How Does Restaking Yield Work: 3 Layers Explained

The Question: Where Does the 8-12% Restaking Yield Come From?

If you hold Ethereum, you can currently earn somewhere between 3% and 4% annually by staking it through a validator. If you restake that same Ethereum through EigenLayer or Symbiotic, the advertised yield rises to somewhere between 8% and 12%, with occasional spikes higher. The question worth answering is where that additional 4 to 8 percentage points actually come from, whether the mechanism producing it is sustainable, and what happens when one of the layers breaks.

The answer is not simple, because the yield is not a single thing. It is three separate revenue streams stacked on top of one another, each with a different source, a different duration, and a different failure mode. The first layer is base Ethereum staking yield, paid by the protocol itself. The second layer is security payments from Actively Validated Services, or AVSs, which are protocols that hire restakers to secure their infrastructure. The third layer is points programs and token incentives, which are temporary and not linked to any durable cash flow. If you see a headline yield of 12%, you are looking at all three layers combined. If you want to know whether that 12% will still be there in a year, you need to decompose it.

This is not the first time financial markets have attempted to extract additional yield from a single asset by reusing it across multiple obligations. The practice is called rehypothecation, and it was common in the years leading up to 2008. Lehman Brothers was particularly aggressive in rehypothecating client collateral, meaning it borrowed against the same assets multiple times to fund additional trades. The advantage was clear: more capital efficiency, more leverage, higher returns. The disadvantage became clear in September 2008, when Lehman collapsed and the web of overlapping claims on the same collateral could not be untangled. Clients who thought their assets were segregated discovered that those assets had been pledged elsewhere, sometimes multiple times over, and there was not enough left to make everyone whole.

Restaking is not identical to rehypothecation, but the parallels are worth understanding. In both cases, the same underlying asset is used to support multiple obligations. In both cases, the additional yield reflects additional risk. And in both cases, the system works smoothly until something breaks, at which point the cascading failures move faster than the institutions managing them. The question is not whether restaking will produce a Lehman-scale collapse. The question is whether the people deploying capital into restaking understand what they are actually earning, where it comes from, and what would cause each layer to stop paying.

The Three Layers: Base Yield, AVS Payments, and Points

Start with the base layer. When you stake 32 ETH to run an Ethereum validator, you earn rewards for participating in consensus. Those rewards currently amount to roughly 3% to 4% annually, though the precise figure fluctuates with network activity and the total amount of ETH staked. This yield comes from the Ethereum protocol itself. It is paid in the form of issuance (new ETH created to reward validators), priority fees (paid by users who want their transactions included faster), and MEV, or maximal extractable value, which is the profit validators can capture by reordering transactions within a block. These three components are explained in detail in our analysis of LST yield sustainability, and they represent the most durable part of the restaking yield stack. As long as Ethereum continues to function, validators will continue to earn this base yield.

The second layer is AVS payments. AVSs are protocols that need their own security but do not want to bootstrap a new validator set from scratch. Instead, they hire restakers through EigenLayer or Symbiotic, effectively renting the security that Ethereum validators already provide. In exchange, the AVS pays those restakers a fee. That fee is the second layer of yield. It is not paid by the Ethereum protocol. It is paid directly by the AVS in the form of its own token, or occasionally in ETH. The largest AVS by total value locked is EigenDA, a data availability layer that helps rollups store transaction data more cheaply than posting it directly to Ethereum. Other AVSs include EigenAI, which provides verifiable AI inference, and EigenCompute, which handles off-chain execution verification. Each of these services specifies its own security requirements and its own payment terms. The restaker opts into those requirements, accepts the additional slashing risk, and collects the additional yield.

The third layer is points programs and token incentives. As of mid-2026, most AVSs are not yet paying meaningful cash-flow-based rewards. Instead, they are distributing points or tokens as a way to attract early participants. These points often convert into governance tokens or airdrop allocations at some future date, but they do not represent a claim on protocol revenue. This layer of yield is temporary by design. It exists to bootstrap liquidity and participation, and it will disappear once the AVS either achieves sustainable revenue or shuts down. The 12% headline yield you see advertised in 2026 often includes this third layer. Strip it out, and the sustainable yield is closer to 5% or 6%. That distinction matters, because the third layer can vanish overnight if a governance vote decides to end the program or if the tokens being distributed lose value faster than they are earned.

The three layers do not carry the same risk. The base staking yield is the safest, because it is paid by the Ethereum protocol and does not depend on any external party remaining solvent. The AVS payment layer is riskier, because it depends on the AVS continuing to operate and continuing to value the security it is renting. The points layer is the riskiest, because it has no underlying cash flow and no enforceable claim. If you are evaluating a restaking opportunity, the first question to ask is which layers you are actually being paid from, and what would cause each one to stop.

How Rehypothecation Worked, and What Broke It

Rehypothecation was legal, common, and widely understood in the years before the financial crisis. A client would deposit securities with a broker as collateral for a margin loan. The broker would then lend those same securities to another party, using them as collateral for a different loan. That second party might lend them again. Each step in the chain created additional liquidity, additional leverage, and additional yield. The client still owned the securities, at least in principle, and the broker was contractually obligated to return them on demand. But in practice, the securities had been pledged multiple times, and the system depended on the assumption that not everyone would ask for their collateral back at once.

Lehman Brothers rehypothecated more than $50 billion in client assets by mid-2008. When the firm filed for bankruptcy in September of that year, those clients discovered that their collateral had been used to secure other obligations, and there was no orderly way to recover it. The legal process took years. Some clients recovered most of their assets. Others did not. The immediate problem was not that rehypothecation was fraudulent. It was that the system had become so complex, and the same assets had been pledged so many times, that when one institution failed, the entire chain of obligations came under stress simultaneously. The risk that had been distributed across the system suddenly concentrated in a very short period of time, and the institutions managing that risk did not have the capital or the operational capacity to resolve it quickly.

The parallel to restaking is direct. When you restake your ETH, you are not lending it to another party, but you are subjecting it to additional slashing conditions. Each AVS you opt into specifies behaviors that, if detected, will result in partial confiscation of your staked ETH. If you restake your ETH with three AVSs, you are now exposed to the slashing conditions of all three. If one AVS is poorly designed, or if its slashing conditions are triggered by an event you did not anticipate, your entire position can be penalized. The risk is not distributed. It is cumulative. And because the same ETH is securing multiple services, a single failure can propagate across all of them.

The clearest example of this risk materializing has not yet occurred in crypto, but the mechanism is well understood. Suppose an AVS experiences a bug that causes it to slash a large number of restakers simultaneously. That slashing event reduces the amount of ETH securing not only that AVS, but every other AVS those restakers had opted into. Other AVSs may now find themselves under-secured, which could trigger further slashing or a loss of confidence in the entire restaking layer. The cascading effect is exactly what happened when Lehman collapsed. One institution’s failure created uncertainty about the solvency of every other institution that had been part of the same rehypothecation chain. The solution in 2008 was government intervention. The solution in restaking, if it becomes necessary, is less clear.

What Could Break Each Layer

Each layer of the restaking yield stack has a different failure mode, and understanding those modes is more important than understanding the headline yield.

The base staking yield is the most robust, but it is not immune to compression. If too much ETH is staked, the protocol reduces the issuance rate to maintain a target level of security. As of mid-2026, more than 4.3 million ETH is restaked through EigenLayer alone, and the total amount staked through all mechanisms continues to grow. If that growth continues, the base yield will fall. The mechanism is designed to prevent the network from becoming over-secured at the expense of capital efficiency, but it also means that the 3% to 4% yield restakers currently earn from Ethereum itself is not fixed. It will decline as more capital enters the system. The decline will be gradual, but it will be real.

The AVS payment layer is more fragile. AVSs pay restakers because they need security, but they do not need as much security as restakers are currently offering. EigenLayer has more than $15 billion in total value locked, but the AVSs operating on top of it will likely need less than 10% of that capital to meet their security requirements. The result is predictable: as more capital flows into restaking, the yield paid by AVSs will fall, because there is simply not enough demand for security to absorb the supply of capital. This is not a failure of the AVS model. It is a feature of any market where supply grows faster than demand. The yield compression will be sharpest for restakers who entered late, because they will be competing with a much larger pool of capital for the same stream of AVS payments. The restaking mechanism itself does not prevent this outcome. It accelerates it.

The points layer has no defense against failure, because it is not based on any underlying cash flow. Points programs are designed to attract early participants, and they work as long as the tokens being distributed retain value. But if the token price falls, or if the airdrop allocation is smaller than expected, the effective yield from points programs can go to zero. This happened repeatedly in 2021 and 2022, when projects distributed governance tokens to liquidity providers and those tokens lost 90% of their value within weeks. The nominal APY looked high, but the realized return was negative once you accounted for the decline in token price. The same dynamic is already visible in restaking. Projects distribute points, participants assume those points will convert into valuable tokens, and then the tokens launch at a valuation that does not support the implied yield. The participants who entered late are left holding a position that no longer pays.

There is a fourth risk, less visible but more dangerous, which is looping. Because liquid restaking tokens (LRTs) are tradable, some users deposit them as collateral on lending protocols, borrow more ETH against them, buy more LRTs with that borrowed ETH, and repeat. Each loop multiplies exposure to the yield, and headline returns of 15% to 20% become possible. But each loop also multiplies exposure to loss. The user now carries a loan that can be liquidated if the price of the LRT falls, or if the LRT temporarily trades below the value of the ETH it represents. During periods of market stress, when liquidity dries up and prices become volatile, looped positions are the first to be liquidated. The liquidations create selling pressure, which pushes the price of the LRT lower, which triggers more liquidations. This is the same feedback loop that destabilized collateralized debt obligations in 2008, and it will destabilize restaking if enough capital adopts the same strategy.

The Numbers: What Restaking Actually Pays in 2026

As of mid-2026, EigenLayer controls 93.9% of the restaking market, with more than $15 billion in total value locked and 4.3 million ETH restaked. The most recent data shows EigenCloud, the successor entity, holding $6.49 billion in TVL, reflecting shifts in the market structure as liquidity moves between protocols. The advertised APY for restaking through EigenLayer currently ranges from 3.8% to 6%, depending on which AVSs you opt into. When you add token incentives and points programs, the combined yield rises to between 8% and 12%, though it can spike higher during periods of elevated AVS demand.

That headline figure is misleading, because most of the yield above 6% comes from the third layer, which is not durable. Strip out the points programs, and the sustainable yield is closer to 5% or 6%. Compare that to the standard 3% to 5% you would earn from traditional Ethereum staking, and the premium for accepting additional slashing risk is roughly 1 to 3 percentage points. That premium compensates you for the risk that an AVS experiences a slashing event, or that the AVS stops paying because it no longer needs as much security, or that the entire restaking layer experiences a cascade of failures when one AVS breaks. Whether that premium is sufficient depends on your assessment of how likely those events are, and how much capital you can afford to lose if they occur.

The fee structure for EigenLayer provides some insight into where the revenue is actually flowing. Under the proposed ELIP-12 governance mechanism, EigenLayer would capture 20% of subsidized AVS rewards and 100% of EigenCloud infrastructure fees, using those revenues to buy back EIGEN tokens. The proposal was scheduled to launch in Q1 2026. If it is implemented, it will create a direct link between AVS activity and token value, but it will also reduce the yield paid to restakers, because 20% of the AVS payments that would have gone to restakers will now go to the protocol instead. This is not unusual. Every protocol that matures eventually shifts from subsidizing participants to capturing revenue for itself. The shift is rational, but it compresses yield for the participants who arrived expecting the early-stage subsidies to continue indefinitely.

The activation of on-chain slashing in early 2026 changed the risk profile in a way that is easy to overlook. Before on-chain slashing went live, restakers faced theoretical slashing risk but no actual enforcement mechanism. Now the risk is fully enforceable at the protocol level. Every AVS you opt into can now penalize your stake if you violate its conditions, and those penalties are automatic. The shift from theoretical to enforceable risk should have reduced the amount of capital willing to restake, but it did not, because the points programs and token incentives masked the change. Participants continued to treat restaking as a higher-yield version of staking, without adjusting for the fact that the risk had become real. That mispricing will correct, either gradually as participants become more informed, or suddenly when a large slashing event demonstrates what the risk actually looks like.

Symbiotic and the Competition for Restaking Capital

EigenLayer is not the only restaking protocol, and the competition for capital is intensifying. Symbiotic crossed $1.7 billion in TVL shortly after raising its deposit limits in mid-2026, and it has since raised $29 million from Pantera Capital and Coinbase Ventures. Symbiotic differs from EigenLayer in two ways. First, it supports a wider range of collateral options, allowing users to restake assets other than ETH. Second, it allows network builders to create and manage their own staking implementations, rather than relying on a single shared security model. Those differences make Symbiotic more flexible, but they also fragment the restaking market, which may reduce the yield available on any single protocol as capital spreads across multiple venues.

The liquid restaking token market reflects the same fragmentation. EtherFi leads with $5.6 billion in TVL and roughly 65% of the LRT market, but it has recently removed nearly all restaking exposure from its main token, weETH. As of August 2026, less than 1% of EtherFi’s assets remain restaked with EigenLayer. That withdrawal signals a shift in how large LRT providers are assessing the risk-return tradeoff. If the premium for restaking is only 1 to 3 percentage points, and the slashing risk is now enforceable, the marginal yield may not justify the marginal risk, particularly for providers managing billions of dollars on behalf of users who do not fully understand the risks they are taking.

Other LRT providers have not followed EtherFi’s lead, and the divergence in strategy creates an opportunity for users who are willing to assess the tradeoffs themselves. Kelp DAO, Renzo, and Puffer Finance continue to offer liquid restaking tokens with full AVS exposure, and their yields remain higher than EtherFi’s. The question is whether that higher yield compensates for the higher risk, and whether the users deploying capital into those tokens understand that the risk is no longer theoretical.

What Happened When Rehypothecation Broke, and What It Means for Restaking

The Lehman collapse in 2008 did not happen because rehypothecation itself was flawed. It happened because the same assets had been pledged so many times, and the obligations built on top of those assets had become so complex, that when Lehman failed, no one could determine quickly who owned what. The legal process took years. The operational process of unwinding the positions took longer. The immediate result was a freeze in the credit markets, because every institution that had lent against rehypothecated collateral suddenly became uncertain whether that collateral still existed. The second-order result was a loss of confidence in the entire financial system, which required central bank intervention on a scale not seen since the Great Depression.

Crypto does not have central banks, and it does not have the legal frameworks that eventually allowed Lehman’s creditors to recover most of their assets. If a large-scale slashing event occurs in the restaking layer, and the same ETH that was securing multiple AVSs is confiscated simultaneously, there will be no government backstop and no orderly resolution process. The participants who opted into the slashing conditions will lose their capital, and the AVSs that depended on that capital for security will need to find alternative arrangements or shut down. The cascading effect will move faster than it did in 2008, because everything in crypto operates at a higher velocity than traditional finance, and the feedback loops are tighter.

The parallel is not exact, but it is close enough to warrant attention. Rehypothecation worked for years before it broke. Restaking has worked for two years without a major failure. The mechanism is sound as long as the underlying assumptions hold: that slashing events are rare, that AVSs remain solvent, and that the capital restaking Ethereum does not all try to exit at once. Those assumptions have held so far. They will not hold forever. The question is whether the participants deploying capital into restaking are pricing the risk correctly, or whether they are treating the additional yield as free money because nothing has broken yet. History suggests the latter. The bond markets in Greece and Portugal paid high yields for years before the sovereign debt crisis, and investors treated those yields as compensation for minimal risk. When the risk materialized, the yields stopped being yields. They became losses that had been accruing all along.

The Takeaway: Three Layers, Three Risks, One Question

Restaking yield is not a single income stream. It is three separate streams stacked on top of one another, each with a different source and a different failure mode. The base staking yield comes from the Ethereum protocol, and it is the most durable. The AVS payment layer comes from protocols renting security, and it will compress as supply outpaces demand. The points layer comes from temporary incentive programs, and it will disappear once those programs end or once the tokens being distributed lose value. If you are earning 12% from restaking, you are earning all three layers combined. If you want to know whether that 12% will still be there in a year, you need to decompose it and ask which layers are actually sustainable.

The historical parallel to rehypothecation is not decoration. It is the analytical tool. Restaking reuses the same capital across multiple obligations, just as rehypothecation did. The additional yield reflects additional risk, and the system works smoothly until something breaks. When it breaks, the cascading failures move faster than the institutions managing them, and the participants who entered late or who did not understand the risks are the first to lose capital. The question is not whether restaking will produce a Lehman-scale collapse. The question is whether you understand what you are earning, where it comes from, and what happens when one of the layers stops paying. The yield is real. The risk is real. The tradeoff is yours to assess.

Frequently Asked Questions

What is restaking and how does it differ from regular staking?

Restaking allows you to use your already-staked ETH to secure additional protocols called Actively Validated Services (AVSs) through platforms like EigenLayer or Symbiotic. Regular Ethereum staking earns 3-4% annually from protocol rewards. Restaking adds a second layer of yield from AVS payments, typically 1-3% more, in exchange for accepting additional slashing conditions from every AVS you opt into. The base ETH remains staked to Ethereum, but it now secures multiple services simultaneously, each with its own penalty conditions.

Where does the 8-12% restaking yield actually come from?

Restaking yield has three layers. Layer one is base Ethereum staking yield (3-4% from protocol issuance, priority fees, and MEV). Layer two is AVS security payments (1-3% from protocols renting validator security). Layer three is points programs and token incentives (variable, temporary, often 3-5%). The headline 8-12% figure combines all three. Only the first two layers are sustainable. The third layer disappears when incentive programs end or when distributed tokens lose value faster than they are earned.

What is the biggest risk in restaking?

Cascading slashing across multiple AVSs. When you restake, you opt into the slashing conditions of every AVS you secure. If one AVS experiences a bug or a penalty event, it can slash your entire staked position. Because the same ETH secures multiple services, a failure in one AVS can reduce security for all others, potentially triggering additional slashing events. Since on-chain slashing went live in early 2026, this risk is fully enforceable. Unlike traditional staking, where penalties are limited to Ethereum protocol violations, restaking exposes you to the operational and design risks of every AVS.

How is restaking similar to the rehypothecation that broke in 2008?

Both mechanisms reuse the same underlying asset to support multiple obligations. In rehypothecation, brokers pledged client collateral multiple times to create leverage and additional yield. When Lehman Brothers collapsed in 2008, the overlapping claims on the same collateral could not be resolved, and clients discovered their assets had been used elsewhere. In restaking, the same ETH secures multiple AVSs. If one AVS fails or triggers slashing, the entire position is at risk, and the failure can cascade across every service that ETH was securing. The additional yield in both cases reflects additional risk, and both systems work until something breaks.

Will restaking yields stay at 8-12% long term?

No. The sustainable portion is 5-6%, not 8-12%. Base Ethereum staking will compress as more ETH is staked, reducing issuance rewards. AVS payments will fall as restaking supply grows faster than AVS demand for security. EigenLayer already has over $15 billion TVL, but AVSs need less than 10% of that capital. Points programs, which account for much of the current headline yield, are temporary and will end. The yield compression has already started. Late entrants compete with a larger capital pool for the same AVS payments, and the marginal return for accepting slashing risk is narrowing.