Public blockchains need a way to decide which version of history is correct without a central authority. Replacing computational work with staked capital changes the economics of that decision throughout.
What the two approaches have in common
Both make participation costly so that attempting to rewrite history requires committing substantial resources.
Both reward honest participants with newly issued units and transaction fees, so the cost of participating is expected to be recovered by following the rules.
The security argument in each case is economic rather than technical: attacking should cost more than it can plausibly return.
Where the cost sits differs fundamentally
Under proof of work, the cost is ongoing. Hardware and electricity are consumed continuously, and stopping means losing position immediately.
Under proof of stake, the cost is capital committed and locked. The resource is not consumed but is exposed, since it can be reduced as a penalty.
This converts a recurring external expense into an opportunity cost on assets held, which is why energy consumption falls dramatically under the second model.
Punishment becomes possible
A proof-of-work attacker who fails keeps their hardware and can try again, so the network can only make attacks expensive rather than punishable.
Staked capital sits inside the system, so protocol rules can destroy part of it when a validator signs conflicting versions of history.
That converts deterrence from a cost of attempting into a direct penalty for misbehaving, which is a materially different security property.
Concentration becomes the central concern
Influence is proportional to stake, so those holding more have more say, and rewards accrue in proportion to existing holdings.
Staking services aggregate holdings from many small participants, which is convenient and concentrates decision-making in a small number of operators.
The concern is not identical to mining pool concentration but rhymes with it, and both are limits on how decentralised a network is in practice rather than in design.
Why withdrawal rules matter
Staked assets are typically subject to delays before they can be withdrawn, which exists so that misbehaviour can be detected and penalised before the stake escapes.
The delay also means stakers carry price risk they cannot exit quickly, which is a genuine cost that the reward is intended to compensate.
Liquid staking arrangements issue a transferable claim against locked stake to address this, introducing an additional layer whose own risks differ from those of the base network.