Every Ethereum staking risk, ranked by how often it costs holders money
This page goes deeper on risk than our beginner guide to Ethereum staking. The first rows affect almost every staker; slashing, downtime and the inactivity leak mainly hit people who run validators, or pool users when an operator makes a mistake.
| Risk | What can go wrong | Evidence | Who it hits | How to reduce it |
|---|---|---|---|---|
| ETH price | Staking pays ETH, so a falling ETH price can wipe out years of yield in a week. | A 2–3% yearly yield is small next to ETH’s past drawdowns | Every staker | Size your position for the drawdowns; see our ETH drawdown history |
| Waiting to get out | Exits go through a protocol queue, so you cannot always sell staked ETH quickly. | Record exit wait 46.6 days (Sept 2025); about 4–11 days on Oct 20, 2026 | Solo stakers, pools and exchanges (liquid tokens can be sold instead) | Keep spending money unstaked; check the live queue before you rely on it |
| Liquid staking token discount | stETH, rETH and similar tokens can trade below the ETH they represent when many holders sell at once. | stETH traded at about a 4% discount to ETH on June 29, 2022 (CoinDesk, Nansen report) | Holders who must sell during the discount | Avoid leverage loops on liquid staking tokens; use the protocol withdrawal when you can wait |
| Platform or custodian | An exchange or lender can freeze withdrawals, fail, or be forced to stop staking. | Celsius froze withdrawals in June 2022; Kraken agreed to end U.S. staking-as-a-service and pay $30 million in an SEC settlement (Feb 9, 2023) | Customers of that platform | Prefer withdrawal addresses you control; read what happens in a bankruptcy |
| Smart-contract bug | Liquid staking and restaking run on smart contracts that can have bugs. | Audits reduce but do not remove this risk; it adds to the risks below | Liquid staking and restaking users | Use long-running, audited protocols; avoid stacking several protocols |
| Slashing | A validator that signs two conflicting messages is penalized and forced out. | 594 validators since 2020; isolated cost today ≈ 0.06 ETH per 32 ETH; correlated cost can reach 100% | Solo stakers, node operators, and pool users if the operator does not cover it | Never run the same keys in two places; keep slashing protection on |
| Downtime | An offline validator misses rewards and pays small penalties. | About 0.0014 ETH a day in penalties plus ≈ 0.0019 ETH of missed rewards per 32 ETH (Oct 2026) | Solo stakers and node operators | Monitor your node; short outages cost little, so do not rush risky fixes |
| Inactivity leak | If over a third of validators go offline and the chain stops finalizing, offline validators lose stake ever faster. | Loss grows with the square of time: about half the stake after ~18 days of non-finality (derived from spec constants) | Validators offline during the leak, which is worst for big operators that fail together | Client and operator diversity; avoid the majority client |
| Tax and rules | Rewards can be taxable when you receive them, and rules can change. | U.S. IRS Rev. Rul. 2023-14: staking rewards are income when you gain control of them | Every staker (varies by country) | Keep records of every reward; see our crypto tax estimator |
How often Ethereum validators get slashed (2020–2026)
Slashing punishes a validator that signs two conflicting messages: two different blocks for the same slot, or two attestation votes that contradict each other. It cannot happen by being slow or offline. From the beacon chain launch on December 1, 2020 to October 20, 2026, 594 validators were slashed on 157 different days; 575 for attestation violations and 19 for proposing two blocks. For scale, 2,390,971 validator indexes have ever been created.
| Year | Validators slashed | Days with a slashing | Most in one day | Proposer slashings |
|---|---|---|---|---|
| 2020 (from Dec 1) | 32 | 9 | 10 (Dec 3, 2020) | 11 |
| 2021 | 135 | 36 | 66 (Feb 3, 2021) | 5 |
| 2022 | 56 | 32 | 20 (Sep 23, 2022) | 1 |
| 2023 | 182 | 29 | 99 (Nov 14, 2023) | 1 |
| 2024 | 65 | 25 | 18 (Dec 5, 2024) | 1 |
| 2025 | 90 | 21 | 40 (Sep 10, 2025) | 0 |
| 2026 (to Oct 20) | 34 | 5 | 21 (Aug 21, 2026) | 0 |
The pattern is clear: slashing comes in bursts when one operator misconfigures many validators at once, not as a steady background risk. Since Pectra went live on May 7, 2025, 120 validators have been slashed, all for attestation violations. Figment counted 21 network slashing events in Q3 2026, which matches the 21 validators slashed on August 21, 2026 in this list.
The biggest slashing incidents and what caused them
| Date (UTC) | Operator | Validators | Cause | Cost and outcome |
|---|---|---|---|---|
| Feb 2, 2021 | Staked (now part of Kraken) | 75 | Validators restarted without a persisted slashing-protection database and signed twice | About 18 ETH of penalties (0.25 ETH each under the launch rules); Staked reimbursed clients [9] |
| Apr 13, 2023 | RockLogic (Lido operator) | 11 | A client update re-imported keys that had been moved to a failover machine (Prysm keymanager bug) | About 13.8 ETH of penalties and missed rewards projected; covered by the operator [11] |
| Oct 11, 2023 | Launchnodes (Lido operator) | 20 | Two validator clients pointed at one remote signer with slashing protection off during a datacenter outage | 28.463 ETH of penalties and missed rewards (incl. downtime of 2,562 other validators); fully compensated [12] |
| Nov 14, 2023 | Bitcoin Suisse | 99 | Newly activated validators double-signed over about two hours | About 1 ETH each (Bellatrix rules), close to $200,000 at the time [14] |
| Sep 10, 2025 | Ankr (on SSV Network) | 39 | Maintenance left keys running in two setups at once; SSV said its protocol was not at fault | First large slashing under Pectra rules: initial penalty only balance ÷ 4,096 [15] |
| Aug 21, 2026 | Unlabelled entity (0xecb5c1… withdrawal address) | 21 | Attestation double votes within minutes; no public post-mortem found | About 0.06 ETH each if 32 ETH validators (formula) [1] |
These six incidents account for 265 of the 594 slashed validators, about 45%. Every one was an infrastructure or key-management mistake; none was an attack. Bitcoin Suisse’s 99 in a day remains the record, and Staked’s 75 in February 2021 was the largest at the time.
What a slashing costs today
A slashed validator pays three things: an initial penalty at once, a correlation penalty about 18 days later that grows with how much other ETH was slashed around the same time, and missed-duty penalties every epoch until it can withdraw about 36 days later. With 43.76 million ETH staked (October 20, 2026):
| Scenario | 32 ETH validator | 2,048 ETH validator |
|---|---|---|
| Only your validator slashed (initial + correlation + missed duties) | ≈ 0.058 ETH (0.18%) | ≈ 4.00 ETH (0.20%) |
| Yours plus 20 other 32 ETH validators (like Aug 21, 2026) | ≈ 0.059 ETH | ≈ 4.09 ETH |
| 100,000 ETH slashed in the same 36 days | ≈ 0.28 ETH (0.9%) | ≈ 17.7 ETH (0.9%) |
| 1 million ETH slashed together | ≈ 2.25 ETH (7%) | ≈ 144.1 ETH (7%) |
| A third of all stake (≈ 14.6M ETH) slashed together | 32 ETH (100%) | 2,048 ETH (100%) |
| Same isolated slashing under the old Bellatrix rules | ≈ 1.05 ETH (3.3%) | n/a (max 32 ETH then) |
The slashed validator also stops earning, so add about 0.068 ETH of missed rewards per 32 ETH over those 36 days. The math matches Lido’s own projection for its October 2023 incident (about 0.06 ETH of attestation penalties per validator), when the initial penalty was still 1 ETH. The point of the correlation penalty is that one operator’s mistake is cheap, while a coordinated attack by a third of the stake is ruinous.
Slashing rules by upgrade
The penalty started low while the beacon chain was new, rose in two steps to 1 ETH per 32 ETH by The Merge, then fell sharply with Pectra, which let validators hold up to 2,048 ETH (EIP-7251) and cut the initial penalty so large validators are not punished harder for a single mistake. See every fork on our Ethereum upgrades timeline.
| Rules | From | Initial penalty | 32 ETH example | Correlation multiplier | Whistleblower reward |
|---|---|---|---|---|---|
| Phase 0 (launch) | Dec 1, 2020 | balance ÷ 128 | 0.25 ETH | × 1 | balance ÷ 512 |
| Altair | Oct 27, 2021 | balance ÷ 64 | 0.5 ETH | × 2 | balance ÷ 512 |
| Bellatrix | Sep 6, 2022 | balance ÷ 32 | 1 ETH | × 3 | balance ÷ 512 |
| Electra (Pectra) | May 7, 2025 | balance ÷ 4,096 | 0.0078 ETH (0.5 ETH at 2,048 ETH) | × 3, no rounding | balance ÷ 4,096 |
In every era a slashed validator is forced to exit and cannot withdraw for 8,192 epochs (about 36.4 days), and the correlation penalty is applied halfway through, at about 18.2 days.
Offline penalties and the inactivity leak
Downtime is the risk solo stakers worry about most, and it is the cheapest. A 32 ETH validator that misses its votes pays about 0.0014 ETH a day in penalties and misses about 0.0019 ETH a day of rewards (October 2026 rates), so a full day offline costs nearly as much as two days online would earn. A weekend outage costs less than 0.01 ETH. Rushing a fix by starting your keys on a second machine is how many slashings happened, so it is safer to stay offline a little longer.
The inactivity leak is different. If more than a third of all stake is offline, the chain cannot finalize, and after four epochs every validator that is still offline loses stake at a rate that grows each epoch. By the spec formula a validator offline for the whole leak would lose about half its balance after roughly 18 days, and validators whose effective balance falls to 16 ETH are ejected. That is why client and operator diversity matters: a bug in a client used by over a third of validators would leak only that client’s users until finality returns.
Who pays when a validator is slashed?
| How you stake | Who absorbs the loss |
|---|---|
| Solo validator | You. Every penalty comes out of your validator balance. [7] |
| Lido (stETH) | The pooled stake, so all stETH holders share it, unless the node operator compensates. In both 2023 Lido incidents the operator paid stakers back in full. [12] |
| Rocket Pool (rETH) | The node operator first, from its ETH bond (and RPL if staked); only a loss larger than that is spread across rETH holders. [17] |
| Coinbase | Coinbase says its user agreement commits to reimbursing slashing caused by Coinbase or its third-party validators, but not losses from hacks, your own actions or protocol bugs. It says it has never been slashed. [18] |
| Staking service with your own keys | Depends on the contract; many offer slashing cover. Read the terms before you deposit. |
For the fees, yields and withdrawal routes of each option, see Lido vs Rocket Pool vs Coinbase staking and the staking comparator.
How to avoid being slashed: lessons from the post-mortems
- Run each key in exactly one place. Staked, Launchnodes, Ankr and RockLogic were all slashed because the same keys signed from two setups.
- Keep slashing protection on and carry it with you. Export and import the slashing-protection history when you migrate; Staked’s incident began when it stopped persisting that database.
- Never run a hot “backup” validator. Redundancy belongs in distributed validator technology, not in a second copy of the keys (SSV’s own lesson from September 2025).
- Wait before restarting keys on a new machine. Stop the old client, confirm it is off, and wait a few epochs or turn on doppelganger protection.
- Prefer downtime to haste. A day offline costs about 0.003 ETH including missed rewards; a panicked failover can cost far more.
Slashing and downtime penalty calculator
Estimate what a slashing or an outage would cost a validator under today’s or the old rules. Add other ETH slashed at the same time to see how fast the correlation penalty grows.
| Initial slashing penalty | 0.00781 ETH | Balance ÷ 4,096, taken at once |
| Correlation penalty (about 18 days later) | 0.00007 ETH | Balance × 3 × 32 ETH slashed ÷ total staked |
| Missed-duty penalties until withdrawable | 0.0501 ETH | 8,192 epochs (~36.4 days) of missed source and target votes |
| Total protocol penalty | 0.0580 ETH (0.18%) | ETH removed from the validator balance |
| Rewards you also miss | ≈ 0.0677 ETH | What it would have earned from attestations in those 36 days |
Estimates from consensus-spec formulas with the total staked you enter (default 43,759,892 ETH, ultrasound.money, Oct 20, 2026). Real losses vary with network participation, block proposals, sync-committee duty and rounding.
Download the data
License: our tables, yearly aggregation, penalty examples and chart are licensed under CC BY 4.0. Credit “GrowYourEthereum, Ethereum Staking Risks, CC BY 4.0” with a link. Slashings are public on-chain events; the operator labels come from beaconcha.in and may be incomplete.
Cite this data
Free to reuse in articles, research, reports and apps under CC BY 4.0. Copy a ready-made citation below; the HTML version includes the credit link the license asks for.
GrowYourEthereum. (2026). Ethereum Validator Slashings and Staking Risks [Data set]. Updated October 20, 2026. https://growyourethereum.com/data/ethereum-staking-risks. Licensed CC BY 4.0.
<a href="https://growyourethereum.com/data/ethereum-staking-risks">Ethereum Validator Slashings and Staking Risks</a> by GrowYourEthereum, licensed under <a href="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0</a>.
@misc{growyourethereum_ethereum_slashings_2026,
author = {{GrowYourEthereum}},
title = {Ethereum Validator Slashings and Staking Risks},
year = {2026},
howpublished = {\url{https://growyourethereum.com/data/ethereum-staking-risks}},
note = {Data set, updated October 20, 2026. CC BY 4.0}
}More datasets, JSON files and embeddable widgets: free Ethereum data for journalists, researchers and builders.
Methodology
- Slashings (measured): the full beaconcha.in slashings list read on October 20, 2026: 592 records, three of which name two validators; one validator appears twice, leaving 594 distinct validators. rated.network also reports 594 keys. We checked a random sample of 10 against a beacon node (all slashed = true). Dates are UTC.
- Incidents: operators, causes and costs come from each operator’s post-mortem or news coverage, linked in the sources. Day counts in the table are all slashings that UTC day.
- Penalties (derived): consensus-spec formulas with 43,759,892 ETH staked (ultrasound.money effective-balance sum): base reward per ETH = 64 ÷ √(total staked in gwei), a slashed validator pays the source and target weights (40/64) each epoch for 8,192 epochs, and the correlation penalty is balance × min(3 × ETH slashed in the window, total) ÷ total. Inactivity-leak loss ≈ balance × epochs² ÷ 2²⁵, an upper bound that ignores the falling effective balance.
- Limits: real losses also depend on missed block proposals, sync-committee duty and rounding, and on the total stake at the time. Labels such as “Whale_0x…” are beaconcha.in groupings by deposit address, not named companies.
Ethereum staking risks FAQ
What are the main risks of staking ETH?
In order of how often they hurt holders: ETH’s price falling, waiting in the exit queue when you want out, liquid staking tokens like stETH trading below ETH, a platform freezing or failing, smart-contract bugs, slashing, downtime penalties and tax. Slashing gets the most attention but is rare: 594 of about 2.39 million validators ever created have been slashed (Oct 20, 2026).
How often are Ethereum validators slashed?
Rarely. From December 1, 2020 to October 20, 2026, 594 validators were slashed on 157 different days, about 1 in 4,000 of all validators ever created. Most came from a few operator mistakes: 99 Bitcoin Suisse validators on November 14, 2023, 75 Staked validators in February 2021 and 39 Ankr validators on September 10, 2025. There were 34 slashings in 2026 so far, 21 of them on August 21.
How much ETH do you lose if you get slashed?
Under today’s Pectra rules, an isolated slashing of a 32 ETH validator costs about 0.058 ETH: 0.0078 ETH at once (balance ÷ 4,096), about 0.05 ETH of missed-duty penalties over the 36 days until it can withdraw, and a tiny correlation penalty. Before May 7, 2025 the initial penalty alone was 1 ETH. If many validators are slashed at the same time the correlation penalty grows, and if a third of all stake is slashed within about 36 days, every slashed validator loses everything.
What causes slashing?
Only two things: proposing two different blocks for the same slot, or casting two conflicting attestation votes. On mainnet it has almost always been an operator mistake, such as running the same validator keys on two machines during a migration, failover or update, or turning off the slashing-protection database. Of the 594 slashings, 575 were attestation violations and 19 were proposer violations.
Can you get slashed for being offline?
No. Being offline is never slashable. An offline validator only misses rewards and pays small attestation penalties, about 0.0014 ETH a day for 32 ETH in October 2026, roughly what it would have earned. The exception is an inactivity leak: if more than a third of validators are offline and the chain stops finalizing, offline validators lose stake faster and faster until finality returns.
Do Lido or Rocket Pool users lose money if a validator is slashed?
With Lido, penalties hit the pooled stake that backs stETH, so all holders share them unless the node operator pays them back, as Launchnodes and RockLogic did in 2023. With Rocket Pool, the node operator’s bond absorbs losses first; rETH holders only share a loss bigger than the operator’s bond. Both protocols also carry smart-contract risk, and their tokens can trade below ETH.
Will Coinbase reimburse me if my staked ETH is slashed?
Coinbase says its user agreement commits to reimbursing slashing penalties caused by Coinbase or its third-party validators, but not losses caused by a hack, your own actions or a protocol bug, and that it has never had a slashing. Read the current user agreement for the exact terms.
Is staking ETH safer than just holding it?
Staking adds risks on top of holding: lock-up time, slashing and downtime for validators, and smart-contract or platform risk for pools and exchanges. In exchange you earn about 2–3% a year in ETH. For most holders the price risk is much larger than any staking-specific risk, but if you might need the ETH quickly, keep that part unstaked.
How do I avoid getting slashed as a solo staker?
Run each validator key in exactly one place, keep the client’s slashing-protection database when you move or restore, wait a few epochs (or use doppelganger protection) before starting keys on a new machine, never run a “backup” validator with the same keys, and update one machine at a time. Every large mainnet slashing broke the first rule.
Where does this slashing data come from?
Every slashing is an on-chain event. We downloaded the full beaconcha.in slashings list on October 20, 2026 (592 records covering 594 validators), checked the total against rated.network (594 keys) and spot-checked validators against a beacon node (slashed = true). Penalty figures use the consensus-spec constants. The CSVs are free under CC BY 4.0.
Sources
- beaconcha.in: Validator slashings (full list, read Oct 20, 2026)
- rated.network: Slashings on Ethereum mainnet (594 keys)
- Ethereum consensus specs: Electra beacon chain (slashing, EIP-7251)
- Ethereum consensus specs: Bellatrix beacon chain
- Ethereum consensus specs: Altair beacon chain (rewards, inactivity leak)
- Ethereum consensus specs: Phase 0 beacon chain
- ethereum.org: Proof-of-stake rewards and penalties
- EIP-7251: Increase the MAX_EFFECTIVE_BALANCE
- Staked: February 2 ETH2 slashing event post-mortem
- The Block: 75 Eth2 validators slashed due to a Staked bug (Feb 4, 2021)
- Lido: RockLogic GmbH slashing incident post-mortem (Apr 2023)
- Lido: Launchnodes slashing incident post-mortem (Oct 2023)
- Lido research forum: Launchnodes final penalties (28.463 ETH)
- DL News: Ethereum stakers “terrified” of slashing (Bitcoin Suisse, Nov 2023)
- SSV Network: September 10, 2025 slashing post-mortem
- CoinDesk: Ethereum rare mass slashing event (Sep 10, 2025)
- Rocket Pool docs: FAQ (how the protocol protects rETH)
- Coinbase Help: Staking risks (slashing reimbursement)
- Figment: Q3 2026 Ethereum validator report (21 network slashings in Q3)
- CoinDesk: Nansen report links Terra collapse and stETH de-peg (Jun 29, 2022)
- SEC: Kraken to discontinue unregistered staking-as-a-service program (Feb 9, 2023)
- IRS: Revenue Ruling 2023-14 (staking rewards)
- ultrasound.money API: effective balance sum (43,759,892 ETH, Oct 20, 2026)
Educational data page. Not financial advice. GrowYourEthereum is independent and not affiliated with Lido, Rocket Pool, Coinbase, SSV Network, beaconcha.in, rated.network or the Ethereum Foundation. Data as of October 20, 2026. Our tables and CSVs: CC BY 4.0.