Unless you intend to run your inference on a potato battery, you're beholden to AMD/Apple/Nvidia/Intel.