Proof of stake

In brief

Proof of stake (PoS) is a family of consensus designs that select block producers and voters according to stake ownership or deposited capital rather than hash-rate contests. Ethereum’s version, live since The Merge on 15 September 2022, has validators propose or attest in 12-second slots and finalize checkpoints with a two-thirds supermajority. Solana combines delegated stake with Proof of History as a clock and Tower BFT as the vote-and-lockout protocol. Cosmos SDK chains typically elect a bonded validator set and keep stake slashable through an unbonding window. PoS is not “the richest always wins,” not zero-energy, and not automatically more decentralized than proof of work. It trades one scarce resource (work) for another (capital at risk) and adds penalties some designs can enforce on identified keys.

What proof of stake is

ethereum.org’s protocol documentation defines proof of stake as a way to prove that validators have put something of value into the network that can be destroyed if they act dishonestly. In Ethereum, that value is ETH deposited into the protocol. Validators check that new blocks are valid and, when selected, create blocks themselves. Dishonest behavior that the protocol can attribute — for example proposing two blocks for one slot, or sending contradictory votes — can destroy some or all of that stake.

That is one implementation. The shared idea across the family is:

• consensus influence is gated or weighted by stake ownership or bonded stake, depending on the protocol, not by winning a hash lottery;
• the protocol can often identify the signer of a bad message and punish bonded stake where the design supports it;
• rewards pay for being online and following the rules.

Bitcoin-style proof of work cannot slash a miner’s past electricity. It can only refuse invalid blocks and hope honest hash power outruns an attacker. PoS designs that support slashing add a different lever: destroy the attacker’s deposit. Not every PoS network slashes today. Solana’s official staking docs state that in-protocol slashing is not currently implemented, though a slashing roadmap exists.

Why stake instead of work

Proof of work makes Sybil identities expensive by requiring real-world energy and hardware for each attempt. Proof of stake ties consensus influence to scarce coin ownership or bonded stake; some designs lock that stake and can destroy part of it for attributable misconduct.

Common reasons protocols give for the switch:

• Energy. There is no Hashcash-style search. Nodes still use electricity to run computers and stay online; they do not need a mining farm. ethereum.org’s PoS page describes this as better energy efficiency relative to proof of work. That is a comparison, not “no energy.”
• Issuance. If security does not have to pay for a global hash race, the protocol may issue fewer new coins to keep participants honest. That is a design choice, not a law.
• Attribution. A validator key that double-signs can be punished. A miner who mined two competing blocks cannot have last month’s power bill clawed back.

None of this implies that PoS is automatically more decentralized. Stake can concentrate in exchanges, liquid-staking protocols and professional operators. Protocol validator counts can also hide common operators, clients, relays and hosting providers; Brown and Bautista-Gomez’s Ethereum study treats those correlations as distinct decentralization dimensions. Hardware barriers fall; capital and operational barriers remain.

A family of designs, not one algorithm

Do not treat Ethereum’s validator set as “how PoS works.”

| Design family | Who produces blocks | How disagreement is resolved | Typical punishment | | --- | --- | --- | --- | | Ethereum (Gasper: LMD-GHOST + Casper-FFG) | One randomly selected proposer per 12-second slot; committees attest | Fork choice by attestation weight; checkpoints justified/finalized at 2/3 of stake | Missed duties lose rewards; slashable offenses can destroy stake | | Solana (PoH + Tower BFT + delegated stake) | Scheduled leader for a slot; others vote | Votes and exponentially growing lockouts; PoH orders time | Rewards depend on voting performance and commission; protocol slashing not live as of official staking docs | | Cosmos SDK / CometBFT-style BFT | A bonded, often size-capped validator set | Round-based prevotes/precommits; fast finality when 2/3+ vote | Unbonding windows keep stake slashable after exit starts; parameters are chain-specific |

Validators, proposal and attestation

A validator is an identified participant that the protocol expects to produce or vote on blocks. It is not a clerk who “approves each payment.” Users still authorize their own transactions with keys. Validators package valid transactions, extend a history, and vote on which history is canonical.

On Ethereum, 32 ETH remains the minimum balance needed to activate a validator. Since Pectra activated EIP-7251, a validator configured for compounding can have an effective balance up to 2,048 ETH; that maximum is not the entry requirement. The operator runs three cooperating programs: an execution client, a consensus client and a validator client. Time is divided into slots (12 seconds) and epochs (32 slots). Each slot, one validator is selected to propose; a committee attests. Every active validator attests each epoch, not each slot.

On Solana, stake is usually delegated to a validator. More delegated stake makes a validator more likely to be scheduled as leader. Delegators and the validator share rewards after a commission the validator sets. The validator is still a vote-and-produce node, not a custodian of the delegator’s spending key if the stake account is set up correctly — though wallet and custodian products can change that.

On Cosmos SDK chains, holders typically delegate the chain’s staking token to a validator in a bonded set. Leaving the set, or undelegation, starts an unbonding period. The Cosmos Hub’s validator FAQ specifies a 3-week period for ATOM, while the SDK makes UnbondingTime a chain parameter. During unbonding, tokens generally do not earn rewards and can remain slashable for offenses from the bonded period.

Selection is not “the richest always wins”

Ethereum selects a slot’s proposer pseudo-randomly (RANDAO in ethereum.org’s walkthrough), weighted by effective balance. Before Pectra, each validator’s effective balance was capped at 32 ETH, so large operators used many validator IDs. EIP-7251 raised the per-validator maximum effective balance to 2,048 ETH while keeping the 32 ETH activation minimum. More effective balance raises selection probability; it does not let the largest depositor win every slot.

Solana weights leader schedule by delegated stake. A large operator is scheduled more often. That is influence, not a monopoly on every block.

CometBFT/Tendermint-style systems give proposers weighted turns among the active set. More than two-thirds of voting power must precommit to commit a block. The richest single validator cannot do that alone unless it already controls that supermajority.

Rewards, penalties and slashing

Rewards pay for duties done: proposing, attesting, voting. Sources of payment differ — new issuance, priority fees, sometimes MEV-related payments to a fee recipient. They are variable. They are not a savings-account rate.

Penalties for being offline are usually small and continuous: missed attestations, missed votes. ethereum.org distinguishes these from slashing.

Slashing is a larger, attributable punishment for proven contradictory behavior. On Ethereum, two primary slashable patterns are proposing multiple blocks in one slot (equivocation) and submitting contradictory attestations. The amount destroyed can include an initial penalty and a correlation penalty that grows if many validators are slashed together — so a coordinated attack is designed to be more expensive than an isolated mistake. ethereum.org describes a forced-exit timeline around those penalties.

Slashing is protocol-specific. Absence of live slashing (Solana, as officially documented) does not mean there is no operator or custody risk. Presence of slashing does not mean delegators cannot lose money some other way.

Finality and weak subjectivity

Finality means a protocol treats a block as settled unless its consensus assumptions fail. Ethereum gives that idea an economic penalty: it finalizes checkpoints (the first block of an epoch) with Casper-FFG. If two-thirds of staked ETH vote for a pair of checkpoints, the newer becomes justified and the older can become finalized. Conflicting finalized histories would make at least one-third of staked ETH slashable. If one-third of stake goes offline and finality stalls for more than four epochs, an inactivity leak drains the non-participating side until the majority can finalize again.

Tendermint-style chains often finalize each block when 2/3+ precommit. There is no long “wait six blocks” story of the Bitcoin kind; the tradeoff is liveness if the supermajority cannot talk.

Weak subjectivity is a beginner-accessible PoS caveat. Because the “right” chain is the one with the right historical votes, a very old alternative history signed by keys that have since withdrawn can confuse a node that has been offline a long time. Ethereum addresses this with weak subjectivity checkpoints: recent states that clients treat as revert limits, closer together than the withdrawal period, so a long-range fork is rejected. New or long-offline nodes need a recent checkpoint from a trusted source (client software, explorers, other nodes). That is extra social input compared with Bitcoin’s “download from genesis and check work,” not a claim that Ethereum has no objective rules after the checkpoint.

Economic security, briefly

An attacker who controlled a majority of attesting stake on Ethereum could try to make their fork the one with the most attestation weight. ethereum.org argues the community could socially refuse that fork and destroy the attacker’s stake — a defense proof of work does not have in the same form. That is a possible social response, not a guarantee written into physics.

Security still depends on:

• how concentrated stake is;
• whether slashing actually fires;
• whether users and exchanges follow the honest chain;
• client diversity (one buggy client can look like a supermajority).

PoS does not delete the 51% problem. It changes the asset the attacker must acquire and the ways the rest of the network can respond.

Delegation, pools and liquid staking

Many holders never run a validator. They delegate or join a pool. That is still protocol staking only if the coins are actually bonded to consensus. A marketing page that “stakes” your tokens into a lending market is not this article’s subject — see Staking.

Liquid staking issues a token that represents a claim on staked coins plus rewards. It adds smart-contract and operator risk and can concentrate voting power in a few protocols. It is an application layer on top of PoS, not a consensus algorithm.

How this differs from proof of work

Proof of work (see the Batch C article) spends energy on a lottery whose winner proposes a block; nodes follow accumulated work. Proof of stake assigns proposal and vote rights to bonded identities and, where implemented, slashes proven contradictions. Ethereum used the first from 2015 to 15 September 2022 and the second afterward. Bitcoin still uses the first. Solana never used Bitcoin-style mining for consensus.

They are alternative answers to “who may extend the ledger, and what does a rewrite cost?” They are not moral grades.