Zama

Zama develops open-source Fully Homomorphic Encryption (FHE) tooling and operates the Zama Confidential Blockchain Protocol, which lets smart contracts compute on encrypted data on existing public chains. Zama describes it as "not a new L1 or L2, but rather a cross-chain confidentiality layer sitting on top of existing chains", so applications stay on their host chain instead of migrating to a dedicated network. Ethereum mainnet is live; further EVM chains and Solana are on the published roadmap.

Zama – Confidential Blockchain Protocol (Fully Homomorphic Encryption on existing chains)

Fits with patterns

Not a substitute for

  • Graph privacy: sender and recipient addresses remain public, so unlinkability needs composition with stealth addresses
  • Privacy rollups that move execution and state off the host chain entirely
  • Multi-party computation or Trusted Execution Environments as general-purpose off-chain compute

Architecture

Contracts import the FHEVM Solidity library and use encrypted types (euint8 to euint256, ebool, eaddress). Each encrypted value is referenced on-chain by a 32-byte handle. Zama's documentation places the ciphertext itself "off-chain, with the coprocessor" and the handle on-chain as the identifier pointing to it. FHE operations run symbolically on the host chain, generating new handles and emitting events, while a network of coprocessors performs the computation using TFHE-rs.

The protocol combines the FHEVM Solidity library; host contracts that enforce access control and emit events; coprocessors that verify inputs, execute FHE operations and store ciphertexts; the Gateway, a dedicated Arbitrum rollup that orchestrates input validation, decryption and bridging; and a threshold multi-party computation Key Management Service. Relayer and oracle services connect users and contracts to the Gateway and are not part of the trusted base.

Privacy domains

  • Fully Homomorphic Encryption at the contract and variable level: amounts and balances stay encrypted end to end
  • Programmable access per ciphertext through an Access Control List (allow, allowThis, allowTransient, makePubliclyDecryptable), including delegation of user decryption
  • Hybrid by design: zero-knowledge proofs validate encrypted inputs, multi-party computation performs threshold decryption

Enterprise demand and use cases

  • Confidential payments and stablecoin transfers where amounts and balances stay hidden from other participants
  • Tokenization and real-world assets where holdings and allocation sizes are confidential but settlement stays on a public chain
  • Sealed-bid auctions, confidential distributions and governance, where inputs stay private until a defined reveal

Technical details

  • FHEVM Solidity library (@fhevm/solidity), with a relayer SDK and a Hardhat plugin
  • Confidential tokens follow ERC-7984, implemented in OpenZeppelin's confidential-contracts library
  • Encrypted inputs carry a zero-knowledge proof of knowledge, validated through FHE.fromExternal
  • Conditional logic uses FHE.select instead of plaintext branching, so the taken branch is not revealed
  • Protocol fees apply to input verification, decryption and bridging. They are priced in USD and can be paid by the end user, the application or a relayer

Strengths

  • Host contracts deploy onto existing chains, EVM or non-EVM, so users do not move to a dedicated network
  • Composable between confidential contracts and with non-confidential ones, and contracts are written in plain Solidity rather than a new language
  • Every FHE computation and input verification carries a commitment and a signature, so anyone can recompute the result and check it independently
  • No single party ever holds the decryption key: it is secret-shared across the threshold network, and each decryption is authorized by the Access Control List
  • Compliance rules are defined by each application in its own contracts rather than by the protocol
  • Audited by Trail of Bits and Zenith, and the FHE and multi-party computation layers are post-quantum

Risks and open questions

  • The trust model is layered: decryption runs through a threshold Key Management Service of 13 multi-party computation nodes, with a documented threshold example of 9 of 13 and tolerance up to one third malicious; those nodes run inside AWS Nitro Enclaves, so verifiability rests partly on hardware assumptions; and coprocessor correctness rests on a majority-honest assumption backed by staking and slashing. Zama states the goal is to add zero-knowledge proofs to the multi-party computation protocol to remove the hardware dependency
  • Operators can pause the protocol and blacklist addresses, and the Access Control List carries an account deny list
  • Open source under a dual license: free for non-commercial use, with a commercial license required for use outside the Zama Protocol
  • Cost and latency stay above plaintext execution, and protocol support beyond Ethereum mainnet is roadmap rather than shipped

Links