Restaking in 2026: How It Works and the Great Unwinding

Restaking spent two years as the loudest idea in Ethereum. The pitch was simple and seductive: take capital that already secures the network and rent it out a second time to secure other services, then collect a second yield on top of the first. By early 2026 tens of billions of dollars had flowed toward that idea. In August 2026 the market delivered its verdict, and it was not the one the sector wanted.

On 7 August, ether.fi, the largest liquid restaking protocol by deposits, stripped restaking out of weETH, its flagship token, and turned it back into a plain liquid staking token. Restaking now lives in a separate, opt-in token called weETHs, built not on EigenLayer but on a rival called Symbiotic. That opt-in token holds roughly $18 million, about 0.5% of ether.fi’s $3.3 billion staking base, according to The Defiant. “End of an era. Sad,” wrote founder Mike Silagadze, before adding that he still believes “restaking will come back in one form or another, I think it was just a bit too early.” Less than 1% of ether.fi’s assets are still restaked, down from about half at the start of the year, and the company plans to hold none by the end of the third quarter.

The strange part is that the tokens are up. As ETH ripped past $2,500 in a broad macro-liquidity rally, EIGEN and ETHFI both jumped more than 30% in a week, moving in lockstep with the wider market rather than on any restaking-specific news, per CoinGecko. That is the tension worth understanding. The plumbing still runs, the tokens still trade, and yet the product restaking was sold as, extra yield for reused capital, has been quietly dismantled by the people who built it. This is what restaking is, how it works, and why 2026 became the year of the great unwinding.

What Restaking Actually Is

Restaking is the practice of pledging already-staked ETH a second time to help secure services other than Ethereum itself. When you stake ETH, your capital backs a promise to behave honestly as a validator; misbehave and part of your stake is destroyed, a penalty called slashing. Restaking extends that same enforcement mechanism to third parties. You agree that your stake can also be slashed if you break the rules of, say, an oracle network, a bridge, or a data-availability layer. In exchange, that service pays you a fee. EigenLayer, the protocol that coined the term, calls these services Actively Validated Services, or AVS.

The appeal to the demand side is that a new protocol does not have to bootstrap its own token and its own validator set from scratch. Instead it rents security that already exists, borrowing the economic weight of billions of dollars of staked ETH. The appeal to the supply side, the stakers, is a second stream of income on capital that would otherwise earn only Ethereum’s base reward of roughly 3%. Sreeram Kannan, EigenLayer’s founder, framed it in 2023 as an efficiency gain rather than a new danger, telling CoinDesk that “anything that restaking can do, already liquid staking can do, so I view restaking as a lesser risk than liquid staking.”

Not everyone agreed. Vitalik Buterin published a warning the same year titled “Don’t overload Ethereum’s consensus,” arguing that “any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator” (vitalik.eth.limo). That disagreement, growth through reuse versus creeping fragility, has run underneath every phase of the restaking story since.

The Staking Stack: Where Restaking Sits

To see why restaking is both powerful and precarious, it helps to lay out the stack it sits on top of. Each layer adds a feature and, with it, a new failure mode.

The bottom layer is base staking: lock 32 ETH, run a validator, earn consensus rewards. The second layer, liquid staking, hands you a tradeable receipt token (an LST) such as Lido’s stETH or Rocket Pool’s rETH, so your capital stays liquid while it earns; we compare the leaders in our look at Lido, Rocket Pool and Frax. Restaking is the third layer: you point your staked ETH or your LST at one or more AVS and accept their slashing conditions for extra yield. The fourth layer, liquid restaking, wraps the restaked position in yet another tradeable token, a liquid restaking token or LRT, such as ether.fi’s eETH or Renzo’s ezETH, so even your restaked capital stays liquid and can be reused again elsewhere in DeFi.

Each wrap increases capital efficiency and, at the same time, stacks one contingent claim on top of another. By the time a single unit of ETH is staked, liquid-staked, restaked and liquid-restaked, it is simultaneously securing Ethereum, backing an LST peg, insuring several AVS, and serving as collateral for a loan somewhere. That is the design restaking’s critics kept pointing at, and it is the design 2026 put to the test.

Inside EigenLayer: AVS, Operators and Shared Security

EigenLayer is the protocol that turned restaking from a whitepaper into an economy. It works through three roles. Stakers deposit ETH or LSTs and choose to restake. Operators run the actual software for the services being secured, and stakers delegate their restaked capital to operators they trust. AVS are the services that buy security: they register with EigenLayer, define the tasks operators must perform and the conditions under which stake gets slashed, and pay rewards in return.

Slashing went live on mainnet in April 2025 on an opt-in basis, the moment restaking stopped being a rewards-only game and became a system with real, enforceable downside. By 2026 EigenLayer, since rebranded EigenCloud, counted thousands of operators and a long roster of AVS in development, though only a fraction were live and fewer still were paying meaningful fees. The flagship AVS is EigenDA, EigenLayer’s own data-availability service; others include zero-knowledge proving networks and interoperability hubs. The pattern that matters is that supply of security arrived long before demand for it.

Total value locked tells that story in numbers, if you read it carefully. EigenLayer’s TVL rebounded past $12.9 billion by mid-August, up about 11% in a week, per Cryptopolitan, but that jump tracked the ETH price almost exactly rather than any net new deposits, and other trackers using different accounting showed figures closer to $6.5 billion (DefiLlama). When capital is denominated in ETH and ETH rallies a third in a week, dollar TVL rises even as coins leave. Active-address counts, which cannot be inflated by price, had fallen to a fraction of their spring peak.

The Yield Story, and Why It Ran Out of Road

Restaking’s growth was powered less by fees than by points. Through 2024 and into 2025, protocols handed out points and the promise of future token airdrops to anyone who deposited, and LRT issuers layered their own points on top of EigenLayer’s. Headline yields of 8% to 12% were common, but most of that came from token emissions and airdrop speculation, not from AVS actually paying for security. When the airdrops landed and the points stopped, the incentive that had pulled in tens of billions of dollars weakened.

The distinction between emissions and real yield is the whole game. Emissions are the protocol printing its own token to pay depositors, which dilutes holders and cannot last. Real yield is outside revenue, in this case AVS fees, flowing to stakers. Restaking’s problem is that AVS fees have stayed small. EigenLayer’s own tokenomics reform, ELIP-012, passed in December 2025, routes 20% of AVS rewards and 100% of EigenCloud service fees into buying back and reducing the supply of EIGEN (GitHub), a sensible design that runs into a hard wall: protocol revenue on the order of a few million dollars a month is far too small to move a token or to justify the capital parked against it.

That gap between promised yield and delivered yield is why the segment looked fragile well before ether.fi acted. Custodial exchanges offering simple, insured staking yields, which we put through their paces in our Coinbase, Binance, Kraken and OKX staking comparison, started to look like the better trade for ordinary users: less yield on paper, far less that could go wrong.

Slashing and the Risk That Was Never Priced

The core promise of restaking is enforceable security, which means the core risk is slashing. In plain staking, slashing is narrow and well understood: you lose stake for going offline for long periods or for signing two conflicting blocks. Restaking multiplies the surface. Every AVS your operator secures can add its own slashing conditions, some of them tied to complex off-chain behavior that is hard to specify and harder to prove. A bug in one AVS’s slashing logic, or a disputed judgment about whether a rule was broken, can in principle destroy stake that was also securing Ethereum and backing an LRT.

There is also correlation risk. If many operators secure the same handful of AVS, and one of those AVS suffers a slashing event, losses can hit a large share of the restaked set at once. That is the too-big-to-fail worry Buterin flagged: a slashing cascade large enough that the community faces pressure to intervene at Ethereum’s base layer to prevent contagion, which is exactly the sort of social-consensus overload he warned restaking could invite. Insurance markets and slashing-cover products emerged to price this, but a thinly traded cover market cannot absorb a correlated, systemic hit. The forensic industry that grew up to autopsy these failures, chronicled in our profile of Halborn, has stayed busy.

The Kelp-Aave Contagion: When Rehypothecation Bit Back

The clearest warning came in April 2026, and it did not even involve slashing. On 19 April, attackers exploited Kelp DAO’s cross-chain bridge to mint about 116,500 rsETH, Kelp’s liquid restaking token, worth roughly $292 million, out of thin air, per CoinDesk. They then deposited the counterfeit tokens as collateral on Aave and borrowed against them, leaving the lending protocol with around $196 million in bad debt and dragging its total value locked down by billions of dollars over a single weekend.

Nothing in that chain of events was a restaking failure in the narrow sense; it was a bridge exploit. But it only became a systemic event because an LRT had been accepted as blue-chip collateral across DeFi without anyone pricing the risk of the bridge that minted it. Aave founder Stani Kulechov moved fast, freezing rsETH’s borrowing power and rallying a rescue coalition; he told CoinDesk that “rsETH has been frozen on Aave V3 and V4, the asset does not have any borrowing power as a measure due to KelpDAO bridge exploit that happened outside of Aave,” and pledged personal ETH to the recovery. By June the shortfall had been closed, with more than 117,000 rsETH burned and backing restored.

The lesson stuck: rehypothecation, the reuse of the same collateral across many protocols, is exactly what makes restaking capital-efficient and exactly what turns a local failure into a shared one. After Kelp, risk teams looked harder at every LRT sitting inside a lending market, and appetite for stacking claims cooled.

The Great Unwinding: ether.fi Walks Away from Restaking

Which brings us back to August. ether.fi was not a skeptic; it was the standard-bearer. Its eETH and weETH tokens made it the largest LRT issuer, and its brand was synonymous with restaking as a consumer product. So when ether.fi removed restaking from weETH and reverted it to a plain liquid staking token, it was the sector’s most committed builder conceding that most of its users never wanted the extra risk in the first place.

Restaking did not vanish at ether.fi; it was quarantined. It now lives only in weETHs, an opt-in token that, tellingly, ether.fi built on Symbiotic rather than on EigenLayer, as CoinDesk reported. As of the split, weETHs held about $18 million against a weETH float of 1.72 million tokens, so the opt-in restaking product accounts for roughly half a percent of the base (The Defiant). Less than 1% of ether.fi’s assets remain restaked, down from about 50% earlier in 2026, with a target of zero by the end of the third quarter and removal of the EigenPod withdrawal credentials that tie its validators to EigenLayer by the fourth.

Silagadze’s epitaph, “End of an era. Sad… I think it was just a bit too early,” is worth taking at face value. He is not saying restaking was a fraud; he is saying the demand was not there yet and the risk was not worth it for a mass-market product. Where ether.fi is putting its energy instead makes the point louder. On 13 August the company rolled out what it openly calls a neobank: tokenized stocks through xStocks, portfolio-backed loans routed through Aave at rates around 4%, tokenized metals as collateral, and fiat accounts across more than 30 currencies, serving roughly 500,000 users against about $3.5 billion in value locked, per The Block. Silagadze’s stated goal is to “replace the traditional bank for most users.” The firm that did the most to popularize restaking now earns its keep selling banking, credit and payments. In 2026 that is not a footnote to the restaking story; it is the story.

Where the Money Went: The Survivors

De-restaking did not vaporize the capital; it redistributed it. Some flowed back into plain liquid staking, some into the survivors that never over-promised, and some, ironically, into the very rival ether.fi chose for weETHs. Symbiotic, which accepts any ERC-20 as collateral and runs slashing from day one, grew its deposits toward $1.5 billion even as EigenLayer’s user activity thinned, partly because de-restaking flows like weETHs landed on it. The table below reads deposit figures from DefiLlama and aggregator trackers, which diverge and should be treated as approximate.

Two things stand out. First, the LRT issuers that survived did so by shrinking, diversifying, or both; none is chasing peak-era growth numbers. Second, the token prices tell you how the market rates the equity of this business. EIGEN trades around $0.23 for a market cap near $200 million, down roughly 96% from its December 2024 high, and ETHFI near $0.63, down about 93% from its 2024 peak (CoinGecko), even after a weekly bounce of more than 30% that was really just the whole market moving together.

Shared Security, Four Ways

Restaking is not one design but several, and they disagree about the fundamentals: what collateral to accept, when to enable slashing, and whether a token is even necessary. The four leading models in 2026 sort out like this.

EigenLayer bet on ETH-denominated security, a native token, and a broad AVS marketplace. Symbiotic bet on flexibility and no public token, letting each network choose its own collateral and its own arbiter for disputes, whether UMA, Kleros, or a committee. Babylon bet on Bitcoin. SSV’s based applications, built on distributed-validator technology, bet that shared security works best when the 32 ETH principal is never at risk and only separately delegated capital can be slashed. The through-line of 2026 is that after the year’s failures, designs that limit downside earned more trust than designs that maximize reuse.

The Demand Problem: Who Pays for Rented Security?

Every restaking design shares one dependency: someone has to actually want to buy the security. This is where the sector has struggled most. Building the supply of restaked ETH was easy, because points and airdrops pulled it in. Building durable demand from AVS willing to pay real fees has been slow. A handful of services, EigenDA among them, plus some proving and interoperability networks, use restaked security in production, but the fees they pay are modest, and many projects concluded they could bootstrap their own token incentives more cheaply than renting security they did not fully control.

The economics resemble a security budget, the same concept that governs Bitcoin mining, where the value spent securing a network has to be justified by the value being secured; we unpack that idea in our look at hashprice and Bitcoin’s security bill. In restaking, the security on offer vastly exceeded the security anyone was willing to pay for, so the price of that security, the AVS fee, stayed near zero. A market with abundant supply and thin demand clears at a low price, and a low price cannot support the token valuations or the deposit base that the points era built. Strip away the emissions and you are left with a product whose customers have not yet shown up in the numbers.

EigenCloud’s Second Act: Betting on Verifiable AI

EigenLayer’s answer to the demand problem is to change what it sells. Rebranded EigenCloud, it is repositioning restaked security as the trust layer for verifiable computation, with a particular focus on AI. The pitch is that a restaked network can attest that an off-chain computation, an AI model’s output or a cloud job, ran correctly, and back that attestation with slashable economic security. In June 2025, a16z put $70 million behind the pivot with a token purchase to support the EigenCloud launch, per CoinDesk, and Kannan has framed the goal as letting developers build with what he calls crypto-grade verifiability.

Whether verifiable AI is a real market or the next narrative in search of demand is unresolved. It sits in a crowded field of projects promising trustworthy off-chain compute, and the same question applies: will paying customers show up? For now it is a bet, not a business, but it is a rational one, because it tries to manufacture the demand that restaking-for-AVS never found. It is worth watching precisely because it is the clearest tell of what the people closest to restaking now think it is good for.

Bitcoin Restaking: Babylon’s Different Wager

Not all restaking runs on Ethereum. Babylon lets Bitcoin holders lock BTC using Bitcoin’s own timelock scripting, with no wrapping and no bridge, and use that locked BTC to help secure proof-of-stake chains that opt in. Because the coins never leave Bitcoin, Babylon sidesteps the bridge risk that turned the Kelp incident systemic. Its dashboard reports roughly 56,853 BTC locked, worth around $5.6 billion, which makes it the largest Bitcoin staking system by that measure, per Babylon Labs, though the figure has looked static for weeks and should be read as a slow-updating dashboard number rather than a live counter.

Babylon shares restaking’s core demand problem, since it still needs proof-of-stake chains willing to pay for Bitcoin-backed security, but its risk profile is genuinely different, and it taps the largest pool of idle capital in crypto. If restaking has a second life, the Bitcoin version, built on a chain that does exactly one thing and does it conservatively, is one of the more plausible places for it to happen.

What US Regulators Say About Staking and Restaking

The regulatory picture in the United States shifted in favor of staking just as the market cooled on restaking. In May 2025, the SEC’s Division of Corporation Finance said that protocol staking, the act of participating in a proof-of-stake network, is administrative or ministerial in nature and not a securities transaction under the Howey test (SEC). In August 2025 the staff extended similar reasoning to liquid staking and to staking receipt tokens like stETH (SEC). Commissioner Hester Peirce summarized the view with the line that providing security is not a security.

Restaking sits in a grayer zone. The staff’s protocol-staking relief explicitly excludes arrangements with intrinsic economic properties such as generating passive yield, and restaking is sold precisely as extra yield. A liquid restaking token that markets an enhanced return, managed by an issuer that decides which AVS to secure and how, looks less like ministerial staking and more like a managed product. These are staff statements, not rules, and a future SEC could reverse them. The deeper fight, over whether writing something into a smart contract puts it beyond a regulator’s reach, is one we examine in our piece on DeFi compliance and the gatekeeper. For custodial staking and restaking services offered to retail, the safest assumption is that securities and consumer-protection rules can still apply.

Restaking After the Reset: How to Think About It Now

So is restaking dead? No, and the framing misleads. What died in 2026 was restaking as a default consumer product, the assumption that every staker should automatically restake for extra yield. What remains is restaking as an opt-in, specialist tool for the minority who understand and want the risk, which is roughly what the more careful voices said it should be all along.

A few markers will show whether the reset holds. Watch AVS fee revenue, not TVL: durable demand shows up as fees paid for security, and until that number grows the sector is still supply in search of demand. Watch whether any restaked service becomes genuinely load-bearing, an AVS that real applications depend on and pay for. Watch Symbiotic and Babylon, the models that either avoided a token or avoided bridge risk, to see whether restraint outperforms reuse. And watch EigenCloud’s AI pivot for the first paying customers that would prove the demand thesis.

For most readers the practical takeaway is simpler. If you hold an LST and someone offers you a few extra points of yield to restake it, understand that you are selling insurance on services you have never audited, backed by capital you also need for something else. In 2024 that trade was dressed up as free money. In 2026, after ether.fi’s exit, it is finally being priced as what it always was: a real risk for a small and uncertain reward.

Frequently Asked Questions

What is restaking in simple terms?

Restaking means pledging ETH that is already staked on Ethereum a second time, so it also helps secure other services such as oracles, bridges or data-availability layers. In return you earn an extra fee, but you also accept extra ways to lose your stake if one of those services decides you broke its rules.

Is restaking safe?

Restaking adds risk on top of ordinary staking. Each service you secure can impose its own slashing conditions, and liquid restaking tokens have been reused as collateral across DeFi, which can turn one failure into a shared one, as the April 2026 Kelp DAO exploit showed. It is best understood as a specialist, opt-in strategy rather than a default, and only with capital you can afford to put at risk.

Why did ether.fi remove restaking from weETH?

In August 2026 ether.fi turned weETH back into a plain liquid staking token and moved restaking into a separate opt-in token, weETHs, built on Symbiotic. Founder Mike Silagadze said restaking was a bit too early, and the company plans to hold almost no restaked assets by the end of the third quarter, focusing instead on a neobank offering loans, tokenized stocks and payments.

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 tradeable token, such as stETH, that represents your staked ETH so it stays usable. Restaking takes staked ETH or a liquid staking token and points it at additional services for extra yield and extra slashing risk; a liquid restaking token then makes even that position tradeable.

Does restaking still have a future?

Yes, but a narrower one than the 2024 hype implied. The technology and the largest protocols, EigenLayer (now EigenCloud), Symbiotic and Bitcoin-based Babylon, are still running, and EigenCloud is repositioning restaked security toward verifiable AI. The open question is demand: whether enough services will pay real fees for rented security to justify the capital behind it.

Yuki Tanaka covers Ethereum staking, restaking and on-chain credit for HOGE Wire.