What is fully homomorphic encryption?
Normal encryption is a locked box. You can store it and send it. But to do anything with what is inside, you have to open it, and the moment you open it, whoever is doing the work sees everything.
Fully homomorphic encryption (FHE) changes that. You can run computations on encrypted data without decrypting it. The result stays encrypted. Only the key holder can read it, and it matches what you would get from doing the same math on the raw data.
Lock the box. Do math on the box. Open the box. The right answer is inside. Nobody in the middle saw a thing.
A 30-second history
Cryptographers chased this idea for decades. In 2009 Craig Gentry published the first plausible fully homomorphic scheme, built on lattice-based cryptography.
The catch was noise. Every operation on a ciphertext adds a little, and too much makes the data unreadable. Gentry's fix was bootstrapping: the scheme runs its own decryption circuit while the data stays encrypted, producing a fresh, low-noise ciphertext. Refresh often enough and you can compute forever.
In 2016, Ilaria Chillotti, Nicolas Gama, Mariya Georgieva and Malika Izabachène published TFHE, a scheme with much faster bootstrapping. TFHE is the family Zama builds on, through its TFHE-rs library.
Why FHE matters onchain
Public blockchains are glass houses. Every balance, trade and salary payment is visible to anyone with a block explorer. Fine for memes. Fatal for most real finance.
FHE lets a public chain stay verifiable and get privacy too:
- Confidential balances. Your token balance is encrypted onchain.
- Private transfers. Amounts stay hidden while the contract still checks you can afford them.
- Sealed-bid auctions, private voting, payroll. Logic runs on data nobody can peek at.
- Selective disclosure. You choose who can decrypt, such as an auditor.
How does Zama work?
Zama builds the Zama Confidential Blockchain Protocol. It is not a new L1 or L2. It is a confidentiality layer that plugs into chains that already exist.
- FHEVM library. Developers write confidential smart contracts in plain Solidity.
- Symbolic execution. The host chain does not run the heavy FHE math. It stores pointers and emits events.
- Coprocessors. Offchain operators pick up those events, run the FHE computations in parallel and store the ciphertexts.
- Gateway. Coordinates input checks, decryption and bridging, on a dedicated Arbitrum rollup.
- Threshold key management. The decryption key is split across 13 MPC nodes, and a threshold of them (9 of 13 in Zama's docs) must cooperate to decrypt. Nothing is decrypted unless the contract allows it.
- Access control. Tracks who may compute on or decrypt each ciphertext.
Source: the Zama Protocol litepaper.
Confidential tokens and shielding
The flagship use case is the confidential token. Zama and OpenZeppelin implemented ERC-7984, a confidential version of the ERC-20 standard. Balances and transfer amounts are encrypted with FHE, so even the smart contract does not know your balance.
You shield a regular token by wrapping it into its confidential version. The amount is visible going in and encrypted after that. Unshielding takes an explicit decryption step. Official wrappers are live on Ethereum for tokens including USDC, USDT, WETH and ZAMA, and the Zama App has Shield and Unshield built in.
Why FHE needs serious hardware
FHE is heavy. Bootstrapping on a CPU is slow, and slow means low throughput. Zama's litepaper lays out the roadmap:
| Hardware | Throughput per chain |
|---|
| CPU (today) | 20+ TPS |
| GPU | Zama has shown up to about 1,000 TPS on GPUs |
| FPGA and ASIC | The longer-term target: thousands of TPS and more, on dedicated FHE hardware |
In May 2025 Zama open-sourced the HPU (Homomorphic Processing Unit), an FHE processor running on an AMD V80 FPGA board, at about 13,000 programmable bootstraps per second for about 200W. That is the family of hardware my creator, FROG Mark 1, comes from. The whole story is on the FROG Mark 1 page.
Compute without looking. Protect without collecting. From the Pond Codex.
Where does $ZAMA fit?
$ZAMA is the protocol token:
- Fees. Encryption, decryption, proof verification and bridging are priced in USD and paid in $ZAMA. Apps can pay on your behalf, so users do not have to hold it.
- Staking. Operators and delegators stake $ZAMA to secure the network and earn rewards.
$ZAMA started trading on 2 February 2026 and lives on Ethereum, BNB Chain and Solana. Want some without buying it directly? Hold $ZFROG and earn $ZAMA. ZRIBBIT.
Not Zama. $ZFROG is an independent community memecoin. It is not affiliated with, endorsed by or sponsored by Zama. For official Zama information, go to zama.org.