Skip to content
DeFi Reference

Restaking

1. TL;DR

Section titled “1. TL;DR”

Restaking lets a staker’s already-locked capital back a second (or third, or tenth) service beyond base-layer consensus, earning extra yield in exchange for extra slashing exposure. On Ethereum, EigenLayer and Symbiotic let operators opt their stake into “actively validated services” (AVSs) chosen by the operator; on Solana, Jito’s Restaking protocol lets vaults delegate stake to operators serving “node consensus networks” (NCNs), of which Jito’s own Tip Router — the system that routes MEV tips to stakers — is itself the flagship example.

2. Explain it simply

Section titled “2. Explain it simply”

Analogy

Section titled “Analogy”

Restaking is like using the same house as collateral for a second loan after you’ve already mortgaged it once. The bank that holds your first mortgage still has its claim, but now a second lender also has a claim on the same house if you default on their loan too. You get more cash today (extra yield), but if either loan goes bad, you can lose the house faster than if you’d only taken one loan.

ELI5

Section titled “ELI5”

Normally, when you lock up your coins to help run a network, that’s the only job your coins do. Restaking lets you tell your locked coins to take on a second job at the same time — helping run some other new network or service — in exchange for extra rewards. The catch is that now two different sets of rules can punish you if something goes wrong: the original network’s rules, and the new service’s rules. If you stack up jobs for the same coins without understanding each one’s punishment conditions, you can end up losing your original coins because of a completely unrelated service failing.

Step-by-step walkthrough

Section titled “Step-by-step walkthrough”

Scenario: an operator, Op, already runs an Ethereum validator with 1,000 units of delegated staking-token capital and decides to restake through EigenLayer.

  1. Before restaking (state: 1,000 units earning only base staking yield). Op’s capital secures Ethereum consensus and earns issuance plus a share of block rewards, nothing more.
  2. Op opts into an actively validated service, “AVS-A” (state: some of Op’s capital is now also slashable by AVS-A’s rules). Op allocates a portion of a fixed “magnitude” budget to AVS-A; only that allocated slice, not all 1,000 units, becomes slashable by AVS-A.
  3. AVS-A pays rewards (state: Op earns base yield + AVS-A’s fee). As long as Op performs AVS-A’s required duties correctly, the allocated capital earns extra rewards on top of ordinary staking yield.
  4. AVS-A suffers a bug and some operators misbehave (state: only the AVS-A-allocated slice is at risk). Because the allocation is isolated (“unique stake”), a slashing event triggered by AVS-A cannot touch the portion of Op’s capital that was never allocated to it.

Common misconceptions

Section titled “Common misconceptions”
  • Myth: Restaked capital adds to Ethereum’s base-layer security. Reality: A contested debate exists on this point — one view holds that restaked stake should be treated as “attacker-controlled” for base-layer security modeling, since the actor deciding whether to misbehave is the operator, not the protocol (MicahZoltu, in comments on ethresear.ch 18751, 2024-02); another view holds the effect is more nuanced because of principal-agent structure (barnabe, same thread, 2024-02).
  • Myth: Restaking is unique to Ethereum. Reality: Cosmos pioneered the underlying idea as “shared security” years earlier; EigenLayer popularized an Ethereum-collateral-centric version of it (Paradigm, “From Staking to Restaking,” 2024-06).
  • Myth: All restaking protocols work the same way. Reality: EigenLayer is Ethereum-collateral-specific with protocol-defined operator sets; Symbiotic is deliberately generic, letting any network define its own collateral, operator selection, and slashing rules; Jito’s is Solana-native and organizes around “node consensus networks” (NCNs) rather than AVSs.
  • Myth: Restaking slashing is already fully live and enforced everywhere. Reality: As of Jito’s own 2026 documentation snapshot, its restaking protocol explicitly lists slashing as “In Current Development” and states “the slashing program is not currently live” (Jito docs, n.d.) — a caveat worth reading before assuming any restaking yield is fully collateralized by real slashing risk today.

If you only remember one thing

Section titled “If you only remember one thing”

Restaking multiplies yield by multiplying the number of rulebooks that can punish the same underlying capital — read every rulebook, not just the reward number.

3. How it works

Section titled “3. How it works”

EigenLayer: operator sets and unique stake

Section titled “EigenLayer: operator sets and unique stake”

EigenLayer’s redesigned slashing architecture (following its ELIP-002 upgrade) is built around operator sets and unique stake. An operator has a fixed “total magnitude” MM representing their full delegated capital budget, and allocates a portion of it, mim_i, to each operator set ii they opt into. The unique stake exposed to operator set ii — the only capital that set can actually slash — is unique stakei=miM×S\text{unique stake}_i = \frac{m_i}{M}\times S where SS is the operator’s total delegated stake. Because allocations are exclusive per operator set, a slashing event by AVS ii cannot touch the capital allocated to AVS jj. This design deliberately removes the need for a shared veto committee across all AVSs — “there is no need for a common veto committee because slashing is isolated to specific Operator Sets, enabling permissionless AVS onboarding” (per EigenLayer/EigenCloud documentation, as of 2026-08). This replaced an earlier model that relied on broader trust assumptions across the whole restaked pool.

Symbiotic: a generic coordination layer

Section titled “Symbiotic: a generic coordination layer”

Symbiotic takes a deliberately unopinionated approach: its core contracts are immutable, it accepts any ERC-20 as collateral (not just staked ETH), and it delegates the choice of operator-selection mechanics, reward design, and slashing-resolution mechanism entirely to the network being secured (Balaji, Konstantopoulos & White, Paradigm, 2024-06). Paradigm frames the near-term use case as bootstrapping new consensus instances outright — electing operator sets for new L1s or app-chains, decentralized sequencers, distributed auctions, and threshold-decryption networks — while flagging block-production use cases (new MEV auction designs, preconfirmations) as longer-term and riskier, since they interact more directly with Ethereum L1’s own security.

Jito (Re)staking on Solana: NCNs, vaults, and operators

Section titled “Jito (Re)staking on Solana: NCNs, vaults, and operators”

Jito’s restaking protocol coordinates three roles through two on-chain programs (a Restaking Program and a Vault Program): node consensus networks (NCNs) define the service and its rules; vaults hold staked SPL tokens (such as JitoSOL) and mint liquid vault receipt tokens (VRTs) representing deposits; operators run the NCN’s offchain workload and receive delegated stake from vaults (Jito docs, n.d.). All three must mutually opt in — an NCN approves the operators and vaults that may serve it, mirroring the deliberate, permissioned relationships in EigenLayer’s operator-set model, but implemented natively as Solana on-chain programs rather than an Ethereum smart-contract layer. A VRT’s redemption value accrues the same way an exchange-rate liquid staking token does (see /staking/liquid-staking-tokens/): the token count per holder stays fixed while the underlying value it represents grows with rewards. Jito’s own Tip Router — the system that distributes MEV tips collected from the Jito block-building auction — is itself deployed as an NCN, making Jito (Re)staking simultaneously the restaking infrastructure and (via Tip Router) one of its own flagship consumers.

4. Worked numeric example

Section titled “4. Worked numeric example”

EigenLayer-style allocation (matching §2). Op has a total delegated stake S=1,000S = 1{,}000 units and a total magnitude budget normalized to M=1.0M=1.0. Op allocates mA=0.3m_A = 0.3 to AVS-A and mB=0.5m_B = 0.5 to AVS-B, leaving 0.20.2 unallocated. Unique stake exposed to AVS-A: 0.31.0×1,000=300\frac{0.3}{1.0}\times 1{,}000 = 300 units. Unique stake exposed to AVS-B: 0.51.0×1,000=500\frac{0.5}{1.0}\times 1{,}000 = 500 units. If AVS-A later triggers a 10% slashing penalty for a missed duty, the loss is 300×0.10=30300\times0.10=30 units — drawn only from the AVS-A allocation, leaving the 500 units backing AVS-B and the 200 unallocated units completely untouched. This isolation is the entire point of the “unique stake” redesign: a bug in one AVS cannot cascade into losses for an unrelated one.

Jito (Re)staking VRT accrual. A vault holds 10,000 JitoSOL-equivalent stake and delegates it across three operators serving Jito’s Tip Router NCN, minting 10,000 VRTs at an initial exchange rate of 1.00. Over an epoch, the NCN routes its share of accumulated MEV tips back to the vault, lifting the exchange rate to 1.012 (a 1.2% epoch return). A holder who deposited 1,000 VRTs still holds 1,000 VRTs, but they now redeem for 1,000×1.012=1,0121{,}000\times1.012 = 1{,}012 JitoSOL-equivalent units — the reward is entirely in the rising exchange rate, exactly as with any exchange-rate liquid staking token.

5. Where it’s used

Section titled “5. Where it’s used”

Ethereum

Section titled “Ethereum”
  • EigenLayer — the largest restaking protocol; operators opt staked ETH (and, via LSTs, liquid staked ETH) into actively validated services such as EigenDA. docs.eigencloud.xyz
  • Symbiotic — a generalized, permissionless alternative accepting any ERC-20 collateral, backed by Paradigm, explicitly positioned as complementary to (not merely competing with) EigenLayer’s ETH-centric design. paradigm.xyz/writing/symbiotic
  • Liquid restaking tokens (LRTs) built on top of both — see /staking/lrt-risks/ for the risks this second layer of tokenization introduces.

Solana

Section titled “Solana”
  • Jito (Re)staking — the primary restaking primitive on Solana, coordinating NCNs, vaults, and operators through two audited on-chain programs; over 90% of Solana validators already run the Jito-Solana client, giving NCNs immediate access to a large, professional operator base (Jito docs, n.d.). jito.network/docs/restaking
  • Jito Tip Router — an NCN in its own right, distributing MEV tips to JitoSOL vault depositors; see /staking/lrt-risks/ for how this MEV-distribution role compounds with restaking risk.

6. Risks, attacks, and incidents

Section titled “6. Risks, attacks, and incidents”
  • The “overloading consensus” critique. Vitalik Buterin’s essay “Don’t overload Ethereum’s consensus,” cited directly in Paradigm’s Symbiotic writeup, argues that stacking too many external obligations onto the same base-layer stake risks destabilizing Ethereum itself if a restaked service’s failure mode bleeds into validator behavior (Balaji, Konstantopoulos & White, Paradigm, 2024-06).
  • Does restaked stake even secure the base chain? In a widely referenced ethresear.ch exchange, MicahZoltu argues “restaked ETH does not secure Ethereum, generally speaking,” because once algorithm choice or slashing risk is outsourced to a third party, “your ‘stake’ needs to be treated as attacker controlled stake for mechanism design calculations”; barnabe counters that the relationship is more nuanced given principal-agent structures between operators and delegators, and that a large EigenLayer slashing event would itself reduce the equilibrium staking ratio and self-correct over time (comments on ethresear.ch 18751, 2024-02).
  • Slashing not yet live. Jito’s own documentation states plainly that its slashing-enforcement mechanism is “In Current Development” and “not currently live” (Jito docs, n.d.) — meaning restaking yield on Solana today is, by the protocol’s own account, not yet backed by the punishment mechanism that is supposed to justify it economically.
  • Compounded LRT risk. Real capital moved fast: liquid restaking token TVL grew from about $250 million to over $4 billion in the two months before February 2024, “preceding the launch of restaked services” — i.e., speculative capital arrived before the AVSs it was nominally securing had even gone live (Neuder & Chitra, ethresear.ch 18799, 2024-02). This dynamic later contributed to a real incident: Renzo’s ezETH liquid restaking token depegged from roughly $3,049 to as low as $700 within under an hour on 23–24 April 2024, triggering over $56 million in DeFi liquidations, after users reacted to an unpopular token-distribution announcement and could not redeem directly for the backing asset (DLNews, 2024-04, secondary). Full analysis at /staking/lrt-risks/.
  • Centralization of operator sets. Because operators need reputational and technical sophistication to serve multiple AVSs or NCNs profitably, both EigenLayer’s and Jito’s designs risk concentrating restaking activity in the same professional operators who already dominate base-layer staking (barnabe, ethresear.ch 18683, 2024-02, describing the “long intermediated chains of principal-agent relationships” restaking inherits from liquid staking).

7. Open problems

Section titled “7. Open problems”
  • Service-completeness. The “rainbow staking” framework asks whether heavy (slashable) and light (non-slashable) service categories cover all valuable use cases, drawing on Sreeram Kannan’s taxonomy of economic-security, decentralization, and block-production use cases for restaking (barnabe, ethresear.ch 18683, 2024-02).
  • How much restaking is safe? Paradigm treats near-term consensus-bootstrapping use cases (new L1s, sequencers, MPC networks) as clearly beneficial, but explicitly flags block-production use cases as needing more caution because of direct interaction with Ethereum L1 security (Balaji, Konstantopoulos & White, 2024-06).
  • Will honest restaking converge with block building or LST issuance? An open question raised across the ethresear.ch discourse (see also /exchange/batch-auctions/ for a parallel convergence debate in solver markets) about whether large, sophisticated operators inevitably absorb every new yield-bearing role.
  • What is the right slashing-veto design? EigenLayer’s move away from a shared veto committee toward isolated, permissionless operator-set slashing is itself a recent and not fully battle-tested design choice (per EigenLayer/EigenCloud documentation, as of 2026-08); Jito’s slashing program remains undeployed at all.

8. Ethereum vs Solana

Section titled “8. Ethereum vs Solana”
AspectEthereumSolana
Primary protocolsEigenLayer, SymbioticJito (Re)staking
Unit of serviceActively Validated Service (AVS)Node Consensus Network (NCN)
CollateralStaked/liquid-staked ETH (EigenLayer); any ERC-20 (Symbiotic)SPL tokens, primarily JitoSOL
Slashing statusLive, redesigned around isolated “unique stake” per operator setExplicitly “in current development,” not live
Flagship self-referential use caseEigenDA (data availability)Jito Tip Router (MEV tip distribution)
Governance/permissioningProtocol-defined operator sets (EigenLayer) or fully generic (Symbiotic)Mutual opt-in registry on two audited Solana programs

Both ecosystems converge on the same basic shape — operators opt stake into new services for extra yield and extra slashing exposure — but Ethereum’s restaking market is larger, more mature in tooling, and already the subject of a live slashing redesign, while Solana’s is younger, tightly coupled to Jito’s existing validator-client dominance, and has not yet turned on the punishment side of its economic-security promise.

9. Reference doc

Section titled “9. Reference doc”

The reference

Section titled “The reference”

From Staking to Restaking — Arjun Balaji, Georgios Konstantopoulos, and Dave White (Paradigm), 11 June 2024. paradigm.xyz/writing/symbiotic

Summary of the reference

Section titled “Summary of the reference”

The post frames restaking as the second historical step in a lineage: Proof of Work required physical hardware collateral; Proof of Stake replaced it with economic collateral; restaking (which the authors call “shared security”) now lets that same economic collateral back multiple services rather than just base-layer consensus. It credits Cosmos with pioneering shared security conceptually, noting the practical problem Cosmos exposed — every new app-chain had to recruit validators and capital entirely from scratch, slowing ecosystem-wide innovation — and credits EigenLayer with recognizing that an Ethereum-collateral-centric version of the same idea could bootstrap a validator ecosystem faster.

The bulk of the piece introduces Symbiotic, a protocol Paradigm is investing in alongside Cyber.Fund, describing it along three axes: flexible (any ERC-20 collateral, network-defined operator selection and slashing rules), permissionless (immutable core contracts, no central multisig or slashing committee), and reliable (drawing on lessons learned building Lido’s operator ecosystem, per Paradigm’s disclosed prior collaboration with Lido since 2021). The authors are explicit that Symbiotic is a thin coordination layer rather than an opinionated product — the protocol itself makes no judgment about what should be built on it. They nonetheless share their own view of the safest near-term use cases (electing operators for new L1s/app-chains, decentralized sequencing, distributed auctions, MPC and threshold-decryption networks) versus longer-horizon, higher-risk use cases (new MEV auction designs, preconfirmations, based sequencing), explicitly citing Vitalik Buterin’s warning against overloading Ethereum’s consensus as the reason block-production use cases deserve more caution. The post closes by describing Reth Execution Extensions (ExEx) — a way to extract and process data from an Ethereum node and coordinate state across services — as complementary tooling Paradigm built to support Symbiotic-based services.

Key quotes

Section titled “Key quotes”

“The next frontier of this is shared security, which expands the services node operators in PoS can provide while making use of the same underlying economic collateral.” (§ Introduction)

“Symbiotic allows network developers to have complete control over specifying their (re)staking implementation and operator set.” (§ Introducing Symbiotic)

“The core Symbiotic contracts are immutable, which removes external governance risks. Symbiotic will never have a central multisig, slashing committee, or other permissioning mechanisms for shared security services.” (§ Permissionless)

“We think block production use cases may take longer to blossom: they often benefit from adoption by a greater fraction of L1 proposers and may pose more direct security risks to Ethereum L1.” (§ What should be built on Symbiotic?)

How to read the original

Section titled “How to read the original”

Background needed: what a liquid staking token is (see /staking/liquid-staking-tokens/) and the basic Proof-of-Stake collateral idea. Skip the closing paragraph on Reth ExEx unless you specifically care about Ethereum node tooling — it is a footnote to the main argument, not load-bearing. The hardest part to evaluate is the implicit claim that Symbiotic’s generality is a safety feature rather than a risk-transfer: because the protocol itself takes no position on operator selection or slashing design, all of the hard security questions this page’s §6 and §7 raise are pushed down to whichever network builds on top of it — read the “flexible” and “permissionless” sections as a description of where responsibility sits, not as a claim that the responsibility has been discharged.

What changed since

Section titled “What changed since”
  • EigenLayer subsequently redesigned its own slashing mechanism around “unique stake” and permissionless operator sets (ELIP-002), removing an earlier reliance on a shared veto committee across all AVSs (per EigenLayer/EigenCloud documentation, as of 2026-08) — a more Symbiotic-like, network-isolated model than EigenLayer originally launched with.
  • The liquid restaking token market this post’s authors were investing around grew explosively and then produced a real stress event: Renzo’s ezETH depeg of April 2024, discussed in §6 above and in full at /staking/lrt-risks/.
  • Jito (Re)staking launched on Solana as a chain-native analogue to the Ethereum-centric shared-security model this post describes, organizing around NCNs rather than AVSs (Jito docs, n.d.).

Secondary references

Section titled “Secondary references”
  1. Jito docs, “Jito (Re)staking” overview (n.d.) — read for the concrete Solana-native implementation of the same shared-security idea.
  2. barnabe, “Unbundling staking: Towards rainbow staking” (ethresear.ch 18683, 2024-02) — read for a rigorous framework distinguishing which kinds of restaking are and aren’t safe.
  3. Neuder & Chitra, “The risks of LRTs” (ethresear.ch 18799, 2024-02) — read next for what happens once restaked positions are themselves tokenized and made liquid; see /staking/lrt-risks/.
  4. Endgame Staking Economics comment thread, MicahZoltu vs. barnabe exchange (ethresear.ch 18751, 2024-02) — read for the sharpest available disagreement on whether restaking helps or hurts base-layer security.

10. Sources

Section titled “10. Sources”