Validator Concentration
How much of a proof-of-stake network's voting power sits with a small number of operators or staking services.
In Byzantine-fault-tolerant designs, thresholds matter mechanically: roughly one third of stake acting together can prevent finality, and roughly two thirds can finalize conflicting histories, so an entity approaching either share is a structural concern regardless of its intentions. Pooling makes concentration hard to see, because liquid staking providers and exchanges run validators on behalf of very large numbers of small holders, and thousands of separate validator identities can share one operator, one cloud region, or one set of signing keys. Common measures are the share of stake per entity and the Nakamoto coefficient, the minimum number of entities needed to reach a critical threshold. Correlated slashing is the related failure, where one operator's misconfiguration penalizes every validator it runs at once.
In practice
Networks publish stake distribution by operator, and a low Nakamoto coefficient means only a handful of entities would need to cooperate, or fail together, to disrupt finality.
The common misunderstanding
That counting validator nodes measures decentralization, when a very large number of nodes can be operated by a handful of firms or funded through a single staking pool.