Nvidia created the CUDA parallel-computing platform, which was first released in 2006.
CUDA stands for Compute Unified Device Architecture. It lets developers use Nvidia graphics processing units for general-purpose parallel computing rather than limiting them to rendering images. Nvidia introduced the platform to make its hardware useful for scientific computing, engineering, finance, and later artificial-intelligence workloads.
CUDA became a major strategic advantage as researchers used graphics processors to accelerate neural-network training. Nvidia’s chips and software ecosystem consequently became central to modern generative AI infrastructure, helping the company become one of the world’s most valuable public companies by market capitalization.
CUDA is not a separate company and is not the same thing as a graphics card. It is a parallel-computing platform and programming model associated with Nvidia hardware. AMD offers competing GPU-computing software, including ROCm.