Ethereum Foundation Abandons Poseidon for Conventional Hashes in Its Post-Quantum Plan

Justin Drake, a researcher at the Ethereum Foundation, announced that the foundation is dropping the Poseidon hash function for Ethereum's base layer and will use SHA or BLAKE instead. He called the decision a milestone for post-quantum security on Ethereum.

Poseidon belongs to a family of hash functions designed to be cheap to prove inside a succinct non-interactive argument of knowledge (SNARK), the compact proof that lets one party confirm a computation without repeating it. Conventional hashes such as SHA-2 and BLAKE were considered too expensive to prove at the volumes a hash-based post-quantum design would require, so the foundation spent years on the alternative. Drake described the effort as "an epic 8-year, 8-figure" programme in post-quantum cryptography, and summarised the reversal by saying the answer turned out not to be proof-friendly hashes but hash-friendly proofs.

The practical argument is that SNARK performance improved faster than the cryptanalysis of new hash designs settled. Conventional hashes have decades of public attack effort behind them. Poseidon and its relatives are newer, use algebraic structure that makes them fast inside proofs, and have a thinner cryptanalytic record. Choosing the older primitive trades raw proving speed for a security assumption that far more researchers have tried and failed to break.

The change applies to Ethereum's future base-layer architecture, not to code running today. Rollups, virtual machines and applications that already use Poseidon are not required to replace it because of the roadmap decision. Drake put a production-ready leanVM, the minimal proof-oriented virtual machine at the centre of the lean Ethereum plan, at 2027, with deployment across the consensus, data and execution layers targeted for 2028. Those are long timelines, and Ethereum roadmap dates have moved before.

For the wider zero-knowledge sector the signal is commercial as much as cryptographic. Proving teams, hardware accelerator projects and rollup designers have optimised around algebraic hashes for years. If the largest smart-contract platform standardises on SHA or BLAKE at layer one, the performance targets that matter for future proving hardware change with it.