Proof of stake: the bond a validator posts
Staking swaps an external energy cost for capital locked inside the system that the protocol can destroy when a validator signs two histories.
Proof of stake replaces an external cost with an internal one. Instead of proving that electricity was burned outside the system, a block producer posts capital inside it, in a form the protocol itself can confiscate if the producer signs two conflicting versions of history. The security argument is the same shape as proof of work — attacking has to be expensive — but the expense is a forfeited bond rather than a power bill.
What is actually at stake
A validator is not a person; it is a deposit plus a signing key plus a piece of software that is expected to be online. To join, an operator locks a fixed quantity of the native asset — 32 ether on Ethereum, other amounts or no fixed size on other chains — and the deposit becomes subject to the protocol's penalty rules. Exiting is not instant: withdrawal queues exist specifically so that a validator cannot misbehave and leave before the evidence is processed.
Slashing is the mechanism that gives the bond teeth, and it applies to a narrow set of provable offenses rather than to poor performance generally. Signing two different blocks for the same slot, or two attestations that contradict each other about the chain's history, produces a permanent, publicly verifiable record of equivocation, and the protocol destroys part of the deposit and ejects the validator. Simply being offline is treated differently: the penalty is a small drip of missed rewards, scaled up under an inactivity mechanism only if the chain is failing to finalize because too many validators are absent at once.
Choosing who proposes
Turns are assigned rather than won. A pseudorandom process, seeded from values the protocol itself produces, selects a proposer for each slot with probability proportional to stake, and assigns other validators to committees that attest to what they see. Ethereum runs twelve-second slots grouped into epochs of thirty-two slots; a chain using a classical Byzantine fault tolerant design instead has every validator vote on each block before it is considered done; a delegated proof of stake chain elects a small fixed set of producers by token vote. Some designs add a separate clock, such as proof of history, so that validators can agree on time before agreeing on content.
Because turns are scheduled, the marginal cost of producing a block is close to zero. That is the source of the design's efficiency and also of its central difficulty: with no physical cost per attempt, nothing physical stops a validator from supporting several competing histories at once.
Nothing at stake, and the answer to it
The classical objection is called nothing-at-stake. Under proof of work, hash power spent on one branch cannot also be spent on another, so miners are forced to choose. Under naive proof of stake, a validator could sign on every branch and collect rewards whichever one wins. Slashing answers this directly: signing on two branches produces the evidence that destroys the bond, so the cost of hedging is made explicit rather than left to physics.
A second and less advertised consequence is weak subjectivity. A node syncing a proof-of-work chain from scratch can identify the correct chain purely from data, because work is objectively measurable. A node syncing a proof-of-stake chain from the genesis block after a long absence cannot always do so, because validators who have since withdrawn their stake could sign an alternative history at no cost to themselves. The practical fix is a recent checkpoint obtained from a source the operator trusts. This is a real, small, permanent trust assumption, and it is worth knowing that it exists rather than discovering it later.
Delegation: who holds the bond, and who holds the risk
Most holders do not run validators. Delegation lets them assign their stake's weight to an operator who runs the infrastructure and takes a commission. Liquid staking goes further, issuing a transferable receipt token representing the staked position, which can then be used elsewhere while the underlying asset stays locked. Restaking extends the same bond to secure additional systems, which adds their slashing conditions to the original ones.
Each layer separates economic exposure from operational control, and each concentrates it. The Nakamoto coefficient and validator count describe the distribution of that control; validator concentration is the risk term for what happens when a handful of operators run a large share of the deposits. Concentration matters for censorship and for correlated failure — a bug or an outage at one large operator can affect many deposits simultaneously — not only for the classical majority-attack scenario.
Staking yield is not a dividend and not interest
The number published as nominal staking yield has two components with different economics. Part of it is newly created supply: issuance paid to validators for doing the job. That portion is not income from an outside party. It is a transfer from every holder who is not staking to every holder who is, achieved through dilution, and its arithmetic is closer to a stock split paid selectively than to a dividend. A company's dividend distributes cash earned from customers; protocol issuance creates units from nothing and hands them to a subset of holders.
The other part comes from transaction fees and priority payments made by actual users, which is external revenue in the ordinary sense. Real staking yield subtracts the supply growth from the nominal figure to isolate what is left after dilution, and it can be materially lower than the headline. Two further caveats belong with any yield figure: it is denominated in the native asset, so it says nothing about value in any other currency, and it is gross of operator commissions, slashing losses, and lock-up constraints. None of that makes staking good or bad; it makes the headline number a different quantity from the one an equity investor would assume.
The staking ratio shows what share of supply is bonded, which links this lesson to both security and liquidity. Continue with the lesson on finality to see what those bonds actually promise about settlement, and use the staking pages and supply data to see how issuance and lock-up are measured for individual networks.