(
Compute
Unified
Device
Architecture-
Quantum) NVIDIA's open source platform for hybrid classical-quantum computing. Launched in 2024, CUDA-Q enables developing software for mixed GPU and QPU (quantum processing unit) operations using traditional code such as C++ and Python.
CUDA-Q supports quantum error correction by using classical processing to detect errors, a major problem with quantum computing. See
quantum coherence,
Alice & Bob and
CUDA.
NVIDIA Quantum Link
Introduced in 2025 to augment CUDA-Q, NVIDIA Quantum Link (NVQLink) provides a high-speed interface between QPUs and NVIDIA GPUs. Using RDMA over Ethernet (RoE) at a throughput of 400 Gb/s, NVQLink provides the speed necessary to address error correction in real time. See
quantum computing and
RoCE.