Staking locks capital. Liquid staking unlocks it by issuing a token that represents the staked position — deposit ether, receive a receipt that accrues staking rewards and can be traded, lent or posted as collateral while the underlying stays bonded. The idea is straightforward and the consequences are not.
The first consequence is that the receipt has a market price, and that price is not the same thing as the value of the underlying. Redemption ultimately enforces the relationship, but redemption is gated by the network's exit queue, which is measured in days when many validators leave at once. In the interval, the receipt trades where the market puts it, and it has traded at a discount during periods of stress. A leveraged position collateralised by a receipt token can therefore be liquidated by a discount that later closes, which is a loss caused by the wrapper rather than by the asset.
The second is concentration. Liquid staking has strong network effects — the largest receipt token is the most liquid and the most widely accepted as collateral, which attracts more deposits. A single protocol controlling a large share of a network's stake is a governance and censorship concern for that network, whatever its operators intend, and it is the reason the largest providers have publicly discussed self-limiting.
Restaking extends the idea in a different direction. Rather than making staked capital liquid, it makes it reusable: the same stake secures the base protocol and, simultaneously, additional services — oracle networks, data availability layers, bridges — that lack the capital to bootstrap their own validator set. Operators opt in, and each service can impose its own slashing conditions on the stake.
The efficiency is real and so is the new risk, and it has a specific shape. Slashing conditions written by a service are not reviewed by the base protocol, so a bug in one service's slashing logic can destroy stake that was securing the base chain. Because the same capital backs several obligations, a fault in one place produces losses in several at once — the definition of correlated risk. And because restaked positions are themselves wrapped into receipt tokens and used as collateral, a slashing event propagates into lending markets that never opted into the underlying service.
The comparison that keeps being drawn is to rehypothecation in traditional finance, and it is apt in one respect: the same collateral supporting multiple obligations is efficient until the moment several of them are called. The mitigations being built — caps per service, delays before slashing takes effect, insurance layers, conservative collateral factors on receipt tokens — are recognisable as the same mitigations that were built after previous episodes elsewhere. For anyone holding a receipt token, the question worth asking is not what it yields but which slashing conditions the capital behind it is now exposed to, and whether they could all be triggered by one event.