You are correct in the sense that they can stop work in a way many generic server use cases can't (which is seen in lowering power supply reliability requirements as the article mentions), but running expensive servers at 50% utilization would dramatically affect the revenue generated per capital invested - IE you couldn't afford to buy the servers.
If you invest into chips that deprecate in value really fast, not utilizing them to their full capacity because of power constraints would be counter productive.
I'm hoping the kind of gas turbines being installed on these datacenters are capable of quickly responding to a load change, meaning the primary source of energy could be solar during the day and complement when there isn't enough energy.
But I haven't looked into where these datacenter are being placed, I'm assuming that although solar is cheap now, the surface needed would make the purchase of nearby land probably not worth it. These new categories of datacenters are becoming very energy dense...
you're just incorrect. You probably missed 2 points :
- battery storage
- and in the article
"However, AI labs and some hyperscalers have relaxed those requirements as there is now a lower uptime tolerance applied to both inference and training, not just training. Many of Meta’s self-built AI datacenters, for example, target just two nines of uptime and forgo backup generators entirely, as detailed in our Industrials Model."