zlacker

Do those people actually _repair_ hardware, or do they swap bad hardware for good chips?

Can SpaceX not just say "OK, GPU #7 on satellite #15872 is broken, don't use it" and just accept that they're now overbuilt on power/cooling for that sat?

>>lavery+(OP)
From what I understand it is in part swapping it, in part upgrading it. Some of it is preventative some of it is reactive. They could overbuild the hardware and slowly disable capacity I think that will not really work. Data centers are static in infrastructure but not in the systems running within them. Actually they are constantly changing to meet the needs.

Overbuilding also comes with a cost when talking about space, it is still very costly to get stuff up there and there is limited bandwidth downstream, you want to balance those two. So if you're overbuilding it costs a lot to get up there, if you disable what's up there you don't fully utilize the bandwidth.

For example AI data centers now use very different hardware compared to 5 or 10 years ago that upgrade path is just a lot harder when your data center is in space.