it seems a bit long for achieving just running your own AI models locally, given it seems to be largely a question of vram and that you already _could_ do it to

it seems a bit long for achieving just running your own AI models locally, given it seems to be largely a question of vram and that you already _could_ do it today with a handful of graphic cards. 20y ago we had the GeForce Ti4400, current graphics cards now come with 100x the vram and 50-60x the bandwidth.