Developer file written for CUDA, an NVIDIA architecture created for parallel processing on nVidia GPUs; contains source code written in the "C for CUDA" language, which is similar to the standard C language with a few CUDA-specific extensions.

CU files are called "kernels," which are processing units that can run in parallel on NVIDIA's CUDA-based GPUs. Kernels are compiled with nvcc, a compiler program included with the NVIDIA CUDA Toolkit. NOTE: CUDA applications are often used in scientific or academic settings for large-scale computations. CUDA-based GPUs are also referred to as GPGPUs, which stands for General Purpose Graphical Processing Units.