Restaking’s 2026 Paradox: Shared Security, Two Operators

Restaking was supposed to be Ethereum’s great multiplier. Pledge your staked ETH a second time, the pitch went, and you could rent the same economic trust out to secure oracles, bridges, rollups and data layers, earning extra yield while spreading Ethereum’s security across a whole new economy of services. Two years on, the arithmetic has run the other way. The capital that was meant to fan out across hundreds of decentralized services has drained away and bunched up instead, and by late October 2026 the slice of Ethereum that is still restaked natively sits mostly with two companies.

That is the paradox at the center of restaking’s 2026. A design sold on decentralization and permissionless participation ended the year more concentrated than the staking it was built on top of. The timing sharpens the point. With the Federal Reserve widely expected to hold its target range at 3.75% to 4.00% at its October 28 meeting and another hike penciled in for December, the risk-free rate that restaking’s thin premium has to beat is sitting near 4%. Against that backdrop, paying a slashing-exposed service to earn a fraction of a percent extra has stopped making sense to almost everyone.

What restaking promised: renting Ethereum’s trust

Start with the base. When you stake 32 ETH to run an Ethereum validator, you lock capital that can be slashed if the validator misbehaves, and in exchange you earn a protocol reward of a little under 3% a year. Liquid staking tokens, such as Lido’s stETH or Rocket Pool’s rETH, hand you a tradable receipt for that staked position so your capital keeps working in DeFi while it secures the chain.

Restaking adds a floor beneath the building. Coined by EigenLayer, it lets you take staked ETH, or a liquid staking token, and pledge it a second time to secure third-party services known as Actively Validated Services, or AVS: data-availability layers, oracle networks, cross-chain bridges, rollup sequencers and the like. Each AVS you opt into pays an extra reward, and each adds its own slashing conditions. Liquid restaking tokens (LRTs) then wrap the restaked position in yet another tradable receipt, so even twice-pledged capital can be recycled through lending markets. We walked through the full ladder in our companion guide to renting Ethereum’s trust. The short version is a stack of claims on the same underlying ETH, each rung promising a little more yield for a little more risk.

There are two ways onto that floor, and the difference matters for how hard it is to leave. Native restaking points an Ethereum validator’s withdrawal credentials at an EigenLayer contract called an EigenPod, so the staked ETH itself backs the AVS. LST restaking instead deposits a liquid staking token such as stETH. Native restaking is the deeper commitment, which is why the native figure is the truest gauge of conviction, and why unwinding it, as the rest of this piece shows, turned out to be the slow part.

The idea had serious backing and a serious critic from the start. Sreeram Kannan, the founder of EigenLayer’s parent Eigen Labs, argued the risk was overstated. “Anything that restaking can do, already liquid staking can do, so I view restaking as a lesser risk than liquid staking,” he told CoinDesk in 2023, reasoning that a staking receipt can already be pushed into risky venues, so restaking merely organizes an exposure that exists anyway.

Vitalik Buterin saw it differently. In an essay warning against overloading Ethereum’s consensus, he cautioned that “any expansion of the ‘duties’ of Ethereum’s consensus increases the costs, complexities and risks of running a validator,” and worried that restaking could pressure the base layer into bailing out services that lean on it, per his May 2023 post. The 2026 story is less about either prediction coming true than about a third possibility neither dwelt on: that the demand to rent all this trust would simply never show up.

The AVS economy, and the demand that never came

For the ladder to pay, somebody has to rent the security. That somebody is the AVS, and this is where restaking’s story breaks. EigenLayer still lists well over 130 services in development, and its own data-availability layer, EigenDA, runs real throughput of around 15 MB/s. But a service in development is not a paying customer. Through 2026 the count of AVS actually paying meaningful fees for restaked security stayed small, and an October review of the ecosystem found no new high-profile AVS had come online that month to change the picture, per Baltex. Supply of rentable trust ballooned; demand for it did not.

The services do exist, and some are genuinely useful. EigenDA posts data for rollups, eOracle runs a restaked oracle network, Lagrange and Brevis sell zero-knowledge coprocessing, Omni Network routes messages between rollups, and AltLayer’s Mach offers fast finality. The problem is not the absence of AVS; it is that securing them is cheap to the point of being nearly free. An oracle or a data layer needs a credible slashing threat, not a vault the size of a national bank, so even a healthy roster of services throws off only a trickle of fees next to the tens of billions that chased points into the system. A two-sided market clears only when both sides are roughly matched, and restaking spent 2026 with a warehouse full of supply and a thin line of buyers.

So why did supply balloon at all if nobody was paying? Points. Through 2024 and into 2025, EigenLayer and the LRTs stacked on top of it handed out points, speculative claims on future token airdrops, to anyone who deposited. Depositors farmed EigenLayer points, then LRTs issued their own points on top of those, and capital poured in chasing tokens that did not yet trade. It was a loan taken against a future valuation. When EIGEN finally listed and then fell more than 95% from its peak, the loan came due in reverse, and the deposits that had chased points had no fee income to hold them in place.

The great unwinding, measured

The reversal is stark in the native numbers. According to an on-chain investigation by Bitquery, native ETH restaked on EigenLayer peaked near 3.5 million around July 2025 and had fallen to roughly 643,000 by September 30, 2026, a drop of about 82%. Counting liquid staking tokens as well as native ETH, EigenLayer’s total value locked sits around $5.1 billion against a peak of $22.06 billion in August 2025, per figures compiled by Blockhead.

One depositor drove most of the exit. On March 30, 2026, ether.fi’s node firms held 71% of all native restaked ETH, 2.26 million of 3.18 million. Over the following 78 days, from April 21 to July 8, ether.fi pulled 2.25 million ETH back out, with 745,594 ETH yanked from three node firms in under two hours on the night of July 7. When the largest customer is also the largest supplier, its decision to leave is not a data point; it is the market.

The exit came in waves rather than one dump, which is part of why the full picture took months to register. ether.fi trimmed in April and May, added some stake back in June, then pulled the bulk out in early July, a cadence that let headline TVL drift down rather than crater in a day. By the time observers tallied it, native restaked ETH had already fallen by four-fifths, and the services still depending on EigenLayer were secured by a fraction of the capital they launched with.

The centralization paradox: what is left sits on two operators

Here is the part that should give restaking’s designers pause. Of the roughly 643,000 native ETH still restaked on September 30, Bitquery found that Kraken’s node operation held 43% and P2P.org held 18%. Two operators, in other words, back more than six out of every ten ETH that still secures EigenLayer’s AVS. The number of node firms holding at least 1,000 restaked ETH fell from 46 in March to 27 by late September. The remnant did not just shrink; it clustered.

Restaking was pitched as the opposite of this. The promise was permissionless, decentralized shared security: many operators, many services, trust spread thin enough that no single failure could cascade. What the great unwinding left behind is the inverse. The capital that stayed is concentrated in a handful of professional operators, which means the security of every AVS still leaning on EigenLayer now depends on two companies continuing to show up. The decentralization restaking was built to deliver has quietly become its own single point of failure. For anyone who thinks hard about where risk actually lives in crypto, the lesson rhymes with the one we drew about following the chokepoints: control pools at the narrow points, whatever the marketing on the label says.

Concentration is not automatically catastrophe. Kraken and P2P.org are among the most professional operators in the business, and a smaller, better-capitalized set is in some ways easier to hold accountable than a sprawling long tail. But it is the precise outcome the architecture was supposed to prevent, and it changes the risk calculus for anyone relying on an AVS: a correlated failure, a compromised key or a regulatory action aimed at one large operator now carries far more of the system than it would have when the base was broad.

It helps to picture what an operator actually is. A node operator runs the servers that sign for a validator and opts that stake into AVS on the restaker’s behalf; the restaker supplies capital, the operator supplies uptime and keys. When two firms run the validators behind most of the remaining restaked ETH, a single misconfiguration, a compromised signing key or a coordinated outage stops being one operator’s problem and becomes the shared-security layer’s problem. Spreading trust was the entire point; the remnant does the reverse, pooling it exactly where a correlated failure would hurt most.

Kraken retires the product it dominates

The concentration comes with a twist that is hard to read as anything but a verdict. Kraken, the single largest remaining restaking operator, is closing its retail restaking product. The exchange’s own support page states plainly that it is retiring the ETH restaking strategy and automatically moving customers’ restaked ETH into bonded ETH staking, with no action required, starting with Australia and the United Arab Emirates on October 21 and other regions to follow. The reason it gives is blunt: EigenLayer has ended EIGEN rewards for restaked ETH, so the strategy no longer earns anything extra, per Kraken’s notice.

Sit with that for a moment. The company that holds the largest share of what is left of native restaking is simultaneously telling retail customers the product is not worth keeping. The exchange-and-wallet layer, where most ordinary users actually touch staking as custody and trading converge, has concluded that restaking yield no longer pays for the complexity. When the biggest operator and the main distribution channel both step back at once, the concentration that remains starts to look less like a vote of confidence and more like a measure of how few participants are left to spread the load.

Why the premium melted: a 4% risk-free rate

None of this happened in a vacuum. Restaking’s entire value proposition was a premium: a point or two of extra yield on top of base ETH staking, in return for taking on extra slashing risk. That premium only matters if the alternatives pay less. In 2024, when dollars earned little and on-chain yield was scarce, an extra point or two looked worth chasing. In late 2026 it does not.

The Federal Reserve raised its target range to 3.75% to 4.00% in September, its first hike since 2023, under hawkish chair Kevin Warsh. Markets expect the Fed to hold at its October 28 meeting, with prediction venues pricing a hold near 93%, then to hike again in December, where the odds sit around three in four, per DeFi Rate and a central-bank preview from ING. A poll of economists found nearly all of them expecting a hold this month and a move in December. We mapped that calendar and what it means for prices in our year-end outlook.

With the risk-free rate near 4%, restaking’s thin premium has nowhere to hide. Worse for restaking, that risk-free rate now exists on-chain with none of the slashing. Tokenized US Treasury products hold around $14.9 billion and pay roughly 3.6%, and Sky’s on-chain savings rate pays about the same, both composable in DeFi the way a liquid restaking token is but without an AVS slashing condition bolted on, per rwa.xyz. When you can earn close to 4% on tokenized T-bills with no slashing tail, nobody pays a premium to rent out slashing-exposed ETH for a fraction of a point more. The demand death that the deposit charts show is, at bottom, a cost-of-capital story.

The money question: $99,977 a week

Follow the fees and the collapse stops being a matter of sentiment. Restaking as a category held about $10.02 billion in value in early September yet generated just $99,977 in fees over a week, according to data from DefiLlama reported by Blockhead. Plain liquid staking, with $51.87 billion locked, earned $27.35 million in the same week. Per dollar secured, ordinary staking out-earned restaking by roughly 53 times.

Put the weekly figure on an annual footing and the mismatch is almost comic. Restaking’s $99,977 a week annualizes to a little over $5 million a year against more than $10 billion in value secured, a gross yield on the order of five hundredths of a percent before any costs. That is the return a restaker is asked to accept a stacked slashing risk to capture. Set beside a tokenized Treasury paying close to 4% with no slashing at all, the trade does not survive contact with a spreadsheet.

The LRT issuers that were supposed to be the industry’s profit engine tell the same story. The five largest remaining ones, Renzo, Kelp, Swell, Puffer and Bedrock, booked a combined $953,350 in gross profit in the second quarter of 2026, down from $2.18 million three quarters earlier. For a sector that once carried tens of billions in deposits, that is rounding error. The twice-pledged ETH was doing the work; it simply was not getting paid for it.

Liquid restaking tokens and the Kelp lesson

If the reward side of LRTs disappointed, the risk side delivered at least once. Liquid restaking tokens let the same ETH appear in several places at once: staked, restaked, and posted as collateral in a lending market. That rehypothecation is efficient when everything holds and dangerous the moment one link breaks.

It broke in April 2026. Attackers exploited Kelp DAO’s cross-chain bridge to mint roughly 116,500 rsETH out of thin air, around $292 million, then used it as collateral on Aave to borrow against value that did not exist, leaving close to $196 million in bad debt before the position was frozen, per CoinDesk. The underlying deposits were never drained, and a coalition of DeFi protocols restored the backing over the following weeks, but the episode showed how an LRT accepted as collateral imports the risk of every bridge and contract standing behind it. The pattern will be familiar to anyone who read our breakdown of what 830 bridge bugs reveal: the weak point is rarely the base chain, it is the plumbing bolted to the side of it.

ether.fi walks away

The clearest verdict came from the company with the most to lose by delivering it. Through 2026 ether.fi unwound restaking from weETH, its flagship LRT, reverting it to a plain liquid staking token and isolating restaking in a separate, opt-in token, weETHs, built not on EigenLayer but on the rival Symbiotic. By late 2026 less than 1% of ether.fi’s assets were still restaked, and the firm said it would remove its EigenPod withdrawal credentials entirely by year-end, per The Defiant.

Mike Silagadze, ether.fi’s founder and chief executive, was unsentimental and sentimental at once. “End of an era. Sad,” he wrote, adding that he still thought restaking would return “in one form or another” and had simply arrived too early. On the economics he was direct: “There were no meaningful yield opportunities in restaking and there was some perceived risk from stakers, so we decided it made sense to exit,” he told CoinDesk, as reported by Blockhead. The market has rewarded the pivot: ETHFI trades around $0.65, roughly 92% below its peak but far above where EIGEN sits near $0.19 and down more than 96%, because investors are now paying ether.fi for its move into neobank-style cards and tokenized assets, not for restaking, per CoinGecko.

The pivot is not cosmetic. ether.fi now leans on card spending, tokenized assets and lending rather than restaking fees, and card fees alone climbed from 17% of monthly revenue in January to 46% by July, with the firm saying the new lines had fully replaced the revenue lost from restaking and a lower ETH price, per Blockhead. The brand that once stood for restaking now earns its keep looking like a bank. It is the loudest signal in the sector: the company best placed to make restaking pay chose to become something else.

The survivors, and EigenLayer’s new bet

Not everyone in the category is shrinking the same way, and the differences are instructive. Symbiotic, EigenLayer’s main competitor, never launched a public token and in mid-2026 pivoted its Core V2 toward composable collateral markets. “Collateral markets are not a new financial idea,” co-founder Misha Putiatin said on the release. “What is new is making them composable, enforceable, and capital-efficient onchain,” per GlobeNewswire. Having no token means no speculative wrapper to collapse, and it was Symbiotic, not EigenLayer, that caught ether.fi’s restaking flows when they moved.

Babylon took a different route entirely, letting Bitcoin holders lock BTC through Bitcoin’s own scripting to secure proof-of-stake chains without wrapping or bridging. Its total value locked, around $3 billion by DefiLlama’s count, makes it the second-largest restaking venue, even though its BABY token sits more than 90% below its peak, a reminder that total value locked and token price measure different things.

Karak, a third venue backed by Andalusia Labs, offers multi-asset restaking but holds only a minor share of what remains. The pattern across the survivors is worth stating plainly: the platforms that never launched a speculative token, Symbiotic and Karak, have no wrapper to collapse, while the two that did, EigenLayer and Babylon, carry tokens down more than 90% even where the underlying network kept growing. Babylon is steering its Bitcoin collateral toward lending, with a native-BTC integration for Aave that reached testnet in 2026, but like much of the sector that demand is still a promise rather than a receipt.

EigenLayer itself has stopped waiting for AVS demand and started manufacturing its own. Rebranded EigenCloud, it is pushing verifiable cloud services: EigenDA for data, and EigenCompute and EigenAI for verifiable off-chain computation pitched at AI agents that need to prove what they ran. a16z backed the pivot with a $70 million token purchase in 2025, per CoinDesk. The wager is that restaked ETH finds its demand not from a thin market of external AVS but from a verifiable-compute business EigenLayer builds and consumes itself, a cousin of the on-chain compute economies we tracked in our decentralized inference scoreboard. It is a reasonable pivot and an admission at the same time: if the open market will not rent your security, you become your own tenant.

Slashing, exits, and the friction of leaving

For all the talk of slashing risk, the catastrophic cascade the skeptics feared has not arrived, partly because so little is left to cascade. EigenLayer turned on opt-in slashing on mainnet in April 2025, and the risk is real on paper: an operator that misbehaves across several AVS can be penalized by each. But with demand thin and the base shrinking, the more pressing friction in 2026 has been getting out, not getting punished.

Leaving native restaking is slow. Rotating or exiting a validator means queueing on Ethereum’s beacon chain, where the entry side has stretched to weeks at a time, and EigenPods keep enforcing restaking accounting even after a withdrawal is queued. A draft governance proposal, ELIP-018, nicknamed RETIRE for Retirement Enabling Terminal, Irreversible Restaking Exit, would let a pod owner permanently disable restaking and reclaim stranded ETH, per the EigenLayer governance forum. That a protocol built to attract deposits is now drafting a clean exit door tells you which way the capital is flowing. The slashing and audit risk that made restaking worth scrutinizing in the first place is the same reason security auditors spent the cycle hardening the tooling around it, from formal verification to slashing simulators.

How the SEC sees restaking

The regulatory picture in the United States softened over 2025 in a way that matters here. The SEC’s Division of Corporation Finance said in a May 29, 2025 staff statement that protocol staking is an administrative or ministerial activity, not a securities transaction under the Howey test, and followed on August 5 with similar comfort for certain liquid staking arrangements. The carve-out is narrow. It leans on the activity staying administrative, and it explicitly steps around arrangements where a provider guarantees a return or where the asset is engineered to throw off passive yield of its own.

That narrowness is exactly where restaking gets awkward. A pure, non-custodial interaction with a decentralized protocol looks like the activity the staff described. A custodial, yield-promising restaking product offered by an intermediary looks less protected, and a liquid restaking token that markets a stacked yield looks less protected still. The concentration documented earlier cuts both ways for regulators: when most of what is left runs through a few identifiable operators and exchanges, the points a supervisor can actually reach are fewer and clearer. Staff statements are not rules, carry no force of law, and can be withdrawn by a future commission; they describe a posture, not a promise.

What to watch into year-end

Restaking is not dead, but it is being repriced in public, and a handful of signals will tell you whether late 2026 was a bottom or a waypoint.

• The rate path. If the Fed hikes again in December as expected, the risk-free hurdle restaking has to clear only rises, and the premium stays underwater.
• ether.fi’s final cut. Removing its EigenPod credentials completes the largest supplier’s exit and sets a floor you can measure rather than guess.
• A paying AVS. One service that actually pays real fees at scale would be the first genuine demand signal in a year.
• EigenCloud revenue. If restaked ETH is going to find a buyer, the most likely buyer is now the protocol itself, so its verifiable-compute income is the number that matters.
• Operator count. Whether the node-firm base keeps thinning past 27, or stabilizes, decides whether the centralization paradox hardens or eases.

Until one of those turns, restaking in 2026 reads as a market that built an enormous supply of trust for rent and found almost no tenants, then watched what remained gather in the hands of the few operators large enough to stay. The multiplier became a concentrator. Whether the next cycle remembers that as a cautionary tale or a false start depends, as it did at the beginning, on whether anyone ever actually needs to rent the trust.

Frequently Asked Questions

What is restaking in simple terms?

Restaking lets you take ETH that is already staked to secure Ethereum and pledge it a second time to help secure other services, such as oracles, bridges and data layers, in exchange for extra yield and extra slashing risk. EigenLayer introduced the idea, and liquid restaking tokens wrap the position so it can still be used in DeFi.

Why did restaking collapse in 2026?

Demand never matched supply. Very few of the services meant to rent restaked security paid meaningful fees, so the extra yield was tiny while the risk was real. With dollar and on-chain risk-free rates near 4%, the thin premium stopped being worth it, and large depositors such as ether.fi withdrew, cutting native restaked ETH by about 82% from its peak.

Who holds Ethereum’s restaked ETH now?

After the unwinding, the remaining native restaked ETH is concentrated. On-chain data from Bitquery showed that as of September 30, 2026, Kraken held about 43% and P2P.org about 18% of the roughly 643,000 native ETH still restaked on EigenLayer, and the number of sizeable operators fell from 46 in March to 27.

Is restaking still worth it for yield?

For most holders, the math is hard to justify in late 2026. Restaking as a category earned roughly 53 times less in fees per dollar secured than plain liquid staking, and tokenized Treasuries and on-chain savings rates pay close to 4% with no slashing. Kraken is even retiring its retail restaking product and moving customers to ordinary staking.

What is the difference between restaking and liquid staking?

Liquid staking gives you a tradable receipt for ETH staked to secure Ethereum, carrying one slashing condition. Restaking re-pledges that stake to secure additional services, each adding its own slashing condition, in return for extra yield. EigenLayer’s founder has argued restaking is a lesser risk than liquid staking, but in 2026 liquid staking earned far more in fees.

By Yuki Tanaka, DeFi and on-chain correspondent, HOGE Wire. Figures cited were accurate as of October 27, 2026; nothing here is investment advice.