Restaking in 2026: A Complete Guide to the Yield and the Risk

Restaking is a simple promise wrapped around a complicated machine. The promise is that the same staked ETH which already secures Ethereum can secure other software at the same time, and earn a second stream of rewards for the trouble. From 2023 through the end of 2024 that promise made restaking the fastest-growing category in decentralized finance, pulling tens of billions of dollars into a system most depositors could not fully describe.

In 2026 the direction reversed in public. ether.fi, the protocol that came to stand for restaking, spent the summer pulling almost all of its money back out. This guide explains what restaking is, how the pieces fit together, where the yield actually came from, what can go wrong, and how to read a market that piled in for two years and then spent much of 2026 backing away. Prices and figures below are current as of late August 2026.

Restaking in one sentence

Here is the whole idea in one line: restaking is re-pledging staked ETH, or a token that represents it, so that it also secures third-party services, in exchange for extra rewards and extra ways to be penalized. The term was coined by EigenLayer, the protocol that launched the category. The word says it plainly. You stake, then you stake the same capital again.

The third-party services carry a clumsy name, Actively Validated Services, usually shortened to AVSs. An AVS is any piece of software that needs its own set of honest validators but does not want to build one from scratch: an oracle feeding prices into a lending market, a bridge moving assets between chains, a data-availability layer for a rollup, or a co-processor running computation off-chain. Instead of bootstrapping a fresh token and a fresh validator set, an AVS rents security from people who have already staked. Those stakers collect a second yield; in return they accept a second set of slashing rules, so misbehavior on the AVS can cost them capital.

The stack underneath: staking and liquid staking

Restaking only makes sense once you can see the layers below it. The base layer is ordinary Ethereum staking. Lock 32 ETH, run a validator, help produce and attest to blocks, and earn a base reward that has drifted down to roughly 2.7 percent a year as more validators crowded in. By late August 2026 about a third of all ETH was staked across roughly 900,000 validators, according to live validator-queue trackers.

The trouble with base staking is that the capital is frozen and lumpy. You need a full 32 ETH, and until you exit you cannot sell it or put it to work. Liquid staking solved that. Deposit any amount with a protocol like Lido or ether.fi, receive a liquid staking token such as stETH or weETH that accrues the staking reward, and keep a freely tradable asset you can lend, borrow against, or sell whenever you like. Liquid staking is now the default way most people hold staked ETH, and it turned staking into a packaged product with an exchange-traded wrapper, a shift we covered in our look at how the SEC approved crypto yield through staking ETFs.

Restaking sits on top of that stack. You can restake native ETH directly, or restake a liquid staking token you already hold. Each layer adds yield, and each layer adds something new to trust.

What EigenLayer actually does: renting out security

EigenLayer is best understood as a marketplace with three sides. Stakers deposit ETH or liquid staking tokens and choose which services to back. Operators run the actual software each service requires, and stakers delegate to them. AVSs, the services themselves, pay for the security they consume. EigenLayer matches the supply of staked trust with the demand for it, and enforces the slashing rules that make the trust credible in the first place.

For a long time the enforcement half was missing. Slashing, the mechanism that lets an AVS actually punish a misbehaving operator, only went live on Ethereum mainnet on 17 April 2025, and even then as an opt-in, per EigenLayer’s own release notes. Until that point restakers were promised extra risk in theory while earning mostly points in practice. The first AVS to run was EigenLayer’s own data-availability layer, EigenDA; others include Lagrange, which builds zero-knowledge proof committees, and Omni, an interoperability network. The company has since rebranded to EigenCloud, signalling a wider ambition than restaking alone.

By 2026 the marketplace had breadth on the supply side and thinness on the other. Hundreds of services were in various stages of development, yet only a small number ran live on mainnet while paying meaningful fees, even as operators numbered in the thousands. That imbalance, plenty of security available for rent and few customers actually renting it, is the single most important fact about restaking’s economics, and it recurs throughout the rest of this guide.

Not everyone welcomed the design. Ethereum co-founder Vitalik Buterin published an essay in 2023 warning against overloading the chain’s consensus, arguing that “any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator,” in his post Don’t overload Ethereum’s consensus. His worry was never a single bad AVS; it was the slow accumulation of systems that grow too large and too interconnected to be allowed to fail.

Liquid restaking tokens: the leverage layer

Most people who “did restaking” never touched EigenLayer directly. They used a liquid restaking token, or LRT. The pattern mirrors liquid staking: deposit ETH or an LST with a protocol like ether.fi, Renzo, Kelp DAO or Puffer, receive a token (eETH, ezETH, rsETH, pufETH) that represents the restaked position, and carry that token back into the rest of DeFi to lend, borrow against, or farm. The same capital could now, in theory, earn a staking reward, an AVS reward, and a lending yield all at once. That is the leverage layer, and it is where retail actually participated.

ether.fi became the runaway leader, at times holding roughly three quarters of the entire LRT segment. The liquid restaking sector as a whole peaked near 18 billion dollars at the end of 2024 before shrinking through 2026, per The Defiant. The table below shows the main issuers and roughly where they stand in late August 2026, drawing on DefiLlama and protocol dashboards; treat the totals as approximate, because trackers count restaked, liquid-staked and idle deposits differently.

Where the yield actually came from

The headline numbers on LRT dashboards often read 8 to 12 percent, comfortably above the roughly 2.7 percent base staking reward. It is worth being precise about where that gap came from, because the answer explains a great deal of what happened next.

Stack the sources. First, the base staking reward, a real cash flow paid by Ethereum, around 2.7 percent. Second, fees paid by AVSs for the security they rent. Third, points and token incentives: protocol emissions and the expectation of a future airdrop. For most of restaking’s growth the third bucket dwarfed the other two. AVS fees, the one component that represents a real customer paying for a real service, stayed close to negligible. A large share of the advertised yield was subsidy and speculation on a token that had not launched yet, dressed up as a sustainable rate.

EigenLayer tried to repair the incentive with a tokenomics overhaul, ELIP-012, merged in December 2025. It pointed EIGEN emissions toward “productive stake” that actively secures live, fee-paying services, and routed a share of protocol fees into buying back the token, per the EigenLayer improvement proposal on GitHub. The redesign was sensible. The problem was arithmetic: protocol revenue running at only a few million dollars a month is far too small to move a token or to replace the points that had drawn the deposits in the first place.

The points era: how airdrops built restaking, then unbuilt it

To understand why so much money arrived and then left, you have to understand points. Before EigenLayer, ether.fi, Renzo and their peers had tokens, they had points: off-chain tallies that rewarded early depositors and hinted, without ever quite promising, at a future airdrop. Deposit early, farm points, and you might be paid in tokens later. It worked spectacularly well. Billions of dollars chased points into contracts whose actual service, securing AVSs, barely existed yet.

The dynamic layered on itself. LRT issuers farmed EigenLayer points on behalf of their depositors, then issued their own points on top, so a single deposit could be chasing two or three speculative rewards at the same time. Yield dashboards quoted rates that quietly assumed those points would convert into valuable tokens. When EIGEN finally became transferable and the other tokens listed, the conversion mostly disappointed. EIGEN peaked at 5.65 dollars in December 2024 and has since fallen about 96 percent, per CoinGecko, and the reflexive loop points had powered began to run in reverse.

The lesson that carried into 2026 is that a points program is a loan against a token’s future, not a business. It can bootstrap deposits, but it cannot manufacture the paying customers, the AVSs, that would justify those deposits staying. Once the points stopped, much of the capital had no reason to remain, which is a large part of why the unwinding described below arrived as quickly as it did.

The risks: slashing, rehypothecation, and correlated failure

Restaking multiplies risk in three distinct ways, and it helps to keep them separate.

• Slashing risk. Every AVS you back adds its own conditions under which your capital can be cut. One operator securing a dozen services carries a dozen ways to lose your delegated stake, some of them defined by young teams whose code has not been battle-tested.
• Rehypothecation risk. Because LRTs circulate through DeFi as collateral, the same underlying ETH can back a staking position, an AVS, a loan and a leveraged loop at the same moment. The layers are individually reasonable and collectively fragile; a wobble at the bottom cascades up through every position built on top.
• Correlated failure. If one LRT or one operator grows large enough, its problems stop being local. A depeg, a bridge exploit or a mass exit can move markets that never opted into restaking at all.

The people building restaking were candid that the first two risks are not new. Sreeram Kannan, the founder of EigenLayer, argued in 2023 that “anything that restaking can do, already liquid staking can do, so I view restaking as a lesser risk than liquid staking,” in an interview with CoinDesk. His point was that liquid staking already concentrates trust, and restaking simply makes the concentration explicit and priced. Critics answered that pricing a risk is not the same as removing it, and that the third risk, correlation, is exactly the kind markets tend to misprice until it arrives.

Case study: the Kelp and Aave contagion

The clearest illustration arrived on 19 April 2026. An attacker exploited Kelp DAO’s cross-chain bridge and minted roughly 116,500 rsETH out of nothing, worth about 292 million dollars, then deposited the fraudulent tokens as collateral on the lending market Aave and borrowed real assets against them. Aave was left with around 196 million dollars of bad debt, and its total value locked dropped by billions over a single weekend, according to CoinDesk’s reporting.

The mechanism is the whole lesson. Nothing was wrong with Ethereum, or with staking, or even with Aave’s core code. The failure was that an LRT had been accepted as blue-chip collateral without anyone pricing the risk of the bridge that minted it. Restaked collateral inherits the weakest link in a long chain, and here the weak link was a single misconfigured bridge. A coalition of protocols eventually coordinated a rescue, burning the fraudulent rsETH and restoring backing by around June 2026, but the episode is now a standing example in every serious crypto exploit post-mortem. It also fed straight into the wider reckoning over who is really pricing collateral risk in DeFi lending.

Restaking without a token: Symbiotic and Karak

EigenLayer was never the only design. Symbiotic, which went live in January 2025, takes a more permissionless approach: it accepts almost any ERC-20 as collateral rather than only ETH and its derivatives, enforced slashing from day one, and pointedly has never launched a public token. Karak is a third venue, multi-asset and much smaller. The comparison matters in 2026 for a reason few would have predicted a year ago. As capital drained out of EigenLayer, some of it landed on Symbiotic.

By mid-August 2026 Symbiotic had grown to around 1.5 billion dollars in value locked even as the category shrank overall, per DefiLlama. Its Core V2 redesign, live since 1 July 2026, hands the terms of each vault to professional risk “curators” such as the London asset manager Fasanara Capital, per The Block. The reason is direct: when ether.fi rebuilt its restaking product, it built the new version on Symbiotic, not EigenLayer. The table below sketches the four main venues and how differently they are put together.

One theme runs through the table. The two platforms with the most valuable tokens, EIGEN and BABY, are also the ones whose tokens have fallen furthest from their highs, while the tokenless venues quietly kept operating. It is a fair question whether a restaking token helps the protocol at all, or simply hands speculators something to sell.

Bitcoin restaking: Babylon’s different bet

Restaking is not only an Ethereum idea. Babylon brings the concept to Bitcoin, and it does so with a design that sidesteps the exact risk that sank Kelp. Rather than wrapping BTC into an ERC-20 and trusting a bridge, Babylon lets holders lock native coins using Bitcoin’s own timelock scripting, on Bitcoin itself, and that locked BTC then helps secure proof-of-stake chains that opt in. No wrapping, no custodian, no smart-contract bridge holding the base asset hostage.

Babylon launched its genesis mainnet and BABY token on 10 April 2025, according to The Block. Its dashboard has shown roughly 56,850 BTC locked, worth on the order of 5.6 billion dollars, for months, a figure that updates slowly enough to treat as a rough marker rather than a live counter, on Babylon’s own site. The appeal is a version of a trade Bitcoin already makes on its security side, where idle capital earns by protecting a network, a dynamic we unpacked in our explainer on how hashprice measures a mining day. The catch is the same one that haunts Ethereum restaking: a proof-of-stake chain secured by Bitcoin still has to pay enough for that security to matter, and real demand for rented Bitcoin trust remains mostly speculative.

The 2026 reversal, and why the rebound is mostly the ETH price

That brings the story to the summer of 2026 and the event that reframed the whole category. ether.fi, the largest LRT issuer, removed restaking from weETH, its flagship token, and reverted it to a plain liquid staking token with no EigenLayer exposure. Restaking did not vanish from ether.fi; it was quarantined into a separate, opt-in token called weETHs, built on Symbiotic. The size of that opt-in tells the story. weETHs held only around 9,136 tokens, roughly 18 million dollars, about half a percent of ether.fi’s staking base, per The Defiant. Less than one percent of ether.fi’s assets are still restaked, with the figure targeted to reach zero in the third quarter and the underlying EigenLayer credentials to be removed by the fourth.

Mike Silagadze, ether.fi’s founder and chief executive, did not dress it up. “End of an era. Sad,” he wrote of the change, adding, “I still think restaking will come back in one form or another, I think it was just a bit too early,” in comments reported by The Defiant. The weETH split was the market’s verdict rendered by the company that had done the most to sell restaking to retail. ether.fi has since leaned into a neobank strategy instead, adding tokenized stocks, portfolio loans and payment rails for a base of roughly 500,000 users, according to The Block.

So why do some headlines still say restaking is recovering? Because of a measurement trap. Dollar-denominated value locked did bounce this summer, with one reading putting EigenLayer around 12.9 billion dollars after an 11 percent weekly jump, per Cryptopolitan. But that move tracked a sharp ETH rally almost exactly; priced in ETH rather than dollars, deposits kept falling, and by stricter methodologies the same protocol counts closer to 5 billion dollars, down from a peak above 22 billion dollars in August 2025, per The Defiant. The demand signal that cannot be faked by price is usage, and usage collapsed. Active addresses fell from a peak around 33,470 in May to a few dozen to a few hundred a day by early August, on DefiLlama data. The token prices say the same thing. EIGEN trades near 20 cents, down about 96 percent from its 2024 high, per CoinGecko; ETHFI has held up better, but on the strength of the banking pivot, not restaking.

EigenLayer’s pivot: from restaking to verifiable cloud

EigenLayer read the same demand problem and answered it by changing what it sells. Under the EigenCloud banner it now pitches verifiable cloud services: EigenDA for data availability, EigenVerify for dispute resolution, and EigenCompute for confidential, verifiable off-chain computation, aimed increasingly at AI workloads that need to prove they ran honestly. Andreessen Horowitz backed the shift with a 70 million dollar token purchase in mid-2025, per CoinDesk.

The pivot is telling in two directions. It concedes that renting security to third-party AVSs did not generate enough paying demand on its own, so the protocol is manufacturing its own demand by becoming a cloud provider. And it moves restaking from a financial primitive toward infrastructure, where the question is no longer “what yield does my ETH earn” but “does this service have real customers.” Whether verifiable AI supplies those customers is the open bet of restaking’s next chapter.

Restaking and the SEC

Because restaking pays a yield, it lives close to a securities question, and 2026 was the year United States regulators drew clearer lines. The Securities and Exchange Commission’s Division of Corporation Finance said in a 29 May 2025 statement that ordinary protocol staking is administrative or ministerial rather than a securities transaction, and followed with a 5 August 2025 statement extending similar comfort to certain liquid staking arrangements.

Restaking is harder to place. The staff carve-out leans on the idea that staking rewards are payment for validation work, not profit from someone else’s effort. A restaking or liquid restaking token that markets a packaged yield, sets rates, and depends on a sponsor’s management fits the classic definition of an investment contract far more comfortably. None of the 2025 statements are rules, and staff guidance can be reversed; the broader direction of travel, from enforcement toward written rulemaking, is the theme we tracked in how the SEC swapped lawsuits for rules. For now the safest reading is that non-custodial, do-it-yourself restaking looks like staking, while a managed LRT sold as a yield product carries real regulatory tail risk.

So is restaking dead? What to watch

Restaking is not dead, but the version that raised tens of billions on the promise of easy stacked yield is finished. What remains is smaller, more honest, and split along a clean line. On one side is restaking as a financial product, LRTs sold for yield, which the market has largely rejected for as long as the yield is subsidy rather than fees. On the other is restaking as infrastructure, shared security and verifiable compute sold to applications that have real customers, which is unproven but still standing.

A few signals will tell you which way it breaks. Watch AVS fee revenue, not TVL: the day services pay meaningfully for security is the day the model works, and that day has not arrived. Watch whether Symbiotic’s tokenless, permissionless design keeps attracting the capital leaving EigenLayer. Watch Babylon’s Bitcoin experiment for any proof-of-stake chain that genuinely needs and pays for rented BTC. Watch whether EigenCloud’s verifiable-AI customers actually materialize. And watch the SEC, because a managed yield token’s fate may be decided in Washington as much as on-chain. Silagadze may yet prove right that restaking was early rather than wrong. In markets, being early usually costs about the same as being wrong.

For an ordinary holder the practical takeaway is narrower. If you only want the base staking reward, liquid staking already delivers it with far less to trust. If you want the extra restaking yield, ask one blunt question first: is that yield coming from fees a real service pays, or from points and emissions that depend on a token going up. Through 2026 the honest answer was almost always the latter, and a yield that depends on the next buyer is not really a yield, it is a position. That distinction, more than any single protocol’s fate, is what the whole episode taught DeFi.

Frequently Asked Questions

What is restaking in crypto?

Restaking is re-using staked ETH, or a token that represents it, to secure additional services beyond Ethereum itself, in exchange for extra rewards. The pledged capital keeps securing Ethereum while also backing things like oracles, bridges and data-availability layers, known as Actively Validated Services, and it accepts extra penalties if those services are attacked or misbehave. EigenLayer launched the category in 2023.

Is restaking safe?

Restaking adds risk on top of ordinary staking. Each service you back adds its own slashing conditions, and liquid restaking tokens used as collateral across DeFi can build leverage that cascades when something breaks. The April 2026 Kelp DAO incident, where a bridge exploit minted 292 million dollars of fraudulent tokens that landed as bad debt on Aave, is the standard cautionary tale. Non-custodial restaking you run yourself is lower risk than a managed yield token you do not control.

How much can you earn from restaking?

Advertised restaking yields ran roughly 8 to 12 percent during the growth years, against a base Ethereum staking reward near 2.7 percent. Most of that gap came from points and token incentives rather than fees paid by real customers, so much of it was temporary subsidy. As those incentives faded in 2026, the sustainable, fee-based portion of restaking yield proved to be very small.

What is the difference between staking, liquid staking, and restaking?

Staking locks ETH to secure Ethereum for a base reward. Liquid staking gives you a tradable token, such as stETH or weETH, that represents that staked ETH, so your capital is not frozen. Restaking pledges staked ETH or a liquid staking token again, to secure extra services for extra yield and extra risk. Each layer adds return and adds something new to trust.

Is restaking dead in 2026?

No, but it has shrunk and changed shape. In 2026 ether.fi, the largest liquid restaking issuer, removed restaking from its flagship weETH token and isolated it in a tiny opt-in product, and less than one percent of its assets remain restaked. Restaking as a yield product has been largely rejected, while restaking as shared security and verifiable compute infrastructure continues on a smaller scale. Its future depends on whether real customers ever pay for rented security.

By Yuki Tanaka, staking and DeFi correspondent, HOGE Wire.