Comment by dangerlibrary | Hacker News Reader

Comment by dangerlibrary | Hacker News Reader

Parent Full thread

I'm pretty sure there are plenty of payloads and demand to put interesting stuff in space for SpaceX to have a profitable business. I don't think "datacenters in space" is a necessary condition for their success.

Which is good, because "data centers in space" has never made sense. Running a data center requires a ton of energy (because chips are functionally just heaters) and a way to recapture / dissipate that energy to keep the chips at an optimal operating temp. All the terrestrial arguments around data centers and water use are about cooling.

As far as I'm aware, space remains a vacuum. You can radiate some amount of heat off into a vacuum, but nowhere near enough to run modern hardware 24/7 at peak flops or whatever. Also, without an atmosphere, all the hardware you put up there needs to be radiation hardened. Also I know people whine about how few jobs data centers support during ongoing operations but the people who _are_ there are pretty much just swapping out failed drives and components because all these things have finite lifespans.

I've never heard someone explain what bottleneck exists on earth that is slowing data center development/deployment where the solution is "what if we put it in space, instead?" Is it some military defense thing about live target inference for rocket defense or something? I don't get it.

Watch the Scott Manley video where he does a very deep dive on the physics of cooling a data centre in orbit. He takes into account reflected heat off the earth and a LOT more.

TLDW summary: it’s entirely doable. Anyone that says it isn't simply hasn’t looked at the physics and math.

Will it be financially beneficial is a totally different question. Of course there were endless experts saying reusable rockets and electric cars never would be.

> I've never heard someone explain what bottleneck exists on earth

1. Power.

2. Political will (current massive pushback over data centre builds).

Feasable for sure, humans can solve this. But Elon Musk doesn't seem to really throw himself on solving power.

I find that very weird.

He 'makes' (panasonic) batteries on mass and had Tesla Solar roofs. So why instead of talking about a dyson sphere or DC in space, he wouldn't just build a solar city in the dessert, put some fiber around and start putting solar panels and his batteries around? AI is not that latency dependend than a lot of other things

Tesla are now refining their own lithium, and SpaceX have just started to build a monster solar panel factory.

Against my better judgment, I let someone on the internet tell me to do my own research by watching a 30 minute Youtube video. Parts of it, at least. I regret it.

Scott Manley literally waves his hands at thermals, claiming that because there exist 28kW SpaceX satellites in space (that operate only intermittently, because they need to cool down), it is feasible to efficiently run modern datacenters that are ~three orders of magnitude more power-hungry 24/7.

He does not do any math. He acknowledges that cooling systems are heavy, take up huge amounts of space, and that small, existing satellites need to schedule downtime to cool off. I'm no physicist, but I'm pretty sure radiant cooling systems scale with surface area, which is bad news for anyone trying to scale existing cooling systems up 1000x.

[0] https://www.youtube.com/watch?v=DCto6UkBJoI

>that operate only intermittently, because they need to cool down

Do you have a source on that?

All the information in my comment is discussing the content of the linked video.

The payload increased from 1000t to 3000t only because of starlink. What we will get is a lot more starlink like satelites which might also then need to be replaced every 5 years and poisening our atmosphwere while doing so.

Reply on news.ycombinator.com