Staking is locking capital into a proof-of-stake protocol as a bond, in return for the right to validate blocks and a share of the protocol's reward flow. On Ethereum the bond is 32 ETH per validator, and the protocol can destroy part of it for provable misbehavior. The reward arrives from two separate places: new issuance the protocol creates, and priority fees plus MEV that users are already paying. Which of those two funds the rate is the question that decides whether a staking design holds up.
Rewards paid out of issuance are paid by every holder through dilution. Rewards paid out of fees are paid by people using the network. A staking design that cannot say which one it is running on has not been designed yet.
Scroll to see the full diagram
What the 32 ETH is actually doing
Proof-of-stake makes dishonesty expensive in a way the chain itself can settle. Ethereum's documentation puts it as proving that validators have put something of value into the network that can be destroyed if they act dishonestly.1 The deposit is not a subscription and it is not an entry ticket. It is collateral posted against a short list of behaviors the protocol can prove you committed.
Running one validator means depositing 32 ETH into the deposit contract and running three pieces of software: an execution client, a consensus client and a validator client. The deposit then joins an activation queue that rate-limits how fast new validators enter.1 None of those three is optional, and the third is where most operational incidents begin.
EIP-7251 gets misread on this point constantly. It introduced MIN_ACTIVATION_BALANCE at 32 ETH and raised MAX_EFFECTIVE_BALANCE_ELECTRA to 2048 ETH, so one validator can now hold and compound up to 2048 ETH of effective balance instead of running 64 separate keys.4 It raised the ceiling. The floor stayed where it was, and the EIP states the reason plainly: the lower bound was kept to encourage solo-staking.4
Where the reward money actually comes from
Two streams, funded by two different parties. The consensus stream is issuance. Ethereum computes a base reward per validator as effective balance multiplied by a base reward factor of 64, divided by four times the square root of total active balance across the network.2 Read the shape rather than the constants. Total issuance grows with the square root of the validator count while the reward per validator shrinks with it, so every additional staker dilutes the rest.2
The execution stream is different money entirely. Priority fees and MEV come out of transaction flow, paid by users and block builders, and no new ETH is created to cover them. Large staking protocols keep the two apart in their accounting for exactly that reason: Lido's core contract collects withdrawals, priority fees and MEV from dedicated vaults before it applies its fee and distributes rewards.6
This split is the first thing we ask about on any staking design, on Ethereum or anywhere else. Issuance-funded rewards are a transfer from every holder to the subset that stakes. Fee-funded rewards are external revenue arriving from people using the product. Both can be defensible. Only one of them survives the subsidy schedule ending, and a protocol quoting a headline rate without separating the two is quoting a number it does not control.
The reward is a weighted sum, and most validators sit below the top of it
Ethereum splits the reward into five components with fixed weights that sum to 64: timely source vote 14, timely target vote 26, timely head vote 14, sync committee participation 2, and block proposal 8.2 A validator earns the fraction of the base reward matching the weights it actually captured in that epoch.
Work the arithmetic and the design intent shows up. A validator that votes correctly and on time but is neither proposing a block nor sitting in a sync committee captures 54 of the 64 weight units, about 84 percent of the base reward, every epoch, indefinitely.2 Proposals and sync duty are the remaining 16 percent and they arrive rarely and at random.
The penalty structure is asymmetric in a way that matters more than the reward structure. Missing the source or target vote costs exactly what the reward would have paid. Missing the head vote costs nothing, because head votes are rewarded and not penalized, and failing to propose a block carries no penalty either.2 The practical reading for anyone sizing a staking operation: this system pays for boring, uninterrupted attestation uptime and it barely pays for anything else.
The entry queue is capped by a constant, not by demand
EIP-7514 added MAX_PER_EPOCH_ACTIVATION_CHURN_LIMIT with a value of 8 and reached Final status in the Deneb fork.3 Before it, activation churn scaled with the validator set, and the EIP's authors projected that a permanently full deposit queue would take Ethereum to 50 percent of supply staked by May 2024 and 100 percent by December 2024.3 The cap was written to buy time for a reward-curve discussion, not to solve it.
The arithmetic is worth doing once, because it sets a hard ceiling on how fast any staking product can grow. An epoch is 32 slots of 12 seconds, so 384 seconds, which gives 225 epochs a day. At 8 activations per epoch that is 1,800 new validators a day. At the 32 ETH minimum activation balance, roughly 57,600 ETH a day can enter the active set, no matter how much capital is waiting.
If you are building a pooled staking product, that number is your throughput limit during exactly the periods when inflows spike. Deposits arrive instantly. Activation does not. Whatever you issue to depositors in the meantime is a claim on stake that is not yet earning, and how you price that gap is a design decision, not an accounting detail.
Getting out is a queue, then a wait, then a sweep
Exiting is a three-stage process and only the first stage is under the staker's control. A voluntary exit is initiated either by signing and broadcasting an exit message through the validator client, which costs no gas, or by triggering the exit from the withdrawal address on the execution layer, which costs gas but does not require the validator signing key.5 Ethereum's own documentation states that how long the exit takes depends on how many others are exiting at the same time.5
Once the account is marked withdrawable, no further action is needed. Block proposers continuously sweep the validator set, building a withdrawal queue of up to 16 eligible withdrawals per block, which works out to 115,200 validator withdrawals a day assuming no missed slots.5 That sweep is automatic and competes for no execution-layer block space.
EIP-7514 names the structural problem in its own security considerations: the change breaks the symmetry between the entry and exit queues, because entry is bounded by the churn constant while exit is not.3 That asymmetry is deliberate and it cuts both ways. The network can be left quickly, which is good for the individual staker and bad for anyone whose product promised liquidity out of a queue that lengthens under stress. The liquid staking token entry picks this up from the product side rather than repeating it here.
What is at risk, ranked by what it actually costs
The immediate slashing penalty is small and was made small on purpose. Ethereum documents an immediate burn of 0.0078125 ETH for a 32 ETH validator, scaled linearly with active balance.2 EIP-7251's rationale explains why: the initial penalty scaled linearly on effective balance, so consolidating validators would have concentrated that risk, and making the initial penalty negligible removed the disincentive to consolidate.4
The expensive part comes later. A slashed validator enters a 36-day forced removal during which its balance bleeds, and at the midpoint on day 18 it takes a correlation penalty whose size scales with the total effective balance of every validator slashed in the preceding 36 days.2 Isolated operator error costs roughly one percent of stake. A mass event that catches many validators at once can destroy the full stake.1
Two exposures get left off most risk pages. The inactivity leak activates whenever the chain fails to finalize for more than four epochs and bleeds stake from validators voting against the majority until finality returns.1 And the exit queue itself is a risk position: capital that cannot leave for an unknown number of days is capital exposed to whatever caused everyone else to want out. Penalty construction in detail belongs to the slashing entry, which is where the correlation formula is worked through.
What we settle before a staking layer ships
Four things, in this order. Name the funding source and its schedule, split between issuance and fees, and write both lines separately in the documentation. State the exit path with its actual constraints, including any queue the staker inherits from the base chain. Quote rates on one consistent basis with the compounding assumption stated, which is the whole argument of the APR versus APY entry. And publish what happens to a depositor's claim when a validator behind it is penalized.
The failure pattern we see most is a staking layer bolted onto a token that has no fee flow underneath it. The rate is entirely issuance, the documentation calls it a reward, and the design is a countdown. Staking is a supply sink that works when holders have a reason to stay after the subsidy is cut. If nothing about the product changes for a staker at zero emissions, nothing is holding them.
One boundary on all of the above. This page describes mechanisms and the design questions they raise. It is not investment advice, not a recommendation to stake or not stake any asset, and the tax and regulatory treatment of staking rewards varies by jurisdiction and belongs to your counsel.
Common questions
How much ETH do you need to stake on Ethereum?
32 ETH activates one validator. EIP-7251 introduced MIN_ACTIVATION_BALANCE at exactly 32 ETH and raised the maximum effective balance to 2048 ETH, which lets one validator compound rewards instead of forcing operators to run many 32 ETH keys.4 The minimum was deliberately left alone to keep solo staking viable. Pooled and liquid staking products exist for smaller amounts, at the cost of adding an operator and a smart contract between you and the stake.
Where do staking rewards actually come from?
Two sources. Consensus rewards are newly issued ETH, computed from a formula in which total issuance rises with the square root of the validator count while the per-validator amount falls with it.2 Execution rewards are priority fees and MEV, paid by users and block builders out of transaction flow with no new issuance. Protocols account for the two separately, and the mix determines whether a rate is funded by dilution or by revenue.
Can you lose your staked ETH?
Yes. Provable offenses such as proposing two blocks for one slot trigger slashing: an immediate burn of 0.0078125 ETH for a 32 ETH validator, then a 36 day forced removal with a correlation penalty on day 18 that scales with how much stake was slashed in the same window.2 An isolated incident costs roughly one percent of stake; a correlated mass event can destroy all of it.1 Downtime alone is penalized far more mildly.
How long does it take to unstake ETH?
It depends on how many people are exiting at the same time, because the exit queue has no fixed length.5 After a voluntary exit clears, the account is marked withdrawable and block proposers sweep it automatically, processing up to 16 withdrawals per block or about 115,200 validator withdrawals a day.5 The sweep is fast. The queue in front of it is the variable, and it lengthens under exactly the market conditions that make people want to exit.
Is staking the same as yield farming?
No, and the difference is the source of the money. Staking rewards come from protocol issuance and from fees users are already paying to transact, both of which continue for as long as the chain runs. Yield farming rewards come from an incentive budget a treasury has chosen to spend, which is bounded by construction. The mechanics of locking capital look similar. What funds the rate, and how long that source lasts, does not.
See LST and LRT Tokenomics Design for how this applies in practice.
Sources
- Proof-of-stake (PoS)
ethereum.org, Ethereum Foundation, 2026
Stake as destroyable collateral, the 32 ETH deposit and three-client requirement, the activation queue, mass slashing up to 100 percent of stake, and the inactivity leak. - Proof-of-stake rewards and penalties
ethereum.org, Ethereum Foundation, 2026
Base reward formula and constants, the five reward weights summing to 64, the reward and penalty asymmetry, the 0.0078125 ETH initial slashing burn, and the 36 day removal with a day 18 correlation penalty. - EIP-7514: Add Max Epoch Churn Limit
Ethereum Improvement Proposals, 2023
Final. Sets MAX_PER_EPOCH_ACTIVATION_CHURN_LIMIT to 8, projects the staked-supply trajectory it was written to slow, and states the resulting asymmetry between the entry and exit queues. - EIP-7251: Increase the MAX_EFFECTIVE_BALANCE
Ethereum Improvement Proposals, 2023
MIN_ACTIVATION_BALANCE at 32 ETH and MAX_EFFECTIVE_BALANCE_ELECTRA at 2048 ETH, plus the rationale for keeping the floor and for making the initial slashing penalty negligible. - Staking withdrawals
ethereum.org, Ethereum Foundation, 2026
Voluntary exit paths with and without gas, variable exit duration, the automatic sweep, and the 16 withdrawals per block limit equating to 115,200 a day. - Lido core contract documentation
Lido Docs, 2026
Primary evidence that execution-layer income, meaning priority fees and MEV, is collected from dedicated vaults and accounted for separately from consensus-layer rewards.
Last reviewed 2026-08
Know the terms but not sure how they apply to your project? That is what an engagement is for. We design, document, and stress-test the whole token economy inside the Tokenomics Data Room.
100+ projects advised. Complete tokenomics in 4 to 6 weeks.