Glossary · Consensus

Proof of Stake (PoS)

Published 23 May 2025 · Updated 9 Nov 2026 · 15 min read

Proof of stake (PoS) is a way for a blockchain to choose who proposes and confirms blocks in proportion to coins staked, instead of computing work. Many protocols lock that stake as collateral and can destroy it for provable misbehavior. Ethereum, Cardano, Solana, Algorand and the Cosmos Hub use it with different locking and penalty rules.

  • Stake weights proposals and votes, while every full node still checks every transaction.
  • Slashing punishes provable double signing, while going offline usually costs small penalties.
  • Ethereum's 2022 switch cut electricity use by about 99.95% but did not add capacity by itself.

What proof of stake means

A public network must stop one person from posing as thousands of voters. Proof of stake makes influence depend on committed coins rather than electricity. Its block validators propose blocks and cast weighted votes. These roles sit within blockchain consensus mechanisms, which determine agreement. Some designs lock collateral, while others allow participation without a lock.

SourceDateDefinition, in its own wordsWhat it adds
QuantumMechanic, BitcointalkDate11 July 2011Definition, in its own words“Instead of your ‘vote’ on the accepted transaction history being weighted by the share of computing resources you bring to the network, it's weighted by the number of bitcoins you can prove you own”What it addsThe earliest known use of the term: weight by coins owned, not by computing power
Sunny King and Scott Nadal, Peercoin paperDate19 August 2012Definition, in its own words“Coin age consumed by a transaction can be considered a form of proof-of-stake.”What it addsThe first deployed design. It weighted minting by coins times holding time
Kiayias, Russell, David and Oliynykov, OuroborosDateCRYPTO 2017Definition, in its own words“They instead run a process that randomly selects one of them proportionally to the stake that each possesses according to the current blockchain ledger.”What it addsLeader election in proportion to stake, with formal security proofs
Vitalik Buterin, Proof of Stake FAQDate31 December 2017Definition, in its own words“Proof of Stake (PoS) is a category of consensus algorithms for public blockchains that depend on a validator's economic stake in the network.”What it addsA short, network-neutral definition built on a deposit
NIST IR 8202, Blockchain Technology OverviewDateOctober 2018Definition, in its own words“A consensus model where the blockchain network is secured by users locking an amount of cryptocurrency into the blockchain network, a process called staking.”What it addsLocking: once staked, coins generally cannot be spent
ethereum.org, Proof-of-stake (PoS)DateUpdated 7 October 2026Definition, in its own words“Proof-of-stake is a way to prove that validators have put something of value into the network that can be destroyed if they act dishonestly.”What it addsStake that can be destroyed, for provable misbehavior such as conflicting votes

Holding more stake gives more weight in proposing and voting, not a say over which transactions are valid. Every full network node checks every block and rejects an invalid transaction. Penalties apply only to acts each protocol defines, such as double signing. Both proof of stake and proof of work resist fake identities and select block authors. Neither mechanism makes an invalid transaction valid.

Coin age means currency amount multiplied by its holding period. Peercoin introduced that weight in 2012, but it is not a general PoS rule. Its current client weights the chain by proof-of-stake difficulty, keeping coin age only for minting chance and rewards (Peercoin source code, read 8 November 2026). Ethereum, Cardano, Solana, Algorand and the Cosmos Hub weight selection by amount, without a holding-time term.

From deposit to an available balance

Ethereum illustrates a bonded PoS lifecycle, from activation through withdrawal. Other networks change the locks, queues and penalties. Since Pectra on 7 May 2025 (EIP-7251), one validator can hold up to 2,048 ETH of effective balance, which sets its weight.

Number 1

Lock or delegate

An Ethereum validator needs 32 ETH to activate. The effective balance ceiling is 2,048 ETH following Pectra in May 2025. Protocol staking differs elsewhere. Holders bond on the Cosmos Hub or delegate without locking on Cardano.

Number 2

Wait for activation

validatorqueue.com showed 944,812 ETH waiting and a wait of about 16 days 21 hours on 9 November 2026. Ethereum's activation and exit churn limit is 256 ETH per epoch at that reading.

Number 3

Get chosen

Effective balance weights the proposer draw and each attestation vote. Attestation committees are shuffled without weighting. Holding coins outside the participation rules is not enough. The next section separates lotteries from deterministic schedules.

Number 4

Propose and vote

One validator proposes per 12-second Ethereum time slot. Committees supply votes called attestations. Every active validator attests once per epoch, comprising 32 slots or 6.4 minutes, rather than every slot.

Number 5

Earn and lose

Issuance, tips and MEV reward validators. At 44.4 million ETH staked, full participation would issue at most about 1.1 million ETH a year (page team calc). Missed attestations incur small penalties. Missed proposals incur none. Slashing conditions punish equivocation.

Number 6

Exit and withdraw

On 9 November 2026 the exit queue meant a wait of about 1 day 3 hours. After exit, withdrawability takes 256 epochs, about 27.3 hours. The withdrawal sweep then pays out within about 7.1 days.

Exit rules differ across proof-of-stake networks. Unbonding on the Cosmos Hub takes 21 days in the 8 November 2026 reading. Delegators who unbond stay slashable for their validator's earlier offenses. ADA stays in the holder's wallet, and Algorand neither locks nor delegates funds. Since July 2026, Polkadot nominators unbond in about two days and are no longer slashable.

How networks choose block producers

Networks assign block production rights differently, even when validators commit the same share of all stake. This proof-of-stake model compares five rule sets side by side. Select a network and set its share between 0.01% and 10%.

Figure 1 · Page team model of published protocol rulesFive ways to draw a producerAdjust the share to compare expected slots, rounds or blocks led daily, the chance of none and how far ahead the schedule is known.
Network
1%
On Ethereum: about 444,042 ETH of the 44.4 million ETH staked on 8 November 2026On Cardano: about 213 million ADA of the 21.3 billion ADA staked on 8 November 2026On Solana: about 4.34 million SOL of the 434 million SOL staked on 8 November 2026Algorand's online stake was not read for this page. Stake there is neither locked nor delegated.On Cosmos Hub: about 3.41 million ATOM of the 341 million ATOM staked on 8 November 2026
Ethereum: RANDAO and effective balancePublic schedule
Slot
12 seconds. An epoch is 32 slots, 6.4 minutes
Draw
A RANDAO seed and a sample by effective balance pick one proposer for each slot
Weight
Effective balance, 32 to 2,048 ETH per validator since May 2025
Known in advance
For the rest of this epoch and all of the next, 32 to 64 slots, since December 2025
Empty or several
Exactly one proposer. A slot stays empty if it misses: 0.705% of slots since the Merge (page team calc)
Expected per day72 slotsOf 7,200 slots a day, each drawn by effective balance
Chance of none in a daybelow 0.01%Each slot is a separate draw
Known in advance6.4 to 12.8 minCurrent and next epoch, EIP-7917
Cardano: a private lottery for every slotPrivate draw
Slot
1 second. An epoch is 432,000 slots, exactly 5 days
Draw
Each pool runs a private VRF lottery for every slot and leads it when p < 1 - (1 - f)^σ, with f = 0.05
Weight
The pool's share of stake σ, from a snapshot taken at an earlier epoch boundary
Known in advance
Only to the pool itself. Others see it when the block arrives with its proof
Empty or several
About 95% of slots have no leader and a few have two, settled by chain selection (page team calc)
Expected per day44.3 leader slotsOf 86,400 one-second slots a day, about 4,320 with a block
Chance of none in a daybelow 0.01%Each slot is a separate private lottery
Known in advanceOnly to the poolOthers learn it from the block's VRF proof
Solana: a stake-weighted schedule per epochPublic schedule
Slot
200 ms target since 9 October 2026. An epoch is 432,000 slots, about a day
Draw
A stake-weighted random schedule for the whole epoch, seeded with the epoch number
Weight
Activated stake delegated to each validator's vote account
Known in advance
About one epoch ahead. Alpenglow's Votor, live since 21 October 2026, kept this schedule
Empty or several
One leader per slot in windows of 4 slots. A leader can miss and leave slots empty
Expected per day4,320 leader slotsOf about 432,000 slots a day at 200 ms, in windows of 4
Chance of none in a daybelow 0.01%Each 4-slot window is a separate draw
Known in advanceAbout a dayComputed from the state one epoch earlier
Algorand: private sortition every roundPrivate draw
Round
About 2.75 seconds on average, 5 to 8 November 2026 (page team calc). No epochs
Draw
A private VRF sortition each round. Several accounts can win, and nodes keep the lowest VRF hash
Weight
Online stake: the more Algo in an account, the greater its chance
Known in advance
No. A winner is known only when it reveals its proof with the proposal
Empty or several
Several proposers can win a round, and one block is certified
Expected per day314 roundsOf about 31,400 rounds a day at about 2.75 seconds
Chance of none in a daybelow 0.01%Each round is a separate draw
Known in advanceNot at allRevealed with the proposal
Cosmos Hub: weighted round robinRound robin
Block
About 5.7 seconds on average, 8 to 9 November 2026 (page team calc). No epochs
Draw
Weighted round robin: priority grows by voting power each round, and the highest priority proposes
Weight
Voting power, the ATOM bonded to a validator. Only the top 180 are active
Known in advance
Yes. Any node with the validator set computes the same proposer
Empty or several
One proposer per round. A failed round passes to the next proposer, and each height commits one block
Expected per day151 blocksOf about 15,100 blocks a day, only inside the top 180 validators
Chance of none in a day0%No luck involved once in the active set
Known in advanceAlwaysAnyone can compute the order
All five networks at this share
NetworkExpected per dayChance of none in a dayKnown in advance
EthereumExpected per day: 72 slotsChance of none in a day: below 0.01%6.4 to 12.8 min
CardanoExpected per day: 44.3 slotsChance of none in a day: below 0.01%Only to the pool
SolanaExpected per day: 4,320 slotsChance of none in a day: below 0.01%About a day
AlgorandExpected per day: 314 roundsChance of none in a day: below 0.01%Not at all
Cosmos HubExpected per day: 151 blocksChance of none in a day: 0%Always

Effective balance weights one validator's selection per 12-second slot. The schedule is known 6.4 to 12.8 minutes ahead.

A private lottery runs each one-second slot with f = 0.05. Slots can have no leader or multiple leaders.

A stake-weighted schedule is fixed about a day ahead. Each leader gets four consecutive 200 ms slots.

Each round of about 2.75 seconds uses a private VRF draw. The lowest VRF hash wins, and no one knows the winner in advance.

Deterministic weighted round robin rotates through 180 active validators. Everyone knows the order, with roughly 5.7 seconds per block.

This page team model follows each protocol's published rules, using stake readings from 8 November 2026.

Ethereum draws validators using RANDAO and effective balance. Solana samples leaders per epoch. Ouroboros Praos and Algorand use private VRF lotteries. The Cosmos Hub uses CometBFT's deterministic weighted round robin instead. Network epoch lengths differ: Ethereum has 32 slots of 12 seconds, while Cardano has 432,000 one-second slots, or five days. Solana has 432,000 slots, about a day at its October 2026 target. Algorand's rounds average about 2.75 seconds between 5 and 8 November 2026.

Public schedules support validator planning but also expose the next proposer to targeting. Private draws reveal winners with their blocks. Fusaka fixed Ethereum's proposers 6.4 to 12.8 minutes ahead from 3 December 2025 through EIP-7917. Cardano permits empty or multiple-leader slots. Solana's slot target reached 200 ms on 9 October 2026. Alpenglow's Votor replaced TowerBFT on 21 October 2026, retaining four-slot leader windows.

Main proof-of-stake design families

Designs combine selection and agreement differently. Ethereum's Gasper pairs LMD-GHOST with Casper FFG. Delegated proof of stake (DPoS) elects fixed producers, as on Tron and Vaulta, formerly EOS. Polkadot's nominated proof of stake uses Phragmén-style elections, BABE for slots and GRANDPA for finalization.

DesignHow producers are chosenWhen a block is finalExamplesTrade-off
Chain-basedHow producers are chosenA lottery for each time slot, weighted by stake: Peercoin's coin-age kernel, Cardano's private VRFWhen a block is finalProbabilistic, deeper blocks are safer. Cardano never rolls back more than 2,160 blocksExamplesPeercoin, CardanoTrade-offKeeps making blocks under heavy faults, finality grows only with depth
BFT-styleHow producers are chosenA proposer for each round, then rounds of votes weighted by stake: CometBFT round robin, Algorand VRF committees, AptosBFTWhen a block is finalAt commit, once more than two thirds of the stake has votedExamplesCosmos Hub, Algorand, AptosTrade-offFinal blocks within seconds, but a third or more offline can stop new blocks
GasperHow producers are chosenRANDAO picks one proposer per 12-second slot by effective balance, committees attestWhen a block is finalCasper FFG, after two to three epochs, roughly 13 to 19 minutesExamplesEthereumTrade-offKeeps making blocks when finality stalls, and the inactivity leak restores it
Delegated (DPoS)How producers are chosenHolders vote a fixed set of producers into office: the top 21 on Vaulta (formerly EOS), 27 Super Representatives on TronWhen a block is finalOn Vaulta, 15 of the 21 producers make a block irreversibleExamplesVaulta (formerly EOS), TronTrade-offFast, predictable rounds with few producers chosen by token votes
Nominated (NPoS)How producers are chosenNominators back validators and an election based on Phragmén-style methods spreads stake evenly. BABE assigns 6-second slotsWhen a block is finalGRANDPA, once more than two thirds of validators attestExamplesPolkadotTrade-offEven stake across validators. Nominators are no longer slashable since July 2026
Liquid delegationHow producers are chosenHolders delegate to pools without a lock, and pools win slots in proportion to delegated stakeWhen a block is finalAs in chain-based Ouroboros PraosExamplesCardanoTrade-offNo lockup and no slashing, so security rests on an honest majority of stake

Read the families by participation rights, confirmation rules and penalty exposure. Chain-based systems resolve competing histories through chain selection. Designs built around Byzantine fault tolerance instead use voting rounds to commit blocks. PoS penalties are not uniform. Double signing on the Cosmos Hub incurs 5% slashing and permanent tombstoning in the 9 November 2026 parameters. A tombstoned validator is permanently excluded. Cardano's liquid delegation has no lock or slashing, so voting weight does not imply identical financial risk.

Attacks proof of stake must answer

Nothing at stake concerns competing histories. A long-range attack instead uses old keys. PoS also needs a liveness response when too many participants stop voting. The walkthrough separates these problems and the defenses that address them.

Figure 2 · Page team walkthrough of published designsThree threats and their defensesChoose a scenario, toggle its matching defense, and follow four steps showing how penalties, checkpoints or balance reductions change the outcome.
Attack
Slashing
1 of 4
1. Two forks appearForkA network split or an attacker leaves two competing chains. Validators could support either one.
Rule that applies
Fork choice follows the chain with the most stake behind it
2. Validators back bothNo costWith rewards and no penalties, a validator gains most by building on every fork at once, just in case.
Rule that applies
Early chain-based designs such as Peercoin paid rewards with no penalties (Buterin, 2017)
3. No fork loses supportUnresolvedSigning costs nothing, so every fork keeps its backing and the chain has no reason to settle on one history.
Rule that applies
A pragmatic validator is equally happy to back any or several forks (ethereum.org)
4. The attacker gainsAttack worksHonest but pragmatic validators lend weight to every fork, including one built by an attacker.
Rule that applies
Without a penalty, signing twice is free
2. Two signatures are evidenceEvidenceTwo conflicting signed blocks or votes from one key prove misbehavior, and anyone can submit the proof.
Rule that applies
Ethereum slashes two blocks in one slot, double votes and surround votes (consensus specs)
3. The deposit is cutSlashedThe validator loses 1/4096 of its effective balance at once, more if many are slashed together, and is ejected.
Rule that applies
Correlation penalty of three times the slashed share on day 18, withdrawable on day 36 (Electra spec)
4. One chain winsDefendedValidators sign one block per slot and one vote per target, so support gathers on one fork and it becomes final.
Rule that applies
Finality conditions and slashing keep one canonical chain (ethereum.org)
1. Validators exitKeys without stakeValidators withdraw their deposits. Their old signing keys still exist, but nothing is at stake behind them any more.
Rule that applies
Once the deposit is withdrawn, nothing stops a vote on a long-range fork (Buterin, November 2014)
2. Old keys build a forkAttack worksAn attacker gathers those old keys and builds another history that starts far back, before the keys left.
Rule that applies
Validators who withdrew could still produce a conflicting fork (ethereum.org)
3. A new node joinsConfusedA node that starts from genesis, or returns after a long time offline, sees two histories and both look valid.
Rule that applies
New or long-offline nodes may not know that the attackers withdrew (ethereum.org)
4. It follows the wrong chainAttack worksWith no outside information the node can be tricked, and slashing cannot reach validators who already left.
Rule that applies
A fork from arbitrarily long range is a basic threat to proof of stake (Buterin, 2014)
3. Start from a checkpointCheckpointThe new node starts from a recent trusted state, a weak subjectivity checkpoint, from its client, an explorer or other nodes.
Rule that applies
Clients let users enter a weak subjectivity checkpoint at startup (consensus specs)
4. The old fork is ignoredDefendedBlocks before the checkpoint cannot change and clients never reorg finalized blocks, so the long-range fork is rejected.
Rule that applies
Weak subjectivity period of about 3,535 epochs, about 15.7 days at the 8 November 2026 stake (page team calc from the spec)
1. A third goes offlineOfflineValidators holding a third or more of the stake stop voting, through an outage or on purpose.
Rule that applies
Finality needs votes from at least two thirds of the stake
2. Finality stopsNot finalizingBlocks are still proposed, but no checkpoint gets a two-thirds vote, so nothing new becomes final.
Rule that applies
With a third or more not attesting, a two-thirds supermajority cannot exist (ethereum.org)
3. The math stays stuckStuckIf the offline stake still counts in full, the online validators can never reach two thirds of the total again.
Rule that applies
Finality depends on the share of all active stake that votes
4. Finality never returnsNot finalizingBlocks keep coming, but anything that waits for final blocks keeps waiting for as long as the outage lasts.
Rule that applies
No checkpoint can be finalized while the online stake stays below two thirds
3. The leak startsLeakingAfter four epochs without finality, the stake of validators that do not vote starts to bleed away, a little each epoch.
Rule that applies
The leak starts once the finality delay passes 4 epochs (consensus specs)
4. Finality returnsRecoveredOnce the inactive stake falls below a third of the total, the online validators hold two thirds and finalize again.
Rule that applies
Inactive stake bleeds away until it is under one third (ethereum.org)

Without penalties, validators can collect rewards across competing forks. Supporting every branch leaves each fork with backing.

Conflicting validator signatures provide evidence for slashing. Cutting the deposit removes the incentive to support competing histories and supports one canonical chain.

A long-range attack uses old keys to rebuild history. A new node cannot identify the genuine chain from those signatures alone.

A recent checkpoint excludes the old fork. Refusing to reorganize finalized blocks keeps that alternative history outside the accepted chain.

With a third or more offline, checkpoints cannot gather the required two-thirds support. Finality stops even while blocks continue.

After four epochs without finality, inactive balances shrink. Once active participants regain the required two-thirds weight, finalization can resume.

This page team walkthrough explains published designs. It is not a simulation of a real attack.

Buterin's January 2014 Slasher proposal and 2017 FAQ describe penalties for nothing at stake. Signed validator equivocation supplies evidence, unlike mere disagreement. His November 2014 post explains weak subjectivity as an answer to a long-range attack. Ethereum nodes obtain a recent checkpoint from client software, explorers or other nodes. Using the 8 November 2026 balance reading, the page team calculates a weak subjectivity period of about 15.7 days. The checkpoint anchors which history the node accepts.

Cardano uses key-evolving signatures, deleting old key material as keys advance. The Cosmos Hub's light client verification uses a trusting period shorter than its 21-day unbonding period. Ethereum's inactivity leak addresses missing votes rather than forged history. After four epochs without finality, inactive validator balances gradually decline. This changes relative voting weight until active participants can resume finalization. A temporary halt therefore has a different defense from conflicting signatures or a fabricated old chain.

What an attacker's stake share allows

An attacker's influence depends on validator voting rules, not just a percentage. Move the share from 1% to 70% to compare three networks. The proof-of-stake model translates that share into coin amounts, capabilities, limits and penalties for double signing.

Figure 3 · Page team model of protocol thresholds and penaltiesThresholds change the possible damageCompare the three networks at the same share, then inspect which actions become possible and which remain excluded.
34% of all stake
Ethereum (Gasper)Cannot stop finalityCan stop finalityCan control fork choiceCan finalize its own chain
Stake at this share
15.1 million ETH of 44.4 million
What it can do
Withhold its votes or try short reorgs, which the honest majority outvotes
What it cannot do
Stop finality, control the fork choice or change finalized blocks
What it can do
Stop finality by not voting. With network delays, 34% could cause double finality by signing twice
What it cannot do
Avoid the inactivity leak, which drains its offline stake until the rest holds two thirds
What it can do
Dominate the fork choice, censor transactions and reorder recent blocks
What it cannot do
Finalize its own chain alone, or revert finality without being slashed
What it can do
Finalize the chain it prefers, and revert finality by double voting
What it cannot do
Stop the community from adopting an honest fork by social consensus
If caught double signing
100% of the slashed stake, about 15.1 million ETH
ethereum.org, proof-of-stake attack and defense (updated 13 April 2026), and the Electra spec. 44.4 million ETH staked on 8 November 2026.
Cosmos Hub (CometBFT)Cannot halt or forkCan halt the chainCan halt the chainCan commit any block
Stake at this share
116 million ATOM of 341 million
What it can do
Miss votes, at the risk of a 0.01% downtime slash and a 10-minute jail
What it cannot do
Halt the chain or commit conflicting blocks
What it can do
Halt new blocks by refusing to vote, since a commit needs more than two thirds
What it cannot do
Commit blocks on its own. Signing two blocks at one height leaves evidence
What it can do
Halt new blocks by refusing to vote. CometBFT has no special one-half threshold
What it cannot do
Commit blocks on its own, without the honest third
What it can do
Commit any block, because more than two thirds of the votes is all a commit needs
What it cannot do
Hide double-signing evidence from the rest of the network
If caught double signing
5% of the bonded stake, about 5.8 million ATOM, and the validator is tombstoned for good
CometBFT spec and docs, and the Hub's slashing parameters read on 9 November 2026. 341 million ATOM bonded on 8 November 2026.
Cardano (Ouroboros Praos)Honest majority holdsHonest majority holdsCan outgrow the honest chainCan outgrow the honest chain
Stake at this share
7.25 billion ADA of 21.3 billion
What it can do
Win its share of slots and cause short forks that chain selection resolves
What it cannot do
Outgrow the honest chain while honest stake holds the majority
What it can do
Win its share of slots. Praos has no special one-third threshold
What it cannot do
Outgrow the honest chain while honest stake holds the majority
What it can do
Build a longer chain than honest pools over time and censor transactions
What it cannot do
Make online nodes roll back more than 2,160 blocks
What it can do
Control which blocks join the chain. Nothing new happens at two thirds
What it cannot do
Make online nodes roll back more than 2,160 blocks
If caught double signing
Nothing. Cardano has no slashing and delegated ADA is never at risk
Ouroboros Praos abstract and the Cardano Developer Portal (updated 31 July 2026). 21.3 billion ADA of active stake, epoch 666.

Below one third, no network loses finality or safety. Ethereum double signers holding 10% of stake would each lose about 30% (page team calc).

At one third, an attacker can stop Ethereum's finality or halt the Cosmos Hub. Cardano still has an honest majority.

Above one half, Ethereum's fork choice can be dominated. Cardano's honest chain can be outgrown from one half.

From two thirds an attacker can finalize its chain on Ethereum, and with more than two thirds it can commit any block on the Cosmos Hub.

This page team model applies protocol thresholds and penalty rules using stake readings from 8 November 2026.

Gasper and BFT-style voting hinge on thirds, while Ouroboros Praos requires an honest majority. The page team calculates Ethereum's one-third threshold at about 14.8 million ETH on 8 November 2026. The correlation penalty can consume each validator's entire deposit when a third is slashed together. The Cosmos Hub applies 5% plus tombstoning for double signing in its 9 November 2026 parameters. Cardano applies none. For severe Ethereum attacks, ethereum.org notes that the community can adopt an honest fork.

Proof of stake versus proof of work

Proof of work commits energy, while proof of stake commits coins under protocol rules. Their different costs shape participation and recovery from attacks. The comparison separates resource use from transaction capacity, rather than treating either consensus mechanism as a complete scaling design.

TopicProof of stakeProof of work
Scarce resourceProof of stakeCoins staked as collateral, which many protocols lock and can destroyProof of workEnergy and specialized hardware
Who makes the blockProof of stakeA stake-weighted draw or rotation, see figure 1Proof of workThe first miner to find a valid hash
FinalityProof of stakeEconomic finality on Ethereum and BFT-style chains, probabilistic on chain-based designsProof of workProbabilistic, growing with each block on top
EnergyProof of stakeEthereum after the Merge: 2,600.86 MWh a year (CCRI, September 2022)Proof of workEthereum before the Merge: 22,900,320 MWh a year (CCRI, September 2022)
HardwareProof of stakeSolo staking on Ethereum: a 4 TB NVMe SSD, 64 GB of RAM and steady uptime (ethereum.org)Proof of workSpecialized mining equipment
Entry costProof of stakeStake: 32 ETH for one Ethereum validator, or a share in a poolProof of workHardware and power, no coins needed
Attack and recoveryProof of stakeAttacker stake can be slashed, leaked away or left behind by a social forkProof of workAn attacker can repeat the attack while its hash power lasts
RewardsProof of stakeIssuance, tips and MEV on Ethereum, with the base fee burnedProof of workBlock subsidy plus fees, with a 3.125 BTC subsidy per block on Bitcoin since April 2024
LockupsProof of stakeEntry and exit queues and unbonding periods, 21 days on the Cosmos HubProof of workNone in the protocol
Centralization pressureProof of stakeStake pools, liquid staking and custodians hold large sharesProof of workEconomies of scale in mining hardware and power

CCRI's September 2022 estimate puts Ethereum's annual electricity use at 22,900,320 MWh before the Merge transition and 2,600.86 MWh afterward. The figure on ethereum.org is about 99.95%. That consensus change does not itself add throughput. Compared with proof of work, PoS brings lockups, delegation risk and weak subjectivity. The same site calls proof of stake younger and less battle-tested than proof of work, and more complex to implement. Validators still need reliable hardware and steady uptime.

How proof of stake developed

The idea moved from a forum post in 2011 to Ethereum's consensus in September 2022. Research on penalties, randomness and finality shaped that transition, while other networks adopted different ways to select producers and agree on blocks.

  1. July 2011

    An alternative to computing work

    The earliest known use of the term is a Bitcointalk post of 11 July 2011. It proposed weighting votes by provable bitcoin ownership rather than computing resources. The proposal challenged proof of work's resource model.

  2. August 2012

    Peercoin combines two mechanisms

    Peercoin launched in August 2012, combining proof of work with coin-age proof of stake. Its paper defined coin age as currency amount multiplied by holding period. Nxt launched as a proof-of-stake currency on 24 November 2013.

  3. January 2014

    Penalties and trusted checkpoints

    Slasher proposed penalties for signing on two chains on 15 January 2014. Buterin's November 2014 post later explained why withdrawn deposits created a need for weak subjectivity.

  4. October 2017

    Research formalizes different designs

    Casper FFG appeared on arXiv on 25 October 2017. Ouroboros was presented at CRYPTO 2017, following Algorand's arXiv paper of 5 July 2016.

  5. March 2019

    A voting-based network launches

    The Cosmos Hub launched on 13 March 2019 using Tendermint BFT proof of stake. Its voting-based design provided a different route to agreement than chain-based selection.

  6. December 2020

    Ethereum starts its parallel chain

    Ethereum's Beacon Chain started on 1 December 2020. Polkadot moved to nominated proof of stake in June 2020, and Cardano's Shelley fork introduced pools in July 2020.

  7. September 2022

    Ethereum changes its consensus

    The Merge moved Ethereum from proof of work to proof of stake on 15 September 2022. Withdrawals of staked ETH followed on 12 April 2023.

  8. October 2026

    Solana changes its voting protocol

    Solana replaced TowerBFT with Alpenglow's Votor on 21 October 2026. The upgrade retained four-slot leader windows, separating the voting change from the existing producer schedule.

Limits of proof of stake

Participation involves more than block rewards. Proof of stake exposes different parties to custody, timing and software risks. These limits depend on the network and service arrangement, so pooled participation does not carry one uniform risk profile.

Concentrated voting weight

One liquid staking protocol accounts for 21.20% of Ethereum's stake on 9 November 2026. Liquid staking as a whole holds around a third of all staked ETH, according to ethereum.org.

Custody changes control

ethereum.org warns that not everything marketed as staking is protocol participation. A custodian controls withdrawal keys and assets. Insolvency or frozen withdrawals can leave nothing redeemable onchain.

Exits take time

On 9 November 2026 an Ethereum exit meant about 1 day 3 hours in the queue and 27.3 more hours. The withdrawal sweep then took up to about 7.1 days.

Shared penalty exposure

Delegators on the Cosmos Hub share their validator's slashing losses, and liquid staking losses typically spread across token holders. Cardano has no slashing. Polkadot nominators are exempt following the July 2026 change.

Checkpoints anchor new nodes

New or long-offline nodes need a trusted recent checkpoint. Ethereum's weak subjectivity period is about 15.7 days in the page team's 8 November 2026 calculation.

Risks can accumulate

Liquid staking tokens depend on smart contracts and can trade below their backing. Restaking adds exposure because slashing in one service can weaken others.

Building products on proof of stake

We work on staking smart contracts and platforms, protocol-level consensus, and applications on proof-of-stake networks. These are different engineering tasks. Application logic pools tokens and distributes rewards, while consensus code determines agreement between nodes. The distinction helps define whether a project changes a product's behavior or the underlying protocol.

PixelPlex case icon

PixelPlex was founded in 2007 and has built blockchain software since 2013. The Qtum implementation case documents our protocol-level proof-of-stake work on a Bitcoin fork. Its UTXO-based platform uses PoS v3.0 consensus, distinct from a smart contract that distributes application rewards.

  1. Platforms and reward logic

    Our DeFi staking platform development covers staking logic, reward distribution and user interaction. This is application-level work, not network consensus.
  2. Protocol-level consensus

    For Qtum, we implemented PoS consensus on a Bitcoin fork. The work concerns a UTXO-based blockchain and dApp platform.
  3. Committee consensus design

    We designed EchoRand for our Echo protocol. It selects a random, balance-weighted committee from network actors for a single block.
  4. Client staking pools

    For a community-governed DeFi platform, we wrote and fixed contracts and deployed pools on Ethereum, BNB Smart Chain and Polygon.
  5. Applications and wallets

    Our Ethereum development services include application-level staking. Our Substrate development services include web wallets with staking capabilities.

Discuss your staking product or protocol requirements

We can discuss the participation model, reward rules and withdrawal behavior your product or protocol needs to support.

Discuss your project

Proof of stake FAQ

What is proof of stake in simple terms?

In networks such as Ethereum, participants lock coins as collateral. The network uses that stake to choose block producers and weight votes, instead of relying on computing work. Provable cheating, such as signing conflicting blocks, can cause some or all of the deposit to be destroyed.

How does it differ from proof of work?

Proof of work ties block production to computing effort and energy expenditure. PoS ties participation weight to committed coins. Proof of work offers probabilistic confirmation, while Ethereum uses crypto-economic finality. Neither label alone describes a network's capacity, hardware needs or complete security design.

Does Ethereum staking always require 32 ETH?

No, pools allow participation with less than the amount needed to activate an individual validator. Activation requires 32 ETH. Since Pectra on 7 May 2025, a validator can carry up to 2,048 ETH of effective balance under EIP-7251. Pool arrangements introduce their own terms and risks.

Can participants lose coins they have staked?

Yes, but the rules depend on the network and the service. A validator's equivocation can trigger slashing, while missed duties can cause smaller penalties. Cardano has no slashing. Custody and smart contract exposure are separate risks, even when protocol penalties do not apply.

What does the nothing at stake problem mean?

A validator may benefit from supporting every competing chain when doing so brings rewards without penalties. This is the nothing at stake problem. Conflicting signed messages let some PoS protocols punish that behavior, making support for incompatible histories economically costly.

Which blockchain networks use proof of stake?

Ethereum, Cardano, Solana and Algorand use different PoS designs. Polkadot uses nominated proof of stake, while Tron and Vaulta, formerly EOS, use elected producers. Their selection rules, lockups and penalty arrangements differ, so the shared label does not imply identical operation.

Is staking considered a security in the US?

In March 2026, the SEC issued an interpretation, joined by the CFTC, that certain protocol staking activities do not involve securities offers or sales. It addresses solo, self-custodial, custodial and liquid staking arrangements as described, but not restaking. It followed staff statements of May and August 2025. This page is not legal advice.
Sources (35)

Get updates about blockchain, technologies and our company

We will process the personal data you provide in accordance with our Privacy policy. You can unsubscribe or change your preferences at any time by clicking the link in any email.

Follow us on social networks and don't miss the latest tech news

  • facebook
  • X
  • linkedin
  • instagram
Stay tuned and add value to your feed