Proof-of-stake replaces proof-of-work's electricity bill with bonded capital. Validators lock the chain's own token, get pseudorandomly selected to propose and attest to blocks, and lose part of that stake if they sign conflicting messages. Attacking costs money the attacker already placed inside the system rather than hardware bought outside it. That is the whole trade: cheaper to run and faster to finalize, with the collateral denominated in the asset the chain issues.
Slashing is what makes a stake mean anything. Remove the penalty that destroys capital for provable misbehaviour and stake weight is just a rich list, offering nothing that proof-of-work does not already give you at a higher electricity cost.
What a validator actually commits
Ethereum is the reference implementation most people mean by proof-of-stake, so the numbers are worth quoting exactly rather than paraphrasing. A validator deposits 32 ETH into the deposit contract and runs three separate pieces of software: an execution client, a consensus client and a validator client. The deposit then joins an activation queue that rate limits new entrants.1 The consensus spec names that floor MIN_ACTIVATION_BALANCE and sets it at 32,000,000,000 Gwei.4
Time is fixed rather than found. Slots run 12 seconds, an epoch is 32 slots, one validator is randomly selected to propose in each slot, and a randomly chosen committee attests to what it sees.1 Since the Electra upgrade the spec also permits a compounding validator to hold an effective balance up to MAX_EFFECTIVE_BALANCE_ELECTRA, 2,048 ETH, consolidating into one entry what previously took 64 separate keys.4 That is a decentralization question dressed as an ergonomics feature, and it is worth tracking rather than assuming.
Finality is voted, not accumulated
Under proof-of-work you wait for a probability to decay. Under Ethereum's proof-of-stake you wait for a vote to clear a threshold. The first block of each epoch is a checkpoint. When a pair of checkpoints attracts votes representing at least two thirds of total staked ETH the more recent one becomes justified, and it becomes finalized once its own direct child is justified in turn.1
The property behind that has a name and a number. Casper's accountable safety property states that two conflicting checkpoints cannot both be finalized unless at least one third of validators violated a slashing condition, which means at least one third of the total deposit is lost.2 Ethereum's documentation restates it operationally: reverting a finalized block commits the attacker to losing at least one third of all staked ETH.1 That is a priced attack. Proof-of-work cannot offer the equivalent, because a miner who attacks and loses still owns the hardware.
Scroll to see the full diagram
The penalty is the mechanism
Slashing is a burn plus an eviction, not a fine. The spec's initial slashing penalty is the validator's effective balance divided by MIN_SLASHING_PENALTY_QUOTIENT, set at 128, so a first order slashing removes roughly 0.78 percent of a full stake before any correlation term applies.3
The correlation term is what matters at design time. PROPORTIONAL_SLASHING_MULTIPLIER scales the additional penalty by how much other stake was slashed in the same window, so one solo operator's mistake is survivable while a correlated failure across a large operator or a shared client implementation is not.3 A validator whose balance falls to EJECTION_BALANCE, 16 ETH, leaves the active set.3 The mechanism is built to be much harsher on concentration than on error, which is the most useful fact on this page if you are choosing a staking provider.
Written against proof-of-work
Both mechanisms make attacking expensive and they disagree about where the expense should sit. Proof-of-work spends outside the system on hardware and power, which is unforgeable and recurring. Proof-of-stake spends inside it, on capital the protocol can destroy, which is far cheaper to run and lets the protocol quote an attack price precisely.
Circularity is what that elegance costs. The collateral securing the chain is the asset the chain issues, so a severe loss of confidence hits the security budget and the token together, while a mining rig keeps hashing regardless of sentiment. Ethereum moved from proof-of-work to proof-of-stake in September 2022, and its documentation still lists proof-of-stake as younger and less battle tested on the same page where it argues the switch improved economic security.1 Publishing both is the correct posture, and it is the posture to copy when you write your own.
Where the pressure lands on your token
Validator rewards are paid in new issuance. Staking yield is therefore a transfer from holders who do not stake to holders who do, and the security budget is a permanent line in the supply schedule rather than a launch cost. If you are designing a token that will live on a proof-of-stake chain, none of this is yours to set, but the chain's issuance is still an input to your own dilution model and belongs in it.
If you are designing the staking layer itself, three pressures show up in nearly every engagement we run. Yield has to stay low enough that issuance is defensible to holders and high enough that participation holds. Liquid staking concentrates operator share, because convenience aggregates. And the minimum stake sets your decentralization floor: raise it and you get fewer, larger operators, lower it and you get a validator set your consensus client has to carry.
What we tell founders to settle
Three questions decide most of it. First, what finality does your product actually need? On Ethereum a checkpoint is justified after one epoch and finalized when its child is justified in the next, and at 12 second slots across 32 slot epochs that puts normal finality near 13 minutes.1 Second, what is your exposure to operator concentration on the chain you chose, and does it appear on the risk register investors read? Third, if your own token carries a staking mechanism, is it securing anything, or is it a yield feature wearing consensus vocabulary?
One caution covering all of it. Whether a staking arrangement you design is a security in any given jurisdiction is fact specific and jurisdiction specific, and that call belongs to your counsel. This page is reference material for design work. It is not legal advice and it is not a recommendation to buy, sell or hold anything.
Common questions
How much ETH do you need to run a validator?
32 ETH deposited into the deposit contract, plus an execution client, a consensus client and a validator client running alongside each other.1 The consensus spec sets that floor as MIN_ACTIVATION_BALANCE, and since Electra a compounding validator may hold up to 2,048 ETH in a single entry.4 Below 32 ETH the practical routes are a staking pool or a liquid staking token, which move your counterparty exposure rather than removing it.
What is slashing in proof-of-stake?
Slashing destroys part of a validator's stake for provable misbehaviour such as proposing two blocks for one slot or signing contradictory attestations. The spec's initial penalty is the effective balance divided by 128, with an added penalty scaled by how much other stake was slashed in the same window, and a validator falling to 16 ETH is ejected from the active set.3 Ordinary downtime is penalized separately and far more gently.
Is proof-of-stake more secure than proof-of-work?
It is differently secure. Proof-of-stake prices an attack explicitly: reverting a finalized Ethereum block commits the attacker to losing at least one third of all staked ETH.1 Proof-of-work prices it implicitly, through hardware and power nobody can fake. Ethereum's own documentation calls proof-of-stake more economically secure and, on the same page, younger and less battle tested.1 Both statements are accurate, which is the honest answer.
How long does Ethereum take to finalize a transaction?
Two epochs. A checkpoint is justified after one epoch and finalized when its child is justified in the next, and with 12 second slots and 32 slots per epoch that lands normal finality near 13 minutes.1 Whether your product waits is your decision. Many applications treat a single attested block as sufficient for small amounts and hold larger transfers to finality.
See LST and LRT Tokenomics Design for how this applies in practice.
Sources
- Proof-of-stake (PoS)
Ethereum Foundation
Validator duties, 32 ETH deposit, 12 second slots and 32 slot epochs, checkpoint finality at two thirds of staked ETH, and the one third cost of reverting finality. - Casper the Friendly Finality Gadget
Vitalik Buterin and Virgil Griffith, arXiv:1710.09437, 2017
Accountable safety: two conflicting checkpoints cannot both be finalized unless at least one third of validators violate a slashing condition. - Phase 0, The Beacon Chain (consensus-specs)
Ethereum Foundation, consensus-specs repository
MIN_SLASHING_PENALTY_QUOTIENT of 128, PROPORTIONAL_SLASHING_MULTIPLIER, EJECTION_BALANCE of 16 ETH, and SLOTS_PER_EPOCH of 32. - Electra, The Beacon Chain (consensus-specs)
Ethereum Foundation, consensus-specs repository
MIN_ACTIVATION_BALANCE at 32 ETH and MAX_EFFECTIVE_BALANCE_ELECTRA at 2,048 ETH for compounding validators.
Last reviewed 2026-08
More in Blockchain Fundamentals
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.
80+ projects advised. Complete tokenomics in 4 to 6 weeks.