Nvidia developed the CUDA parallel-computing platform used extensively for modern AI software.
Nvidia introduced CUDA in 2006 as a software platform and programming model for using its graphics-processing units for general-purpose computing. Before CUDA, GPUs were primarily associated with rendering images and video, but their many parallel cores also made them useful for scientific calculations and machine-learning workloads.
CUDA became especially important because it gave developers libraries, compilers, and programming tools tailored to Nvidia hardware. Deep-learning frameworks such as PyTorch and TensorFlow can use CUDA-enabled GPUs for accelerated computation, although the frameworks themselves are separate projects.
A common mix-up is confusing CUDA with a physical chip. CUDA is software and an ecosystem, while Nvidia's GPUs are the hardware that executes CUDA workloads. The platform's adoption helped Nvidia expand from gaming graphics into data centers and artificial intelligence, supporting its exceptional market value.