I’ve used both and found both to work well for me. There are a variety of other gpu cloud options that aren’t good, though. I listed a bunch here, some are good, some seem to be worse on all dimensions (price, capacity, and UX) - https://gpus.llm-utils.org/alternative-gpu-clouds/

Tl;dr of the good ones: FluidStack and Lambda for H100s (1x instances), Runpod for A100s.