He was the cofounder, president and CEO since 1993. Nvidia hired/created CUDA in 2004 and made it central to its strategy. That’s almost 20 years before the current AI boom.
GPUs, as you know, are just Graphics processors and there is no reason to use them for AI. The only reason for their popularity in AI is CUDA. Because in 2012 Alex Krizhevsky, along with Sutskever and Hinton trained AlexNet to win the ImageNet competition and started the current AI revolution. One of their secrets was using a graphics accelerator (GPU) as a supercomputer to train larger neural networks. This was only possible thanks to CUDA. After 14 years CUDA is still the industry standard for training neural networks.
That strategy converted the 90s 3D accelerator startup into the most valuable company of the world today. But after today, they’re just a standard chipmaker with no vision.