid rewards and the tokens that pay them were not addressed, and staff statements are not rules. Read the originals on the SEC's statements page rather than a summary. On tax, IRS Revenue Ruling 2023-14 treats staking rewards as income when the taxpayer gains dominion and control, which for Merkle-claimed restaking rewards most conservatively means at claim. Points have no guidance at all, which is a gap, not a permission.

Post-airdrop seasons. Every restaking yield curve in this sector has the same shape: high while points accrue, a cliff at token generation, then a long flat stretch at whatever the fee revenue actually supports. If you are being quoted a yield during a points campaign, you are being quoted the top of that curve.

Here is the calculation that decides whether any of this is worth doing, and it takes about twenty minutes. Take the rewards distributed through the RewardsCoordinator over a recent month. Strip out everything paid in the restaking protocol's own token and everything paid in an AVS's own token, leaving only ETH and stablecoins. Annualise what remains. Divide by the restaked value backing those operator sets. That ratio is the fee yield: the part that survives the emissions ending.

It is a small number. It has been a small number since the first AVS went live, for the unremarkable reason that the services being secured are early-stage infrastructure with modest revenue, while the collateral pledged against them is measured in billions. EigenDA remains the flagship example, with genuine external users including L2s that pay for data availability, and its fee take is still a rounding error against the capital standing behind it. That is not a criticism of EigenDA, which works. It is an observation that the supply of restaked capital arrived years before the demand for it, and price does what price does when that happens.

So the honest case for restaking is narrow and still real. If you hold ETH you intend to stake for years, if you are comfortable with a fortnight-plus exit, if you choose an operator whose allocations you monitor, and if you treat the emissions component as a lottery ticket rather than a yield, the incremental return is positive and the incremental risk is bounded and legible. The dishonest case is the one where a double-digit APR is quoted against a leveraged LRT position, and the leverage, the depeg risk, the slashing exposure and the redemption queue are all described as capital efficiency. The same enthusiasm for stacking one yield source on another shows up in liquidity pool returns and their real drivers, and it ends the same way when the emissions stop.