The restaking thesis is simple to state and hard to price. Ethereum validators pledge their stake to secure additional services, those services pay for the borrowed security, and everyone earns more than they would alone. The pitch is capital efficiency. The reality is a market for correlated risk, and correlated risk is exactly the kind that markets historically misprice.

This piece is an attempt to map that risk properly. Not to argue against restaking, which we hold positions in and continue to underwrite, but to separate the yield that compensates for real risk from the yield that merely obscures it.

Three layers of exposure

When a validator restakes, it takes on three distinct exposures, and it is worth being pedantic about the difference:

The math the yield has to clear

A useful discipline: take the incremental yield a service pays, subtract the expected annualized slashing loss, subtract a liquidity discount for the wrapper you actually hold, and ask whether the remainder still beats plain staking on a risk-adjusted basis. On our estimates, for the median service today, it clears the bar, but by a far thinner margin than headline rates suggest. Roughly a third of advertised restaking yield is, in our view, compensation for risks the average holder has not examined.

Yield that cannot be traced to a payer is not yield. It is either emissions, or it is premium collected for insurance you did not know you were writing.

What we look for as investors

We remain constructive on the category, with three filters. First, services whose fault conditions are objectively attributable on-chain, because subjective slashing is governance risk wearing a costume. Second, operators with segregated infrastructure per service, since correlation lives in operations, not in cryptography. Third, protocols that publish their security budget as a cost line rather than an emissions schedule, because a service that cannot afford its own security at market rates does not have product-market fit; it has a subsidy.

Shared security is a genuine primitive, and we expect it to underwrite much of the next generation of middleware. But primitives get priced correctly only after their first crisis. Our preference is to be positioned for that repricing rather than surprised by it.