I think you're both right. It's a bit inflated as AWS services are overpriced in general. On the other hand, we're speaking specifically about significant use of Nvidia GPUs which are in demand right now so it's not like Amazon can just use their spare capacity.
AWS has its own line of AI accelerators called Inferentia that offers reduced cost versus Nvidia, but it's often a pain to make stuff work because everything is written for CUDA. Inferentia team spends a lot of time porting OS models to their hardware for customers to use (e.g. through SageMaker).
Amazon reduces its profits, and pays less taxes, true.
They also do get sales growth.
But nowhere is this "AWS' own money". AWS has to pay Nvidia for the chips. Nvidia is worth 1 trillion for a reason.