I bet they want to do this to prevent AI from writing code for platforms that compete with CUDA. Because that's how nVidia is going to become a victim of their own success.