The same word covers three arrangements with three different risks. What you are actually doing depends on who holds the coins.

Staking is the clearest source of yield in this subject, because you can point at what it pays for: the network needs people to lock coins and help produce blocks honestly, and it pays them for doing so.

What varies enormously is how you take part, and the word covers three arrangements that are not remotely the same.

What the network is actually paying for

On a proof-of-stake network, producing blocks requires putting coins at stake. Behave correctly and you are paid from newly issued coins and transaction fees. Behave badly — sign contradictory blocks, or go offline for long stretches — and part of your stake can be destroyed. That penalty is called slashing, and it is what makes the whole arrangement work.

So the return is real, it is paid in the network's own coin, and it typically runs from low single digits to low double digits depending on the network and how many people are staking.

Slide the picture sideways →

Native staking: you keep the keys

You delegate your coins to a validator without giving up custody. The coins stay at your address; you grant the validator the right to stake on your behalf, and they take a commission.

This is the arrangement closest to what the network intended. Nobody can spend your coins. You choose the validator and can change your mind.

The costs are real too. Your coins are locked, and unlocking takes an unbonding period — often days, sometimes weeks — during which they earn nothing and cannot be moved. If your validator misbehaves or goes offline badly, your delegated stake shares the penalty. And you have to choose a validator, which is a genuine decision rather than a formality.

Exchange staking: simplest, and custodial

You click a button on a platform. They stake on your behalf, keep a cut, and credit you the rest.

The convenience is real: no validator to choose, often no unbonding period because the platform manages a pool, and it works from an interface you already use.

But this is not really staking from your point of view. It is a deposit with a company that stakes. Your coins are theirs to move, your return is a promise on their books, and everything in the page on custody applies without modification. If the platform fails, the fact that the coins were "staked" changes nothing about your position as a creditor.

Liquid staking: a token that stands in for the stake

You deposit coins into a protocol, which stakes them and gives you a token representing your position. The token accrues the staking reward and can be moved, traded or used elsewhere while the underlying coins stay locked.

This solves the illiquidity problem cleverly, and it adds two risks that native staking does not have.

The first is contract risk: the protocol is software holding a large amount of value, and software has flaws.

The second is that the token can trade below what it represents. It is supposed to track the value of the staked coins, and usually does — but in a rush for the exit it can drop below, because everyone wants out at once and the underlying coins are still locked. That gap is the exact moment you would want to sell, and the exact moment the discount is worst.

Comparing them honestly

1. Who holds the coinsNative: you. Exchange: the platform. Liquid: a smart contract.
2. What kills the depositNative: severe slashing, which is rare. Exchange: the company failing. Liquid: a contract flaw, or selling the token at a discount.
3. How fast you get outNative: unbonding period. Exchange: usually instant, at their discretion. Liquid: instant if you sell the token, at whatever price it fetches.
4. What you have to decideNative: which validator. Exchange: whether you trust the platform. Liquid: whether you trust the code.

If you stake natively, choose the validator deliberately

Commission is the visible number and the least important one. What matters more is uptime history, whether they have ever been slashed, and whether they run their own infrastructure or rent it from the same provider as everybody else.

There is also a point most guides skip: spreading stake across smaller validators is better for the network than piling onto the largest, because a network where a handful of operators control most of the stake is less resistant to exactly the pressure decentralisation is for.