I avoid using cloud whenever I can on principle. For instance, OpenAI recently indicated that they are working on some social network-like service for ChatGPT users to share their chats.

Running it locally helps me understand how these things work under the hood, which raises my value on the job market. I also play with various ideas which have LLM on the backend (think LLM-powered Web search, agents, things of that nature), I don't have to pay cloud providers, and I already had a gaming rig when LLaMa was released.