Anyone know? Feel like I should get educated on this so I don’t look silly when I talk about it. I guess we can support it much more easily than I thought.
Anyone know? Feel like I should get educated on this so I don’t look silly when I talk about it. I guess we can support it much more easily than I thought.
Uh… You literally just said we can’t do the thing and then said we’ll do the thing. Gas turbines from South Korea are the components you need to build a power plant. Sure, you can build it behind the meter, but it still requires engineers and other specialists and as you said, we don’t have enough of those to go around. So, no, the idea of making independent power plants to run these data centers is just as hard as making grid-attached power plants
There’s a difference between building more grid capacity and building a private generator.
One requires more engineering expertise than the other.
But it’s the same type of engineering expertise, of which we have a shortage. For example, Amazon said they’re going to build a 7.5GW gas power plant for their data center cluster. The largest gas power plant in the world is 8.5GW and it took 40 years to build. It sort of doesn’t matter if Amazon’s plant is going to be grid-attached or not, it’s going to be a massive engineer effort.
Easier to engineer for one customer than 20,000, maybe.
Yes, the grid itself is infrastructure that is failing regardless of the state of power plants, and going behind the meter eliminates that one complexity. But that’s not as much a rarified engineering problem as it is a straight up under-developed infrastructure problem.
There’s still a 3+ year waiting list for gas turbines. There’s still a massive shortage of engineers that can build a power plant. The power plants are still incredibly difficult. And a.i. workloads are so volatile that it requires entirely new solutions to power management.