I guess I'd assume that the premise driving this would be that there will eventually be enough business in space that it's necessary for space-centric use, and that terrestrial use is just a fringe benefit or loss leader or something.
But oddly this doesn't seem to be how the concept is typically framed.
My second level curiosity is how much cheaper/competitive it'd be if we had space elevators.
I suspect that this orbital data centers isn't entirely about dollars (No doubt dollars are important).
I suspect it is about the regulatory environment. The regulatory environment on data centers is moving quickly. Data centers used to be considered a small portion of the economy and thus benign and not worth extorting/controlling. This seems to be changing, rapidly.
Given that data centers only exchange information with their consumers they are a natural candidate for using orbit as a way to escape regulators.
Further, people are likely betting that regulators will take considerable time to adjust since space is multinational.
> ...we should be actively goading more billionaires into spending on irrational, high-variance projects that might actually advance civilization. I feel genuine secondhand embarrassment watching people torch their fortunes on yachts and status cosplay. No one cares about your Loro Piana.
I 100% agree with this. There are ~2,600 billionaires in the world and we should encourage all of them to spend their money. Even buying a superyacht is a benefit to the economy. But the best billionaires, like Bill Gates and Elon Musk, are actually trying to advance the tech tree.
We are honestly lucky that Musk is wired funny. Any normal human being would retire and hang out on the beach with supermodels after all the abuse he has taken. But he takes it all as a personal challenge and doubles down. That is both his worst quality and his best.
This is very well done! I love including all the sliders so that we can play with the (reasonable looking) assumptions the author has made. Like the author, I share their surprise that the result did not come out even more in favor of terrestrial GPUs.
Will these space-based data centers run on rad-hard silicon (which is dog slow compared to anything on Earth) or just silently accept wrong results, hardware lockups and permanent failure due to the harsh space environment? Will they cool that hardware with special über-expensive high-temperature Peltiers that heat the radiators up to visible incandescence so that the heat can be shed with any efficiency? There's zillions of those issues. The whole idea is just bonkers.
When Starcloud put together that whitepaper the first thing I looked at was the launch costs[1]. It references a $5M cost to launch, which right away made absolutely no sense to me. Just a cursory search shows launch costs are around $50M per launch, if not more.
It's great that this site drills down even further to demonstrate that there is absolutely no point at which the launch costs ever make this economical or viable, so I really don't understand what people are doing.
Especially because this site was harping for years about the cost of launches and putting things in to orbit, the whole reason why SpaceX got started and has grown as it has. As soon as that became an inconvenient number, we now just make things up (Just pretend that launch costs are 10% of what they actually are to get people to invest?).
If I were to guess, my first bet would be grand PR damage control for all the Mexicans, Irish, and what have you as in “don’t worry, we’ll soon be in space and out of your backyard” (no, they won’t).
I realize terrestrial data centers have environmental risks, but are the risks greater for an orbital data center? I would think space debris, solar flares, or a bad actor satellite with a laser could do a lot of damage. Good luck repairing the orbital data center.
What really worries me is that I keep hearing "cooling is cheap and easy in space!" in a lot of these conversations, and it couldn't be farther from the truth. Cooling is _really_ hard and can't use efficient (i.e. advection-based air or water cooling) approaches and are limited to dramatically less efficient radiative cooling. It doesn't matter that space is cold because cooling is damned hard in a vacuum.
The article makes this point, but it's relatively far in and I felt it was worth making again.
With that said, my employer now appears to be in this business, so I guess if there's money there, we can build the satellites. (Note: opinions my own) I just don't see how it makes sense from a practical technical perspective.
I love the sliders, but note that the numbers on this site literally came from ChatGPT, so there is plenty of room for disagreement.
Seems like according to this analysis it all hinges on launch cost and satellite cost. This site's default for Starship launch cost is $500/kg, but SpaceX is targeting much lower than that, more like $100/kg and eventually optimistically $10/kg (the slider doesn't even go that low). At $100/kg (and assuming all the other assumptions made on the site hold) then you break even on cost vs. terrestrial if you can make the satellites for $7/watt (excluding GPUs, as the whole analysis does).
I’d love to see these variables fitted to learning curves. That would give you a forecast for when, if learning continues as predicted, the economics could be competitive. (If it doesn’t, you need a new paradigm first.)
I'm not really interested in the problems that can come with orbital compute. We've seen them listed ad nauseam.
Have we seen any benefits to orbital computing by launching a cluster of raspberry pis to LEO? Surely this isn't an impossible task to test out on a smaller scale?
"Datacenters in space" make for a catchy narrative and an interesting demo, but the math simply doesn't work.
When considering factors like launch cost, maintenance complexity, and the cost of high-bandwidth communications (latency included), there is no realistic set of economic and engineering assumptions under which orbiting datacenters become cost-competitive with simply building conventional nuclear-powered (or renewable energy-powered) datacenters on the ground.
In fact we're off by 50-100x. Dramatic launch cost reductions still won't make it work. And of course if you invest a lot in specific lines of tech to make it work you then have to consider that the same can also be invested in better ground-based nuclear, bringing the cost of power down for everyone.
I have no idea what this Grok assisted article is trying to say. But the data-center-in-space hype is irrational. It hand-waves cooling and bit flip errors. It does not explain why we need chat bots in space (we don't need them on earth either).
It is a nice talking point for the U.S. Saudi Investment Forum. The Saudis apparently buy anything:
Did a similar back-of-the-napkin and got 5x $ / MW of orbital vs. terrestrial. This article's analysis is ~3.4x.
I do wonder, at what factor of orbital to terrestrial cost factor it becomes worthwhile.
The greater the terrestrial lead time, red tape, permitting, regulations on Earth, the higher the orbital-to-terrestrial factor that's acceptable.
A lights-out automated production line pumping out GPU satellites into a daily Starship launch feels "cleaner" from an end-to-end automation perspective vs years long land acquisition, planning and environment approvals, construction.
More expensive, for sure, but feels way more copy-paste the factory, "linearly scalable" than physical construction.
AFAIK compute heavy datacenter in space don't work. But if you already have a vast fleet of laser connected LEO satellites throwing some efficient SSDs into them can make a lot of sense. A large portion of the traffic is fairly static, e.g. video based content or even model weights. Caching that will save you ground to space side of the transmission. This will let you put more user beams on satellites and use less ground stations.
The short of it is that cooling is likely the biggest problem, given you will need to pump the heat to the backside and radiate it away, and the amount of mass you will need to dedicate to cooling works against deployments and increases the cost per unit significantly. Not to mention, the idea of these huge deployments runs into potential space debris issues.
Whenever one of these ventures actually manages to launch a proof of concept, I think we'll be able to quickly discern if there is actually a near-future here.
> This is all to say that the current discourse is increasingly bothering me due to the lack of rigor; people are using back-of-the-envelope math, doing a terrible job of it, and only confirming whatever conclusion they already want. Calculating radiation and the cost of goods is not difficult. Run the numbers.
> References: Gemini, Gemini, ChatGPT, ChatGPT, Gemini, ChatGPT, Gemini, ChatGPT, Grok, Gemini (There are sub-references from these services in the GitHub.)
I think, if you're going to make statements like this - especially from a position of expertise, you should be personally verifying the numbers and citing their sources directly. What good is asking the reader to trust an AI on your behalf? They should trust you.
(To be clear, I suspect the conclusions drawn are still correct.)
I sometimes wonder if there are people out there who just read too much Neuromancer, and they think they can construct their own Tessier-Ashpool orbital dynasty.
I suppose there are several other Oligarchs In Space stories and movies since then, but if the point of the space station is to host AI, that narrows it down a bit.
Or perhaps it's performative, designed to spook gullible politicians into changing laws to "keep" businesses that were never actually going to go somewhere else anyway.
this does not makes sense from dollars pov to me, I ran a back of napkin session with claude & gemini on this and the short of it is you need a magical weightless radiator for cooling and even then it wont work because the launch costs need to be sub $100 before this can be feasible. this does not even factors the 5y amortization and LEO orbit drag correction.
It then occurred to me that they (all major AI companies) know all of these facts but still pushing for it so there must be another reason. Then I recalled the offhand statement from the openAI lady about govt backstop for infra, which was strongly opposed by public and AI czar. this might be be a backdoor way of injecting that backstop capital in terms of subsidies now for results in 5 years or so. and needless to say after pilot programs those will fail spectacularly.
56 comments
[ 4.4 ms ] story [ 56.6 ms ] threadBut oddly this doesn't seem to be how the concept is typically framed.
My second level curiosity is how much cheaper/competitive it'd be if we had space elevators.
I suspect it is about the regulatory environment. The regulatory environment on data centers is moving quickly. Data centers used to be considered a small portion of the economy and thus benign and not worth extorting/controlling. This seems to be changing, rapidly.
Given that data centers only exchange information with their consumers they are a natural candidate for using orbit as a way to escape regulators.
Further, people are likely betting that regulators will take considerable time to adjust since space is multinational.
I 100% agree with this. There are ~2,600 billionaires in the world and we should encourage all of them to spend their money. Even buying a superyacht is a benefit to the economy. But the best billionaires, like Bill Gates and Elon Musk, are actually trying to advance the tech tree.
We are honestly lucky that Musk is wired funny. Any normal human being would retire and hang out on the beach with supermodels after all the abuse he has taken. But he takes it all as a personal challenge and doubles down. That is both his worst quality and his best.
It's great that this site drills down even further to demonstrate that there is absolutely no point at which the launch costs ever make this economical or viable, so I really don't understand what people are doing.
Especially because this site was harping for years about the cost of launches and putting things in to orbit, the whole reason why SpaceX got started and has grown as it has. As soon as that became an inconvenient number, we now just make things up (Just pretend that launch costs are 10% of what they actually are to get people to invest?).
[1]: https://starcloudinc.github.io/wp.pdf
https://taranis.ie/datacenters-in-space-are-a-terrible-horri...
—
If I were to guess, my first bet would be grand PR damage control for all the Mexicans, Irish, and what have you as in “don’t worry, we’ll soon be in space and out of your backyard” (no, they won’t).
https://www.nytimes.com/2025/10/20/technology/ai-data-center...
The article makes this point, but it's relatively far in and I felt it was worth making again.
With that said, my employer now appears to be in this business, so I guess if there's money there, we can build the satellites. (Note: opinions my own) I just don't see how it makes sense from a practical technical perspective.
Space is a much harder place to run datacenters.
Putting data centers under water makes way way more sense than into space.
Seems like according to this analysis it all hinges on launch cost and satellite cost. This site's default for Starship launch cost is $500/kg, but SpaceX is targeting much lower than that, more like $100/kg and eventually optimistically $10/kg (the slider doesn't even go that low). At $100/kg (and assuming all the other assumptions made on the site hold) then you break even on cost vs. terrestrial if you can make the satellites for $7/watt (excluding GPUs, as the whole analysis does).
Have we seen any benefits to orbital computing by launching a cluster of raspberry pis to LEO? Surely this isn't an impossible task to test out on a smaller scale?
"Datacenters in space" make for a catchy narrative and an interesting demo, but the math simply doesn't work.
When considering factors like launch cost, maintenance complexity, and the cost of high-bandwidth communications (latency included), there is no realistic set of economic and engineering assumptions under which orbiting datacenters become cost-competitive with simply building conventional nuclear-powered (or renewable energy-powered) datacenters on the ground.
In fact we're off by 50-100x. Dramatic launch cost reductions still won't make it work. And of course if you invest a lot in specific lines of tech to make it work you then have to consider that the same can also be invested in better ground-based nuclear, bringing the cost of power down for everyone.
It is a nice talking point for the U.S. Saudi Investment Forum. The Saudis apparently buy anything:
https://xcancel.com/elonmusk/status/2000603814249079165#m
I do wonder, at what factor of orbital to terrestrial cost factor it becomes worthwhile.
The greater the terrestrial lead time, red tape, permitting, regulations on Earth, the higher the orbital-to-terrestrial factor that's acceptable.
A lights-out automated production line pumping out GPU satellites into a daily Starship launch feels "cleaner" from an end-to-end automation perspective vs years long land acquisition, planning and environment approvals, construction.
More expensive, for sure, but feels way more copy-paste the factory, "linearly scalable" than physical construction.
https://www.youtube.com/watch?v=d-YcVLq98Ew
The short of it is that cooling is likely the biggest problem, given you will need to pump the heat to the backside and radiate it away, and the amount of mass you will need to dedicate to cooling works against deployments and increases the cost per unit significantly. Not to mention, the idea of these huge deployments runs into potential space debris issues.
Whenever one of these ventures actually manages to launch a proof of concept, I think we'll be able to quickly discern if there is actually a near-future here.
> References: Gemini, Gemini, ChatGPT, ChatGPT, Gemini, ChatGPT, Gemini, ChatGPT, Grok, Gemini (There are sub-references from these services in the GitHub.)
I think, if you're going to make statements like this - especially from a position of expertise, you should be personally verifying the numbers and citing their sources directly. What good is asking the reader to trust an AI on your behalf? They should trust you.
(To be clear, I suspect the conclusions drawn are still correct.)
I suppose there are several other Oligarchs In Space stories and movies since then, but if the point of the space station is to host AI, that narrows it down a bit.
Or perhaps it's performative, designed to spook gullible politicians into changing laws to "keep" businesses that were never actually going to go somewhere else anyway.
It then occurred to me that they (all major AI companies) know all of these facts but still pushing for it so there must be another reason. Then I recalled the offhand statement from the openAI lady about govt backstop for infra, which was strongly opposed by public and AI czar. this might be be a backdoor way of injecting that backstop capital in terms of subsidies now for results in 5 years or so. and needless to say after pilot programs those will fail spectacularly.