Skip to main content
All terms
Hardware & Systems

CUDA

NVIDIA's platform for running general-purpose parallel programs on GPUs.

Definition

CUDA is NVIDIA's parallel computing platform and programming model that lets developers run general-purpose code on GPUs. Nearly all modern AI training and inference runs on CUDA under the hood: frameworks like PyTorch dispatch tensor operations to CUDA kernels. Its dominance is a major reason NVIDIA hardware is the default for AI, since most optimized libraries (cuDNN, cuBLAS, TensorRT) are CUDA-based.