Google und SpaceX verhandeln über den Start von Orbital-Rechenzentren. Google-CEO: „Für mich besteht kein Zweifel daran, dass wir es in etwa einem Jahrzehnt als einen normaleren Weg zum Bau von Rechenzentren betrachten werden.“
https://finance.yahoo.com/sectors/technology/articles/google-spacex-talks-launch-orbital-171757141.html
36 Kommentare
How are they letting heat escape then?
What the fuck are they talking about
Why? Seems like all the down side and no upside? Expensive af, can’t fix it, terrible data transfer speeds, one time use…
Here’s my question. What does this solve? Are they thinking people might attack data centers because of AI job loss? Expecting service interruptions due to riots? Want unrestricted power from the sun? A way to get money into space to start private space stations? Thinking they’ll just run out of space?
This is a space sub, so imo we should be excited. If Google/SpaceX can solve this problem, will this not help improve future space missions?
I think its a cool problem they are trying to solve. The R&D will produce new tech that will be beneficial imo. Anyways, they are spending from their budget – I would rather see it go to solving impossible sounding problems, instead of dividends to shareholders or something.
Cool. So all of this hardware is expendable too? Forget repurposing this hardware. Forget recycling it. And say goodbye to our precious resources as we burn them up in the atmosphere en masse!
And you thought pollution was bad now? We will be sending hundreds of tons of computer hardware to orbit on a regular basis. It will inevitably face malfunctions and be deorbited. And then we get to destroy the atmosphere with even more rocket launches!
This is all coming from someone who is extremely pro space avidly follows starships development. What the hell are these people smoking?
None of my complaints even get into the engineering problems associated with this.
That’s awesome! I’m always up for putting more in space.
You mean like we’d be living on Mars in 2026? 🤣 Another grift to get more money. I’m not buying it.
This is Elon looking for ways to shift funds from what’s about to be an incredibly lucrative SpaceX IPO to xAI, which is bleeding money at a horrific rate even by his standards.
Im not understanding why exactly. You dont have a solid internet connection, so itll have to be satellite. You also cant radiate heat away nearly as easy because there is no air.
Other than being able to use a nuclear reactor to generate power, and no country to pay taxes to, whats the actual benefit here?
Anyone else concerned that we are just shoving AGI down human society’s throat and some of the only kill switches will be in space?
Lots of angry people in these comments and I’m really not sure why lol. The literal largest company in the world just released a whole new chip platform designed for use in space. They obviously know more than the average redditor.
I can’t find the prior comment I made on this but it’s provable nonsense just using Stefan-Boltzmann without needing to point out how idiotic it is logistically.
H200x8 server has TDP of 10.2KW. Even with a coating like PDMS you need 18m^2 of cooling per server (extremely idealized, likely much higher). 2-3 times that for powergen. Unless you are inventing very high temperature superconductors soon or exotic materials tech so NFRHT works this isn’t happening in a decade. It will like remain impossible through the end of the century.
You also can’t shield the ICs from particle interactions. RAD ICs all use very large nm nodes (200nm+ is typical) because there is an unfixable problem with very small traces where particles will activate multiple traces at the same time damaging components on it. Unless you are flying it inside lead or a comet you can’t protect against this. This is why a PowerPC design from 1997 is the most popular space processor.
This fraud train needs to stop.
Cool so when the server crashes its literally crashing into earth. great plan guys
Maybe learn to build it on earth first?
I mean if we can somehow take the computing power needed to sustain late in space then sure…
I still haven’t seen a decent hardware description. Since when were we making radiation hardened GPUs at scale? I get they are probably not going to be launching the equivalent of a server rack each time bc then they’d have to deal with massive radiators but, even then. It’s not that this is impossible but, it’s not a trivial level of investment and we don’t even know for certain that scaling compute will continue to give us gains, if anything the current rate of improvement in AI models has slowed to the point you would be forgiven for thinking we speedran moore law for AI
What exactly is the utility in these data centers whether they be on land or in space, serious question? How do they benefit me or society at large? I fail to see how any of this technology is benefiting the working class or poor people. It just seems to me that this will be used to exploit, manipulate, surveil, propagandize, and control people while enriching the top one tenth of one percenters.
This is someone’s AI-gets-loose fantasy, isn’t it.
Don’t worry, at some point it will sink in that AI is just another advancement and not the next coming, we’ll see a market correction to the AI hype train we’ve been on, people will sober up and then we can stop seeing these stupid articles about nonsensical space datacenters.
Isn’t cooling stuff a big deal with data centers?
Isn’t cooling stuff hard in space?
The laws of thermodynamics are screaming and sobbing into a pillow.
The only way this would work is with manned stations.
Right now zero day security vulnerabilities are being weaponized by bad actors at an alarming rate. I am keenly aware of it because the IT space I work in is being impacted at an accelerating rate.
When 100+ unmanned data centers get put into orbit, get pwned, then it is going to be ridiculously expensive to manually bring them down for reset and patching.
I’m not a cybersecurity guy, but I imagine those in that area are mostly shaking there heads in wonder/terror at this idea.
This is Hyperloop levels of „isn’t going to happen“.
I can definitely see it – in a decade we should definitely be in a place where we’ve solved the cooling mechanics in space. Easy peasy
You just have to make the data center not be hot then you won’t need to cool it. Problem solved.
Acronyms, initialisms, abbreviations, contractions, and other phrases which expand to something larger, that I’ve seen in this thread:
|Fewer Letters|More Letters|
|——-|———|—|
|[ITAR](/r/Space/comments/1tbn8vt/stub/oli154m „Last usage“)|(US) International Traffic in Arms Regulations|
|[LEO](/r/Space/comments/1tbn8vt/stub/oli7b3s „Last usage“)|Low Earth Orbit (180-2000km)|
| |Law Enforcement Officer (most often mentioned during transport operations)|
|[SEE](/r/Space/comments/1tbn8vt/stub/oli45yn „Last usage“)|Single-Event Effect of radiation impact|
|[SLS](/r/Space/comments/1tbn8vt/stub/olhze81 „Last usage“)|Space Launch System heavy-lift|
|[SSO](/r/Space/comments/1tbn8vt/stub/oli4evu „Last usage“)|Sun-Synchronous Orbit|
|Jargon|Definition|
|——-|———|—|
|[Starlink](/r/Space/comments/1tbn8vt/stub/oliboz6 „Last usage“)|SpaceX’s world-wide satellite broadband constellation|
|[kerolox](/r/Space/comments/1tbn8vt/stub/oli8sdw „Last usage“)|Portmanteau: kerosene fuel, liquid oxygen oxidizer|
|[methalox](/r/Space/comments/1tbn8vt/stub/oli8sdw „Last usage“)|Portmanteau: methane fuel, liquid oxygen oxidizer|
Decronym is now also available on Lemmy! Requests for support and new installations should be directed to the Contact address below.
—————-
^([Thread #12412 for this sub, first seen 13th May 2026, 04:02])
^[[FAQ]](http://decronym.xyz/) [^([Full list])](http://decronym.xyz/acronyms/Space) [^[Contact]](https://hachyderm.io/@Two9A) [^([Source code])](https://gistdotgithubdotcom/Two9A/1d976f9b7441694162c8)
This will be an entertaining comment section to come back to in 10 years
What the actual fuck. Let’s put the data further away so that the latency is higher. Literally the further you put it away the harder it is and the slower it is going to be to access your data. This is why AI Data centers are going in all over the place. You want the data as close to the users as possible the the best performance. Also consider how this mission to the moon barely could support the bandwidth for a single HD video stream. Source: I’ve been a network engineer for over 20 years.
Putting data centers in space is one of the dumbest things I’ve ever heard of, even if it is is *technically* possible.
I love how many “space data centre” experts this sub has, lol.
I hope everybody realizes that the reason this is being done is to make it more difficult to disrupt them… there’s legitimately no other practical reason you would put a data centre in space. The incredible cost involved in getting it there and then servicing it means it makes absolutely no sense unless the core purpose is to make sure they’re absolutely safe from the general population.
These guys know less about space than me, a normal earth human
I am still curious as to how they are going to handle the heat problem and the maintenance problem. I get redundant systems, but any “fix” will be a years long process.
Still waiting on that hyperloop. Oh and the mars mission. Oh and self driving cars. Oh and vr. Oh and….
There’s no latency issues with the data being in space if they build Elysium next.
Absolutely amazing to me these “experts” assume that the current approach to AI will scale beyond thermodynamic limits. It’s obvious that intelligence can be implemented on self repairing hardware running at about 37C.
Every single one of these grifters have such limited horizons despite claims of innovation and industry. Are they actually thinking we’ll still be using GPU-based 10 years from now?