Restaking: Why Stacked Risks Are Not Independent

Restaking lets staked capital be pledged as security to additional services beyond the network it originally secured. Each service can impose its own slashing conditions on that collateral. The pitch is that the capital works harder. The complication is that the conditions under which it can be destroyed multiply, and they do not fail independently.

The reward side is simple. The risk side is not.

Adding a second source of payment to the same capital is easy to reason about: two streams instead of one, minus fees. If the added yield is a few percentage points on capital you had already committed, the offer looks close to free.

The risk side resists that treatment because the natural intuition — small independent probabilities compound into a still-small probability — does not apply. If two services could each slash your collateral, and the events were genuinely independent, the combined exposure would be modest. They are not independent, and the reasons are structural rather than incidental.

Four ways the correlation arrives

Shared operators. The same professional operators run infrastructure for many services, because operating is a specialised business with few participants. An operator failure hits every service that operator runs, and your collateral may be pledged to several of them.

Shared software. Services built on a common framework inherit common bugs. A defect in a widely-used component is a single event that triggers penalties across every service using it.

Shared conditions. Slashing is usually triggered by liveness or correctness failures, and the things that cause those — network partitions, cloud region outages, a bad client release — are exactly the events that hit many services at once.

Shared incentive to exit. When one service has a problem, restakers reassess all of them. Withdrawal pressure appears across services simultaneously, which is when queues lengthen and receipt tokens trade at discounts.

Each of these means the same underlying event shows up as several penalties. The distribution of outcomes is not a gentle curve of small independent losses; it is mostly nothing, occasionally a lot.

The opacity problem

There is a second issue specific to how restaking is packaged for ordinary depositors. Protocols that make restaking accessible generally do so by abstracting away which services the collateral is pledged to and which operators run them. That is the product: you deposit, and someone else makes the allocation.

The consequence is that the depositor cannot enumerate their own slashing exposure. They know they are exposed to a set of conditions; they frequently cannot say which, how many, or how correlated. Assessing an exposure you cannot list is not risk management, and no yield figure resolves it.

The protocols that score best on this are the ones that publish the allocation and the operator set, and that let a depositor see what their collateral is actually securing. That disclosure is not a nicety; it is the difference between a priced risk and an accepted one.

Where the extra yield comes from

It is worth applying the same question here as to any yield: who is paying, and why.

Restaking rewards are paid by services that need economic security and would otherwise have to bootstrap their own. That is a genuine service with a genuine buyer, which puts it in a better category than a return funded by token emissions. But the size of the payment depends on those services having sustainable revenue of their own, and many are early enough that their payments are subsidised by their own token issuance rather than by customers.

So a restaking return can be real external revenue, or it can be a second layer of emissions dressed as revenue, and the composition changes as the services mature. The question of who is paying is answerable service by service and is worth answering.

How to assess a restaking position

• Can you enumerate every service your collateral secures. If not, you cannot assess the exposure.
• How concentrated is the operator set across those services. Shared operators are the main correlation channel.
• Do the services pay from revenue or from their own token issuance. The second is emissions with extra steps.
• Has the withdrawal path been exercised at size, or only in principle.
• Does the protocol retain your withdrawal credentials, or do you. That decides who can ultimately move the principal.

None of this makes restaking a bad instrument. It makes it a leveraged one: the same capital, more claims against it, and a loss distribution with a longer tail than the headline suggests. That is a legitimate position to take deliberately and a poor one to arrive at because the yield was a little higher.