Restaking in 2026: How It Works and Why the Boom Cooled

Restaking was the defining Ethereum story of 2024 and early 2025. The pitch was simple and seductive: the same ether that already secures Ethereum could be pledged a second time to secure everything built around it, and collect a second yield for the trouble. Billions of dollars chased that idea. By the autumn of 2026, the story has a very different ending. EigenLayer, the protocol that coined the term, holds roughly $5.1 billion in total value locked, down from a peak above $22 billion in August 2025, according to figures from DefiLlama cited across the industry. Its token, EIGEN, trades near $0.24, about 96% below its December 2024 high, per CoinGecko.

And the single biggest customer the sector ever had, the liquid restaking protocol ether.fi, has spent most of the year walking out the door. This guide explains what restaking actually is, how the machinery fits together, where the yield came from, what can go wrong, and why a technology that looked inevitable 18 months ago is now being quietly unwound. The mechanics still matter, because restaking is not finished; but understanding it in 2026 means understanding both the design and the reckoning that followed.

What Restaking Actually Means

Start with plain staking. To help run Ethereum, you lock up ether as collateral and run, or delegate, a validator that proposes and attests to blocks. In return you earn newly issued ETH plus a share of transaction fees and maximal extractable value, usually shortened to MEV. If your validator misbehaves, part of your stake is slashed. That bonded capital is what makes attacking Ethereum expensive, and therefore what makes the network trustworthy in the first place.

Restaking asks a further question. If your 32 ETH is already locked and mostly idle, why not let it back a second promise at the same time? In EigenLayer’s design, a staker opts in to additional slashing conditions on behalf of a third-party service. That service, called an Actively Validated Service, or AVS, gets to borrow Ethereum’s economic security instead of bootstrapping its own token and validator set from scratch. The staker takes on extra risk and, in theory, earns an extra reward. Restaking, in one line, is the reuse of staked capital as rented security for other systems. The appeal is obvious: security is expensive to build and Ethereum had a surplus of it, so why not sell the surplus?

The Staking Stack: From ETH to a Second Yield

Restaking sits at the top of a stack of financial wrappers, each one adding liquidity and leverage on top of the layer below. It helps to see the whole ladder at once, because almost every risk in the system comes from how these rungs connect.

Each rung promised a little more yield and a little more risk than the one below. The genius, and the danger, is that an LRT can be posted as collateral in other DeFi protocols, which can themselves be leveraged or looped again. Capital that started as a single deposit of ether could be counted as backing for a staking reward, an AVS, a lending position, and a points campaign all at once. That is efficient when everything works and combustible when one link breaks.

A worked example makes the leverage concrete. Suppose you deposit one ether. You receive a liquid staking token, then wrap it into a liquid restaking token, then post that LRT as collateral to borrow a stablecoin, and finally use the stablecoin to buy more ether and start again. Four protocols now treat the same underlying coin as backing. Each hop adds a slice of yield and a slice of counterparty, oracle, smart-contract and slashing risk. In a rising market the loop prints returns; in a falling one, a single bad price feed or a frozen token can unwind every layer at once. Restaking did not invent this kind of stacking, but it added a fresh rung near the top of an already tall ladder.

The base yields involved are modest. Roughly a third of all ether is staked, close to 35% of supply, and the native reward sits in the high twos as a percentage, around 2.6% a year before MEV, which lifts the all-in figure closer to 3%, according to StakingRewards. Restaking was supposed to add a few more points on top of that thin base. Whether it ever really did is the question the rest of this article keeps returning to.

Who Built It, and Why Vitalik Pushed Back

EigenLayer was created by Sreeram Kannan, a former University of Washington professor, and brought restaking to Ethereum mainnet in 2023. The thesis was that Ethereum had a surplus of economic security and a shortage of ways to put it to work, so a marketplace that let new services rent that security would unlock a wave of innovation. Investors agreed in size; EigenLayer became one of the most heavily funded protocols of the cycle. Kannan was careful about how he framed the risk. In a September 2023 interview with CoinDesk, he argued that “anything that restaking can do, already liquid staking can do, so I view restaking as a lesser risk than liquid staking.”

Not everyone was convinced. Weeks earlier, Ethereum co-founder Vitalik Buterin published an essay titled Don’t overload Ethereum’s consensus, warning that “any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator.” His concern was systemic. If enough value depended on restaked security, a failure in some third-party service could spill back into Ethereum itself, and the community might feel pressure to bail out a project that had grown too big to fail. The design question at the heart of restaking has always been that tension: how much extra work can you pile onto validators before the base layer you were trying to protect becomes more fragile, not less.

What Is an AVS, and What Does It Actually Pay?

An Actively Validated Service is any system that needs its own trust guarantees but would rather rent them than build them. In practice the catalogue of AVS falls into a handful of buckets:

• Data availability, such as EigenDA, EigenLayer’s own high-throughput layer for rollups to publish transaction data.
• Oracles, such as eOracle, which pay operators to report prices and other off-chain facts honestly.
• Zero-knowledge coprocessors and proving networks, such as Lagrange and Brevis, which generate verifiable computation for other chains.
• Interoperability and messaging hubs, such as Omni Network, that move state between rollups.
• Fast-finality and sequencing services, such as AltLayer’s Mach, and watchtower-style monitors, such as Witness Chain.

These are real systems, and some of them, especially the proving networks, sit in genuinely useful territory. Verifiable off-chain computation is a hard problem, and the economics of who gets paid to produce proofs is a live debate across the industry, as our look at zk-ML proving markets explored. The trouble is not the technology. The trouble is the bill. By late 2026 more than 130 AVS were listed as in development, yet only a few dozen were live, and the money they paid stakers for security was startlingly small. According to data from DefiLlama reported by Blockhead, the entire restaking category secured about $10.02 billion in assets but produced just $99,977 in fees in a single week, while plain liquid staking secured $51.87 billion and earned $27.35 million over the same period. On a per-dollar-of-security basis, ordinary staking earned roughly 53 times more. Restaking had built an enormous supply of rentable security and almost no paying tenants.

That imbalance is the core of the whole story. Restaking is a two-sided market: stakers supply security, AVS demand it, and a price clears in between. The supply side scaled into the billions of dollars very quickly, because points and airdrop expectations pulled deposits in. The demand side never caught up. Most AVS are early-stage projects with little revenue, and many would rather pay operators in their own tokens than in hard fees, which recycles speculation rather than generating real income. A marketplace with abundant supply and thin demand does not produce a rich yield; it produces a glut, and a glut is exactly what the fee numbers describe.

Liquid Restaking Tokens and the Rehypothecation Game

If restaking locked your capital, liquid restaking tokens set out to unlock it again. Deposit ETH or an LST with a restaking manager and you receive a tradable receipt, an LRT, that you can trade, lend, or use as collateral while the underlying position keeps earning. This is where most retail money actually sat, because the LRT hid the complexity of choosing operators and AVS behind a single token.

The LRT era ran on points. Users deposited to farm EigenLayer points, the LRTs issued their own points on top, and speculators layered one anticipated airdrop over another. For a while the implied yields looked spectacular, but very little of that number was cash flow; most of it was the expectation of future tokens. The unit economics were brutal once the music slowed. Blockhead, again citing DefiLlama, reported that the five largest LRTs, Renzo, Kelp, Swell, Puffer and Bedrock, earned a combined gross profit of only $953,350 in the second quarter of 2026, down from $2.18 million three quarters earlier. A billion-dollar sector was splitting under a million dollars in quarterly profit between its biggest players.

Slashing: The Risk You Are Paid to Take

The entire premise of restaking is that stakers accept extra slashing conditions in exchange for extra reward. Slashing on EigenLayer went live on mainnet in April 2025 as an opt-in feature, so for the first two years restaking paid points with no enforceable penalty at all, which flattered the apparent risk. Once slashing is active, the risk is real and specific: if an operator you are delegated to breaks an AVS rule, you can lose principal, not just forgo yield.

The subtler danger is correlation. A single operator can run many validators across many AVS, so one bug or one malicious act can trigger penalties across a whole cluster at once. Stack an LRT on top, use that LRT as collateral in a lending market, and a slashing event does not stay contained; it marks down a token that other protocols are relying on. This is the scenario that worried Buterin, and it stopped being hypothetical in the spring of 2026.

The Kelp and Aave Contagion of April 2026

On 19 April 2026, an attacker exploited Kelp DAO’s cross-chain bridge and minted roughly 116,500 rsETH out of thin air, inflating the token’s supply by around 18% and creating about $292 million in fake restaking tokens, as CoinDesk reported. The attacker then deposited that counterfeit rsETH as collateral on Aave and borrowed real assets against it, leaving the lending protocol with roughly $196 million in bad debt. Aave’s total value locked fell by billions of dollars over a single weekend as users rushed for the exits.

The response became a case study in how DeFi handles a crisis. Aave froze rsETH so it could not be borrowed against further, and a coalition of protocols coordinated a rescue; Aave founder Stani Kulechov personally pledged 5,000 ETH toward covering the shortfall. By the middle of 2026 the hole had been closed, the fraudulent tokens burned, and Kelp’s bridge rebuilt with stronger attestation requirements. The lesson stuck, though: the exploit happened in a bridge, not in EigenLayer’s core contracts, yet the damage flowed straight into a blue-chip lending market because that market had accepted a restaking token as collateral without fully pricing the risk hidden underneath it. Restaking did not just add a yield; it added a web of dependencies that most users never saw.

The Challengers: Symbiotic, Karak, and Bitcoin’s Babylon

EigenLayer was never alone. Symbiotic took a more flexible approach, accepting almost any ERC-20 as collateral and enforcing slashing from day one, with disputes settled by a chosen resolver rather than a single built-in rule. It raised a $29 million Series A led by Pantera in April 2025 and, tellingly, never launched a public token. In July 2026 it shipped Core V2, repositioning the protocol around what it calls collateral markets. Co-founder Misha Putiatin framed the pivot in the launch announcement carried by GlobeNewswire: “Collateral markets are not a new financial idea. What is new is making them composable, enforceable, and capital-efficient onchain.” Karak, built by Andalusia Labs, which raised a $48 million Series A in December 2023 co-led by Lightspeed Venture Partners and Mubadala Capital, rounded out the top three, though it stayed smaller.

The most interesting challenger does not touch Ethereum at all. Babylon lets Bitcoin holders lock their coins using Bitcoin’s own timelock scripting, with no wrapping and no bridge, and use that locked BTC to secure proof-of-stake networks that opt in. Its mainnet and BABY token launched in April 2025, and by its own dashboard it remains the largest Bitcoin staking system, with tens of thousands of BTC committed (that figure updates slowly, so treat it as a running total rather than a live counter). Babylon is a useful counter-example: demand for Bitcoin-backed security is a different market with different customers, and it did not unwind the way Ethereum restaking did, even though the BABY token still trades more than 90% below its 2025 high.

The Great Unwinding of 2026

The clearest verdict on restaking came from the protocol that embodied it. Over 2026, ether.fi, the largest LRT issuer, pulled almost all of its capital out of EigenLayer. According to an investigation by Bitquery, the native ether restaked directly through EigenLayer peaked at just under 3.5 million ETH in mid-2025 and had fallen to roughly 643,000 ETH by 30 September 2026, about one-fifth of the peak. ether.fi alone withdrew 2.25 million ETH over 78 days between late April and early July, including more than 1.1 million ETH pulled in under a single day on 7 and 8 July. Its flagship token, weETH, was stripped of restaking and reverted to a plain liquid staking token, with restaking isolated in a separate, opt-in token built on Symbiotic rather than EigenLayer, holding only a tiny fraction of the protocol’s assets.

Mike Silagadze, ether.fi’s founder, was blunt about why. Speaking to CoinDesk as reported by Blockhead, he said: “There were no meaningful yield opportunities in restaking and there was some perceived risk from stakers, so we decided it made sense to exit.” When the split was first announced, he put it more wistfully to The Defiant: “End of an era. Sad. I still think restaking will come back in one form or another, I think it was just a bit too early.” The company that most defined restaking now earns its keep elsewhere, rebuilding as a crypto neobank with a spending card, portfolio loans, and tokenized stocks; card fees rose from 17% of monthly revenue in January to 46% by July, per reporting from The Block.

The exit left an awkward footnote for a technology sold on decentralization. With ether.fi gone, the remaining native stake concentrated fast: Bitquery found that the node operator Kraken held about 43% of the native stake still on EigenLayer, and P2P.org another 18%, so two firms controlled more than 60% of what was left. The security budget had not just shrunk; it had bunched into a few hands.

Why the Yield Melted: Points, Fees, and a 4% Risk-Free Rate

Three forces drained restaking’s appeal at roughly the same time. The first was the realization that the yield had always been mostly emissions, not fees. Stakers were paid in points that converted into tokens, and when the tokens arrived and fell, the loop ran in reverse. EigenLayer’s own tokenomics tried to fix this with a mechanism introduced in December 2025, documented in ELIP-012, that routes 20% of AVS fees and all EigenCloud fees into buying back EIGEN. The idea is sound, but with the whole category earning around $100,000 a week, the buyback is far too small to offset the token’s scheduled monthly unlocks of tens of millions of EIGEN.

It is worth being precise about how the points machine worked, because it explains both the boom and the bust. A points program is effectively a loan against a token that does not exist yet. Depositors chase points, which lifts total value locked, which makes the eventual token look more valuable, which attracts more deposits. On the way up the loop is self-reinforcing and the paper yields look enormous. On the way down it reverses just as fast: once EIGEN and the LRT tokens had listed and started falling, the implied yield evaporated, deposits had no further airdrop to farm, and the only thing left was the thin trickle of real AVS fees. Strip out the emissions and restaking was paying a fraction of a percent, well below plain staking, for strictly more risk.

The second force was demand that never showed up. Daily active addresses on EigenLayer collapsed from tens of thousands in the spring to a few dozen or a few hundred a day by late summer, according to DefiLlama data, even as deposits stayed large. A two-sided market with plenty of supply and almost no buyers cannot clear at a high price.

The third force was macro. Through 2026 the US Federal Reserve kept short-term rates elevated, and short-dated Treasury bills yielded around 4%. Tokenized versions of those bills became available on-chain, offering a similar composable, dollar-denominated yield with no slashing risk at all. When the risk-free rate is near 4%, almost nobody will accept slashing-exposed exposure for a premium of a point or two, especially against the uncertain, deleveraging market backdrop that defined late 2026, which we covered in our year-end outlook and around the Black Monday anniversary. Put together, the premium melted. EIGEN now trades about 96% below its all-time high of $5.65 set in December 2024, and touched an all-time low near $0.15 in April 2026, per CoinGecko.

EigenCloud: A Bet on Verifiable AI

Faced with tenants that would not pay, EigenLayer changed what it is selling. The protocol rebranded as EigenCloud and now markets itself less as a restaking marketplace and more as a platform for verifiable cloud services, organized around three pillars: EigenDA for data availability, EigenVerify as a dispute-resolution layer, and EigenCompute for confidential, verifiable computation. The pitch leans heavily into artificial intelligence, the idea being that AI agents need a way to prove their work was executed faithfully, and crypto-economic security plus trusted hardware can provide it. The venture firm a16z backed the pivot with a $70 million token purchase reported by The Block.

The strategic read is revealing. By building its own services, EigenCloud is, in effect, becoming its own tenant, manufacturing the demand for security that outside AVS never generated. That may prove to be a smarter business than renting security to third parties who could not afford it. But it is a very different company from the one that launched restaking, and it is a tacit admission that the original two-sided marketplace did not clear.

Skeptics point out that verifiable compute for AI is itself an unproven market, and that EigenCloud’s revenue remains close to zero while it builds. The competitive field is crowded, spanning zero-knowledge machine learning, trusted execution environments and optimistic verification, each with its own trade-offs between cost, speed and how much you have to trust the hardware. EigenCloud’s wager is that real economic security, the thing restaking was always about, becomes genuinely valuable the moment autonomous software starts spending money and signing transactions on its own. That may turn out to be right. But it is a bet on a future demand curve, not evidence of one today, and it asks investors to value the token on a story rather than on current fees.

How the SEC Sees Staking and Restaking

For readers in the United States, the regulatory picture softened in 2025. The SEC’s Division of Corporation Finance issued a statement on certain protocol staking activities on 29 May 2025, taking the view that ordinary protocol staking is an administrative or ministerial activity rather than a securities transaction under the Howey test. It followed up on 5 August 2025 with a companion statement on liquid staking, concluding that liquid staking activities likewise do not involve the offer and sale of securities, provided the provider stays administrative and does not guarantee or set rewards.

Restaking sits in a greyer zone. The staking carve-out is explicitly narrow: it excludes arrangements whose tokens carry intrinsic economic properties such as generating passive yield. A restaking reward token that pays holders a stream of income simply for holding it could plausibly fall outside the safe harbor, and nothing the SEC has said changes the fact that these are staff statements, not rules, and can be revised or withdrawn by a future Commission. Anyone building a yield-bearing restaking product for US users should treat the comfort as provisional rather than permanent.

Is Restaking Dead, and What to Watch Next

Restaking is not dead, but the version that defined 2024 is over. The exit is even being formalized in code: a July 2026 draft proposal on the EigenLayer governance forum, ELIP-018, nicknamed RETIRE, would let a staker permanently and irreversibly disable restaking on their position, acknowledging that leaving cleanly had become a real pain point. ether.fi, for its part, plans to remove its last EigenPod withdrawal credentials by year-end, severing the final structural tie.

What survives tells you where the value really was. The platforms doing comparatively well are the ones that did not lean on a speculative token: Symbiotic, with no public token and a pivot to collateral markets, and Babylon, serving a separate Bitcoin market. EigenCloud is trying to become a real business by selling verifiable compute rather than renting idle security. And the honest takeaway for any investor is the one Silagadze landed on, that restaking may well return in a different, more sober form once there are services that genuinely need rented security and can afford to pay for it. The next time around, the test will be simple: is the yield coming from fees that real customers pay, or from tokens the protocol prints? In 2026, for most of restaking, the answer was tokens, and the market finally noticed.

For anyone still weighing a restaking position, a few practical questions cut through the noise. Where does the advertised yield actually come from, fees or emissions, and what is the figure once the points are stripped out? Which operators are you delegated to, and how concentrated are they, given that two firms now hold most of EigenLayer’s native stake? What slashing conditions have you opted into, and could a single operator fault cascade across several services at once? If an LRT is involved, who accepts it as collateral elsewhere, and what happens to those lenders if it depegs? None of these questions were easy to answer at the peak of the mania, which is precisely why so much capital learned the answers the hard way.

Frequently Asked Questions

What is restaking in simple terms?

Restaking is reusing ether that is already staked on Ethereum to also help secure other services, such as data-availability layers, oracles and bridges. The staker accepts extra slashing conditions in exchange for an additional yield on top of normal staking rewards.

How is restaking different from staking and liquid staking?

Plain staking secures Ethereum itself. Liquid staking gives you a tradable receipt token, such as stETH, for your staked ETH. Restaking adds a further layer: it pledges that staked capital to secure third-party services as well, which raises both the potential reward and the risk.

What are the main risks of restaking?

The two big ones are slashing, where misbehaviour by an operator can cost you part of your stake, and rehypothecation risk, where the same capital backs many positions at once. The April 2026 Kelp and Aave incident showed how a single exploit can spread across protocols that accept restaking tokens as collateral.

What is a liquid restaking token (LRT)?

A liquid restaking token is a tradable receipt for a restaked deposit, such as ether.fi’s weETH or Renzo’s ezETH. It lets holders keep their capital liquid and reuse it in other DeFi applications while it earns restaking rewards, which also stacks additional risk on top.

Why did EigenLayer’s token and TVL fall so sharply in 2026?

Demand for rented security never matched supply. Paying customers were scarce, so fees stayed tiny while most of the yield came from token emissions. As those emissions faded and short-term interest rates rose, capital left: EigenLayer’s TVL fell from more than $22 billion to around $5 billion, and the EIGEN token dropped roughly 96% from its high.

By Yuki Tanaka, DeFi and on-chain correspondent at HOGE Wire.