Unfortunately I think you're overestimating how many people care enough about privacy to go through the effort of running LLMs locally and likely buying a GPU