Nvidia managed to hit the right spot to win the market - developers. I wrote shaders for depth detection before CUDA, and it was already great back then, but was very graphics centric. Then CUDA came and pushed everything even further forward. I completely agree - it was a software victory.

Yes, Khronos, Intel and AMD managed to make a mess out of OpenCL.

Which is one of the reasons why Apple stop caring about OpenCL, apparently they have repented giving it Khronos, and how it was managed afterwards.