local RAM used to run local AI models, whichl he argues is unused most of time

Cue in that interview from Nadella saying he has unused GPUs because they haven’t got datacenters to put them in. Maybe Altman should give him a call about his unused RAM…