A CFO briefing in three numbers.
What's live on mainnet today, what's blocking the rest from mainnet, and what each operation costs once it ships.
Three products. Three different gates.
fhe-disperse
Singleton + per-user OZ clones, each user calls register(token) once, the singleton deploys that caller a dedicated wallet pair (two ERC-1167 clones, wallet0 and wallet1), then disperse({ mode: 'wallet' }) batches confidential payouts.
- ·Audited by OpenZeppelin (May 2026); live in mainnet production.
- ·Zama relayer dependency on every encrypt-side call.
fhe-vesting
Factory + LibClone deterministic clones with 87-byte packed immutable args. factory.createManager({ token, userSalt }) deploys an audited manager clone per creator; recipients receive schedules via manager.createVesting() / batchCreateVesting() inside that clone.
- ·Mainnet factory not yet deployed; no mainnet DEPLOYED_ADDRESSES entry until the CREATE3 deploy lands.
- ·Mainnet KMS readiness, Zama's KMS must operate against mainnet with the same threshold posture.
- ·OpenZeppelin's audit (May 2026) cleared factory-clone determinism + packed immutable args with 0 critical / 0 high findings.
fhe-airdrop
Factory + EIP-712 gated claims, admin signs a per-recipient authorization off-chain; recipient submits the signature against the campaign clone to unlock.
- ·Mainnet factory not yet deployed; no mainnet DEPLOYED_ADDRESSES entry until the CREATE3 deploy lands.
- ·Mainnet KMS readiness, same gate as fhe-vesting.
- ·OpenZeppelin audited the EIP-712 domain separator + signature replay surface (Jan 2026, 1 medium resolved, 0 critical / 0 high).
Four moving parts a CFO needs to be able to name
1. Zama relayer dependency
Every encrypt and user-decrypt round-trip hits Zama's relayer. Outage degrades the interactive surface; pending on-chain ciphertexts already written remain decryptable when the relayer returns.
2. KMS / key-management story
The contract never reconstructs cleartext. KMS only signs decrypt-proofs for parties the contract granted ACL to via FHE.allow. No SDK-side key custody.
3. Mainnet gas + HCU envelope
FHEVM measures FHE-op cost in HCU; the per-tx ceiling is 5M HCU. The SDK targets ≤ 2 FHE ops per vesting-style calculation to stay well under the cap. ~2–3× plaintext equivalent at scale.
4. Per-claim ETH fee
fhe-airdrop charges a per-claim ETH gas fee; fhe-disperse charges an ETH gas fee only in its gas-fee modes (mode: "wallet" and mode: "direct"), while mode: "wallet-token-fee" charges a token BPS fee on-chain instead (ethValue 0n, non-payable entrypoint, collected via _collectTokenFeeOnTotal) and attaches no ETH. This mirrors fhe-vesting, which charges an ETH gas fee only under FeeType.Gas, while FeeType.DistributionToken deducts the fee on-chain in the distribution token and attaches no ETH (the clone reverts with UnexpectedETH if msg.value > 0). None collect at the factory. fhe-disperse takes the gas fee in the singleton's payable disperse calls (_collectExactGasFee); fhe-vesting and fhe-airdrop bake the fee into each clone's immutable args and collect it in the clone's own claim(). The fee keeps relayer-side compute economically rational at claim-volume scale.
The four gates
The audit gate is already cleared. When the remaining three close, mainnet addresses for fhe-vesting + fhe-airdrop land in DEPLOYED_ADDRESSES in a minor SDK release.
- 1
Audit publication
Complete. OpenZeppelin audited all three products with 0 critical and 0 high findings across both reports; reports are linked from Resources › Audits.
- 2
Mainnet KMS readiness
Zama's KMS operates against Ethereum mainnet with the same threshold posture as the Sepolia deployment.
- 3
Design-partner pilots
At least three pilots running confidential vesting on Sepolia for a full quarter against production-like volume, surfacing telemetry into the SDK adapter.
- 4
CREATE3 deployment
Deploy from the audited deployer, same address on mainnet as on every other chain the same deployer publishes from.
See also
Resources · deployed addresses + audit posture · SDK telemetry · the relayer round-trip event your team alerts on · The encryptor source pattern · the three encryptor variants the relayer powers.