>>20930558They were in the right place at the right time and the first generation of deep learning frameworks were all written to use their GPUs. Honestly, I think the only reason why CUDA stuck is because the very first hardware-accelerated neural net prototype in 2013 ran on CUDA, and techjeets fundamentally only know how to copy things. Even in back in 2013 there were software alternatives to CUDA like OpenCL, and Google's TPU hardware was considered superior to GPUs as early as 2015. The people FOMOing into Nvidia are normalniggers that don't know what they're getting into, and Nvidia's lead won't last more than a few years at most.