Only reason to spend a bunch of money on hardware to run LLMs locally is if it's a hobby to you to an extent that even renting the GPUs temporarily won't satisfy you.