Yes, because nobody replicated the pragmatism and power of CUDA. OpenCL is much uglier and lower level. So AMD decided to do something about that... and invente

Yes, because nobody replicated the pragmatism and power of CUDA. OpenCL is much uglier and lower level. So AMD decided to do something about that... and invented ROCm, which is somehow even uglier and more low-level! A reference FFT implementation in CUDA is about 150 lines of code... and it is almost 10 times more in ROCm. It was always about software. Granted, CUDA is not the best and most elegant platform in the world... but AMD seems not to be able to reach even that level.