Nvidia developed the CUDA parallel-computing platform used widely for GPU applications.
Nvidia introduced CUDA in 2006 as a parallel-computing platform and programming model for its graphics processing units. It allowed developers to use Nvidia GPUs for general-purpose calculations rather than limiting them to rendering images. CUDA supports languages and tools that let software divide suitable workloads across many GPU processing cores.
The platform became important in scientific computing, engineering, machine learning, and artificial intelligence. Training and running large neural networks often require huge numbers of simultaneous mathematical operations, a task for which GPUs are well suited. Nvidia’s hardware, software libraries, developer tools, and data-center products consequently became a major part of its business.
CUDA is proprietary to Nvidia, which distinguishes it from cross-vendor standards such as OpenCL. AMD develops competing GPU technologies, including ROCm. Nvidia’s market capitalization has been strongly influenced by demand for AI computing, but the company’s earlier roots were in graphics processors for personal computers and video games.