Item — HN Reader

This video is the perfect example of why Scott Manley is incredibly irritating.

He spent an hour demonstrating why data centers in space are impossible, failed to mention half of the issues that make it impossible. Then closed the video saying it "remains challenging"....So he can do another five videos about it over the next few months.

Satellites and the Space Station have computers. So we have "data centers" in space.

But what is being discussed are 100 MW space data centers. Make the calculations....Those need a 20 x 20 mile radiator and a solar farm the same size. Those are space constructions at scifi level maybe for the next century.

The type of data centers we aiming to substitute at 1 GW...

Finishing the video saying this "remains challenging" its like spend an hour talking about Musk wanting to land on the Sun, listing half the issues and instead of identifying the craziness concluding it ...."remains challenging..."

2 comments

> But what is being discussed are 100 MW space data centers. Make the calculations....Those need a 20 x 20 mile radiator and a solar farm the same size. Those are space constructions at scifi level maybe for the next century.

No one is talking about building a single massive structure. SpaceX has said they’re targeting 100kW satellites. A lot of them. Which they’re already pretty good at.

Technically it’s pretty straightforward. The only real challenge is making it cheap enough.

Lots of 100 kW satellites, does not remove the problem. For 100 MW you need about 1000 of them, each with power, batteries, radiators, pointing, shielding, communications, propulsion, spares, and replacement. GB200s are not space rated and depreciate in 3 years... And a 100 MW data center is a tiny thing, will hardly make a dent on the compute landscape.

Do you see the massive white panels of this massive structure called the ISS? : https://i0.wp.com/www.nationalreview.com/wp-content/uploads/...

Those are radiative panels who do at a max ...drum roll...120 kw

Saying "the only challenge is making it cheap" is like going back 10 years ago and talking to Tesla FSD defenders.... Just ignore the hard parts, call them implementation details, then declare victory...

> what is being discussed are 100 MW space data centers

Thats... not how engineering anything works.

20kw gives you a reasonable amount of compute, about a GB200 rack. That's pretty good for serving inference workloads. Starship is designed to deploy 50-100 of these satellites per launch. The claim from this video is the existing design can do it.

If the next generation of chips run hotter, require less energy to do more things, suddenly you have a data center worth of compute per launch. And those are exactly the goals for the upcoming few generations.

Maybe you misunderstood the point of the video? It was to address the poorly reasoned arguments about thermodynamics making cooling data centers in space impossible. Not that you can yeet a data center today.

A GB200 rack is 120 kW. That is all a massive structure like the ISS can do. The white panels: https://i0.wp.com/www.nationalreview.com/wp-content/uploads/...

The problem with Musk chimeric ideas, is that Engineering gets in the way...