[sbrother]

Oh, 100%. But you can do some pretty amazing things with fine-tuning LLMs too, and that is very compute intensive. Not to mention it's ridiculously hard even getting access to a cloud GPU instance nowadays.