Data Centers in Space

Pixelbrick said:
Once the data centers are up there, we should ban golf inside the orbit of Mars too. Would solve many problems and make for great telly.
Click to expand...
Well, I can certainly support banning golf, at least for executive management. One of the dumbest tech business deals that I know of in our day and age came about, quite literally, as snap a decision made by two CEOs playing golf with each other.
 
While I'm sure they chose the name 'Lumen' becaus it's the name for a unit of light intensity , are they also aware that 'lumen' is also the name for the cavity of a body part and / or organ . . .
 
Nik said:
Uh, why must you immerse a data centre module in *salt* water ??
Click to expand...
Generally the proposals to do so are based on putting them close to major cities to minimise latency. As the cities in question tend to be on the coast, that means salt water.
 
A similar concept:

An out-of-this-world idea: Placing data centers in space could pave the way for sustainable computing with unlimited solar energy and free cooling, say scientists from Nanyang Technological University, Singapore (NTU Singapore). Their study shows how satellites equipped with advanced processors could harness unlimited solar energy and the cold vacuum of space to power a greener digital future. https://www.nature.com/articles/s41928-025-01476-1

From:
https://techxplore.com/news/2025-10-scientists-carbon-neutral-centers-space.htmlhttps://techxplore.com/news/2025-11-eyes-space-power-hungry-centers.htmlhttps://forum.nasaspaceflight.com/index.php?topic=63819.0https://services.google.com/fh/files/misc/suncatcher_paper.pdfhttps://arstechnica.com/google/2025...googles-plan-to-put-ai-data-centers-in-space/https://www.nextbigfuture.com/2025/...cher-to-put-tpus-for-ai-in-space-in-2027.htmlhttps://www.nextbigfuture.com/2025/...sla-future-chips-spacex-starship-are-key.html
Perhaps VAST could play a role
https://arstechnica.com/space/2025/...-station-startup-now-has-a-foothold-in-space/
More
https://twitter.com/x/status/1986617644246847664
View: https://x.com/PhilipJohnst0n/status/1986617644246847664

https://starcloudinc.github.io/wp.pdf
China responds
https://spacenews.com/chinas-starde...omputing-and-space-domain-awareness-services/
Discussion
https://www.nextbigfuture.com/2025/...level-ai-in-orbit-moon-industrialization.html
On space sustainability
https://www.thespacereview.com/article/5090/1
 
Last edited:
robunos said:
While I'm sure they chose the name 'Lumen' becaus it's the name for a unit of light intensity , are they also aware that 'lumen' is also the name for the cavity of a body part and / or organ . . .
Click to expand...
It got renamed Starcloud a while ago.
 
Seems they aren’t the only ones thinking this.

As artificial intelligence drives the need for vastly more computing storage and processing power, interest in space-based data centers has spiked.

Although several startup companies, such as Starcloud, have begun to address this problem, the idea has also attracted the interest of tech barons. In May, it emerged that former Google chief executive Eric Schmidt acquired Relativity Space due to his interest in space-based data centers. Then, earlier this month, Amazon founder Jeff Bezos predicted that gigawatt-scale data centers will be built in space within the next 10 to 20 years.

Now, Elon Musk, whose SpaceX owns and operates significantly more space-based infrastructure than any other company or country in the world, has also expressed interest in the technology.
Click to expand...
https://arstechnica.com/space/2025/10/elon-musk-on-data-centers-in-orbit-spacex-will-be-doing-this/
 
Justo Miranda said:
In my opinion this thing does not seem very safe to preserve or exchange data, regardless of temperature changes it would need a very heavy radiation shielding... What is the advantage?
Click to expand...
Avoiding terrestrial power and water consumption as well as the associated heat generation.
 
iverson said:
I have read that the works of L. Ron Hubbard have been inscribed on stainless-steel LP records and buried in bomb-proof bunkers (along with the requisite record players) for the edification of future evolved humans returning from the stars to the California coast.
Click to expand...
Wonderful - a protected repository of disinformation and mind pollution...
 
martinbayer said:
I've believe I've heard of a thing called human spaceflight...
Click to expand...
Ohhh. Self-loading cargo, you mean. Let's see you expense the travel expenses for a physical reboot. On the other hand, the Runzheimers mileage reimbursement rate for a POV is at 70 cents a mile...
 
aim9xray said:
Ohhh. Self-loading cargo, you mean. Let's see you expense the travel expenses for a physical reboot. On the other hand, the Runzheimers mileage reimbursement rate for a POV is at 70 cents a mile...
Click to expand...
Expensive? Quite possibly, though large orbital infrastructure projects like that could actually *help* to bring launch costs down. *Categorically impossible* (as in "that would rule out visiting")? Absolutely NOT.
 
>>An out-of-this-world idea: Placing data centers in space could pave the way for sustainable computing with unlimited solar energy and free cooling, say scientists from Nanyang Technological University, Singapore (NTU Singapore). Their study shows how satellites equipped with advanced processors could harness unlimited solar energy and the cold vacuum of space to power a greener digital future. https://www.nature.com/articles/s41928-025-01476-1<<

But how "free" is the cooling? As far as I know, space is neither hot nor cold. Objects that are exposed to sunlight heat up quite a lot in space. Objects in shadow get very, very cold.

But, to cool an object, you still have to transfer the heat from the sunlit parts to the shadowed parts. A vacuum (or near vacuum like space) contains too little matter to transfer heat by the familiar processes of conduction and convection. Excess heat has to radiate away as IR light. So you typically have to build some sort of external radiator that is reliably on the cold side of the object.

That is just to cool the object--the spacecraft itself. The computers and storage arrays are a separate problem. IT equipment requires large cooling plants and lots of airspace around the equipment and under the floors. Air has generally been the approved cooling medium for all but special situations. Liquid cooling is more expensive, presents electrical risks (shorts due to leaks or condensation on piping), is less reliable (leaks, corrosion), needs more maintenance, etc.

So how do we get the heat out of the space data center and out into space via the radiator? Even small data centers generate enormous quantities of heat (as those of us who have worked in them can testify). We can't just have the data center components in a vacuum, radiating heat willy nilly. Some parts will get too cold. Others will get too hot. We need a heating and cooling medium to distribute needed heat evenly and carry off the excess--just like on Earth. The extra volume required for a circulating air- or other gas-based cooling (and heating) system will probably be an unaffordable luxury on a space data center: equipment will be densely packed, greatly exacerbating the heat problem. In space, moreover, air/gas will present even more problems with leaks, maintenenace, and reliability that liquid cooling does on earth. So I think you'll need liquid cooling in a space application, which will have to be kept from freezing, boiling, evaporating/sublimating, and leaking.

So, unless I am missing something, a space-based data center presents creates temperature regulation problems over and above those encountered down here on Earth--and more rather than less expensive.

The logical fix for the runaway environmental impacts of huge numbers of huge data centers is simple: don't build them. The demand for these things is not driven by any obvious practical need. They are being built for crypto, which is just silly and useless, and so-called large language-model AI, which, at present, is much the same.

When your laptop makes sudden, really huge demands for processing power, memory, and storage, one of your programs has died and left a zombie process running away with your machine. Crypto and current AI are the macro-scale equivalents: zombie management fads that need to be killed before they consume the planet.
 
BlueAbyssal said:
It got renamed Starcloud a while ago.
Click to expand...
I hadn't seen that . . . now I'm expecting that they will get a visit from SpaceX's lawyers , SpaceX having copyrighted every combination of pairs of words with 'Space' at the front (like Apple did with 'i') . . . /s

cheers,
Robin.
 
An idea I have seen:

One obvious evolution path is to incorporate some "data center" capabilities into a large GEO comsat.
https://up-ship.com/blog/?p=49953https://forum.nasaspaceflight.com/index.php?topic=63177.msg2732320#msg2732320
The idea of Orbital Antenna Farms has returned..."dead thoughts live new and oddly bodied." ("The Festival")

Now, have them double as powersats, data centers, sunshades.
One structure--many uses:
https://www.nextbigfuture.com/2025/...-megawatts-of-solar-in-space-exist-today.html

As per NSF, we are looking at a: "20,000 ton cluster about 1km in diameter, with each element of the cluster around 5-10t."
https://forum.nasaspaceflight.com/index.php?topic=63812.100
When I advocated large structures in space--I was roundly called an idiot. Strange what a few hundred billion dollars does to public opinion:
https://forum.nasaspaceflight.com/index.php?topic=63804.0
On orbit assembly
https://www.sciencedirect.com/science/article/pii/S0094576525007544https://www.sciencedirect.com/science/article/abs/pii/S009457652500757X
Small satellites have problems with flashover?
https://arc.aiaa.org/doi/10.2514/1.A36161
Electric propulsion refueling
https://arc.aiaa.org/doi/10.2514/1.A36441https://arc.aiaa.org/doi/10.2514/1.A36386https://arc.aiaa.org/doi/10.2514/1.A36403https://arc.aiaa.org/doi/10.2514/1.A36077
Now to get Buffett to fund Shuttle 2 :)
https://www.google.com/imgres?q=Lun...d=2ahUKEwjznpGaj-GQAxWhF1kFHctLFEAQM3oECB4QAAhttps://finance.yahoo.com/news/warren-buffett-keeps-selling-stocks-144553267.html
He's stepping down like Bezos did after all.....Come on Warren--jocks have Cuban!

Space advocates need a patron who ISN'T a tech-bro or something out of Death On The Nile....
 
Last edited:
iverson said:
>>An out-of-this-world idea: Placing data centers in space could pave the way for sustainable computing with unlimited solar energy and free cooling, say scientists from Nanyang Technological University, Singapore (NTU Singapore). Their study shows how satellites equipped with advanced processors could harness unlimited solar energy and the cold vacuum of space to power a greener digital future. https://www.nature.com/articles/s41928-025-01476-1<<

But how "free" is the cooling? As far as I know, space is neither hot nor cold. Objects that are exposed to sunlight heat up quite a lot in space. Objects in shadow get very, very cold.

But, to cool an object, you still have to transfer the heat from the sunlit parts to the shadowed parts. A vacuum (or near vacuum like space) contains too little matter to transfer heat by the familiar processes of conduction and convection. Excess heat has to radiate away as IR light. So you typically have to build some sort of external radiator that is reliably on the cold side of the object.

That is just to cool the object--the spacecraft itself. The computers and storage arrays are a separate problem. IT equipment requires large cooling plants and lots of airspace around the equipment and under the floors. Air has generally been the approved cooling medium for all but special situations. Liquid cooling is more expensive, presents electrical risks (shorts due to leaks or condensation on piping), is less reliable (leaks, corrosion), needs more maintenance, etc.

So how do we get the heat out of the space data center and out into space via the radiator? Even small data centers generate enormous quantities of heat (as those of us who have worked in them can testify). We can't just have the data center components in a vacuum, radiating heat willy nilly. Some parts will get too cold. Others will get too hot. We need a heating and cooling medium to distribute needed heat evenly and carry off the excess--just like on Earth. The extra volume required for a circulating air- or other gas-based cooling (and heating) system will probably be an unaffordable luxury on a space data center: equipment will be densely packed, greatly exacerbating the heat problem. In space, moreover, air/gas will present even more problems with leaks, maintenenace, and reliability that liquid cooling does on earth. So I think you'll need liquid cooling in a space application, which will have to be kept from freezing, boiling, evaporating/sublimating, and leaking.

So, unless I am missing something, a space-based data center presents creates temperature regulation problems over and above those encountered down here on Earth--and more rather than less expensive.

The logical fix for the runaway environmental impacts of huge numbers of huge data centers is simple: don't build them. The demand for these things is not driven by any obvious practical need. They are being built for crypto, which is just silly and useless, and so-called large language-model AI, which, at present, is much the same.

When your laptop makes sudden, really huge demands for processing power, memory, and storage, one of your programs has died and left a zombie process running away with your machine. Crypto and current AI are the macro-scale equivalents: zombie management fads that need to be killed before they consume the planet.
Click to expand...
Basic rules of spacecraft design - solar radiation absorbing power generating photovoltaic panels are as much as possible oriented to the maximum perpendicular area exposed to our local star, whereas radiation heat rejecting panels aimed at deep space are oriented exactly 90 geometric degrees from that orientation with respect to minimizing solar insolation.
 
Last edited:
martinbayer said:
Basic rules of spacecraft design - solar radiation absorbing power generating photovoltaic panels are as much as possible oriented to the maximum perpendicular area exposed to our local star, whereas radiation heat rejecting panels aimed at deep space are oriented exactly 90 geometric degrees from that orientation with respect to minimizing solar insolation.
Click to expand...

The perfect solution would be T-shaped satellite, where solar panels oriented perpendiculary to sun, provide shade for radiators oriented at 90 degrees.
 
robunos said:
I hadn't seen that . . . now I'm expecting that they will get a visit from SpaceX's lawyers , SpaceX having copyrighted every combination of pairs of words with 'Space' at the front (like Apple did with 'i') . . . /s

cheers,
Robin.
Click to expand...
I must be getting old , I meant to type 'Star' rather than 'Space' ( although they've probably done it with 'Space' as well . . . ) .
 
Michel Van said:
The company Lumen Orbit want to put Data servers,
to use the abundant solar energy, cooling, and the ability to freely scale up in space.
launch of Prototype hardware is set for may 2025 launch of first micro data center in 2026
once Starship goes into commercial service they build the 5GW Data server in Low orbit
Click to expand...
Have they considered the cooling problem?


aim9xray said:
Ohhh. Self-loading cargo, you mean. Let's see you expense the travel expenses for a physical reboot. On the other hand, the Runzheimers mileage reimbursement rate for a POV is at 70 cents a mile...
Click to expand...
Well, it's ~500 miles up and another 500 miles down, and then some 27,000 miles per orbit. 500 miles up is a 100min orbit.

Wait, no. It's a lot more than 500 miles up/down, as that's just vertical distance. Isn't orbital insertion burn at about halfway around the globe from the launch location? So a good 10,000 miles up and back. And however many orbits as to get you back to wherever you took off from. 5 days seems to be the easiest round number to get your landing at a reasonable time. ~72 orbits.

~1.38 million in travel reimbursement.



iverson said:
But, to cool an object, you still have to transfer the heat from the sunlit parts to the shadowed parts. A vacuum (or near vacuum like space) contains too little matter to transfer heat by the familiar processes of conduction and convection. Excess heat has to radiate away as IR light. So you typically have to build some sort of external radiator that is reliably on the cold side of the object.
Click to expand...
Not a problem, those big solar panels happen to make excellent sunshades.



iverson said:
So how do we get the heat out of the space data center and out into space via the radiator? Even small data centers generate enormous quantities of heat (as those of us who have worked in them can testify). We can't just have the data center components in a vacuum, radiating heat willy nilly. Some parts will get too cold. Others will get too hot. We need a heating and cooling medium to distribute needed heat evenly and carry off the excess--just like on Earth. The extra volume required for a circulating air- or other gas-based cooling (and heating) system will probably be an unaffordable luxury on a space data center: equipment will be densely packed, greatly exacerbating the heat problem. In space, moreover, air/gas will present even more problems with leaks, maintenenace, and reliability that liquid cooling does on earth. So I think you'll need liquid cooling in a space application, which will have to be kept from freezing, boiling, evaporating/sublimating, and leaking.
Click to expand...
Mineral oil seems to work really well as a liquid cooler unit. Have you seen the "aquarium PCs"?



martinbayer said:
Basic rules of spacecraft design - solar radiation absorbing power generating photovoltaic panels are as much as possible oriented to the maximum perpendicular area exposed to our local star, whereas radiation heat rejecting panels aimed at deep space are oriented exactly 90 geometric degrees from that orientation with respect to minimizing solar insolation.
Click to expand...
And conveniently, solar panels make nice sunshades for your radiators.



robunos said:
I must be getting old , I meant to type 'Star' rather than 'Space' ( although they've probably done it with 'Space' as well . . . ) .
Click to expand...
You need more caffeine.
 
TheKutKu said:
On the other hand, the destruction can be done by any motivated and skilled decently sized team no matter the country.
Click to expand...
Destruction of *ANYTHING* has, for better or worse, *NEVER* been a hard brick wall limiter on human activity...
 
Last edited:
Scott Kenny said:
Have they considered the cooling problem?
Click to expand...
They have to, otherwise it make no sense to build that thing

Scott Kenny said:
Wait, no. It's a lot more than 500 miles up/down, as that's just vertical distance.
Click to expand...
This data center in combination with Starlink or Kepler constellation would be powerful !
Data transfer by laser on orbital plane and then direct down to target receiver

however if Starlink install server on version 3 or 4 of satellite, the constellation it self become a data centre...
 
 
Last edited:
https://www.wsj.com/opinion/elon-musk-xai-spacex-solar-powered-data-centers-f78fe95d

The Political Battle for AI in Space​

Will regulatory hurdles give China the edge in orbital data centers?​


Elon Musk last week announced an ambitious goal of launching a million solar-powered AI data centers into orbit. It could be an innovative work-around to America’s permitting headaches and electric-power shortages that are slowing the AI buildout. But get this—some progressives want to make it harder to build in space.

“Space-based AI is obviously the only way to scale,” Mr. Musk wrote after announcing SpaceX’s merger with xAI. “It’s always sunny in space!” Mr. Musk’s mission to launch solar-powered AI data centers in space involves significant technical challenges, but it’s not a journey to Mars. Other companies including Google have the same ambitions. So do the Chinese.

Mr. Musk explained on a podcast Thursday that his plan would avoid the headaches involved in building data centers and electric plants to power them on earth. “It’s actually very difficult to build power plants. You don’t just need power plants, you need all of the electrical equipment. You need the electrical transformers to run the AI transformers,” he said.

There’s a shortage of all of the above on earth. Operating data centers in space would avoid an environmental thicket and require less water and energy for cooling. “It’s actually much cheaper to do in space,” he said. “My prediction is that it will be by far the cheapest place to put AI. It will be space in 36 months or less. Maybe 30 months.”

Mr. Musk has a history of making overly optimistic projections. Still, it’s notable that SpaceX has applied to the Federal Communications Commission for permission to deploy its new non-geostationary orbit satellite system. FCC Chair Brendan Carr immediately opened the application to public comment with the aim of getting the project off the ground soon.

Mr. Carr has been on a mission to fast-track approvals for satellites. The number of applications has increased three-fold over the past decade as more companies have sought to commercialize space, especially for telecommunications. Yet applications during the Biden Presidency could linger in the FCC queue for years.

The FCC bottleneck has slowed the deployment of internet satellites that could improve connectivity for rural areas and advance U.S. strategic interests abroad. Iranian protesters used Starlink to communicate when the government shut down the internet. The Ukrainian military has relied on some 50,000 Starlink terminals against Russia.

Texas Sen. Ted Cruz has teamed up with Vermont Democrat Peter Welch on a bill to expedite satellite approvals. Their bill would give the FCC a year and a half to approve or reject applications, after which they would be automatically approved. One goal is to future-proof Mr. Carr’s deregulation so a future Administration can’t revive old regulatory obstacles.

The obstructionists are striking back. At a Senate hearing last week, Washington Democrat Maria Cantwell objected because the bill would help Mr. Musk’s AI space ambition. That’s a bad reason to oppose it, since it would benefit all satellite companies. Democrats should love Mr. Musk’s plan because it would reduce data center CO2 emissions.

Ms. Cantwell also said the FCC should be able to take its merry time when reviewing satellite applications to ensure they won’t interfere with existing spectrum holders. A year and a half is plenty of time. Mr. Carr managed to cut the FCC application backlog in half during his first six months as Chair.

The real danger is political interference that hamstrings America’s space development and gives China an upper hand. Democrats and astronomical groups are also opposing Mr. Carr’s proposal to exempt satellites from environmental reviews. They say satellites could indirectly affect U.S. earthlings by, say, brightening the night-time sky.

But such effects are hypothetical. The Federal Aviation Administration separately evaluates the environmental impact of rocket launches in the U.S., which has in the past delayed satellite launches. Permitting difficulties are America’s economic Achilles’ heel. Let’s hope they don’t get in the way of U.S. space innovation.



Appeared in the February 10, 2026, print edition as 'The Political Battle for AI in Space'.
Click to expand...
 
 
Last edited:
On the whole data center subject and local opposition to their being built. If you oppose data centers you should never use social media, store data on the Cloud ect. You are contributing to the issue.
 
Kevin Renner said:
On the whole data center subject and local opposition to their being built. If you oppose data centers you should never use social media, store data on the Cloud ect. You are contributing to the issue.
Click to expand...
There's a big difference between being a hard NIMBY and opposing any being built, and wanting them built in strategic and purposeful locations that impacts the environment and people in the smallest ways. There are legitimate health risks associated with data centers that some towns are currently dealing with. We shouldn't be paving over prime farm land or housing for data centers either, and we need to consider the electrical needs and cooling before building one wherever you want to.
 
Nobody wants to live by a landfill—but you have to use them.

Space based servers might be the only way powersats get built.

Bass boats and RPGs can’t threaten those like Thermopylae East (Hormuz Strait).
 
publiusr said:
Nobody wants to live by a landfill—but you have to use them.

Space based servers might be the only way powersats get built.

Bass boats and RPGs can’t threaten those like Thermopylae East (Hormuz Strait).
Click to expand...
By its very nature a landfill needs to be by the garbage it takes, unless you want to truck in said garbage every day over large distances. Not very practical given the cost of trucking and that amount you have to move at the speed you can move it.

By its very nature a data center does not. A data center 50 miles away from a town but connected with fiber optics has no appreciable difference is speed or latency or ping to that town than one in the center. To the rest of the world the location of that data center to the town is inconsequential.
 
In-space light computing?
https://phys.org/news/2026-05-sunlight-powered-generation-photon-pairs.html
Using sunlight for SPDC presents clear challenges. Sunlight collected from the ground is inherently unstable, with continuous changes in intensity, angle, and position that interfere with the precise illumination and photon detection required for SPDC experiments. At the same time, sunlight offers a compelling advantage: it removes dependence on lasers and external power sources, opening possibilities for photon-pair generation in remote or extreme environments.

Researchers led by Wuhong Zhang and Lixiang Chen at Xiamen University have now demonstrated a practical solution. For their study published in Advanced Photonics, they developed an experimental system in which sunlight serves as the sole pump source for SPDC.

More
https://phys.org/news/2026-05-physicists-hybrid-particles-interact-strongly.html
Zhen's team has created a quasiparticle that "combines the speed of light with the strong interactions of matter." These quasiparticles, or exciton-polaritons, are made by coupling photons with electrons in an atomically thin semiconductor, allowing light to interact strongly enough for signal switching needed in computation.

The advance could be especially important for AI.

Now server farms in orbit just need sunlight, I hope
 

Similar threads

Back
Top Bottom