So, I did some analysis on where the location of skynet will be.
I started with these 5 locals: ant/arctic, ocean, desert, river, space. Then I ranked them in 6 categories/features. Finally I added the rankings to get an overall location fitness.

[energy availability]
river,space,desert,ocean,arctic

[absolute cooling]
space,arctic,ocean,river,desert

[heat transfer]
arctic, river, space,ocean,desert

[construction cost]
desert,river,arctic,ocean,space

[connectivity]
river,space,ocean,desert,arctic

[defensibility]
arctic,space,ocean,desert,river

naive category ranking results:
river 21, space 21, arctic 19, ocean 15, desert 14

Available energy and cost to build are probably the most important categories, so I doubled their weighting value. Deserts improved quite a bit and was the only ordering change:
river 30, space 26, arctic 23, desert 22, ocean 19

Surprisingly, rivers appear to be the best place to find mega data/computation centers. I had expected the arctic to be better, but lack of energy and cost to build hurt it.

I can safely say that we will see massive data/compute centers (distributed) in orbit. Musk/Starlink has no doubt already reached the same conculsion, so this analysis probably just 'unveils' a hidden function of Starlink that is already implemented. I would guess that each satellite has ~1000x the processing and storage needed to perform their 'official' function.

Fun fact: The arctic is 'colder'* than space.
I realized that when doing this analysis.