There's at least two reasons here why GS would be interested:
1. High frequency trading, if you control the software and make it as fast as possible, then all that is left is the networking between you and the exchange. Controlling the networking is the next step, this is total control, total integration
2. Limit backdoors; if you own the system totally and completely, you can nearly guarantee your system has no backdoors from state actors.
If you're as big as GS, you definitely don't want to own any american made networking hardware, and building it from the ground up is a cheap hedge against whatever lawsuits come down the line
The main problem is that the sun does not produce a whole lot of energy that can be captured on the night side of the earth, and we happen to consume a lot of energy when it's dark. If you overbuild capacity for daytime generation, nighttime generation is mostly solved, the big problem now is not cheap renewable energy, but rather, how to store it. Even if converting water and CO2 in to Ethanol is only 15% efficiency, you're still able to store 15% of your excess grid energy, whereas before you could only store 0% of excess grid energy. These guys are claiming 60% in the lab, which probably means 20-30% at industrial scale, perhaps 40% within our lifetime. It's not 85-90% hydro-electric efficient, but that's pretty dang good for a fuel which has so many uses, stores well for long periods of time and works with existing combustion engines.
A good chunk of that is executive and engineering salaries, R&D costs etc etc
If they fired all but a skeleton crew and ran it out of AWS they could definitely do what was proposed.
The problem is that they've been in "perpetual growth mode" for 8 years or so, but haven't been able to show any growth for 2-3 years. They've been trapped at about 320 million users for well over one year, soon to be two years. Due to... Silicon Valley mentality + brand name appeal(?) they've continued to attract business dollars but I suspect when they do hard analysis of how many engaged human (not bots, not automated scripts) users they have, they realize it's a skeleton service like PR Wire or whatever, that marketers buy in to and the business/journalism community uses as a read-only service, but there's no "social networking" happening by anyone with any money (i.e. adults and college students).
My guess is that Sales Force, etc sign a NDA, do their due diligence and find out that they have less than 150 million engaged verifiably human users. That certainly devalues the stock. Eventually it'll come out, it's only a matter of time.
My buddy bought a second hand, Xeon 16 physical, 32 logical core workstation with 96GB ram for under $1000. It's sandybridge era, but hot damn there's plenty of cores to spread the work around. We estimate that even though his workflow involves six different VMs, he won't need to upgrade his personal machine again for probably five years. I was considering buying one too, but I think I'll settle for a brand new i7 quad core XPS 15 with a 1050 GPU for ~$1800, which ought to last me four years or so. Maybe more. My 2012 era i5 thinkpad laptop (ivy bridge) is still faster than anything I need it for, but the screen needs updating, and to be larger as my eyesight starts to deteriorate as an adult and replaces my old "fire breathing" desktop from 2010. I'm a power user and barely find reason to upgrade my hardware, I can only imagine how long the average user can go between upgrades these days.
Wasn't there something about a PASCAL programmer knowing the value of everything and the Wirth of nothing?