Given my dev machine has 32GB of RAM and 32GB of VRAM that sits mostly idle when I'm not running AI models, this is not that bad of an idea.