The Restaking Stress Test: EigenLayer at £12 Billion and the Limits of Stacked Trust
Introduction
By Q1 2026, EigenLayer secures around £12 billion in restaked assets (about 4.36 million ETH) and commands roughly 93.9% of the restaking market. It reached an all-time high of £16 billion before retreating as April 2025's slashing launch forced the market to price restaking risk for the first time. Symbiotic holds around £720 million (5.5% share); Karak around £82 million (0.6%). The total Ethereum restaking ecosystem sits at approximately £13 billion. Liquid restaking has matured into a full DeFi primitive, with EtherFi leading at around £4.5 billion TVL, followed by Kelp DAO, Renzo and Puffer; collectively the top liquid restaking protocols hold over two-thirds of EigenLayer deposits.
This is real infrastructure. It is also, structurally, the most concentrated trust experiment in DeFi's history: one validator set, multiple slashing conditions, billions of dollars of dependent protocols. Whether it scales without blow-up is one of the most important open questions in the space.
What restaking actually does (and where the risk concentrates)
The mechanic is simple. ETH staked on Ethereum's consensus layer is redeposited into EigenLayer contracts, which delegate the staked capital to operators that run additional services known as Actively Validated Services (AVS). AVS pay operators in fees; operators share fees with restakers. Restakers receive an additional yield (around 3.87% base on top of standard ETH staking yield, depending on AVS demand). In exchange, restakers accept additional slashing conditions specific to each AVS.
The AVS ecosystem has matured into specialised verticals, what EigenLayer now calls Vertical AVS. EigenDA (data availability), EigenAI (verifiable AI inference, mainnet late 2025), EigenCompute (off-chain execution verification, alpha Q1 2026), AltLayer MACH (fast finality for OP and Arbitrum rollups), Brevis (ZK coprocessor), Lagrange (state committees for cross-chain messaging), Omni (cross-rollup messaging) and Witness Chain (physical-location attestation). The "verifiable cloud" framing is now official.
The risk concentration follows directly. A single operator running multiple AVS faces correlated slashing exposure. A liquid restaking protocol pooling delegations across operators concentrates that exposure further. An LRT integrated as collateral across DeFi concentrates it once more. The result is a system where capital efficiency and trust composition are inseparable, and where a misbehaving AVS, a buggy slashing condition or a compromised operator can ripple through three layers of dependent protocols.
What the data has already taught us
April 2025's slashing launch was a useful stress test. EigenLayer TVL halved from peak to around £5.6 billion before recovering. This was not a hack. It was the market correctly repricing the cost of additional slashing exposure once the abstract became concrete. The recovery to £12 billion plus through 2025 to 2026 reflects a clearer risk-yield equilibrium.
Two specific events bear close reading:
- Kelp DAO bridge exploit (April 2026). The £230 million loss on the LayerZero bridge was not a slashing event. It was a smart-contract compromise of a wrapped or bridged restaked asset. But it triggered a cascading £4.3 billion ETH withdrawal from Aave as participants priced in correlated exposure across the restaking stack. This is exactly the cascade scenario that institutional restaking sceptics have warned about.
- EIGEN governance and tokenomics. The EigenLayer Foundation's ELIP-12 proposal, slated for Q1 2026, introduces an Incentives Committee directing emissions toward fee-generating AVS, with a fee model channelling 20% of subsidised AVS rewards and 100% of EigenCloud fees into EIGEN buybacks. EIGEN itself has fallen around 91% from its 2025 peak even as TVL hit a £23 billion record, the disconnect between protocol usage and token value capture has been the cycle's most-discussed governance question.
How a 2026 institutional treasurer should think about restaking
Three rules:
One: distinguish native restaking from LRTs from LRT-on-LRT. Native EigenLayer restaking carries one set of risks: additional slashing conditions, 7-day withdrawal queue. LRTs add a second set: smart-contract risk in the LRT protocol, oracle risk for the LRT token, secondary-market liquidity risk. Yield aggregators that loop LRTs into further DeFi positions add a third. The yield stack and the risk stack grow in lockstep; institutional allocators should not accept stacked yield without explicit acknowledgement of stacked risk.
Two: the question is not EigenLayer vs Symbiotic vs Karak. It is operator and AVS selection. A diversified delegation across multiple operators, with explicit caps on AVS exposure per operator, is the institutional baseline. A single-operator delegation across high-yield AVS is uninsurable in any traditional sense.
Three: liquidity is asymmetric. EigenLayer has a 7-day exit queue. LRT secondary-market liquidity is good in normal markets and very thin in stressed markets. The April 2026 Kelp event saw secondary liquidity gap immediately. Position sizing should assume stressed-market liquidity, not normal-market liquidity.
The contrarian read
Restaking has become "Ethereum's AWS" in the official narrative, a verifiable cloud underneath the modular stack. We think the AWS analogy is misleading. AWS is built on independent failure domains explicitly designed for blast-radius containment. EigenLayer is the opposite: a single shared validator set absorbing correlated risk from heterogeneous services. The right analogy is not AWS; it is the prime brokerage layer of pre-2008 finance, capital-efficient, deeply connected, and structurally fragile under correlated stress.
That does not mean restaking is wrong. It means restaking should be regulated like prime brokerage, with explicit capital requirements, transparent slashing-condition risk reporting, insurance markets for correlated slashing, and operational separation of concerns. Some of this is emerging organically (slashing insurance products are early-stage; institutional-grade operator due diligence is improving). Some of it will require explicit rule-making. We expect the next slashing-driven loss event, and there will be one, to accelerate that rule-making materially.
Conclusion
Restaking is the cycle's most important new primitive and its most fragile. It works. It scales. It will, periodically, lose people money in correlated ways that the marketing materials do not foreground. The institutions that engage with it from a posture of explicit risk decomposition (native vs LRT vs LRT-on-LRT, operator selection, AVS exposure, exit liquidity) will benefit from the yield. Those that buy "stacked yield" without buying "stacked risk" will be the case studies in the next cycle's post-mortems.
Modelling a restaking allocation? Our team builds operator-level and AVS-level risk decompositions. Get in touch.