Restaking in 2026: What It Is and Why the Market Is Unbundling It

On August 7, 2026, ether.fi did something that would have read as heresy two years earlier: it took restaking out of its flagship token. weETH, one of the largest liquid staking tokens in the market, was rolled back to plain Ethereum staking, and the higher-yield, higher-risk restaking exposure was pushed into a separate token called weETHs, built on a rival protocol, according to CoinDesk. The firm holds roughly $3.55 billion in customer deposits, so this was not a fringe experiment. It was the biggest liquid restaking issuer telling its users that staking and restaking should be two separate products, not one bundled together.

That decision is the clearest signal yet of what 2026 has done to restaking. The idea was intoxicating in 2023 and 2024: take the tens of billions of dollars of ETH already staked to secure Ethereum, and recycle that same security to protect oracles, bridges, rollups and data layers that could never bootstrap their own. Capital poured in chasing points and airdrops. Then the reset arrived. This guide explains what restaking actually is, how the machine works, why the market spent this year pulling staking and restaking back apart, and how to think about the whole category now that the froth has drained out of it.

What restaking actually is

Start with plain staking. To help run Ethereum, you lock up ether (32 ETH for a full validator, or a slice of that through a pool) and your capital stands behind the network’s honesty. Behave, and you earn a base yield; break the rules, and part of your stake is destroyed through a penalty called slashing. That base yield sits in the high-2 percent range in 2026, and roughly one-third of all ether is now staked, per CoinDesk. With ether trading near $1,885, down about 62 percent from its August 2025 peak of $4,946 (CoinGecko), that staked pile represents an enormous, mostly idle security budget.

Restaking asks a simple question: if all that ETH is already pledged to secure Ethereum, why not pledge it a second time to secure other things too? A restaker takes staked ETH (or a liquid token that represents it) and opts in to also back third-party services known as Actively Validated Services, or AVSs: oracle networks, cross-chain bridges, data-availability layers, rollup sequencers, keeper networks, coprocessors. In return the restaker earns extra rewards from those services. The catch is that each new service adds its own slashing conditions, so the same stake now has more ways to be penalized. Restaking is, in one line, renting out Ethereum’s economic security for extra yield and extra risk.

How EigenLayer turned staked ETH into rentable security

EigenLayer coined the term and built the first market for it. Its design is a three-sided marketplace. Restakers supply the capital, depositing ETH or liquid staking tokens and choosing which services to back. Operators run the actual software for those services and take delegated stake from restakers. AVSs are the buyers: new protocols that need a credible, slashable security guarantee and pay for it rather than launching a token and recruiting their own validator set from scratch. The pitch to a founder is that instead of spending years building trust, you rent it from Ethereum on day one.

For a long time this was security without teeth, because slashing was not yet switched on. That changed on April 17, 2025, when EigenLayer enabled opt-in slashing on mainnet, giving AVSs the ability to actually punish operators who misbehave, per the project’s own announcement. The company has since rebranded from EigenLayer to EigenCloud, widening its ambition from pure restaking into a broader stack of verifiable cloud services (data availability, off-chain compute and dispute resolution). Its EIGEN token, the asset used for staking and for slashing certain hard-to-attribute faults, trades around $0.17, a market value near $127 million and roughly 97 percent below its December 2024 all-time high of $5.65, according to CoinGecko. That drawdown is not a footnote; it is central to the story of 2026.

It helps to make the AVSs concrete, because the whole model depends on them existing. The best known is EigenDA, EigenLayer’s own data-availability layer, which lets rollups post transaction data cheaply while restaked ETH stands behind its honesty. Others include Lagrange, which produces zero-knowledge proofs about blockchain state for rollups, and Omni Network, which routes messages between Ethereum rollups. Each buys security from the same shared pool instead of minting a token and recruiting a validator set from nothing. When people ask whether restaking has real customers, these are the services they mean, and the honest answer in 2026 is that dozens are live but only a handful pay fees large enough to matter.

The 2026 reset: where the money went

At the peak, restaking looked like the most important new market in crypto. Total value locked across the sector climbed above $15 billion by various trackers in early 2026. By mid-year it had roughly halved. DefiLlama now counts around $7.9 billion across the restaking category, with EigenCloud alone holding close to two-thirds of it, or something near $5 billion, per DefiLlama. Precise figures are genuinely contested between data providers, so treat any single number as directional, but the direction is not in doubt: money left.

The reason is structural, not cosmetic. Most of the capital that arrived in 2024 was farming points, chasing an eventual airdrop rather than earning cash yield. Once the tokens landed and the points programs matured, that mercenary capital had little reason to stay. Underneath the incentives, the market discovered an awkward truth: there was far more staked ETH offering to sell security than there were AVSs willing to pay real fees to buy it. Supply of security dwarfed demand for it. Yields that had looked spectacular were mostly emissions and expectations, and when both faded, so did the headline numbers.

The exodus was not total, and the capital that stayed looks different from the capital that left. As yield farmers rotated out, a slower kind of money rotated in through regulated rails. SharpLink Gaming, a Nasdaq-listed treasury company chaired by Ethereum co-founder Joseph Lubin, deployed roughly $170 million of a planned $200 million ETH treasury into restaking and ether.fi positions, custodied through Anchorage, starting in early 2026, per CoinDesk. Much of that flow ran through Consensys-affiliated infrastructure, so it is not fully independent validation, but it does show that restaking is no longer a purely retail points game. The category is professionalizing even as it shrinks.

Liquid restaking tokens and the rehypothecation problem

Most people never touched EigenLayer directly. They used a liquid restaking token, or LRT. You deposit into a protocol like ether.fi, Renzo, Kelp or Puffer, and receive a liquid token (weETH, ezETH, rsETH, pufETH) that represents your restaked position and keeps accruing rewards while remaining tradable. ether.fi became the runaway leader of this segment. The appeal is composability: you can take that liquid token and put it to work again in DeFi, as collateral for a loan or into a yield strategy.

This is where the risk quietly compounds. The same underlying ether is now doing three jobs at once: securing Ethereum as a validator, securing one or more AVSs through restaking, and backing a loan somewhere in DeFi. Financiers have a word for reusing the same collateral in multiple places: rehypothecation. It is efficient when everything works and dangerous when one layer breaks, because a single shock can transmit through all three at once. The LRT boom stacked leverage on leverage, and 2026 delivered a live demonstration of what happens when the bottom layer cracks.

The great unbundling: why weETH split from restaking

Which brings us back to the ether.fi news, the anchor of this whole story. In August 2026 the firm turned weETH into a straightforward Ethereum staking token and carved restaking out into weETHs, a separate token whose restaking security runs on Symbiotic rather than EigenLayer. ether.fi captured roughly $223 million in annualized fees on about $51 million of annualized revenue across its $3.55 billion of deposits, per CoinDesk, so this is a large, profitable business making a deliberate product bet, not a struggling one lashing out.

The bet is that most users want a clean risk profile, and that bundling restaking into a mainstream staking token forced everyone to hold risk that only some of them wanted. Splitting the two lets a conservative holder own plain staked ETH and lets a yield-seeker opt in to restaking through weETHs, with eyes open. It also amounts to an exit. Less than 1 percent of ether.fi’s assets remained restaked, the restaked share is set to reach zero in the third quarter of 2026, and the firm plans to remove its EigenLayer withdrawal credentials entirely in the fourth quarter, according to The Defiant. The single largest liquid restaking brand is walking away from the protocol that started the category.

The timing collides with a separate fight over Ethereum’s own rewards. Researchers floated a proposal in 2026 to taper new staking issuance toward zero once about 60 million ETH (roughly half the supply) is staked, a move meant to curb over-staking. ether.fi founder Mike Silagadze pushed back hard, arguing the change would push out smaller stakers and weaken the products built on staking rewards, his own among them, again per CoinDesk. Read together, the two stories describe an industry renegotiating how much yield staked ETH should earn and how much extra risk stakers should take to earn more.

Slashing, correlated risk and what Vitalik warned about

The deep objection to restaking is not that any single AVS is dangerous. It is that stacking many of them onto the same stake creates correlated risk. If one unit of ETH is slashable across a dozen services, then one nasty bug, one compromised operator, or one badly written slashing rule can trigger losses in several places at once, and those losses can cascade. Ethereum’s core security was designed to answer one question honestly. Restaking asks it to answer many, and the failure modes multiply.

Vitalik Buterin saw this coming. In his May 2023 essay warning against overloading Ethereum’s consensus, he wrote that any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator, and cautioned against a world where a restaking project grows so large that the community feels obliged to fork Ethereum itself to bail it out. EigenLayer founder Sreeram Kannan has argued the risk is overstated, telling CoinDesk in 2023 that anything restaking can do, liquid staking can already do, so he views restaking as a lesser risk than liquid staking, not a greater one. Both can be true at once: restaking need not be reckless, and it can still concentrate systemic risk if too much value leans on too few operators and contracts.

The mechanism that turns this from theory into tail risk is operator concentration. Restakers delegate to a relatively small set of professional operators, and those operators often run many AVSs at once. If a single popular operator is slashed for a fault on one service, every restaker who delegated to it can be penalized, and if that operator secured a dozen services, the damage radiates outward in one stroke. Slashing on EigenLayer is opt-in and capped per service, which softens the blow, but the underlying shape of the risk is a web of shared dependencies rather than a set of isolated bets. That is why serious due diligence in restaking is really due diligence on operators, not just on protocols.

The Kelp and Aave contagion: a live case study

April 2026 turned the theory into a headline. Attackers exploited Kelp’s cross-chain bridge to mint 116,500 rsETH out of thin air, a position worth about $292 million, then deposited it as collateral on Aave and borrowed against it, leaving the lending market with roughly $196 million of bad debt, as CoinDesk reported. The trouble did not stay inside the restaking protocol. It flowed straight into DeFi credit because a liquid restaking token had been accepted as collateral without fully pricing the risk of the bridge that minted it. This is exactly the rehypothecation failure the previous section described, playing out with real money.

The response showed both the fragility and the resilience of the system. Aave froze rsETH’s borrowing power, and a coalition of DeFi teams organized a rescue. Aave founder Stani Kulechov, whose protocol absorbed the bad debt, said publicly that Aave is my life’s work and we’re working nonstop to find the best possible outcome for users, and personally pledged 5,000 ETH toward the shortfall. By early summer the fraudulent rsETH had been burned and the backing restored. The episode is now a standard reference for anyone weighing whether a restaked token belongs in a lending market, and it fits into the broader picture of collateral risk we cover in our 2026 risk map for on-chain credit.

The challengers: Symbiotic, Karak and SSV

EigenLayer is no longer the only game. Symbiotic took a deliberately different path: it accepts almost any ERC-20 as collateral rather than only ETH and its derivatives, and it has still not launched a token, running a points program instead since June 2024. It has raised about $34.8 million across seed and Series A rounds, per FinanceFeeds, and in mid-2026 repositioned with a Core V2 release aimed at collateral markets rather than pure restaking. The fact that ether.fi chose Symbiotic to host its new weETHs restaking token is a meaningful vote of confidence in the challenger. Symbiotic’s TVL is quoted anywhere from a few hundred million dollars to more than a billion depending on the methodology, so it is best described as a strong number two rather than pinned to a false decimal.

Karak, once counted as the third major player, has drifted out of the category entirely, rebranding to OpenGDP and reorienting toward tokenizing real-world economic activity rather than selling shared security. A fourth model comes from SSV Network, the largest distributed validator technology provider, which splits a single validator key across several independent operators so no one machine holds it whole. Its SSV 2.0 based applications framework proposes a safer form of shared security in which a validator’s 32 ETH principal is never slashable and only optional, separately delegated capital is at risk. SSV Labs founder Alon Muroch has called it the most ambitious project the team has attempted. Taken together, the challengers are all trying to answer the same question EigenLayer opened, whether shared security can be delivered with less correlated downside.

Sources for the figures above: DefiLlama for EigenCloud TVL, FinanceFeeds for Symbiotic’s funding and token status, and Babylon Labs for the Bitcoin staked figure.

Bitcoin restaking: Babylon and native BTC security

Restaking is no longer an Ethereum-only idea. Babylon extends the concept to Bitcoin, and it does so without the wrapped-token bridges that caused the Kelp disaster. A Bitcoin holder locks coins using Bitcoin’s own timelock scripting, keeping the BTC on the Bitcoin chain itself, and that locked stake can then secure proof-of-stake networks that opt in. Babylon’s dashboard reports about 56,853 BTC staked, worth roughly $5.64 billion, per Babylon Labs, which would make it the largest Bitcoin staking system by that measure. Treat the dashboard as slow-moving rather than a live ticker, because the figure has sat unchanged for long stretches.

The most-watched integration is a planned link with Aave V4 that would let native Bitcoin serve as lending collateral through a trust-minimized vault anchored on Bitcoin, using its Taproot scripting to lock coins on-chain. As of this writing that integration has reached testnet and passed an early governance temperature check, but it is not yet live on mainnet, with risk parameters, oracle design and audits still being finalized. Babylon’s approach leans heavily on Bitcoin’s post-Taproot capabilities, and readers who want the mechanics of what that upgrade actually delivered can see our Taproot scorecard. The BABY token, like EIGEN, trades at a small fraction of the value it helps secure, a recurring theme in this sector.

Where does the yield actually come from?

This is the question that decides whether restaking survives as a real business. Restaking yield comes from four distinct sources, and they are not equally durable. The first is native staking yield, the ordinary reward for securing Ethereum, which is real but modest and slowly falling. The second is token emissions and points, freshly minted incentives that inflated 2024 returns and then largely evaporated; these are dilution dressed up as yield. The third is AVS fees, the actual money that services pay for security, which is the only genuinely sustainable source and remains small. The fourth is rehypothecation, reusing a liquid restaking token in DeFi, which is not new yield at all but leverage layered on top of the first three.

EigenCloud has openly restructured itself around the third source. A December 2025 tokenomics overhaul redirected EIGEN emissions toward productive stake, meaning capital that actively secures live, fee-generating services rather than sitting idle. The company’s larger bet is to manufacture fee demand by turning itself into verifiable cloud infrastructure, using restaked security to guarantee off-chain computation for things like AI inference and agent workloads; a16z backed that pivot with a $70 million token purchase in 2025. The thesis is that AI systems will need cryptographically verifiable execution, and that restaking can price and enforce it. That connects restaking to two fast-moving frontiers we cover separately, the effort to make machine-learning results provable in our zkML explainer and the rails that let autonomous software transact in our guide to agentic payments. For now, measured on-chain fee revenue still lags the size of the capital, which is the honest bear case: the real yield is real, but it is not yet large.

The exit problem: withdrawal queues and ELIP-018

Getting into restaking was frictionless. Getting out is not, and that asymmetry explains why ether.fi’s exit stretches across two quarters rather than happening overnight. Unwinding a restaked position can mean waiting through an AVS unbonding period, redeeming a liquid token for its underlying stake, and then queuing to exit the validator itself. Ethereum’s validator queues have swung violently in 2026, and at times the entry queue backed up to roughly six weeks while the exit side ran near empty, so rotating capital in and out is far from instant.

EigenLayer’s own governance has acknowledged the pain. A draft proposal known as ELIP-018, nicknamed RETIRE, would let a validator’s EigenPod owner permanently and irreversibly disable restaking through a new method, clearing the accounting friction that otherwise strands stakers who want to fully rotate keys or leave, per the EigenLayer governance forum. The proposal is still a draft with no mainnet schedule, but its existence is telling: a protocol only builds a formal off-ramp once enough users want the door. Exit design has quietly become one of the most important features in the category, and any product that makes leaving expensive should be viewed with suspicion.

How US regulators treat staking and restaking

The regulatory picture in the United States has shifted decisively in staking’s favor, though restaking sits in a grayer zone. The old benchmark was the SEC’s $30 million settlement with Kraken in February 2023 over its staking-as-a-service program, which treated packaged staking yield as an unregistered securities offering. The stance since then has reversed. In May 2025 the SEC’s Division of Corporation Finance issued a staff statement concluding that protocol staking, whether solo, delegated or custodial, is not itself a securities transaction, and Commissioner Hester Peirce reinforced the point in a companion statement titled Providing Security is not a Security.

The agency went further in August 2025, extending that comfort to liquid staking and to the receipt tokens that represent staked positions, per an SEC statement, and a joint SEC and CFTC interpretation in March 2026 began drawing clearer digital-commodity lines. The guardrails still matter: a custodian that guarantees a fixed return, or exercises discretion over when and how much to stake, can fall back outside the safe harbor, and restaking products that bundle slashing risk with marketed yield are less clearly covered than plain protocol staking. With a broader market-structure bill still grinding through Congress, the treatment of these newer yield products remains unsettled, part of the packed policy calendar we track in our look at crypto’s September regulatory gauntlet.

How to evaluate restaking as a user in 2026

The unbundling is a gift to anyone who wants to make a deliberate choice. Now that issuers are splitting plain staking from restaking, you can decide how much extra risk you actually want rather than inheriting it by default. A short checklist does most of the work. First, ask where the yield comes from: if the advertised return leans on emissions or points rather than fees that services genuinely pay, expect it to compress. Second, map the slashing surface: how many AVSs is your stake exposed to, and how correlated are they, since one shared dependency can turn several small risks into one large one.

Third, look hard at any DeFi reuse: a liquid restaking token used as loan collateral inherits both the protocol’s risk and the lending market’s, as Kelp and Aave demonstrated. Fourth, read the exit terms before you enter, because unbonding periods, redemption mechanics and validator queues decide how fast you can actually leave. Fifth, know your regulatory footing: non-custodial protocol interaction is treated very differently from a custodial product promising a fixed return. None of this makes restaking bad. It makes restaking a specific tool with a specific risk budget, and 2026 is the year the market finally started pricing it that way rather than treating it as free money bolted onto staking.

Frequently Asked Questions

What is restaking in simple terms?

Restaking means taking ETH you have already staked to help secure Ethereum, or a liquid token that represents it, and pledging it a second time to secure other services such as oracles, bridges and data-availability layers. You earn extra rewards from those services, but you also accept extra ways to be penalized, or slashed, if the operator running your stake misbehaves. EigenLayer, now branded EigenCloud, popularized the model starting in 2023.

Is restaking safe?

Restaking adds risk on top of ordinary staking. The same stake can be exposed to slashing across several services at once, so a single bug or bad operator can cause correlated losses, and when restaked tokens are reused as loan collateral, trouble can spread into lending markets, as the April 2026 Kelp and Aave incident showed. It is not inherently unsafe, but the extra yield comes with a larger and more complex risk surface than plain staking.

What is the difference between staking and restaking?

Staking secures one network, Ethereum, and pays a base yield in the high-2 percent range in 2026. Restaking reuses that staked capital to secure additional third-party services for additional rewards and additional slashing conditions. In 2026 several issuers, led by ether.fi, split the two apart so users can choose plain staking or restaking deliberately rather than getting both bundled into a single token.

Why did ether.fi remove restaking from weETH?

In August 2026 ether.fi turned weETH back into a plain Ethereum staking token and moved restaking into a separate token, weETHs, built on Symbiotic. The firm said the split gives users a cleaner choice between lower-risk staking and higher-risk restaking, and it is winding down its EigenLayer exposure, with the restaked share heading toward zero later in 2026. Less than 1 percent of its assets were still restaked at the time.

Does restaking still pay well in 2026?

Much less than during the 2024 points era. Total value locked across restaking protocols fell from a peak above $15 billion to under $8 billion by mid-2026, and headline yields compressed as token emissions and airdrop farming faded. The durable part of restaking yield is the fees that services pay for security, and those remain small relative to the capital on offer, which is a big reason the market is repricing the category.

Written by Adrian Cole, HOGE Wire markets desk.