Rand Hindi, CEO of the fully homomorphic encryption firm Zama, said the company reached 1,000 confidential transfers per second on GPUs, a self-reported benchmark he described as a milestone for the privacy technology.
Fully homomorphic encryption, or FHE, lets computations run on encrypted data without decrypting it, a method Zama applies to confidential token transfers on public blockchains. Hindi posted the figures on X on Monday. Zama has not activated the performance on any live network.
Hindi said the 1,000 tps figure is up from 0.2 tps when the company first demoed an FHE smart contract at ETHcc in 2022, which he called a 5,000x improvement in four years. He said running an FHE node now costs about $0.000004 per ERC7984 transaction, which he characterized as orders of magnitude cheaper than zero-knowledge proofs. That cost comparison is Hindi's own framing and is not independently verified here.
The benchmark is not running in production. Hindi said the performance has not been activated on mainnet yet and is planned for end of year alongside other features and improvements. The figures are self-reported by the company and were not accompanied by published test methodology.
Hindi tied the number to Zama's deployment plans. He said 1,000 tps was the milestone the company needed to feel confident launching on chains faster than Ethereum, where users expect low latency, high throughput and near-zero gas fees.
