The current regular servers proved capabilities that hyperscalers are selling at a drastic loss, most of which can be done with a build run on your PC.
in order for hyperscale AI to break even, it has to make $6 trillion a year. The ambition was to replace most of the skilled workforce with something cheaper or more productive. (And that also means pushing most of society into ghettos and reservations.)
The other interest was government projects, such as commanding drone armies and processing surveillance data.
Six trillion is an awful lot, and its looking unlikely there’s that much business to go around.
There’s no scenario in which building data centers will make the company stop existing, they can just sell or rent what ai data centers they have and exit the business worst case.
What? No. This is ai data centers new build. Nothing about that would melt their current regular servers.
The current regular servers proved capabilities that hyperscalers are selling at a drastic loss, most of which can be done with a build run on your PC.
in order for hyperscale AI to break even, it has to make $6 trillion a year. The ambition was to replace most of the skilled workforce with something cheaper or more productive. (And that also means pushing most of society into ghettos and reservations.)
The other interest was government projects, such as commanding drone armies and processing surveillance data.
Six trillion is an awful lot, and its looking unlikely there’s that much business to go around.
It’s all just hype, the big numbers were all future spending so they’ll just be dialed down.
That money has to come from somewhere. Ultimately the debt service will hit their opex.
There’s no scenario in which building data centers will make the company stop existing, they can just sell or rent what ai data centers they have and exit the business worst case.