Ethereum Researchers Drop the Poseidon Hash From the Chain's Next Base Layer

Ethereum Foundation researcher Justin Drake confirmed on August 13 that the foundation's plans for a future base layer will no longer rely on Poseidon, the hash function designed in 2019 specifically to be cheap to verify inside zero-knowledge proofs. The design will instead use members of the SHA and BLAKE families, the conventional hash functions that have absorbed decades of public cryptanalysis.

The reason is a change in the underlying proving technology, not a flaw found in Poseidon. Proof systems that operate over binary fields have become fast enough to handle the bit-level operations that conventional hash functions perform, the very operations Poseidon was designed to avoid. Reported figures put binary-field systems at roughly one million conventional hash calls proved per second, and joint work between the foundation and the proving company Succinct raised proving speed by about 2.5 times. Once that cost gap closes, the case for using a newer, more specialized hash function instead of a long-studied one closes with it.

The decision sits inside the leanVM effort, the simplified execution and proving design intended to let Ethereum verify large volumes of activity with a small, formally tractable core. A production leanVM is scheduled for 2027, with layer deployments in 2028. Hash functions matter far beyond speed in this design, because post-quantum signature schemes are built almost entirely from hash functions, so the hash chosen for the base layer becomes the security assumption behind the chain's long-term resistance to quantum attack.

No one involved claims Poseidon is broken. The foundation's position, as reported, is that the tradeoff that justified using Poseidon has shifted. That is the conservative choice: consensus-critical code inherits the weakest primitive built into it, and a primitive with seven years of scrutiny carries more unknown risk than one with thirty.