Funnily, who I blame the most for there not being real competition to CUDA is apple. As of late, Apple has been really pushing for vender lock in APIs rather than adopting open standards. The end result is you can get AMD and Intel onboard with some standard which is ultimately torpedoed by apple. (See apple departing from and rejecting everything that comes from the khronos group).

With the number of devs that use Apple silicon now-a-days, I have to think that their support for khronos initiatives like SYCL and OpenCL would have significantly accelerated progress and adoption in both.

We need an open standard that isn't just AMD specific to be successful in toppling CUDA.