Sure. I'm talking about now. I don't know what will happen tomorrow, let alone 1 year or more down the line. If it ever becomes economical to run and maintain

Sure. I'm talking about now. I don't know what will happen tomorrow, let alone 1 year or more down the line. If it ever becomes economical to run and maintain bare metal GPU compute to run LLMs, then that's what will need to be done..