$15.8 billion. That is the current TVL sitting inside EigenLayer — a restaking protocol that launched in 2023, ranks 10th across all of DeFi by total value locked, carries audits from Sigma Prime and Consensys Diligence, and has recorded zero exploits to date. Zero exploits sounds like safety. It is not safety. It is a young track record on a protocol that introduced an entirely new category of slashing conditions to Ethereum, and $15.8 billion is currently betting that the track record holds.
I get asked some version of "what are the real risks of restaking" at least once a week. The honest answer is always the same: it depends. It depends on how much of your portfolio you deposited, whether you restaked natively or through a liquid wrapper, how many Actively Validated Services your operator opted into, and — this is the part almost nobody frames correctly — whether you understand that restaking is not "staking but with extra yield." It is staking with an additional layer of slashing conditions that do not exist in vanilla Ethereum validation. The risk is not additive. It is multiplicative. And what that means for you changes dramatically depending on who you are. So instead of handing you a generic risk list, I am going to walk through three hypothetical profiles. Three people, three positions, three ways the math shakes out. None of them are real — they are composite illustrations built from the kind of questions I see every week. But the numbers are grounded, and every step is reproducible.
Scenario 1: The 5 ETH Passive Holder
Imagine someone — let us call her the passive holder — who has 5 ETH and wants yield. She is not a trader. She does not run nodes. She heard about restaking from a Crypto Twitter thread and liked the idea of earning on top of her staking rewards.
Here is what she probably did: deposited ETH into a liquid restaking protocol that wraps EigenLayer, received a liquid restaking token in return, and moved on with her life. She may not have checked which AVSs her operator opted into. She probably does not know what an AVS is. And that is the first risk nobody discusses — delegation without comprehension.
When you stake ETH natively on Ethereum, your slashing conditions are narrow and well-understood. Double-signing. Surround-voting. The penalties are defined in the protocol spec, and the base rate of accidental slashing for a well-run validator is effectively negligible. When you restake through EigenLayer, you extend those slashing conditions to additional services — Actively Validated Services — each of which defines its own slashing logic. Your operator chooses which AVSs to validate. If the operator misbehaves on any one of them, your restaked ETH is subject to slashing under that AVS's rules, not just Ethereum's.
For someone holding 5 ETH — maybe 60% of her crypto portfolio — the question she should be asking is not "what is the yield." The question is "what is the maximum I can lose, and can I absorb losing it." In EigenLayer, the answer is: up to the full restaked amount, depending on the slashing parameters of the AVSs involved. Total-loss scenario on 60% of her crypto, for a yield that I could not pull for this piece because it changes hourly and varies by operator — but is almost certainly single digits annualized.
And here is the comparison she probably never ran. Binance — daily volume $18.5 billion, CER security score 9.4, proof-of-reserves last audited 2025-03-01 — offers staking as a product. So does Bybit, with its own reserves verified as of 2025-03-12. CEX staking is not trustless and it carries platform risk, no question. But the risk model is understood: you trust the exchange, the exchange runs the validator, slashing exposure is limited to the standard Ethereum conditions, and the exchange absorbs operational errors. For a 5 ETH holder who does not know what an AVS is, the gap between "I understand this risk" and "I do not understand this risk" is doing more work than the gap between staking yields. Risk you understand is manageable. Risk you do not understand is just exposure.
Scenario 2: The 100 ETH DeFi Native
Now picture a different depositor — someone with 100 ETH, deep protocol experience, runs their own validator. This person understands slashing. They have read the EigenLayer documentation. They chose their AVSs deliberately. They are not delegating blindly.
Their risk profile is completely different from scenario one, but not in the direction most people assume.
The DeFi native's central risk is concentration and correlation. EigenLayer holds $15.8 billion in TVL, ranked 10th across all of DeFi. That number sounds like security — big TVL, lots of trust, institutional validation. But TVL concentration is a double-edged condition. If something goes wrong at the protocol level — not an AVS-level slashing event but a smart contract vulnerability in EigenLayer's core contracts — the blast radius is the entire $15.8 billion. Sigma Prime and Consensys Diligence have audited the contracts. Real. Meaningful. But I have been around long enough to know that "audited" and "unexploitable" are not synonyms. The protocol launched in 2023. We are early, and the system complexity — restaking, delegation, operator selection, multi-AVS slashing coordination — is not something two audits fully retire.
For the 100 ETH depositor, I want to run a risk teardown with working shown, because this is the part most restaking content skips entirely.
Think of the risk as a tree with layers.
Layer 1: Ethereum protocol risk. Probability of a consensus-level bug causing mass slashing — extremely low for a well-run, non-malicious validator. Expected annual loss from this layer: effectively zero. We set this aside.
Layer 2: EigenLayer smart contract risk. The contracts hold $15.8 billion. Audited twice. Zero exploits. Protocol age: under 3 years. If we assign — conservatively, hypothetically, for the structural exercise — a 1% annualized probability of a material smart contract event, then on 100 ETH the expected loss from this layer alone is 1 ETH per year. That is not a fee you see. It is the invisible cost of the risk you are carrying.
Layer 3: AVS slashing risk. Varies wildly by operator and by AVS. Let us say our depositor is opted into 3 AVSs. If each carries an independent 0.5% annual probability of a slashing event that costs 10% of restaked ETH, the combined probability of at least one event is: 1 - (1 - 0.005)^3 = approximately 1.49%. If it occurs, the loss is 10 ETH. Expected annual value from this layer: roughly 0.149 ETH.
Aggregate expected annual risk cost, layers 2 and 3: approximately 1.15 ETH on 100 ETH. That is 1.15%.
Now subtract that from whatever restaking yield is being earned. If the yield is 3%, the risk-adjusted yield is closer to 1.85%. If the yield is 2%, you are at 0.85%. And that is before the illiquidity premium you should be demanding for locking ETH in a multi-layer system with withdrawal queues.
I want to be precise about something: those probability inputs — 1%, 0.5% — are illustrative, not observed. Nobody has a calibrated probability distribution for EigenLayer smart contract risk. That is, in fact, the point. If you are not running this kind of model with your own assumptions on your own position, you are accepting a yield without knowing whether the yield compensates you for the risk. And "I do not know whether I am being compensated" is a different statement from "I am being compensated."
Scenario 3: The Institutional Allocator Weighing Custody
Third profile. Imagine a small fund — or a high-net-worth individual working with a custodian — evaluating whether to restake a portion of their ETH allocation through EigenLayer versus keeping it in vanilla staking under a qualified custodian.
This is where the custody angle sharpens the entire conversation. Coinbase Custody operates as a NY DFS Trust Company. Fidelity Digital Assets holds the same designation. Anchorage Digital carries an OCC Federal Trust Charter — the first crypto-native entity to receive one. These are not "places to park crypto." They are regulated fiduciary entities with capital requirements, insurance structures, and audit obligations that exist specifically because institutional money requires a defined liability framework.
When this profile looks at EigenLayer, the risk calculus is not about yield. It is about fiduciary exposure. If they restake and a slashing event occurs, who is liable? The operator? EigenLayer? The custodian? The fund manager who approved the allocation? The legal surface area of restaking in an institutional context is, plainly, uncharted.
And the detail that sharpens the point: EigenLayer's EIGEN governance token introduces a governance layer on top of the economic layer. Governance tokens in DeFi have historically been used to modify protocol parameters — including, potentially, slashing conditions. For an institutional allocator, "can the rules change after I deposit?" is not philosophical. It is a compliance question. And right now, the answer is: yes. Governance can modify protocol parameters. That is how DeFi governance works. Whether your compliance framework can accommodate that — well, that is the conversation most allocators have not had yet.
The institutional depositor might see $15.8 billion in TVL and read it as validation. Smart money is in, audits on file, zero exploits. But smart money has been wrong at scale before. TVL measures what is deposited. It does not measure what is safe. The distinction is load-bearing.
For this profile, the real comparison is not restaking yield versus staking yield. It is restaking yield minus risk-adjusted cost minus compliance cost minus fiduciary liability exposure versus vanilla staking yield through a qualified custodian with clear legal responsibility and defined recourse. When you frame it that way, the yield premium from restaking needs to be meaningfully higher to justify the additional layers. I am not persuaded it is.
What All Three Share
Three depositors. Three position sizes. Three levels of sophistication. The pattern underneath is the same.
Everyone underprices smart contract risk on new protocols. EigenLayer has been audited by Sigma Prime and Consensys Diligence. Zero exploits. But it launched in 2023 and holds $15.8 billion. The ratio of TVL to protocol age is aggressive by any historical standard in DeFi. I am not predicting a failure. I am observing that the market is pricing the probability of failure at approximately zero, and that price is almost certainly too low.
The layered nature of restaking risk is poorly understood. Most people model restaking as "staking plus extra yield." It is actually staking plus extra slashing surfaces plus smart contract dependency on a second protocol plus governance risk on that second protocol plus operator selection risk. Each layer adds a non-zero probability of loss. The probabilities compound. And the yield does not scale with the risk — it is set by market dynamics that have nothing to do with how many layers of exposure you are carrying.
Exit liquidity is assumed, not guaranteed. Whether you restaked natively or through a liquid wrapper, your ability to leave depends on withdrawal queue dynamics, liquid restaking token peg stability, and — in a stress scenario — whether everyone else is trying to leave at the same time. $15.8 billion is not liquid. It is committed. In a crisis, "committed" and "trapped" can feel uncomfortably similar.
Which Scenario Is You
If you recognized yourself in scenario one, the question is direct: is single-digit yield worth tail risk on what might be the majority of your crypto portfolio? If you would not put 60% of your savings into an uninsured 3-year-old financial product in any other context, think about why you are comfortable doing it here.
If scenario two is closer, you probably already know most of this. But knowing the risk and pricing it are different things. Run the layer model I outlined with your own probability assumptions. If your risk-adjusted yield after honest estimates comes out under 1%, ask whether the complexity premium is worth carrying.
If scenario three is your reality, the question is not financial — it is structural. Can your compliance framework accommodate a product where the rules change by governance vote, slashing conditions vary by AVS, and fiduciary liability in a loss event is legally undefined? If the answer is "we have not worked that out," then you have not worked out whether you should be in EigenLayer.
1.15% expected annual risk cost on a 100 ETH position — that was the number from the teardown in scenario two. It is illustrative, not definitive. But it is the number that should force one specific decision: before you deposit anything into restaking, build your own version of that model. Plug in your own position size, your own probability estimates, your own assumptions about smart contract risk on a protocol holding $15.8 billion that is under three years old. If the yield still looks attractive after the math — fine. If you have never run the math at all, you are not restaking. You are speculating on a risk you have not measured. The difference matters. The math is open.