FHE is much less useful and much harder compared to ZK for privacy/crypto related purposes
Compute overhead and key management are perennial issues that won’t be solved any time soon
Just go ZK
let us check on the surrounding crypto FHE landscape
fhenix - raised $22m for a FHE L2 using zama tech - pivoted to own coprocessors - never fully launched
inco - raised $10m for a FHE L1 using zama tech - pivoted to TEEs - never fully launched
mind network - raised $15m for a FHE L1 using zama tech - launched an EVM fork, no FHE was ever added - trades at $5m
zama - raised $150m for open source FHE tech - turned out unusable - pivoted to coprocessors for institutions - unclear if FHE is still used (blackbox) - founder says the protocol offers no privacy - trades at -80% from latest round
privasea - raised $15 million to build FHE for ML - pivoted to depin and “FHE ID” proof of humanity - no FHE in live code - trades at $700k
primus labs - raised $6.5m for “zkFHE” - launched a TEE-based zkTLS network with coprocessors - no FHE in any live product
fair math - raised $2 million for FHE research - pivoted to a “decentralized FHE computer” with an AI generated paper - never launched
sightAI - announced FHE LLM - never launched
FHEra - announced FHE storage - never launched
honorable mentions:
sunscreen - raised $5m - never really promised anything beyond a compiler - delivered
lattica - raised $4M - CUDA innovation in FHE for ML / AI - delivered
anything i’m missing?
bottom line: FHE is HARD!