Nvidia has expanded its open-source CUDA-Q platform by introducing CUDA-Q Logical, a toolkit designed for quantum error correction. This software layer enables researchers and developers to design, simulate, and execute workflows using logical qubits, which are created by encoding information across multiple physical qubits to detect and correct errors.
The platform includes accelerated decoding tools and libraries under the CUDA-QX extension. It is hardware-agnostic and currently supports approximately 75% of publicly available quantum processors. On December 10, 2024, Infleqtion utilized CUDA-Q Logical to conduct a materials science experiment on Sqale’s neutral-atom processor, successfully reducing logical error rates.
Nvidia continues to iterate on its quantum software stack. Version 0.8 of CUDA-Q, released in 2024, increased simulation speeds by 2.4 to 2.9 times for specific variational quantum eigensolver workloads. The company plans to release version 0.4 of the CUDA-QX libraries in 2025 to further enhance error-correction capabilities.
Additionally, Nvidia is developing NVQLink, a technology intended to provide microsecond-latency connections between quantum hardware controllers and GPUs. This development aims to facilitate real-time integration for hybrid algorithms, allowing classical and quantum processors to exchange information efficiently during computation.
