Restaking in 2026: A Security Glut Nobody Is Renting

In the week running up to the end of September 2026, the entire restaking category on Ethereum, roughly ten billion dollars of staked value pledged to secure other people’s software, earned its participants a grand total of $99,977 in fees. Over the same seven days, plain liquid staking, the product restaking was supposed to make look quaint, earned about $27.35 million, according to figures compiled by Blockhead from DefiLlama data. Measured per dollar secured, ordinary staking out-earned restaking by a factor of about 53.

That one statistic is the whole story in miniature. Restaking is the idea that the economic security locked behind Ethereum’s validators can be rented out a second time, to oracles, bridges, data-availability layers and other services that need their own guarantees. By late 2026 the supply of that rentable security was enormous and the demand for it was close to nonexistent. The market built a warehouse the size of a small country and then found almost no one wanted to rent a unit.

The supply numbers are genuinely large. EigenLayer, the protocol that invented the category, had seen deposits cross 5 million ETH (native ETH plus liquid-staking tokens) across roughly 18 active security networks, bitcoinist reported in early September. The demand numbers are the ones that keep collapsing. In late September, ether.fi, the firm that did more than anyone to turn restaking into a retail product, confirmed it would sever its last structural ties to EigenLayer by the end of the third quarter, with fewer than 1 percent of its assets still restaked. A companion piece looked at the same retreat through the lens of interest rates; this guide looks at it as a market that never cleared, and uses that frame to explain what restaking actually is.

None of this means restaking was a fraud or a failure of engineering. The plumbing works. Slashing is live. Billions of dollars really are deposited. What failed, so far, is the business: a two-sided marketplace needs two sides, and only one of them turned up. To see why, you have to start with the thing being sold.

Restaking in one sentence (and the risk the sentence hides)

Restaking is re-pledging ETH that is already staked so it can also secure a second service, in exchange for a second stream of rewards and a second set of ways to lose money. That is the whole concept. Everything else is detail about who holds what, who gets paid, and who gets punished when something breaks.

To see where restaking sits, walk up the stack one rung at a time. At the bottom is base staking: you lock 32 ETH, run a validator, help produce and attest to blocks, and collect a reward of roughly 3 percent a year plus a little extra from transaction tips and maximal extractable value. If your validator misbehaves or goes offline at the wrong moment, the network slashes part of your stake. That is the original bargain: put capital at risk, follow the rules, get paid.

The next rung is liquid staking. Instead of locking 32 ETH yourself, you deposit any amount with a protocol like Lido or ether.fi and receive a token (stETH, weETH and the like) that represents your staked ETH plus its accruing rewards. The token stays liquid, so you can trade it, lend it or use it as collateral while the ETH underneath keeps earning. Liquid staking is now the dominant way ETH is staked, and crucially, it is also the business that still earns real fees.

Restaking adds a third rung. You take staked ETH, or a liquid-staking token, and pledge it again through a restaking protocol to back one or more extra services. Those services pay you for the privilege of borrowing Ethereum’s security, and in return you agree that their rules can slash your stake too. Then comes a fourth rung, liquid restaking, which issues yet another token against the restaked position so it can be reused in DeFi. Each rung promises a little more yield. Each rung also stacks another layer of conditions under which your ETH can be cut.

Read the right-hand column from top to bottom and the design tension jumps out. The reward grows by thin increments as you climb, while the list of things that can destroy your principal grows with every step. That asymmetry is fine if the extra rewards are large and reliable. In 2026 they turned out to be neither.

A marketplace for trust: the supply side and the demand side

The cleanest way to understand restaking is to treat it as a marketplace, because that is exactly what EigenLayer set out to build. On one side are suppliers: people with staked ETH who are willing to accept extra slashing risk in return for extra pay. On the other side are buyers: new protocols that need a large, credibly neutral set of validators to vouch for them but cannot bootstrap that security on their own. Restaking is the exchange that is supposed to match them.

The buyers in this market are called Actively Validated Services, or AVS. An AVS is any system that needs its own decentralized validation but would rather rent Ethereum’s than build its own from scratch. The promise to a new oracle network or a cross-chain bridge is seductive: instead of launching a token, convincing hundreds of operators to stake it, and praying the token holds value long enough to deter attackers, you simply plug into a pool of restaked ETH that is already worth billions and already run by professionals.

This is also the part that Ethereum’s own co-founder warned about before the first dollar was ever restaked. In a May 2023 essay titled Don’t overload Ethereum’s consensus, Vitalik Buterin cautioned that bolting extra duties onto the validator set is not free: “Any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator.” The marketplace framing is useful precisely because it keeps that warning in view. Every unit of security sold is a unit of extra risk bought by the people who stake. The question was always whether enough buyers would pay enough to make that risk worthwhile.

What an AVS actually is, with the real examples

The abstraction becomes concrete once you look at the services that actually launched. EigenLayer’s mainnet opened to external services in the first half of 2024, and the first wave included recognizable names, documented in EigenLabs’ own AVS launch post and catalogued since by trackers like OnchainTimes. They fall into a handful of buckets, and each bucket answers the same question: what kind of guarantee is this service trying to buy?

The flagship is EigenDA, a data-availability layer that lets rollups publish their data cheaply while leaning on restaked ETH to guarantee that the data was really made available. Then there are oracle networks such as eOracle, which deliver price feeds and other outside data on-chain; zero-knowledge and state-proof services like Lagrange and Brevis, which generate verifiable proofs about chain state; interoperability layers like Omni Network, which route messages between rollups; and fast-finality or sequencing services like AltLayer’s Mach and Witness Chain. Oracles are the most instructive case, because an oracle that can be bribed or manipulated is worse than no oracle at all, a failure mode we have watched play out repeatedly, most recently when oracle manipulation went multi-chain. Shared security is meant to raise the cost of exactly that attack.

Here is the catch that the marketplace framing exposes. EigenLayer has advertised a sprawling ecosystem, with OnchainTimes counting more than 190 partners and around 40 services described as live, yet on-chain metrics around the time deposits crossed 5 million ETH pointed to roughly 18 networks actually operating. The gap between a partner, a live service, and a service that pays real money for its security is the gap that eventually broke the business. Plenty of projects were happy to accept subsidized security while the subsidies lasted. Very few were willing to pay market price for it afterward.

Why the supply side ballooned: points, airdrops and a reflexive loop

If demand was always thin, why did 5 million ETH show up to supply it? The honest answer is that most of that capital did not arrive because an oracle network was paying a handsome fee. It arrived chasing points and airdrops, which is a very different thing.

During 2024 and into 2025, EigenLayer ran a points program that rewarded early depositors with a claim on a future token. Liquid restaking protocols then competed to capture those deposits by stacking incentives on top: deposit with ether.fi, Renzo, Kelp or Puffer, the pitch went, and you would earn EigenLayer points, the protocol’s own points, and often a third layer of partner rewards, all at once. A point is really a loan a depositor makes to a protocol against the hope of a token that does not exist yet. For a while the loan looked free, so capital piled in, and the piling-in pushed implied valuations higher, which attracted more capital. That reflexive loop, not fee income, is what inflated the supply side.

Then the tokens listed and the loop ran in reverse. EIGEN, EigenLayer’s token, trades at about $0.2514 as of early October 2026, down roughly 95.5 percent from its December 2024 high of $5.65, with a market capitalization near $244 million, according to CoinGecko. The point of a points program is that it eventually has to settle in something real, and when it settled, the something real was a token that had already given back almost everything.

The protocol has tried to make its token capture value from actual usage. A December 2025 tokenomics change, ELIP-012, imposes a 20 percent fee on AVS rewards that are subsidized by EIGEN incentives and routes all EigenCloud service fees into buying back EIGEN, as reported when the Eigen Foundation outlined bigger rewards for active users. The logic is sound. The arithmetic is not yet there: when the fees being taxed are measured in tens of thousands of dollars a week and fresh tokens keep unlocking on a regular schedule, so the circulating supply keeps climbing, a buyback funded by those fees cannot move the price. Value capture requires value, and the value was supposed to come from the demand side.

Why the demand side never arrived: the fee scoreboard

Return to the number this guide opened with. In the last week of September 2026, the restaking category held about $10.02 billion and generated $99,977 in fees, while liquid staking held $51.87 billion and generated $27.35 million, per the Blockhead tally. Normalize for size and restaking earned roughly $10,000 per billion dollars secured per week; liquid staking earned roughly $527,000. The product that was supposed to be staking with a turbocharger earned about one fifty-third as much per dollar.

The bottom row is as telling as the headline. The five largest liquid restaking tokens left standing, Renzo, Kelp, Swell, Puffer Finance and Bedrock, made a combined $953,350 in gross profit across the entire second quarter of 2026, according to the same Blockhead analysis. That is gross profit, before most real costs, for an entire quarter, for the five biggest names in the business. It is the kind of figure a single mid-sized DeFi lender can clear in a good week.

Why would demand be so weak when the pitch sounded so strong? Three reasons, and they compound. First, security is a cost that every buyer tries to minimize, not a product they want more of; an AVS that can get adequate protection for less will always take the cheaper option. Second, many services discovered they did not need Ethereum-grade security at all, or could bootstrap enough with their own token and a smaller operator set. Third, paying for restaked ETH means paying in real fees, and in a market where most AVS were themselves pre-revenue, there was little real money to pay with. The subsidies came from token emissions, and once emissions slowed, so did the willingness to keep the lights on.

It is worth remembering that EigenLayer’s own founder framed the product modestly from the start. In a September 2023 interview with CoinDesk, Sreeram Kannan said: “Anything that restaking can do, already liquid staking can do, so I view restaking as a lesser risk than liquid staking.” Read in 2026, that line lands differently than it did in 2023. If liquid staking can do what restaking does, and liquid staking is where the fees actually are, the market may simply have agreed with him and stayed one rung down the ladder.

Liquid restaking tokens: leverage on a yield that mostly is not there

Most people who restaked never touched EigenLayer directly. They used a liquid restaking token, or LRT. The mechanics mirror liquid staking one rung up: you deposit ETH or a liquid-staking token with a protocol such as ether.fi (eETH), Renzo (ezETH), Kelp (rsETH) or Puffer (pufETH); the protocol restakes it across a basket of services and hands you a token that represents the restaked position plus whatever rewards it earns. You keep something liquid and tradable while, in theory, three or four income streams accrue underneath.

The appeal and the danger are the same word: reuse. Because an LRT is liquid, you can post it as collateral to borrow against, deposit it in a lending market, or farm it somewhere else. Every reuse is rehypothecation, the same staked ETH backing several obligations at once. When everything holds, the capital efficiency looks miraculous. When one link breaks, the breakage travels along every chain the token was reused in. The sector had an early taste of the fragility in April 2024, when Renzo’s ezETH briefly traded well below its intended value and liquidations cascaded through leveraged positions.

The clearest signal of where the market landed came from the LRT leader itself. In August 2026, ether.fi stripped restaking out of weETH, its flagship token, reverting weETH to a plain liquid-staking token and isolating restaking in a separate, opt-in token called weETHs, built not on EigenLayer but on the competing Symbiotic protocol, as The Defiant reported. The company that had made restaking mainstream quietly gave its biggest product an exit door, and most users walked through it. By late 2026 the restaking-flavored token was a rounding error next to the plain liquid-staking one.

The Kelp and Aave lesson: risk you import without pricing it

If you want a single event that explains why stakers grew wary, it happened in April 2026. Attackers exploited a cross-chain bridge used by Kelp DAO and minted about 116,500 rsETH, worth roughly $292 million, out of thin air, then moved that freshly minted collateral into the Aave lending market and borrowed against it. Aave was left with roughly $196 million in bad debt, and its total value locked fell by billions over a single weekend, as CoinDesk documented. The underlying rsETH deposits were never stolen; the damage came from an inflated supply of a token that other protocols had already agreed to treat as money.

That is the precise danger of rehypothecation. A lending market accepted a liquid restaking token as collateral without fully pricing the risk of the bridge that token depended on. When the bridge broke, the loss did not stay with Kelp; it traveled straight into Aave, whose users had never knowingly taken on bridge risk at all. Using an LRT to borrow is structurally close to using volatile collateral to mint your own dollar, and it carries the same lesson: the quality of what you can borrow is only ever as good as the weakest assumption baked into the collateral.

The ending was better than it might have been. Aave froze rsETH’s borrowing power, and a coalition of DeFi protocols, with Aave founder Stani Kulechov personally pledging 5,000 ETH, worked through the following weeks to burn the illegitimate tokens and restore full backing, a recovery CoinDesk tracked at the time. But the memory stuck. A product class whose whole appeal was reuse had shown that reuse is also how a single failure becomes everyone’s failure.

Slashing: the cost side is very real even when the yield is not

Restaking’s extra yield is optional and, as the fee data shows, thin. Its extra risk is neither. EigenLayer turned on slashing for opted-in services in 2025, described in its own slashing goes live post, which means the penalties stopped being theoretical. When you restake to an AVS, you are agreeing that the AVS’s rules can order a portion of your stake destroyed if its operators misbehave by its definition of misbehavior, a definition Ethereum’s core protocol knows nothing about.

The scenario that worries careful operators is correlated slashing. Picture a single operator running validators for many services, or a bug in a widely used AVS client, or an upgrade that quietly changes a slashing condition. Any of these can trigger penalties across a large slice of the restaked set at once, and because the same ETH may sit behind several services, one incident can compound. This is the practical form of Buterin’s original warning: every extra duty is an extra way for the validator set to be punished for something that has nothing to do with securing Ethereum itself.

A small cover-and-insurance niche grew up to price this tail, with providers like Nexus Mutual offering on-chain protection and some of the newer restaking platforms courting underwriters to supply reinsurance capacity. The existence of that niche is itself a verdict: when a market has to sell insurance against its core product’s accidents, it is telling you the accidents are plausible enough to price. The uncomfortable summary is that a restaker in 2026 was underwriting a real, correlated tail risk in exchange for a fee stream that, sector-wide, rounded to a few hundred thousand dollars a quarter.

ether.fi rings the bell: the supplier that quit

In a two-sided market, the loudest signal is a big supplier packing up. That is what ether.fi did. Having stripped restaking out of weETH in August, the firm confirmed in late September that it would remove its remaining structural ties to EigenLayer by the end of the third quarter of 2026, with fewer than 1 percent of its assets still restaked and its EigenPod withdrawal credentials due to be removed by the fourth quarter, per Blockhead. As this guide publishes, that quarter has closed.

Chief executive Mike Silagadze was blunt about why, in comments carried by Blockhead: “There were no meaningful yield opportunities in restaking and there was some perceived risk from stakers, so we decided it made sense to exit.” It is a line that doubles as an epitaph for the entire points era. Yield too thin, risk too real, so the supplier left. It is hard to describe the fee scoreboard above more efficiently than that.

What makes the retreat sting is where ether.fi went instead. The firm leaned into a consumer-finance pivot, adding tokenized stocks and metals, portfolio loans and payment cards, a strategy reported across the summer. Its token tells the story in miniature: ETHFI trades around $0.7222, up almost 7 percent on the week but still down about 91.5 percent from its 2024 peak, with a market capitalization near $697 million, according to CoinGecko. Whatever strength the token has left rests on the banking app, not on restaking. The brand that personified restaking now earns its keep selling cards and credit.

It is worth separating the token bounce from the business. ETHFI, EIGEN and ETH itself all drifted higher in parts of late 2026, but that was the broad market moving, not restaking demand returning. ETH trades near $2,697.99, up only about 1.6 percent on the week and still down more than 45 percent from its own all-time high, per CoinGecko. When the whole asset class ticks up, restaking tokens tick up with it; that is beta, not a recovery in the thing those tokens are supposed to represent.

The platforms, sorted

Restaking is not one protocol, and the platforms have aged very differently. The spread between them is itself informative: the ones leaning hardest on a token have struggled most, while the ones selling plumbing without a token, or securing a different asset entirely, have held up better. EigenLayer itself now secures about $5.10 billion, down from a $22.06 billion peak reached in August 2025, according to Blockhead.

Symbiotic is the quiet winner of the unwinding, which is ironic, because it never shipped the thing everyone else sold. It has no public token; it raised an early round led by Pantera Capital; and it lets almost any ERC-20 be used as restaking collateral rather than funneling everyone into one asset. When ether.fi needed somewhere to park the restaking it was pulling out of EigenLayer, it chose Symbiotic. Co-founder Misha Putiatin framed the firm’s mid-2026 pivot toward collateral markets this way in a company statement: “Collateral markets are not a new financial idea. What is new is making them composable, enforceable, and capital-efficient onchain.” The absence of a token looks less like a missing feature and more like a deliberate refusal to let speculation front-run the product.

Babylon is the genuine counter-example, and it matters because it attacks a different asset. Instead of restaking ETH, Babylon lets Bitcoin holders lock coins using Bitcoin’s own timelock scripting, with no wrapping and no bridge, and uses that locked BTC to secure proof-of-stake chains that opt in, as described on Babylon’s own site. Its value secured has held in the billions even as its BABY token sank more than 90 percent from its 2025 high, a cleaner adoption-versus-price divergence than anything on the Ethereum side. Karak, the third ETH-side platform, has drifted away from restaking framing entirely toward an execution-layer product called OpenGDP. The pattern across all four is consistent: the less a platform depended on selling a token into the points mania, the better it reads today.

EigenCloud: when the marketplace becomes its own customer

Faced with a demand side that would not grow, EigenLayer did something telling: it decided to become the demand itself. The protocol rebranded to EigenCloud and recast its mission around a verifiable cloud, a suite of services that consume restaked security rather than wait for outsiders to. The pillars are EigenDA for data availability, EigenCompute for confidential off-chain computation, and EigenAI for verifiable machine-learning inference, with the whole stack pitched as a way to run ordinary software with crypto-grade guarantees. Andreessen Horowitz backed the direction with a $70 million token purchase to support the launch, as CoinDesk reported.

Through the lens of this article, the pivot is a fascinating admission. If the open marketplace for rented security does not attract enough buyers, the operator builds its own flagship tenants and rents to them. That is a reasonable strategy, and verifiable compute for AI is a real and growing field, one that overlaps with the decentralized-compute race we covered in Akash versus io.net versus Render. But it is a different business from the original one. Restaking promised a neutral exchange where many services would compete to buy Ethereum’s trust. EigenCloud is closer to a vertically integrated cloud provider that happens to settle its guarantees on-chain. The security is still real; the two-sided market is quietly becoming one-sided, with the house supplying both the security and most of the demand for it.

This also reframes the ELIP-012 buyback. Routing all EigenCloud fees into EIGEN purchases only matters if EigenCloud earns real fees, which puts the token’s future squarely on the success of the compute pivot rather than on the restaking marketplace it was named for. The token is now a bet on a cloud startup with a crypto settlement layer, not on a thriving bazaar of AVS.

What United States regulators have said, and what they have not

Regulation did not kill restaking; economics did. But anyone staking or restaking from the United States should understand where the Securities and Exchange Commission has drawn its lines, because those lines shape what a compliant product can even look like. In 2025 the SEC’s Division of Corporation Finance issued two staff statements that together carved out a cautious safe zone. The first, on certain protocol staking activities in May, said that staking which is administrative or ministerial does not, by itself, involve the offer and sale of a security. The second, on certain liquid staking activities in August, extended similar comfort to liquid staking, provided the provider stays in a ministerial role and does not promise or set returns.

Two caveats matter for restaking specifically. First, the comfort narrows as the product grows more active: the moment a provider exercises discretion over whether, when or how much to stake, or guarantees a return, it steps outside the described safe zone, and restaking plus liquid restaking is nothing if not a stack of discretionary choices. Second, these are staff statements, not rules; they reflect a posture that can shift with the people in the building, a dynamic we traced in SEC crypto enforcement in 2026. A reward-bearing restaking token that markets a yield is a far less comfortable fit for a carve-out written around passive, ministerial activity than a plain validator receipt is. The regulatory path exists; it runs through staying boring, and restaking is not boring.

What to watch next, and whether restaking comes back

The case for a revival is not empty, and it is worth stating fairly. Silagadze, even while leaving, told The Defiant he still thinks “restaking will come back in one form or another,” and that it was simply “a bit too early.” Shared security is a genuinely good idea; the failure was one of timing and pricing, not of logic. So the signals to watch are concrete ones about demand, not about token prices.

Watch for a single AVS that pays real, recurring, market-rate fees for restaked security, the way a tenant pays rent, rather than subsidizing it with its own emissions. Watch whether EigenCloud’s compute and AI services generate enough revenue to make the buyback arithmetic bite. Watch the macro backdrop, because when short-term dollar yields sit near 4 percent, a thin and risky on-chain premium has stiff competition, and the next Federal Reserve meeting on 28 October 2026 will nudge that hurdle one way or the other. And watch Bitcoin: if Babylon’s model proves that locked BTC can secure real services, the most interesting restaking story of the next cycle may not involve Ethereum at all.

For now, the scoreboard is the honest summary. Restaking in 2026 is an elegant machine that produces an abundant supply of security and almost no paying demand for it. The plumbing works, the slashing is live, the capital is parked, and the fees are a rounding error. Until a buyer shows up willing to pay rent, restaking is a warehouse full of trust that very few people want to borrow.

Frequently Asked Questions

What is restaking in crypto?

Restaking is re-pledging ETH that is already staked so it can also help secure a second service, such as an oracle, bridge or data-availability layer, in return for extra rewards. The trade-off is that the restaker takes on extra slashing risk: the second service can penalize the stake under its own rules, on top of Ethereum’s.

Is restaking still worth it in 2026?

For most holders the extra yield is hard to justify. Across the whole restaking category, weekly fees were under $100,000 in late September 2026 while liquid staking earned about $27.35 million, so the premium for taking on restaking’s added risk has largely disappeared. With short-term dollar yields near 4 percent, a thin and slashable on-chain premium faces stiff competition.

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

Staking locks ETH to run or back a validator and earns the base reward of roughly 3 percent. Liquid staking does the same through a protocol and gives you a tradable token like stETH or weETH that stays usable in DeFi. Restaking goes one step further by re-pledging that staked ETH to secure additional services for extra rewards and extra slashing risk.

What is an AVS (Actively Validated Service)?

An AVS is any service that rents its security from a restaking protocol instead of building its own validator set. Examples include data-availability layers like EigenDA, oracle networks like eOracle, and interoperability layers like Omni Network. The idea is that a new service can borrow Ethereum’s economic security rather than bootstrap a token and operator set from scratch.

Why is ether.fi leaving EigenLayer?

ether.fi, the largest liquid restaking provider, decided the returns from restaking were too small to justify the risk its users were taking. Its chief executive said there were no meaningful yield opportunities in restaking and some perceived risk, so exiting made sense. The firm reverted its main weETH token to plain liquid staking and shifted its focus to consumer finance products like cards and loans.

By Yuki Tanaka, staff writer at HOGE Wire covering Ethereum, staking and on-chain infrastructure.