I wonder when we will start seeing small commercial accelerators integrated with NVMes; could potentially use far more bandwidth than typical NVMe drive can push thru the four lanes.
Given that you'd have probably a lot more capacity just to get to required bandwidth it could be used to switch very quickly between models (as all stored there would essentially be "loaded")
There are many of these stream-experts-from-SDDD projects now. What's needed is a guide to all of them and maybe for someone to take the best features of each of them and merge them into one uber project.
Which seems sufficient for mortals who are not running gas town, but want a local model.
Given that you'd have probably a lot more capacity just to get to required bandwidth it could be used to switch very quickly between models (as all stored there would essentially be "loaded")
Maybe next you can run an LLM off of some old MFM drives