we need ML GPU workload standardization, and more options to chose from. Sucks being held hostage by Nvidia. No ML GPUs available in my country, and waitlist goes in years.