I understand that many people on HN prefer open-ish LLM on local hardware, but I think it doesn't make sense sadly for efficient hardware usage perspective. Tra

I understand that many people on HN prefer open-ish LLM on local hardware, but I think it doesn't make sense sadly for efficient hardware usage perspective. Transferring input/output text is almost free cost and local hardware can't be fully utilized by a few people. SaaS make sense here, though I understand that privacy and censorship are matter.