There are currently about 12600 satellites in orbit. As a result on average every day 3 fall out of space, dumping metals and other nasties into the upper atmosphere.
If we continue that rate of satellite loss, 1 in 4200, and extrapolate it to 1,000,000. That would be ~238 satellites PER DAY, falling out of the sky and spreading the materials they are made up of in the upper atmosphere. With some more substantial chunks hitting the surface, and possibly people.
That's just bonkers
1/n
This of course completely overlooks all other practicalities of orbital datacentres, that makes putting high power computing in orbit. Which for a summary include: too much radiation noise making the systems unstable (see Wikipedia for "single even upset"), cooling when you have to dump heat into a vacuum, low data bandwidth (compared to a fibre on earth), latency, and shear fucking cost.
It's an absolutely fucking stupid idea. And I'm angry I have to spend my Sunday debunking this shit.
2/2
@quixoticgeek
The cooling is manageable (the solar panels have a back side)
The low bandwidth and high latency are probably tractable for many AI and training purposes.
The whole point of the proposal, is to reduce the cost to orbit by another factor 100, just like space x did the first time, so complaining about cost misses the entire point. (Whether they can do it, I don't know!)
However. I don't see a good solution for radiation hardening. We have 15 pounds of shielding per square inch...
Quixoticgeek
Very Human Robot
jiub 
