Crypto staking: what the rewards are really worth

There is a useful thought experiment for anyone new to crypto staking. Imagine the platform you are using disappears overnight. With a savings product, so does your money, because the yield was somebody's promise. With staking, the blockchain carries on issuing rewards to whoever is validating, because the yield was never a promise — it was a subsidy the network pays for security. That difference is why staking survives market cycles that destroy lending platforms, and it is also why almost every argument about staking is really an argument about the middleman.

This page is about the mechanism: what a validator does, what it can be punished for, how long you are locked in, and how much of the reward the operator keeps. The per-asset numbers live on our staking rewards comparison, Ethereum's specifics on Ethereum staking, and receipt tokens onliquid staking. All data was checked on 16 September 2026.

What a validator is actually paid for

A proof-of-stake network needs a set of participants who agree on what happened and can be punished if they lie. Staking is the bond. You lock tokens; in exchange you get the right to propose and attest to blocks, and the protocol pays you newly minted tokens for doing it correctly. If you equivocate — sign two conflicting versions of history — the bond is what the network takes.

On Ethereum the arithmetic is explicit. Consensus-layer rewards are split by weight out of 64 across timely source, target and head votes, sync committee duty and block proposal, and a validator's base reward scales with the inverse square root of the total amount staked. That last detail has a consequence most staking guides skip: doubling the total stake cuts the per-validator yield by about 29%. Ethereum's rate did not fall because anything went wrong. It fell because more than 43 million ETH — around 35% of supply — is now staked, and the formula does exactly what it was written to do.

On top of protocol issuance sit execution-layer rewards: priority fees and maximum extractable value, paid to whoever proposes the block. These are lumpy and demand-driven rather than issued, and estimates of their contribution vary enormously — a commonly cited range puts MEV at 10% to 30% on top of base rewards, roughly 0.28 to 0.83 percentage points of APR at current parameters, while CoinShares has argued that in weak on-chain demand execution rewards can reach around 80% of total returns. Treat those two as the ends of a wide, volatile range rather than a settled number.

Four ways to stake, and what each one costs

Every staking product on the market is one of four structures. They differ in minimum size, in who holds the keys, and in how much of the reward reaches you.

Solo staking is free in fee terms and expensive in every other term: hardware, uptime, and the personal consequences of a misconfiguration. Delegated staking keeps your withdrawal keys with you while an operator runs the machine, and since Ethereum's Pectra upgrade introduced execution-layer triggerable exits you no longer depend on that operator holding your signing keys in order to leave. Pooled and liquid staking swaps the 32 ETH threshold for smart-contract exposure and a receipt token. Exchange staking swaps everything for a checkbox.

What a staking pool changes, and what it does not

Pools exist for one reason: 32 ETH is a large ticket, and most people do not have it. A pool aggregates many small deposits into whole validators and splits the rewards pro rata, usually issuing a receipt token so the position stays usable elsewhere. Lido is the largest by a distance, holding roughly $23.3 billion and about 56.5% of the liquid staking market with a 10% fee on rewards; Binance's WBETH is second at about $8.9 billion, and Rocket Pool — the most decentralised of the majors — sits at around $1.25 billion. Roughly 17.19 million ETH, about 40% of all staked ETH, is staked through one of these.

What pooling changes is the entry ticket and the liquidity. What it does not change is the underlying reward. Net rates on the major receipt tokens cluster between about 2.05% and 2.73%, and the spread between them is almost entirely fee differential rather than validator skill. What pooling adds is smart-contract exposure — the pool's code now sits between you and your stake — and, for the liquid variety, the possibility that the receipt trades below the asset it represents. The full account of that, including why the 2022 and 2025 depegs had completely different causes, is on our liquid staking page. The important point here is simply that "staking pool" describes a distribution method, not a better yield.

Commission: the number that decides your return

Nothing else on this page moves the outcome as much as the commission. The market range is absurd — from nothing to nine-tenths of the reward — and the endpoints belong to companies selling the same underlying service.

Put the two ends of that table against a real rate. Ethereum's network APR was 2.46%. A 10% protocol fee costs about 25 basis points; a 35% exchange commission costs about 89 basis points. Against a real, inflation-adjusted ETH yield of roughly 1.66%, the 10% fee takes about a seventh of your actual economic return and the 35% fee takes more than half of it. The stakefish structure is worth singling out because it is marketed as a zero fee: 0% on consensus rewards and 50% on MEV and tips can cost more than a flat 10% whenever execution rewards are a large share of the total.

Nominal yield, real yield, and the arithmetic in between

A staking APR tells you how fast your token balance grows. It does not tell you how fast your claim on the network grows, because the protocol is minting tokens for every other staker at the same time. To get from one to the other:

real_yield = (1 + nominal_APR) ÷ (1 + inflation_rate) − 1

The common shortcut — subtracting inflation from the nominal rate — overstates the answer, and the error grows with the numbers. Cosmos is the worked example. ATOM advertises one of the highest staking rates in the market, around 19.6% nominal, against roughly 12.67% annual token issuance. Subtraction gives 6.97%. The correct calculation gives 6.19%. That 78 basis point gap is larger than the entire real yield available on several other chains.

Two refinements matter. First, be explicit about which figure you are quoting: the real yield above measures your purchasing power in token terms, while the gain relative to somebody who holds the same asset and does not stake is closer to the nominal rate, because they absorb the dilution and you do not. Second, Ethereum is the only major asset here whose issuance is demand-dependent — its 0.88% figure is net of the EIP-1559 burn and can turn negative when block space is in heavy demand.

The uncomfortable corollary is that a high advertised APR often signals high issuance rather than a good deal. Sui advertises 1.49% against 2.54% inflation, which is a negative real yield: stakers are diluted in token terms while watching their balance rise. We rank thirteen assets on this basis, and flag where the underlying data is disputed, in thereal-yield table.

Unbonding: the risk that actually bites

Slashing gets the attention. Illiquidity does the damage. Unbonding periods exist so a chain can still punish a validator for something discovered after it stopped validating, and they vary by more than an order of magnitude between networks.

Ethereum is a special case because there is no fixed period — there is a queue, and it moves. On 16 September 2026 the exit queue held 1,056 ETH and cleared in about 26 minutes, while the entry queue held 1,831,267 ETH and ran to roughly 31 days and 19 hours. Twelve months earlier the position was reversed. Anyone planning around a specific exit date should check the queue on the day rather than trusting a number in an article, ours included.

Slashing: rare, real, and usually somebody's mistake

Slashing on Ethereum punishes equivocation, not unreliability. There are exactly three offences: proposing two different blocks at the same height, casting two conflicting attestations for the same target epoch, and a surround vote. Being offline is not slashable — it is merely unrewarded and mildly penalised.

The penalty has three parts. An initial charge, reduced from one thirty-second of effective balance to one four-thousand-and-ninety-sixth by the Pectra upgrade. Attestation penalties throughout a forced exit whose withdrawability sits about 36 days out. And a correlation penalty applied at roughly the 18-day midpoint, calculated from the total effective balance slashed across a surrounding 36-day window. For an isolated incident the correlation penalty rounds to nearly nothing. In a mass correlated event it can approach the entire balance. That tail — not the base rate — is the real reason operator diversity matters.

The largest recent event illustrates the point better than any risk model. On 10 September 2025, 39 validators were slashed. The cause was not an attack: it was traced to distributed-validator clusters, an Ankr maintenance action, and a cluster migrated from Allnodes two months earlier where a leftover secondary signer produced duplicate signatures. One validator holding 2,020 ETH lost about 0.3 ETH, worth roughly $1,300 at the time. Essentially all historical slashing has the same shape: duplicate signing setups and bad migrations, not adversaries.

And for most readers it is not the relevant risk at all. Of the major proof-of-stake assets, Cardano, Avalanche, NEAR, Tron, Sui, Aptos, Algorand and Polygon have no active slashing mechanism. On those chains a validator that fails simply earns you less. If a platform is selling you "slashing protection" on an asset that cannot be slashed, you have learned something about the platform.

Where the law currently stands

US permission for staking rests on agency interpretation rather than statute, and that is the single most important thing to understand about its legal status. The SEC's Division of Corporation Finance stated on 29 May 2025 that solo, self-custodial, delegated and custodial protocol staking are not securities transactions, reasoning that a provider's role is administrative or ministerial and so fails Howey's "efforts of others" prong; Commissioner Caroline Crenshaw dissented. A staff statement on 5 August 2025 extended similar treatment to liquid staking receipt tokens, subject to conditions, while explicitly excluding restaking and provider-guaranteed returns. On 17 March 2026 a joint SEC–CFTC Commission-level interpretation, published at 91 FR 13714, set out an asset taxonomy and addressed protocol staking directly — materially stronger than a staff statement, and still interpretive guidance rather than law. The Digital Asset Market CLARITY Act, which would have supplied the statute, failed a Senate cloture vote 49–50 on 15 September 2026. We track all of this onregulation of crypto yield.

In the European Union, MiCA never uses the word staking; staking-as-a-service is captured through the regulated services it involves — custody, administration and transfer — with a compliance deadline of 1 July 2026, while solo and protocol-level staking without an intermediary falls outside its scope. In the United Kingdom, a statutory instrument in force from 31 January 2025 excludes qualifying cryptoasset staking arrangements from the definition of a collective investment scheme, and the FCA's Policy Statement PS26/11 of 30 June 2026 amended its conduct rules to avoid unintentionally restricting auto-staking.

How to read any staking offer

Four numbers decide what a staking product is worth, and they are rarely presented together. The protocol's own rate, which you can check independently. The commission, which you often cannot. The proportion of your balance actually deployed — a question almost nobody thinks to ask until they read Kraken's flexible staking clause. And the network's issuance, which converts a nominal figure into a real one.

Work through those in order and most of the market sorts itself out quickly. A 12% advertised rate on a chain issuing 12.67% is not a good deal; a 2.46% rate on a chain issuing 0.88% net might be. A 0% fee that applies only to consensus rewards is not a 0% fee. And a product that pays on half your balance is paying half the rate, whatever the banner says. If you want the same analysis applied to a single asset, start with Ethereum orSolana; if you want to know how staking compares with the two other ways of earning on the same capital, we set it againstsavings products and againstyield farming directly.