(Lemmy failed to send my original reply and I lost it I think, but if another similar one eventually show up... that's why)

I wouldn't rule nuclear out in the not so far future if SMRs properly come online. I think those could scale up demand at the pace they could build these monstrosities in the not so distant future, like give it another 15 years and it might start being okay?

And ya, the renewable way IS possible, and Texas would be a good spot for something of this scale, but it would be a huge undertaking. I imagine we would see amazing improvements in solar/battery tech building something like that out with economies of scale and additional research from the huge amount of funding it would create. Also there's no reason they can't make them do some % with renewables.

This is aslo why Space based stuff maybe isn't as far flung as the idea seems. It obviously requires new tech that doesn't function yet.

e.g Starship needs to get down to its targeted $50-$200/kg) which is not a given.
New chips that can run hotter so the radiator size is reasonable. Cheaper solar panels as space based panels are very expensive right now.

But with that (and all the other things) you get 7.65GW of power being 7.65GW of power and minimal batteries required. Space solar is more effective as well since there's no atmosphere, so the panels will be 30% more efficient or something like that than on land. It really does solve some problems, but it obviously creates others.

(edit: I worked it out once, and at the launch cadence of F9, they could put ~1GW of solar up a year, which would be pretty fast compared to land)

All that being said, while I don't doubt they can launch these satellites, being profitable is an entirely other matter, but being profitable on land is also a huge unknown for something of these scales. Like, do they think there's enough demand for these large quantities of huge datacenters? I'm not convinced this is a, if you build it, they will come situation for LLMs. True AI absolutely, but LLMs not a chance. So ALL of this IMO is betting these insane amounts of money and pollution on something they think they'll eventually build, not something they have today.

At least with these land based ones if they take on building the SMRs or massive solar fields, we can repurpose those after. If they do it with natural gas, that's much less ideal, sure the power might be useful, but we need to move away from it and that is the wrong direction. Space based ones will just flop with no backup benefit.

You're not wrong. One of the newest reactors to come on line is reactor 3 at Plant Vogtle in GA. After as much as 5 years for the approval process, the reactor broke ground in 2009 and didn't start producing power until 2023. That's just shy of 2 decades from start to finish. As much as my wind and solar idea is flawed, nuclear is as much of a non-starter. These folks may want to open their data center a little sooner than 2045.