Not sure that I completely agree here... I think part of the issue is that things have slowed down and even parallel GPUs are about as big as they can physicall

Not sure that I completely agree here... I think part of the issue is that things have slowed down and even parallel GPUs are about as big as they can physically get. The top of the line is pretty much gate-kept and sitting in warehouses (already purchased for AI datacenters that are years away from being built, let alone where the electricity will come from). We're in a kind of weird limbo of overpriced scraps and wait and see, while the all of computer mfg feels like an AI shell game combined with a cartel of memory pirates.