Remix.run Logo
RIMR an hour ago

Let's just put it this way:

The ISS produces about 120 kilowatts of electricity.

An Nvidia Blackwell B200 GPU uses 1.2 kilowatts of electricity.

So, you would need a similar array of solar panels and radiators just to power 100 of them. You probably would need 2-3 launches for a satellite this big, and realistically, you would just make smaller satellites.

That's $4,000,000 worth of GPUs, A couple millon or more of RAM, SSDs, etc., a radiation-proof satellite housing to support all of that hardware, solar arrays, launch costs ($74M per Falcon launch), all for maintenance to be impossible and the hardware to become obsolete in a couple of years.

It's a delusion unless we invent some way to go to space for free.

AnthonyMouse 4 minutes ago | parent | next [-]

I don't really get the obsolete argument.

The thing has two main parts. One, a bunch of solar panels, shielding and radiators. This the heavy / expensive to launch part, but should last for what, decades? Two, a bunch of GPUs. These become obsolete, but so what? They're not that heavy, so in a few years you send up another rocket and swap them out.

asadotzler 3 minutes ago | parent | prev | next [-]

Launches are not $74M. That's retail pricing.

SpaceX's launch cost, the internal spend to put one Falcon 9 Starlink payload in orbit, with a return to launch site booster recovery, is about $15M.

If you're going to make such assertions, do the legwork to make sure your numerical claims aren't off by 500%.

LorenDB an hour ago | parent | prev [-]

SpaceX would be launching these on Starship, which has a much lower targeted launch cost.

RIMR 8 minutes ago | parent [-]

For reasons already outlined, I have doubts about their math. Targeting $250/kg payload costs is ambitious for a rocket that has not yet successfully reached orbit or proven cost-effective.

Even if we do somehow succeed at affordably dumping tons of GPUs into orbit, what do we do about the Kessler Syndrome?