An interesting analysis:
There is a rush for AI companies to team up with space launch/satellite companies to build data centres in space. TL;DR: It's not going to work.
In the interests of clarity, I am a former NASA engineer/scientist with a PhD in space electronics. I also worked at Google for 10 years, in various parts of the company, including YouTube and the bit of Cloud responsible for deploying AI capacity, so I'm quite well placed to have an opinion here.
The short version: this is an absolutely terrible idea, and really makes zero sense whatsoever. There are multiple reasons for this, but they all amount to saying that the kind of electronics needed to make a datacenter work, particularly a datacenter deploying AI capacity in the form of GPUs and TPUs, is exactly the opposite of what works in space. If you've not worked specifically in this area before, I'll caution against making gut assumptions, because the reality of making space hardware actually function in space is not necessarily intuitively obvious.
[Source]: https://taranis.ie/datacenters-in-space-are-a-terrible-horrible-no-good-idea/
(Score: 2, Insightful) by Anonymous Coward on Tuesday March 03, @06:01AM
A phrase used about 25 years ago.
Have all of the grownups left business now as well as government?
(Score: 5, Insightful) by ikanreed on Tuesday March 03, @06:06AM (4 children)
You need to think like a CEO getting rich off the biggest bubble in history and you desperately need it to not pop yet
(Score: 4, Interesting) by Thexalon on Tuesday March 03, @12:01PM (1 child)
Or alternately, think like somebody making a pitch to vulture capitalists who will give you big bucks to work with which you can use to pay yourself a hefty salary for a few years even if you don't produce anything that actually works.
Either way, the goal is to sound like it's something that makes sense, even if it's total nonsense. See also: Theranos.
"Think of how stupid the average person is. Then realize half of 'em are stupider than that." - George Carlin
(Score: 2, Informative) by Anonymous Coward on Tuesday March 03, @02:55PM
The circle-jerk of capturing all of the world's memory and disk drive capacity is
taking this game to the next level.
(Score: 2) by coolgopher on Tuesday March 03, @02:16PM (1 child)
Think not of the insanity of the "solution", think instead of what it might enable. Drop enough compute into orbit, you get autonomy. Couple autonomy with thruster control and an objective, and you have an orbital weapon. If you have a weapon, you have leverage. Leverage equals money.
(Score: 0) by Anonymous Coward on Tuesday March 03, @02:57PM
Like the Jewish space lasers?
At least those funnel all their impossibly high power requirements
into one short shock.
(Score: 4, Informative) by canopic jug on Tuesday March 03, @06:16AM (1 child)
Looks like a duplicate of this one, even down to the main URL:
(2025) Datacenters in Space Are a Terrible, Horrible, No Good Idea. [soylentnews.org]
Money is not free speech. Elections should not be auctions.
(Score: 3, Insightful) by pTamok on Tuesday March 03, @06:27AM
I prefer to have a false negative in the duplicate detection algorithm and see a topic twice rather than a false positive and not see a potentially interesting topic at all.
Comments about duplicates show that "many eyes make dupes shallow." [wikipedia.org].
(Score: 5, Funny) by jb on Tuesday March 03, @07:40AM (2 children)
Given that LLMs mostly produce the wrong (or outright irrelevant) answers anyway, there seems a reasonable probability that when running them unprotected in space, the odd bit flipped by some sort of bizarre unearthly particle or other might actually improve the accuracy of their output.
Then again, for people who think such "antiquated" concepts as reason and logic don't matter (the only kind of people who would ever use LLMs), throwing the whole thing away and just buying one of those "magic 8-balls" from the local junk shop instead would probably achieve a similar level of accuracy and would certainly cost far less than sending anything into space...
(Score: 3, Funny) by sonamchauhan on Wednesday March 04, @12:31AM
A true RNG!
For additional random goodness, don't use ECC memory!
(Score: 2) by driverless on Wednesday March 04, @11:33AM
Also given that you're running your gigawatt-heat-producing data centres insulated by a near-perfect vaccuum you're going to see all sorts of thermal-induced failures even before you get the radiation-induced ones. It's a win-win!
(Score: 5, Insightful) by ledow on Tuesday March 03, @08:17AM (11 children)
I'm not a space scientist, nor do I work for datacentres, and even I know it's among the dumbest ideas ever posited.
Microsoft struggled (and then basically abandoned) an "underwater" datacentre.
Space is orders of magnitude more expensive and difficult than even that. There's just no way to cool anything, and power it all.
And there's absolutely zero advantage to doing this "in space" compared to anywhere else except one thing: Legal jurisdiction. But not being on earth they can just (try to) pretend that the illegal material they're storing and training on is somehow immune to the laws of any particular country. That's what they're really after.
(Score: 5, Insightful) by c0lo on Tuesday March 03, @08:31AM
Theoretically, there are means to power and cool the thingy - involves as much area as needed (for both PV capturing solar light and cooling panels leaking IR) .
Practically, it's extremely costly to build and even more so to maintain - those micrometeorites will have quite an intercept cross-section to impact into.
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 5, Touché) by turgid on Tuesday March 03, @09:08AM (4 children)
Microsoft struggled (and then basically abandoned) an "underwater" datacentre.
To be fair, Microsoft struggles and abandons most things that are not to do with milking the cash cow of their user base they locked in with MS-DOS back in the day. Weren't they going to take over the world of mobile phones, for example? They did manage to kill Nokia, though.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 4, Funny) by c0lo on Tuesday March 03, @10:17AM (3 children)
To be fair, sysadms with a diver certification should be hard to find. I mean, divers should value their time of life beyond wasting it working for Microsoft.
https://www.youtube.com/@ProfSteveKeen https://soylentnews.org/~MichaelDavidCrawford
(Score: 3, Insightful) by turgid on Tuesday March 03, @11:10AM
Quite. Who would put a GUI under the sea where it's hard to click on the buttons?
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 2) by turgid on Tuesday March 03, @12:53PM
I forgot, you can log into Windows remotely now with SSH, if you set it up right. It takes a couple of hours to set up. Windows does however tend to forget its configuration randomly and spontaneously, so you'd still need the diver on standby. And then there's RDP, which works sometimes.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 3, Funny) by khallow on Tuesday March 03, @01:33PM
(Score: 2, Interesting) by khallow on Tuesday March 03, @02:40PM
Here's a simple strategy. Throw up a few moderately useful datacenter satellites. They don't have to have a lot of capability or cost much (relatively speaking). Designate them as "prototypes". After a few weeks, sell them to a shell corporation. The subsequent shenanigans fall onto the shell corporation not the original owner. Launder data through the satellites - that is, conduct illegal activities on Earth, route the output through the satellites so to make it appear as if this were done on the satellite, and then profit. Since they're in space, and the shell corporation is based in some convenient locale, the original owner has plausible deniability in case illegal activities or data are exposed.
(Score: 2) by owl on Tuesday March 03, @03:47PM (1 child)
And that part only lasts until one or more governments get fed up with the "outside our jurisdiction" arguments and imposes a rule that: "you, there in space with your datacenter, you are now under our jurisdiction". Will take years to shack out through the courts, but it will occur.
(Score: 1, Insightful) by Anonymous Coward on Tuesday March 03, @05:01PM
"Will take years to shack out through the courts"
If the wrong people get tired of it, they will just take it out,
hopefully without a rapid disassembly while still in orbit.
(Score: 0) by Anonymous Coward on Tuesday March 03, @06:56PM
And that would be a good thing. We shouldn't let some corrupt government decide what's legal or illegal. We need an indelible and bulletproof internet where tyrants can't meddle. But yeah, space is dumb. It's much safer and secure underground
(Score: 2) by Reziac on Wednesday March 04, @02:42AM
Yonder is Greenland. Coal, cooling, and who the hell really owns the place? Profit!
And there is no Alkibiades to come back and save us from ourselves.
(Score: 4, Informative) by clive_p on Tuesday March 03, @02:08PM (3 children)
I have been a space scientist for many years. I had a bit of hardware in the back of my car once that was part of a spacecraft that is still in orbit. The making of space-qualified electronics is very difficult and expensive and I totally agree that this is an idea that is simply not going to work in the foreseeable future.
(Score: 4, Interesting) by owl on Tuesday March 03, @03:51PM
This is the part that the general public (and the vulture capitalists) don't understand.
Radiation hardened electronics is very different tech than the latest Nvidia GPU. A rad. hardened Nvidia GPU would set the GPU back to a very early stage in GPU performance, requiring even more GPU's to make up the difference.
But the Elongated Muskrat is simply not going to buy commercial Nvidia GPU's and launch them into space and gain some working space data center. He can do all the above, but the result won't work properly.
But then, the Elongated Muskrat is not actually looking to put data centers into space, he is looking to PT Barnum the Vulture Capitalists into throwing more money at him, and "data centers in space" is the latest way to extract Vulture Capitalists money.
(Score: 3, Insightful) by https on Tuesday March 03, @04:23PM
It doesn't need to work. It needs to convince suckers now that Spacex will have growth in the future.
Offended and laughing about it.
(Score: 2) by corey on Wednesday March 04, @10:10PM
I’m working on space electronics and have done so for a few years, including moon rover.
What this dude says in his article is absolutely on the mark. I especially got into radiation tolerance in electronics for the moon mission as it’s out of earth’s protective magnetosphere. This might be the only “yeah maybe”’s for data centres in space, so long as they are in LEO (which they would be for low latency comms away). Thermal management is then a major issue. Radiator paneling is difficult and large, as he says. On the moon, we had to design the top panels to not “see” any lunar surface because even that radiates heat. Solar panels facing the sun were estimated to reach 150C. While paneling away from the sun would be -50C pretty quickly.
Power needs also mean they’d have to be in a sun synchronous orbit (to always be in the sunshine). That places constraints on comms to your data centre because one minute it’s over your land, then it’s over the Arctic, then it’s over Antarctica. Even super expensive triple junction solar panels are mid 30s% efficient, and are susceptible to degradation from TID radiation effects, so you’d need a shitload for the centre and you need a program to constantly go up and replace aged panels.
Lugging all that up there, that’s fun too. SpaceX just upped their launch payload cost (on Transporter missions) to $7k/kg.
To be frank, this is all possible, engineering wise. It’s just going to cost the GDP of a small sized nation. And if that stacks up, I’ll eat my hat.
(Score: 4, Insightful) by Spamalope on Tuesday March 03, @02:41PM (2 children)
Why don't we do some recreational speculation for fun?
Lets say you want to put up a sat constellation with capability that's not public until it's a faite acompli - or even secret abilities.
You'd want a cover story to explain why they need so much extra power, and lots of buzz around the story.
You'd want a backup layer.
Ok, you got me. The dev effort is really because it's easier to get data centers accessible from anywhere for the Moon and Mars if you never have to land them and the comms and data center are the same infrastructure.
Then you can have fun with Austin Powers speculation for the immediate purpose. Space lasers! Microwave power sats? Just disinformation until the cell phone network is built to disrupt telcos?
Go wild, you can dream up a 100 things. Or it's dull... they're about to be able to launch for 1/100th the normal cost, and are dreaming up something to do with that ability.
Or it's just dumb... but, so far all the dumb seeming things from that group have worked (or been actual speculative Mars projects - like the boring company making equipment that just so happens would fit on Starship)
(Score: 0) by Anonymous Coward on Tuesday March 03, @03:04PM (1 child)
Musk isn't that clever.
A lot of high-tech solutions for non-problems the world will
soon not be able to afford. Or, just another smoke screen
and lightning rod to attract the world's news cycles.
(Score: 2) by owl on Tuesday March 03, @03:58PM
Exactly this. If you apply a PT Barnum [wikipedia.org] lens to anything that comes out of the Elongated Muskrat's mouth for any "new tech idea" you'll recognize that his entire aim is to hype something new to obtain Vulture Capitalists money. When "new thing" begins to get old, or somewhat less shiny, the Elongated Muskrat pivots to the next new thing to keep the cycle going.
(Score: 2) by Deep Blue on Tuesday March 03, @04:21PM
if the data is to be consumed in space, like for a example in a space base or something, then it woud make sense. For earth consumption, i agree with the topic.
(Score: 2) by Zoot on Wednesday March 04, @06:56AM (3 children)
I love space and space exploration, especially but not exclusively unmanned exploration, and honestly I don't think there's anything out there that we need that we can't find much more inexpensively here on Earth.
A trillion-ton asteroid made of 100% pure gold/diamond/iron? The cost to move it safely to Earth and somehow get it down here is going to be WAY higher than finding a sufficient amount of the same mineral here on Earth. For the same price you can strip mine all of Texas if humanity somehow is so desperate for the resource in question.
(Score: 2) by KritonK on Wednesday March 04, @01:30PM
A trillion-ton amount of anything would make that material worthless. One would not want to invest money in such an endeavor, unless making that material dirt cheap is the intended goal.
(Score: 2) by gnuman on Wednesday March 04, @06:45PM
Bad example ... gold is $100/g. That means 0.01g/$. $1e12*1e-2g/$=1e10g=10,000tonnes. That's just like a trainload worth. And such a mass you could even land safely, if you don't look up.
As for iron, well, different scale and you are correct. It's $100/T. so, that would be 1e10T. Calculator says that's like a 2.5km diameter ball. Landing *that* safely ... I think you get a movie from that?
(Score: 2) by jb on Thursday March 05, @08:16AM
The biggest problem with mining asteroids is finding a relatively stationary one within range. I'm not convinced there are any. But if you could find one (and its composition was mostly whatever it was you wanted to mine), it could be viable. Yes the actual mining would cost much more if done in space. But a sensible miner would do the processing up there too. That not only solves the "how do I land such a big / heavy thing safely" problem (simply bring down the processed product in smaller quantities) but more importantly means that your refinery or other processing plant is not in the territory of any country, so not subject to any country's planning or environmental laws. Skipping the usually-mandatory EI study (which takes years and costs a packet) and the (usually even larger) cost of implementing all of its recommendations in order to get your permit, would likely even up the scales quite a bit.
Still, I doubt it's like to happen in my lifetime because: (a) the asteroids that pretty much sit still are much too far away (round trip time measured in decades); and (b) the ones that pass by Earth close enough to reach more easily tend to pass by too quickly to get enough steady-state operating time of the "mine" to pay back the up-front cost of establishing it.