May 2025 Summaries

Zama has introduced the first open-source hardware accelerator for Fully Homomorphic Encryption (FHE), named the Homomorphic Processing Unit (HPU), which operates on a Field-Programmable Gate Array (FPGA). Released alongside TFHE-rs v1.2, the HPU is a processor that computes directly on encrypted data, equipped with a register file, control logic, and an arithmetic unit capable of executing instructions over encrypted operands. Unlike traditional accelerators that enhance cryptographic primitives, the HPU performs boolean and integer operations directly on encrypted data, with its instruction set being fully customizable for developers. The integration of the HPU into the TFHE-rs library allows users to execute encrypted operations using the same high-level API previously used for CPUs or GPUs, enhancing performance and energy efficiency. Currently running at 350 MHz on a 7nm FPGA with the capability to process around 13,000 Programmable Bootstrapping operations per second, the HPU is noted for its speed and lower cost per unit compared to CPUs and GPUs. Although not fully production-ready, Zama is working to integrate it into their product line by 2025, alongside developing multi-HPU servers to improve FHE operation efficiency. The open-source nature of the HPU stack, which includes systemVerilog hardware design and TFHE-rs integration, invites developers and researchers to audit, extend, and repurpose the technology for their projects.

Concrete is an open-source Fully Homomorphic Encryption (FHE) Compiler designed to simplify the development of FHE programs by utilizing a TFHE Compiler based on LLVM, making it accessible for developers. In a tutorial, Alexandre Pere from the Zama team introduces "Concrete for Rust," a new feature that enables developers to integrate Python modules within Rust for encrypted computations. The initiative encourages engagement with the community by inviting users to star the Concrete GitHub repository, review documentation, participate in community support channels, and contribute to the advancement of FHE through the Zama Bounty Program.

TFHE-rs is a library designed for developers and researchers interested in utilizing the advanced features of TFHE without dealing with low-level implementation details, aiming to provide a stable, simple, high-performance, and production-ready solution. Agnes Leroy from the Zama team demonstrates in a tutorial how to enhance multiple-GPU throughput by selecting specific GPUs for encrypted operations using TFHE-rs. The library encourages community support through its GitHub repository, offers extensive documentation, and invites participation in the Zama Bounty Program and Grant Program to further advance the field of Fully Homomorphic Encryption (FHE). Additionally, developers are encouraged to contribute feedback via a developer survey.

The article delves into the emerging field of tokenizing private equity and emphasizes the necessity of confidentiality in this process. With predictions from BCG and Citi suggesting that over $4 trillion in private market assets could be tokenized by 2030, the transition from traditional private equity management to blockchain-based tokenization is gaining interest, although hindered by regulatory fragmentation and confidentiality concerns. Tokenization, facilitated by Zama’s fhevm using Fully Homomorphic Encryption (FHE), allows private equity operations like fund shares, capital calls, secondary trading, and distributions to be securely managed on blockchain while maintaining the privacy of sensitive data. Despite early institutional pilots, widespread adoption remains limited due to these concerns. The article highlights that current tools like eFront, which manage confidentiality off-chain, are outdated compared to the potential of public-permissioned blockchains that offer both compliance and innovation. Zama’s fhevm provides a solution by enabling encrypted smart contracts that secure private equity operations onchain, thus merging the benefits of blockchain transparency with the confidentiality required in traditional finance.