Proving costs, once the decisive objection to validity rollups, have fallen sharply as teams combine improved arithmetisation with commodity GPU fleets.
Benchmarks published by two independent groups put the cost of proving a full block in the low cents, down from figures that made per-transaction economics uncomfortable eighteen months ago.
The improvement comes from several directions at once: smaller fields, recursive aggregation that amortises fixed costs, and proving software that finally treats memory layout as a first-class concern.
Decentralising the prover set is the next problem. Most networks still rely on a small number of operators, which is acceptable for liveness but awkward for the censorship-resistance claims teams like to make.
Auditors caution that faster systems are also newer systems. Several of the optimisations have not yet been through a full review cycle, and soundness bugs in proof systems are unusually costly.
