Vitalik Buterin thinks the holy grail of cryptography is closer than most people realize, and he’s putting numbers on it. In a September 6 analysis, Ethereum’s co-founder estimated a 60% probability that three of the most important cryptographic primitives, SNARKs, fully homomorphic encryption, and indistinguishability obfuscation, will achieve sub-10x overhead when factoring in combined energy consumption and amortized computing costs.
He went even further, putting a 33% chance on those same technologies reaching near-zero overhead (expressed mathematically as 1+ε) at large scales. Translation: there’s roughly a one-in-three shot that running encrypted computations could become almost as cheap as running them in the clear.
Three technologies, one ambition
SNARKs (Succinct Non-interactive Arguments of Knowledge) let you prove something is true without revealing the underlying data. They’re the backbone of zero-knowledge proofs, which power everything from Ethereum’s Layer 2 rollups to privacy-preserving transactions.
FHE (Fully Homomorphic Encryption) allows computation on encrypted data without ever decrypting it.
Then there’s iO (indistinguishability obfuscation), which Buterin himself labeled the “final boss” of cryptography in a June 29 analysis. iO lets you scramble a computer program so thoroughly that no one can figure out how it works internally, only what it outputs. Current implementations are described as polynomially bounded yet effectively unusable due to extreme runtime requirements stemming from the layering of various primitives.
SNARKs leading the pack
Of the three technologies, Buterin sees SNARKs as the most mature. He noted that achievable overhead for SNARKs has already reached single digits in specialized scenarios, specifically for dedicated hash functions and certain large language model inference tasks.
Buterin expects SNARKs to hit single-digit overhead multiples more broadly by the end of the decade.
Three roads to the final boss
In his earlier June commentary, Buterin outlined three potential research routes for making iO practical. The first involves optimizing existing lattice-based constructions. The second relies on making stronger cryptographic assumptions, trading theoretical caution for practical speed. The third requires entirely new methods, non-lattice approaches that could sidestep current bottlenecks altogether.
He drew an explicit parallel to how SNARKs evolved post-2010, suggesting iO might follow a similar arc, though the starting point is considerably worse.
Buterin’s projections extend through the end of the decade for SNARKs and likely further for FHE and iO. No protocol changes or development milestones are attached to his estimates, which remain within a personal purview focusing on philosophical implications rather than commercial application.