More Tech Moguls Want to Build Data Centers in Outer Space (msn.com) 90
"To be clear, the current economics of space-based data centers don't make sense," writes the Wall Street Journal.
"But they could in the future, perhaps as soon as a decade or so from now, according to an analysis by Phil Metzger, a research professor at the University of Central Florida and formerly of the National Aeronautics and Space Administration." "Space enthusiasts (comme moi) have long sought a business case to enable human migration beyond our home world," he posted on X amid the new hype. "I think AI servers in space is the first real business case that will lead to many more...."
The argument essentially boils down to the belief that AI's needs are eventually going to grow so great that we need to move to outer space. There the sun's power can be more efficiently harvested. In space, the sun's rays can be direct and constant for solar panels to collect — no clouds, no rainstorms, no nighttime. Demands for cooling could also be cut because of the vacuum of space. Plus, there aren't those pesky regulations that executives like to complain about, slowing construction of new power plants to meet the data-center needs. In space, no one can hear the Nimbys scream. "We will be able to beat the cost of terrestrial data centers in space in the next couple of decades," Bezos said at a tech conference last month. "Space will end up being one of the places that keeps making Earth better."
It's still early days. At Alphabet, Google's plans sound almost conservative. The search-engine company in recent days announced Project Suncatcher, which it describes as a moonshot project to scale machine learning in space. It plans to launch two prototype satellites by early 2027 to test its hardware in orbit. "Like any moonshot, it's going to require us to solve a lot of complex engineering challenges," Pichai posted on social media. Nvidia, too, has announced a partnership with startup Starcloud to work on space-based data centers. Not to be outdone, Elon Musk has been painting his own updated vision for the heavens... in recent weeks he has been talking more about how he can use his spaceships to deploy new versions of his solar-powered Starlink satellites equipped with high-speed lasers to build out in-space data centers.
On Friday, Musk further reiterated how those AI satellites would be able to generate 100 gigawatts of annual solar power — or, what he said, would be roughly a quarter of what the U.S. consumes on average in a year. "We have a plan mapped out to do it," he told investor Ron Baron during an event. "It gets crazy." Previously, he has suggested he was four to five years away from that ability. He's also touted even wilder ideas, saying on X that 100 terawatts a year "is possible from a lunar base producing solar-powered AI satellites locally and accelerating them to escape velocity with a mass driver." Simply put, he's suggesting a moon base will crank out satellites and throw them into orbit with a catapult. And those satellites' solar panels would generate 100,000 gigawatts a year. "I think we'll see intelligence continue to scale all the way up to where...most of the power of the sun is harnessed for compute," Musk told a tech conference in September.
"But they could in the future, perhaps as soon as a decade or so from now, according to an analysis by Phil Metzger, a research professor at the University of Central Florida and formerly of the National Aeronautics and Space Administration." "Space enthusiasts (comme moi) have long sought a business case to enable human migration beyond our home world," he posted on X amid the new hype. "I think AI servers in space is the first real business case that will lead to many more...."
The argument essentially boils down to the belief that AI's needs are eventually going to grow so great that we need to move to outer space. There the sun's power can be more efficiently harvested. In space, the sun's rays can be direct and constant for solar panels to collect — no clouds, no rainstorms, no nighttime. Demands for cooling could also be cut because of the vacuum of space. Plus, there aren't those pesky regulations that executives like to complain about, slowing construction of new power plants to meet the data-center needs. In space, no one can hear the Nimbys scream. "We will be able to beat the cost of terrestrial data centers in space in the next couple of decades," Bezos said at a tech conference last month. "Space will end up being one of the places that keeps making Earth better."
It's still early days. At Alphabet, Google's plans sound almost conservative. The search-engine company in recent days announced Project Suncatcher, which it describes as a moonshot project to scale machine learning in space. It plans to launch two prototype satellites by early 2027 to test its hardware in orbit. "Like any moonshot, it's going to require us to solve a lot of complex engineering challenges," Pichai posted on social media. Nvidia, too, has announced a partnership with startup Starcloud to work on space-based data centers. Not to be outdone, Elon Musk has been painting his own updated vision for the heavens... in recent weeks he has been talking more about how he can use his spaceships to deploy new versions of his solar-powered Starlink satellites equipped with high-speed lasers to build out in-space data centers.
On Friday, Musk further reiterated how those AI satellites would be able to generate 100 gigawatts of annual solar power — or, what he said, would be roughly a quarter of what the U.S. consumes on average in a year. "We have a plan mapped out to do it," he told investor Ron Baron during an event. "It gets crazy." Previously, he has suggested he was four to five years away from that ability. He's also touted even wilder ideas, saying on X that 100 terawatts a year "is possible from a lunar base producing solar-powered AI satellites locally and accelerating them to escape velocity with a mass driver." Simply put, he's suggesting a moon base will crank out satellites and throw them into orbit with a catapult. And those satellites' solar panels would generate 100,000 gigawatts a year. "I think we'll see intelligence continue to scale all the way up to where...most of the power of the sun is harnessed for compute," Musk told a tech conference in September.
Re: (Score:2)
Maybe AI can finally help you add capital letters to your posts.
LIKE THIS?
Re: (Score:1)
i use cusur ai to get me in the ball park. I have yet have it make something actually worked. but when it fucks up relly bad and i tell it did it. it comits sucside.
Re: (Score:1)
do they have any clue the cooling isses they would have, the space shuttle had a 240 watt total engery budget.
DCs in space is just fucking delusional (Score:5, Insightful)
Re: (Score:2, Interesting)
Yeah it seems to me the choice is either
1) a stable orbit with a 50% duty cycle, which means more collectors AND heavy batteries. Expensive.
2) Parking it in something like a lagrange point. Extremely stable forever-orbit, but extremely expensive per kg to get it to that place. And space junk from abandoned shit will never go away
3) geostationary orbits that require a tonne of fuel for constant adjustment burns.
None of this of course factors iin cosmic ray shielding and the enormous amount of infrastructure
Re: (Score:1)
In the meantime, gigaDCs are invading unincorporated and small towns driving up the local cost of water and electricity while adding to traffic and noise pollution to mine BTC or fulfill Zuck's and Sama's fanciful mirage of stonks going up forever by virtue of "if you build it, they will come". It's billionaires and corrupt politicians having Great Gatsby parties and sticking random peo
Re: (Score:3)
The only reasonable way to make these PV behemoths is as heliogyros, otherwise it takes too much structural components.
They will sail the solar wind ... how that will actually look orbit wise, dunno, it's completely new territory and no one is doing the simulations yet.
Cosmic rays should be handled with resilient architecture, not shielding. Everything needs to be extreme light weight, at most you could generate put a couple megavolt between wire electrodes to guide charged particles to avoid the sails (fle
Re: (Score:1)
Viability is questionable regardless, but the $500/kg assumption may be overly pessimistic. The article assumes a time frame of at least a few years from now. SpaceX Starship future cost estimates predict as low as $10/kg to LEO and $50/kg to GEO by then - assuming high reusability is achieved (100 launches per ship).
Re: (Score:2)
Re: (Score:1)
They’re based on engineering estimates from the world’s top rocket payload delivery company (by far), its consistent history of meeting its own estimates, as well as independent analysis of same by industry experts. They’re NOT based on an oddly smug and rather spectacular flip: “person M was the second coming of Jesus until he derailed my side’s self affirming censorship”.
But hey, no worries - the truth is simply performative, eh? Even NPR CEO Katherine Maher agrees that
Re: (Score:2)
When you say 50% duty ... Is that an engineering factor or are you referring to orbiting into Earth's shadow? The latter won't happen because a polar orbit can easily avoid going into Earth's shadow.
BTW: I agree with the insanity of AI spending in general. The big deployments are just doing search engine function.
Re: (Score:2)
There is an opportunity, for a short blackout of a few seconds, a couple of times a months - When crossing behind the Moon's shadow as the Moon orbits on the Sun side of Earth.
So there will need to be some full-load battery capacity.
Re: DCs in space is just fucking delusional (Score:2)
It's pretty easy to power down system and temporarily reduce workload. Especially if you know in advance that your orbital is going to take you into a shadow. Could get away with a battery system capable of a fraction of the load.
But I predict that one lunacy follows another and a space data center will include a fusion reactor
Re: (Score:2)
Black body radiators work pretty well if they can only see the earth. PV will get more power per m2 than you can get rid off, but they are in the same order of magnitude.
Re: (Score:1)
Cooling? (Score:1)
I fail to see how cooling will work.
Cyrille
Re: Cooling? (Score:2)
Re: (Score:2)
Re:Cooling? (Score:5, Informative)
And, sure enough, not a year later did Zalman have a giant copper block heatsink with a heatpipe for cooling Socket A CPUs.
The issue is that this works on Earth because the heat is ultimately transferred to the environment. The heatsink does that by conducting heat to the air contacting it, which is then transferred to the environment through convection.
In orbit you don't have an environment with a huge mass of air that you can use for heat transfer, so you have to resort to radiation, which is much less efficient than conduction/convection. Radiating away significant amounts of heat would require very large surface areas which can quickly become unrealistic.
TL;DR: The current physics of space-based data centers don't make sense.
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Cool, thanks.
I imagine a lot can be done to increase surface area and effectiveness of radiators, ie: protruding fins. And on the sunny side reflect unwanted spectrum so then less is absorbed as heat.
Re: (Score:2)
Re: (Score:2)
Ya, that's where the reflective coating goes. Only allow the desired band, nominally optical, through to the solar cells.
Re: (Score:1)
They don't make financial sense, agreed. I think this and all the other AI bros power generating proposals are not intended to be realistic. It's just for appeasing Trump.
However, on the matter of heat removal, I will point out that radiating heat works really well when the radiators are shaded from the Sun and there is a large differential from the background temperature. Funnily, there is also a need to have large areas of sunshine collecting PV panels. Some of which could be arranged to act very well
Re: (Score:2)
The thing is that while the heat pipes can work in space and may have been used in satellites and then brought to earth, the issue is with the amount of thermal energy and having radiation as the only way to evict heat.
So while the mechanism for heat pipes started in space, the computers are *way* more wattage than the space based applications.
Re: (Score:3)
I fail to see how cooling will work.
Cyrille
Havn't you seen any movies. Stuff instantly freezes in space. Everyone knows that.
/s
Re: (Score:2)
If there's any moisture on it, that might actually happen. Evaporative cooling is very efficient, including in space. Not much good for a data center that would need a constant supply of water, though.
Re: (Score:2)
Solar PV sails would have film with metallised aluminium on the back, a bit of vacuum then more aluminium, then a blackbody HOPG film with liquid cooling channels (for its thermal conductivity at low weight). The silicon would be distributed along the sails to minimise the distance the coolant has to travel.
Obviously.
and the data bill for an link with no cap and no s (Score:2)
and the data bill for an link with no cap and no shaping will eat up what they save.
... generate 100,000 gigawatts a year. (Score:1)
Sounds like Musk doesn't know the difference between a Watt and a Watt-hour.
Re: ... generate 100,000 gigawatts a year. (Score:1)
Re: (Score:2)
Re: ... generate 100,000 gigawatts a year. (Score:1)
Re: (Score:2)
Cooling is easier in a vaccum?? (Score:5, Insightful)
" Demands for cooling could also be cut because of the vacuum of space."
How does that work? Vaccum makes cooling harder, not easier, doesn't it?
Re: (Score:2)
Yeah, when I hit that sentence my brain went "how's that again?"
I think they've seen too many sci-fi TV shows where someone gets sucked out an airlock and is instantly frozen solid.
Re: (Score:2)
Yeah! And the frozen person shatters into smithereens when hit by a protagonist with a hammer.
Re: (Score:1)
Total vacuum makes convective and conductive heat transfer impossible. But it makes radiative easier. And when the background is -100 C (in shadow), radiative becomes highly effective.
Re: (Score:2)
You're right about radiative cooling - I didn't consider that radiative cooling would be that much easier.
I asked claude.ai a bunch of questions.below According to it, purely radiative cooling of a device in space (40-60C) is roughly as effective as very crude conductive cooling of a similar device on earth (internal temp 49 C).
_---------_
Consider a 1 m radius spherical object in LEO. One direction, bathed in direct sunlight is coated with solar panels. The electricity generated runs computers internally
Re: (Score:2)
Vaccum makes cooling harder, not easier, doesn't it?
Indeed it does. Sounds like some CEOs need to go back to space camp.
In space, no one can hear your servers scream
Re: Cooling is easier in a vaccum?? (Score:1)
Perhaps they get their info from those Hollywood movie films, where space pirates freeze solid when they walk the space plank.
Either that, or these turbonerds have just never used a vacuum flask to keep their hot drinks hot.
Re: (Score:3)
Vaccum makes cooling harder, not easier, doesn't it?
Shhhh managers and tech bros as talking. Don't disturb them with your "physics".
Re: (Score:2)
Just let them build them on earth. (Score:2)
The only reason anyone is talking about this is we’ve made it too hard to build new power generating capacity on earth.
Re: (Score:2)
Whether any are realistic or not, none of the proposed options from these guys are short term fixes. Certainly not this one. They're dressing up alternatives to quick and easy renewables so as to appease Trump's aversion to renewables.
human migration beyond our home world (Score:2)
Re: (Score:2)
Speaking of pollution, sending up zllions of satellites with maybe several missions for one is sure to pollute the Earth's atmosphere. The scurvy little bastards that will do it to us won't care as long as it turn them into billionaires.
Can we send them with it? (Score:2)
And maybe make sure those things are in a close solar orbit to make sure they stay warm. I hear space is really cold.
Cooling DCs in space is near impossible (Score:2)
It might be freezing cold in space, the lack of a medium to transport heat spoils all these pipe dreams.
Re: (Score:1)
It's actually easier since it doesn't need a single moving part - heat pipes to quickly transfer to radiators where, when in shade, it radiates really well.
Production Tasks (Score:2)
Capacity Upgrades
Repair down servers
push the damm button
find the failing cable
replace the storage device
backup backup and send backup offsite
replace entire server rack with new new new servers
train new people and replace exiting people
free power may not make enough of a difference
Latency will suck (Score:2)
If they want solar 24/7, the Lagrange points are the only way, but then the latency will suck, 10 seconds round-trip.
I guess we'll get 'Thinking.....' a LOT.
Re: (Score:2)
A polar orbit, at any distance, can have continuous sunshine.
Re: (Score:3)
The polar orbit can have continuous sunshine, but three months later it will be in the shade half the time. Three months after that it will be back in the sun full time.
To keep it in the sun continuously you will have to adjust the orbit regularly and that will take a fair bit of propellant.
Re: (Score:2)
Apparently sun-synchronous orbits without using propellant are possible:
https://en.wikipedia.org/wiki/... [wikipedia.org]
These orbits take advantage of the Earth's oblate shape to move about a degree each day.
Re: (Score:2)
LOL, I was being loose with "polar". It is the general orbit rather than exact. You obviously understand this though, otherwise you wouldn't have nit-picked.
Shady (Score:1)
What are actual applications and requirements (Score:2)
The talk about space being cold, having more light, etc. is pretty much nonsense. More like it has less regulations and bigger budgets. However, I could imagine some kind of low latency compute being needed for applications like:
- Luna: Automated robotic exploration and construction drones wanting 2 second lag, different countries are working on it
- Asteroid mining robots: Onboard compute, or lower latency needed on arrival / when opposite Sun
- Telescope on other side of the sun: Maybe not needed
- Solar sys
Re: (Score:2)
They're not serious proposals. It's just to appease Trump's aversion to dropping down lots more quick and cheap renewables.
Cool much harder in a vacuum. (Score:5, Informative)
Re: (Score:2)
In shade the external temperature is at least below -100 C. That's more than enough to radiate a lot of heat easily. Not only that but it can done with no moving parts, and no power required to transfer it to the radiators.
Re: (Score:2)
The effectiveness of radiative heat transfer depends on the 4th power of the source temperature. Go ahead and look up the formula and the math.
Re: (Score:2)
Which works fine with a decent differential. Funnily, there is a decent differential.
Pretentious? Moi? (Score:2)
I lost interest in everything after that with a surprising quickness.
Why are tech moguls such idiots? (Score:1)
Why the hell would you want to put a datacenter in space where communications latency would make them slow and useless?
Re: (Score:2)
And, a repair mission is going to be a costly thing and takes time to plan and you'd have to launch the repair mission with all conceivable parts onboard, and have to wait for weather to be 'right' for liftoff... and I'll hazard a guess, they'll fix the problematic server, check the rest and they'll be fine, they'll take off and head home, make it about halfway and another server needs fixing.
Redundancy is an option, as long as doubling the size of the craft can still be launched. Are the redundant servers
Easy cooling in space? (Score:3)
Someone needs a physics lesson.
guys, it's already there, on Altair IV (Score:1)
AI NEEDS? (Score:2)
AI's needs are eventually going to grow so great that we need to move to outer space
AI doesn't "need" anything. It's people - madmen tech tycoons - who want to move AI to space so it can continue to grow with unlimited energy until it overtakes us all and kills us off.
That they are interested in this for continuing their fanatical growth of AI, rather than to, say, end human suffering from extreme poverty or solving the climate crisis, speaks to their lack of connection with humanity and the madness of the AI craze.
Re: (Score:2)
Yep... or, on the flipside of the coin, if they gave the LLM-AI a task like perfecting fusion or something instead of just being a chatbot "friend", maybe (just maybe) that'd be a step in the right direction for the climate.
There is a crucial word being overlooked here (Score:2)
I've been studying the possibility of space based data centers for some time. The operative word is 'could' as in, "But they could in the future". The math just doesn't work at the moment. Even with Starship online, the support equipment at the scale needed is ten times harder to service than here on Earth. Unless there is some major breakthrough in cooling technology, this is not going to work.
Of course that wont stop VCs from throwing tons of money at companies claiming to have the answer. Sigh.
Space is h
No regulations? (Score:2)
Plus, there aren't those pesky regulations that executives like to complain about
Because of course there's no such thing as space law [wikipedia.org] that governs what you do in space and allows countries to regulate it.
dream on (Score:2)
Better show me pigs flying first.
Pure Distraction (Score:2)
No, They Really Don't (Score:2)
These plans are as real as the Hyperloop. It is even easier to show that this is engineering and economic nonsense than it was with Hyperloop. This is a PR war -- billionaires looking to look like Tony Stark by selling snake oil (but not snake oil that will ever actually be produced).