Buterin Sees 60% Chance Cheap Crypto Tech Cuts Wall Street Roles

On Sept. 6, Ethereum co-founder Vitalik Buterin assigned a 60% probability that three advanced cryptographic tools-SNARKs, fully homomorphic encryption (FHE) and indistinguishability obfuscation (iO)—will eventually operate at under ten times the cost of ordinary computation. He put the odds at about 33% that the same technologies could approach near-zero additional overhead at sufficient scale, and he cautioned that his timeline is uncertain.

SNARKs, FHE and iO provide different privacy functions. SNARKs let a party prove that a computation or statement is correct without revealing the input data. FHE allows systems to compute directly on encrypted data so multiple parties can interact with shared state without decrypting it. iO aims to hide a program’s internal logic while still permitting users to run and verify its behavior.

Buterin framed these tools as a sequence of capabilities for private computation. Cheaper SNARKs would make private proofs easier to generate. FHE would extend privacy to shared encrypted state and multi-user confidential applications. Practical iO would allow software to run and prove correct behavior without exposing its internal logic.

Ethereum’s research separates privacy work into private reads, private writes and private proving so improvements can arrive incrementally. Engineers are developing anonymous remote procedure call systems, private information retrieval, oblivious RAM techniques and improved light clients to limit what users reveal when querying blockchain data. For transaction privacy, researchers have proposed frame transactions to separate fee payers from action accounts, inclusion lists to require validators to include eligible transactions they see, and keyed nonces to make activity harder to correlate; those mechanisms remain proposals rather than deployed upgrades.

SNARKs are the most developed of the three technologies. Application-specific zero-knowledge systems already support features such as anonymous voting, private claims and shielded withdrawals. Privacy risks persist through patterns like wallet reuse, use of remote procedure call providers, public inputs and small anonymity sets, but SNARKs today allow many private proofs without waiting for broader cryptographic advances.

FHE would enable confidential shared-state applications such as private automated market makers, confidential lending pools and sealed-bid auctions where a system must process inputs from many users without revealing them. Current FHE performance remains far from commodity compute: a 2026 benchmark on a small BERT workload required multi-gigabyte cryptographic artifacts and more than six hours of encrypted computation. Developers are using specialized coprocessors, narrow workloads and threshold-decryption schemes to make encrypted applications practical in limited settings; those workarounds add infrastructure or new trust assumptions.

iO is the furthest from practical use. Buterin has previously described some iO constructions as carrying “galactic” runtimes, while newer theoretical approaches remain unproven. If practical iO were to emerge, software could run on external infrastructure while keeping model mechanics private and disclosing only authorized outputs.

Christopher Inks, founder of Texas West Capital, wrote: “Cheaper cryptography could allow proprietary financial models to run on outside infrastructure without exposing private data, model mechanics or intellectual property.” He added that such systems could weaken business lines built around acting as trusted observers or validators, including auditors, custodians, clearinghouses and exchanges.

Buterin also estimated there is a reasonable chance that at least one of the three technologies-most likely SNARKs-could reach single-digit overhead by the end of the decade, though he did not assign a specific probability to that deadline. He described the privacy roadmap as a progression rather than a single breakthrough and noted that useful protections can arrive through specialized systems even if general encrypted computation remains expensive.