AIPS BOOM

cuPQC

3 min read

cuPQC

Type: Technology Tags: CUDA, NVIDIA, GPU, Post-Quantum Cryptography, Security, Cryptography, CUDA-X Related: NVIDIA-Quantum, NVIDIA-CUDA, NVCC, NVRTC, CUDA-Cpp-Standard-Library, Thrust Sources: https://docs.nvidia.com/cuda/cupqc/index.html, https://docs.nvidia.com/cuda/cupqc/overview/features.html, https://docs.nvidia.com/cuda/cupqc/overview/requirements.html, https://docs.nvidia.com/cuda/cupqc/additional/release_notes.html, https://www.nvidia.com/en-us/solutions/quantum-computing/ Last Updated: 2026-04-29

Summary

cuPQC is NVIDIA’s CUDA SDK for building high-performance cryptographic applications on GPUs. Current documentation describes two device-side libraries: cuPQC-PK for post-quantum public-key algorithms such as ML-KEM and ML-DSA, and cuPQC-Hash for SHA-2, SHA-3, SHAKE, Poseidon2, and Merkle tree operations. The SDK is designed to integrate directly into CUDA kernels so cryptographic work can be fused with GPU computation.

Detail

Purpose

Cryptographic workloads often need both high throughput and low data-movement overhead. cuPQC addresses this by exposing CUDA-kernel-integrated cryptographic building blocks, allowing applications to run post-quantum key/signature operations and hash/Merkle computations inside GPU kernels rather than staging work through CPU-side libraries.

Current SDK shape

  • cuPQC-PK: device-side post-quantum public-key library for ML-KEM key encapsulation and ML-DSA digital signatures.
  • cuPQC-Hash: device-side cryptographic hash library for SHA-2, SHA-3, SHAKE, Poseidon2, and Merkle tree operations.
  • Descriptor/operator model: applications describe the algorithm, function, execution mode, and block dimension in CUDA C++ types.
  • Kernel fusion: cryptographic functions are intended to execute directly in CUDA kernels so they can be fused with other GPU work.
  • Batched throughput: current docs emphasize high-throughput batched operations for both public-key and hash/Merkle workloads.
  • Security posture: NVIDIA documents side-channel protections for cuPQC-PK and a separate security notes section.
  • Packaging: current releases use static libraries such as libcupqc-pk.a and libcupqc-hash.a, with CommonDx headers as a dependency.
  • Release direction: cuPQC 0.4.1 documents the cuPQC-PK/cuPQC-Hash naming split and new Poseidon2/Merkle tree improvements.

Use Cases

  • High-throughput TLS/certificate authority operations migrating to PQC
  • Code signing and firmware authentication at scale
  • VPN and secure channel establishment with quantum-resistant KEM
  • Blockchain, ZK, and digital ledger systems requiring GPU-side hashing or Merkle tree computation
  • HSM (Hardware Security Module) software acceleration
  • IoT device provisioning at scale
  • Research into PQC and cryptographic hash performance benchmarking

Hardware Requirements

  • CUDA Toolkit 12.8 or newer
  • Supported CUDA compiler, with NVCC 12.8+ listed in current requirements
  • C++17-capable host compiler
  • x86_64 or Arm64 CPU
  • NVIDIA GPU architecture sm_70, sm_75, sm_80, sm_86, sm_87, sm_89, or sm_90

Language Bindings

  • CUDA C++ device-side libraries
  • CMake integration through SDK package configuration files

Connections

  • NVIDIA-Quantum - cuPQC is listed in NVIDIA’s current quantum solution map for quantum-safe security.
  • NVIDIA-CUDA - platform context for CUDA kernels, compilers, and GPU execution.
  • NVCC - current cuPQC examples compile CUDA C++ with LTO-oriented NVCC flags.
  • NVRTC - adjacent runtime compilation path for systems that generate CUDA code dynamically.
  • CUDA-Cpp-Standard-Library - C++17 CUDA development context for header/static-library integration.
  • Thrust - adjacent CUDA C++ parallel-programming library often used in GPU-side data processing.

Source Excerpts

  • NVIDIA describes cuPQC as an SDK of high-performance GPU cryptography libraries that integrate directly into CUDA kernels.
  • Current docs split the SDK into cuPQC-PK for ML-KEM/ML-DSA and cuPQC-Hash for SHA-2, SHA-3, SHAKE, Poseidon2, and Merkle tree operations.

Resources

Graph View

Backlinks