Space is a vacuum. i.e. The lack-of-a-thing that makes a thermos great at keeping your drink hot. A satellite is, if nothing else, a fantastic thermos. A data center in space would necessarily rely completely on cooling by radiation, unlike a terrestrial data center that can make use of convection and conduction. You can't just pipe heat out into the atmosphere or build a heat exchanger. You can't exchange heat with vacuum. You can only radiate heat into it.
Heat is going to limit the compute that can be done in a satellite data centre and radiative cooling solutions are going to massively increase weight. It makes far more sense to build data centers in the arctic.
Musk is up to something here. This could be another hyperloop (i.e. A distracting promise meant to sabotage competition). It could be a legal dodge. It could be a power grab. What it will not be is a useful source of computing power. Anyone who takes this venture seriously is probably going to be burned.
It probably increases Elon's share of the combined entity.
It delivers on a promise to investors that he will make money for them, even as the underlying businesses are lousy.
Also the same issue with radiative cooling pops up for space solar cells - they tend to run way hotter than on Earth and that lowers their efficiency relative to what you could get terrestrially.
> As conduction and convection to the environment are not available in space, this means the data center will require radiators capable of radiatively dissipating gigawatts of thermal load. To achieve this, Starcloud is developing a lightweight deployable radiator design with a very large area - by far the largest radiators deployed in space - radiating primarily towards deep space...
They claim they can radiate "633.08 W / m^2". At that rate, they're looking at square kilometers of radiators to dissipate gigawatts of thermal load, perhaps hectares of radiators.
They also claim that they can "dramatically increase" heat dissipation with heat pumps.
So, there you have it: "all you have to do" is deploy a few hectares of radiators in space, combined with heat pumps that can dissipate gigawatts of thermal load with no maintenance at all over a lifetime of decades.
This seems like the sort of "not technically impossible" problem that can attract a large amount of VC funding, as VCs buy lottery tickets that the problem can be solved.
While personally I think it's another AI cash grab and he just wants to find some more customers for spacex, other thing is "you can't copyright infringe in space" so it might be perfect place to load that terabytes of stolen copyrighted material to train data sets, if some country suddenly decides corporation stealing copyright content is not okay any more
It's perfectly possible to put small data centres in city centres and pipe the heat around town, they take up very very little space and if you're consuming the heat, you don't need the noisy cooling towers (Ok maybe a little in summer).
Similarly if you stick your datacentre right next to a big nuclear power plant, nobody is even going to notice let alone care.
That's wise.
However, TFA's purpose in assuming cooling (and other difficulties) have been worked out (even though they most definitely have not) was to talk about other things that make orbital datacenters in space economically dubious. As mentioned:
But even if we stipulate that radiation, cooling, latency, and launch costs are all solved, other fundamental issues still make orbital data centers, at least as SpaceX understands them, a complete fantasy. Three in particular come to mind:Sufficient hype funds more work for his rocket company.
The more work they have the faster they can develop the systems to get to Mars. His pet project.
I really think it's that simple.
Have you done a calculation yourself?
It's a way to get cheap capital to get cool tech. (Personal opinion.)
Like dark fibre in the 1990s, there will absolutely–someday–be a need for liquid-droplet radiators [1]. Nobody is funding it today. But if you stick a GPU on one end, maybe they will let you build a space station.
Those flasks don’t have any space age insulating material - mainly just a vacuum…
Technology from 1892…
What (literally) on earth makes you say this? The arctic has excellent cooling and extremely poor sun exposure. Where would the energy come from?
A satellite in sun-synchronous orbit would have approximately 3-5X more energy generation than a terrestrial solar panel in the arctic. Additionally anything terrestrial needs maintenance for e.g. clearing dust and snow off of the panels (a major concern in deserts which would otherwise seem to be ideal locations).
There are so many more considerations that go into terrestrial generation. This is not to deny the criticism of orbital panels, but rather to encourage a real and apolitical engineering discussion.
But in all seriousness, if there is a possibility of building industrial centers outside of the Earth’s atmosphere, it is surely not here yet. Lots of areas would need improvement.
But now looking back and accounting for the claims he made there's a pattern.
I saw this article:
https://www.wired.com/story/theres-a-very-simple-pattern-to-...
that said... he did jumpstart the EV industry. He has put up satellites every week for years. He is still a net benefit to all of us.
A satellite is quite unlike a thermos in the sense that it is carefully tuned to keep its temperature within a relatively narrow band around room temperature.[1] during all operational phases.
This is because, despite intended space usage, devices and parts are usually tested and qualified for temperature limits around room temperature.
[1] "Room temperature" is actually a technical term meaning 20°C (exceptions in some fields and industries confirm the rule).
1) The heat can be transported by a heat carrier conducting heat standing still.
2) The heat can be transported by a heat carrier in motion.
3) The heat can be transported by thermal radiation.
The first 2 require massive particles, the latter are spontaneous photons.
A thermos bottle does not simply work by eliminating the motile mass particles.
Lets consider room temperature as the outer thermos temperature and boiling hot water as the inner temperature, that is roughly 300 K and 400 K.
Thermal radiation is proportional to the fourth power of temperature and proportional to emissivity (which is between 0 and 1).
Lets pretend you are correct and thus thermally blackened glass (emissivity 1) inside the vacuum flask would be fine according to you. That would mean that the radiation from your tea to the room temperature side would be proportional to 400^4 while the thermal radiation from room temperature to the tea would be proportional to 300^4. Since (400/300) ^ 4 = 3.16 that means the heat transport from hot tea to room temperature is about 3 times higher.
If on the other hand the glass was aluminized before being pulled vacuum the heat transports are proportional to 0 * 400 K ^ 4 and 0 * 300 K ^ 4 . So the heat transport in either direction would be 0 and no net heat transport remains.
If you believe the shiny inside of your thermos flask is an aesthetic gimmick, think again.
You are making a non-comparison.
Imagine comparing a diesel engine car to an electric car, but first removing the electric motor. Does that make a fair comparison???
Then you get people paying much more money to use less-tightly-moderated space-based AI rather than heavily moderated AI.
But SpaceX has lots of real engineers who are very smart. I’m certain they ran the math on it. Which is more than you or I have done.
If they say it can be done, I’m inclined to believe them.
This is just a question. I have no expertise at all with this.
(Space doesn’t help in cooling GPUs in a satellite - space makes cooling worse)
Specifically: Starship makes no economic sense. There simply isn’t any pre-existing demand for the kind of heavy lift capacity and cadence that Starship is designed to deliver. Nor is there anyone who isn’t currently launching heavy payloads to LEO but the only thing holding them back is that they need weekly launches because their use case demands a whole lot of heavy stuff in space on a tight schedule and that’s an all-or-nothing thing for them.
So nobody else has a reason to buy 50 Starship launches per year. And the planned Starlink satellites are already mostly in orbit. So what do you do? Just sell Starship to xAI, the same way he fixed Cybertruck’s demand problem by selling heaps of them to SpaceX.
Thermos has 2 parts responsible for thermal isolation: vacuum and some coating that reflects heat radiation
Also in earth conditions, most of thermal radiation goes to surrounding air, heating it. This is not the case with vacuum, so thermal radiation will be even more efficient
I'm pretty his point is that while cooling is an impossibility, it is not the only one!
https://www.nasa.gov/smallsat-institute/sst-soa/thermal-cont...
But I really hope posts like this don't discourage whoever is investing in this. The problems are solvable, and someone is trying to solve them, that's all that matters. My only concern is the latency, but starlink seems to manage somehow.
Also, a matter of technicality (or so I've heard it said) is that the earth itself doesn't dissipate heat, it transforms or transfers entropy.
Lag for roundtrip: 35ms. But when satelite needs to pass through other satellites as has no ground coverage you add more lag and reduce bandwidth of the whole network.
The best part is jurisdiction safety. Very hard to get raided by govs.
https://research.33fg.com/analysis/debunking-the-cooling-con...
X failing and can't pay its debts? Welp, better give him a government bailout otherwise no more rockets for you!
At earth or above sea, we use cooling to maintain the temperature below 60 degrees, or 80 or 100 or something.
Shadow of space is -157 degrees, the cooling design will be different.
Unfortunately no. The arctic region is too cold and humid. You need way more energy to manage the cooling of a datacenter there than somewhere hotter.
That specific aspect is NOT true in space because there's nothing stopping thermal radiation.
Now you're correct that you can't remove heat by conduction or convection in space, but it's not that hard to radiate away energy in space. In fact rocket engine nozzle extensions of rocket upper stages depend on thermal radiation to avoid melting. They glow cherry red and emit a lot of energy.
By Stefan–Boltzmann law, thermal radiation goes up with temperature to the 4th power. If you use a coolant that lets your radiator glow you can conduct heat away very efficiently. This is generally problematic to do on Earth because of the danger of such a thing and also because such heat would cause significant chemical reactions of the radiator with our corrosive oxygen atmosphere.
Even without making them super hot, there's already significant energy density on SpaceX's satellites. They're at around 75 kW of energy generation that needs to be radiated away.
And on your final statement, hyperloop was not used as a "distraction" as he never even funded it. He had been talking about it for years and years until fanboys on twitter finally talked him into releasing that hastily put together white paper. The various hyperloop companies out there never had any investment from him.
I don’t remember the difference from my science classes, isn’t This the same thing essentially?
That.
Also, am I the only one to remember when SpaceX was supposed to pivot to transporting people from cities to cities, given how cheap and reusable and sure BFF/Starship was going to be ?
Or how we were all going to earn money by pooling our full self driving cars in a network of robo taxis ?
In all seriousness, what is the number of "unrealized sci-fi pipe dreams" that is acceptable from the owner a company ? Or, to be fair, what is the acceptable ratio of "pipe dreams" / "actually impressive stuff actually delivered (reusable rockets, starlink, decent EVs, etc...)" ?
The problem is whether it's worth the launch costs for it.
If you put a pipe with hot gas inside, in space, it will get colder by convection.
Blow air through the pipe.
This. Like it would make far more sense to colonize the poles than Mars.
It could also just be ignorance and talking out of his ass to look smart. Like when he took over Twitter and began publicly spewing wrong technical details as if he knew what he was talking about and being corrected by the people actually working on the product.
e.g. the lack-of-a-thing that makes a thermos great at keeping your drink cold too
His plan here clearly hinges around using robots to create a fully-automated GPU manufacturing and launch facility on the moon. Not launching any meaningful number from earth.
Raises some big questions about whether there are actually sufficient materials for GPU manufacture on the moon... But, whatever the case, the current pitch of earth-launches that the people involved with this "space datacenter" thing are making is a lie. I think it just sounds better than outright saying "we're going to build a self-replicating robot factory on the moon", and we are in the age of lying.
Power input and heat radiation both scale with area so maybe there is some way to achieve this at scale. For instance, maybe it will not look like a traditional data center or even traditional chips.
They might be closer to collapsing than most people think. It's not unheard of that a billionaires net worth drops to zero over night.
I think it's mostly financial reasons why they merged the companies, this space datacenter idea was born to justify the merge of SpaceX and xAI. To give investors hope, not to really do it.
Next up in the equation is surface emissivity which we’ve got a lot of experience in the automotive sector.
And finally surface area, once again, getting quite good here with nanotechnology.
Yes he’s distracting, no it’s not as impossible as many people think.
What about gamma rays? there is a reason why "space hardened" microcontrollers are MIPS chips from the 90s on massive dies with a huge wedge of metal on it. You can't just take a normal 4micron die and yeet it into space and have done with it.
Then there is the downlink. If you want low latency, then you need to be in Low earth orbit. That means that you'll spend >40% of your time in darkness. So not only do you need to have a MAssive heat exchanger and liquid cooling loop, which is space rated, you need to have ?20mwhr of battery as well (also cooled/heated because swinging +/- 140 C every 90 minutes is not going to make them happy)
Then there is data consistency, is this inference only? or are we expecting to have a mesh network that can do whole "datacentre" cache coherence? because I have bad news for you if you're going to try that.
Its just complete and total bollocks.
utter utter bollocks.
This is Musk, yet again, pulling themes from sci-fi books. He has that vision of ushering in the "future" which is good for dragging us forward but also he fails a lot. His open letter cited the Kardashev scale and his vision for getting us forward like in the novel accelerando.
For example: quite apart from the fact of how much rocket fuel is it going to take to haul all this shit up there at the kind of scale that would make these space data centres even remotely worthwhile.
I'm not against space travel or space exploration, or putting useful satellites in orbit, or the advancement of science or anything like that - quite the opposite in fact, I love all this stuff. But it has to be for something that matters.
Not for some deranged billionaire's boondoggle that makes no sense. I am so inexpressibly tired of all these guys and their stupid, arrogant, high-handed schemes.
Because rocket fuels are extremely toxic and the environmental impact of pointlessly burning a vast quantity of rocket fuel for something as nonsensical as data centres in space will be appalling.
Sure, space is cold. Good luck cooling your gear with a vacuum.
Don't even get me started on radiation, or even lack of gravity when it comes to trying to run high powered compute in space. If you think you are just going to plop a 1-4U server up there designed for use on earth, you are going to have some very interesting problems pop up. Anything not hardened for space is going to have a very high error/failure rate, and that includes anything socketed...
Please, no!
Of course it doesn’t fucking make sense to put data centers into space. Even if heating were solved somehow magically, server disks are veeery prone to fail and need replacement. Shoot a rocket up every week to replace failed drives or absolutely burned through GPUs? Yeah, that doesn’t even remotely sound feasible.
There should be some temperature where incoming radiation (sunlight) balances outgoing radiation (thermal IR). As long as you're ok with whatever that temperature is at our distance from the sun, I'd think the only real issue would be making sure your satellite has enough thermal conductivity.
On the other hand Starlink has several thousand satellites up there using solar power to run processors and cooling them with radiators so it's not totally new technology.
Here's a Musk tweet linking some analysis https://x.com/elonmusk/status/2013676764099199156
A satellite as a whole will come to thermal equilibrium with space at a fairly reasonable temperature, the problematic part is that the properties of electricity make it easy to concentrate a good part of the incoming energy in a small area where the GPU is. Heat is harder to move than electricity and getting that heat back out to the solar panels or radiators requires either heavy heat pipes or complex coolant pumps.
1. Getting things to space is incredibly expensive
2. Ingress/egress are almost always a major bottleneck - how is bandwidth cheaper in space?
3. Chips must be “Rad-hard” - that is do more error correcting from ionizing radiation - there were entire teams at NASA dedicated to special hardware for this.
4. Gravity and atmospheric pressure actually do wonders for easy cooling. Heat is not dissipated in space like we are all used to and you must burn additional energy trying to move the heat generated away from source.
5. Energy production will be cheaper from earth due to mass manufacturing of necessary components in energy systems - space energy systems need novel technology where economies of scale are lost.
Would love for someone to make the case for why it actually makes total sense, because it’s really hard to see for me!
1. Solving cost of launching mass has been the entire premise of SpaceX since day one and they have the track record.
2. Ingress/egress aren't at all bottlenecks for inferencing. The bytes you get before you max out a context window are trivial, especially after compression. If you're thinking about latency, chat latencies are already quite high and there's going to be plenty of non-latency sensitive workloads in future (think coding agents left running for hours on their own inside sandboxes).
3. This could be an issue, but inferencing can be tolerant to errors as it's already non-deterministic and models can 'recover' from bad tokens if there aren't too many of them. If you do immersion cooling then the coolant will protect the chips from radiation as well.
4. There is probably plenty of scope to optimize space radiators. It was never a priority until now and is "just" an engineering problem.
5. What mass manufacture? Energy production for AI datacenters is currently bottlenecked on Siemens and others refusing to ramp up production of combined cycle gas turbines. They're converting old jet engines into power plants to work around this bottleneck. Ground solar is simply not being considered by anyone in the industry because even at AI spending levels they can't store enough power in batteries to ride out the night or low power cloudy days. That's not an issue in space where the huge amount of Chinese PV overproduction can be used 24/7.
Elon musk has a history of making improbable-sounding promises (buy a tesla now, by 2018 it will be a self-driving robotaxi earning money while you sleep, humanoid robots, hyperloops).
The majority of these promises have sounded cool enough to enough people that the stock associated with him (TSLA) has made people literal millionaires just by holding onto the stock, and more and more people have bought in and thus have a financial interest in Musk's ventures being seen in a good light (since TSLA stock does not go up or down based on tesla's performance, it goes up or down based on the vibes of elon musk. It is not a car company stock, it is an elon vibes check).
The thing he's saying now pattern matches to be pretty similar, and so given Musk's goal is to gain money, and he gains money by TSLA and SpaceX stock going up, this makes perfect sense as a thing to say and even make minor motions towards in order to make him richer.
People will support it too since it pattern matches with the thing prior TSLA holders got rich off of, and so people will want to keep the musk vibes high so that their own $tsla holdings go to the moon.
Make sense now?
The technical discussions here are distractions to the avoid dealing with the glaring dissonance that for the vast majority none of this will improve life on earth.
Sun-synchronous orbit means solar panels collect the same amount 24/7. I guess that's the #1 benefit. Cheap energy.
Free space optics are much faster than data to/from the ground. If the training workloads only require high bandwidth between sats, this isn’t a real issue.
They don't do RAD hardening on chips these days, they just accept error and use redundant CPUs.
1. every gram you need to send to space is costly, a issue you don't have at ground level
2. cooling is a catastrophe, sure space is cold, but also a vacuum, so the cooling rate is roughly the infrared radiation rate. This means if you are not careful with the surface of a satellite it can end up being very slowly cooked by sunlight alone not including running any higher heat producing component (as it absorbs more heat from sunlight then it emits, there is a reason satellites are mostly white, silver or reflective gold in color). Sure better surface materials fix that, but not to a point where you would want to run any heavy compute on it.
3. zero repair-ability, most long running satellites have a lot of redundancy. Also at least if you are bulk buying Nvidea GPGPUs on single digit Million Euro basis it's not rare that 30% have some level of defect. Not necessary "fully broken" but "performs less good then it should/compared to other units" kind of broken.
4. radiation/solar wind protections are a huge problem. Heck even if you run things on earth it's a problem as long as your operations scale is large enough. In space things are magnitudes worse.
5. every rocket lunch causes atmospheric damage, so does every satellite evaporating on re-entry. That wasn't that relevant in the past, but might become a problem just for keeping stuff like Starlink running. We don't need to make it worse by putting datacenters into space.
6. Kessler Syndrom is real and could seriously hurt humanity as a whole, no reason to make it much more likely by putting things into space which don't need to go there.
Last but not least, wtf would you even want to do it?
There is zero benefit, non nada.
Even this isn't true. It's ~120 degC in daylight in LEO. It only gets cold in the shade, but a solar powered data center is pretty useless in the shade.
And yet journalists at major institutions have been repeating Musk's claims with very little skepticism ("xAI and SpaceX are merging to bring data centers to space").
> 1. every gram you need to send to space is costly, a issue you don't have at ground level
This is a one time cost. Maybe the running costs are cheap enough to offset this.
> 2. cooling is a catastrophe, sure space is cold, but also a vacuum, so the cooling rate is roughly the infrared radiation rate. This means if you are not careful with the surface of a satellite it can end up being very slowly cooked by sunlight alone not including running any higher heat producing component (as it absorbs more heat from sunlight then it emits, there is a reason satellites are mostly white, silver or reflective gold in color). Sure better surface materials fix that, but not to a point where you would want to run any heavy compute on it.
I would assume the people designing this are "very careful" with everything they put in the data center. If achieving the cooling is only very hard and requires careful material engineering, then it can be worked out and they will get it done. If it is impossible, then this will not happen, but I'm a physicist myself and I can't tell without a very involved analysis whether it is impossible or not to get enough cooling power for this in space, considering all, possibly ingenious ways to engineer the surfaces of the data center to dissipate a maximum amount of heat.
> 3. zero repair-ability, most long running satellites have a lot of redundancy. Also at least if you are bulk buying Nvidea GPGPUs on single digit Million Euro basis it's not rare that 30% have some level of defect. Not necessary "fully broken" but "performs less good then it should/compared to other units" kind of broken.
I suppose they could make something like the International Space Station, which would get regular traffic back-and-forth exchanging and servicing hardware as needed.
> 4. radiation/solar wind protections are a huge problem. Heck even if you run things on earth it's a problem as long as your operations scale is large enough. In space things are magnitudes worse.
Again, it's not a question whether this is "problematic"; everything about putting data centers in space is. The question is whether, with huge amount of work and resources, they can engineer a solution to overcome this. If they can, it's again a one time cost for the data center that might be offset by the running costs of the facility.
> 5. every rocket lunch causes atmospheric damage, so does every satellite evaporating on re-entry. That wasn't that relevant in the past, but might become a problem just for keeping stuff like Starlink running. We don't need to make it worse by putting datacenters into space.
> 6. Kessler Syndrom is real and could seriously hurt humanity as a whole, no reason to make it much more likely by putting things into space which don't need to go there.
These are collective problems for the whole of humanity and will not concern an individual actor such as Elon Musk who wants to send more satellites into space.
Then they work backwards, trying to figure out some economic engine to make it happen. "Data centers" are (A) in-vogue for investment right now and (B) vaguely plausible, at least compared to having a space-casino.
It appears to have come out of a crack pipe.
But because that's fiction, Vinge can just handwave away all the hard engineering problems for sci-fi flavor.
Anti satellite weapons are a thing. Besides, the more vulnerable part becomes you as a person rather than the equipment. There's no space colony yet, and even if there is, the supplies can be easily held hostage by an earthly government too.
https://recommentions.com/elon-musk/books/culture-by-iain-ba...
https://www.vox.com/culture/413502/iain-banks-culture-series...
https://fortune.com/2025/12/15/billionaire-elon-musk-say-tha...
> Musk pointed to The Culture series by Iain M. Banks as his best “imagining” of this world. The science fiction novels depict a utopian future where citizens can have virtually anything they want thanks to AI—making money obsolete and leaving citizens free to spend their time doing whatever they love.
So whenever I see here or anywhere else that your ideas mean nothing I just laugh at it. Of course, these come from people who are bland, doesn't have any imagination and they are not creative at all at all, but they have brute force, which is money.
1. The only reason there are 15,000 satellites in space is because SpaceX launched about 9,500 of them (Starlink is 65% of all satellites) on their semi-reusable Falcon 9. If fully-reusable Starship pans out, they will be launching satellites at 10x the rate of Falcon 9 at the very least.
2. You don't need to upgrade the satellites, you just launch new ones. The reason data center companies upgrade their servers is because they can't just build a new data center to hold the new chips. But satellites in space are a sunk cost, so just keep using the existing satellites while also launching new ones.
3. Falling solar panel costs decreases the power costs for both earth-based and space-based, but they're more efficient in space so the benefit would be proportionally greater there.
As I said, I'm skeptical too, but let's be skeptical for good reasons.
- SpaceX just requested a license to launch up to a million satellites.
- the satellites already have some incredible anti collision software, which I believe Elon has now open sourced.
- the cost to launch 1 kg to space has dropped by a factor of 10 in the past few years and is currently less than $1000. It's perfectly reasonable to estimate that over the next 10 years the cost could drop by another factor of 10, if not more, particularly if the heavy rockets are reusable.
1. https://techcrunch.com/2026/01/31/spacex-seeks-federal-appro...
2. https://starlink.com/updates/stargaze
3. https://www.netizen.page/2025/05/cost-per-kilogram-to-low-ea...
Edit: added item 3
For a benchmark - the IIS uses about 4500sqft (420 sqm) of radiators just to keep it's onboard equipment (~70KW) cooled. That's ~150-200 W/sqm.
That means, per GPU, you'd need about 2.5-3.0 sqm of passive radiators.
For a 1MW satellite (~8 datacenter racks of GB200/NVL72) you'd need basically half a football field of bleeding-edge solar panels (that also need to radiate their heat on the reverse side) and a similar sized cooling array of heat radiators for the electronics.
This is on the scale of 40-50 tons - about 10% of the IIS. This should fit on falcon heavy or starship - assuming the solar array and radiators can fold up to fit inside. You could, purely based on weight, launch 2 of these per starship launch.
If you consider the Opex savings (electricity, rent, facilities maintenance) and putting 2 of these on a single starship launch, I still think the ROI would be too long. You're saving about ~$1M per year in Opex but it's costing you $25M to launch it into space and likely an extra ~$50M in satellite equipment (based on starlink satellite costs) on top of the compute. Will those GPUs still be useful in 10 years? Probably not.
I don't think the math is there that justifies the free electricity - even at gigawatt scale (thousands of satellites mass-produced) and at a dramatically lower cost per satellite and per launch. Getting costs down on this would involve tightly integrating the compute and the satellite hardware which would make upgrading the compute independently from the cooling/power infrastructure in the future a significant challenge.
- In the EU, the ASCEND study conducted in 2024 by Thales Alenia Space found that data center in space could be possible by 2035. Data center in space could contribute to the EU's Net-Zero goal by 2050 [1]
- heat dissipation could be greatly enhanced with micro droplet technology, and thereby reducing the required radiator surface area by the factor of 5-10
- data center in space could provide advantages for processing space data, instead of sending them all to earth. - the Lonestar project proved that data storage and edge processing in space (moon, cislunar) is possible.
- A hybrid architecture could dramatically change the heat budget: + optical connections reduce heat + photonic chips (Lightmatter and Q.ANT) + processing-in-memory might reduce energy requirement by 10-50 times
I think the hybrid architecture could provide decisive advantages, especially when designed for AI inference workloads,
How unbelievably crass. "Let's build something out of immense quantities of environmentally-destructive-to-extract materials and shoot it into space on top of gargantuan amounts of heat and greenhouse gas emissions; since it won't use much earth-sourced energy once it's up there, that nets out to a win!"
Insane.
Intentionally causing Kessler Syndrome?
> A hybrid architecture could dramatically change the heat budget: + optical connections reduce heat + photonic chips (Lightmatter and Q.ANT) + processing-in-memory might reduce energy requirement by 10-50 times
It would also make ground-based computation more efficient by the same amount. That does nothing to make space datacenters make sense.
This is only relevant to the compute productivity (how much useful work it can produce), but it's irrelevant to the heat dissipation problem. The energy income is fundamentally limited by the solar facing area (x 1361 W/m^2). So the energy output cannot exceed it, regardless useful signals or just waste heat. Even if we just put a stone there, the equilibrium temperature wouldn't be any better or worse.
How about now? https://www.bbc.com/news/articles/ce3ex92557jo
Once you have solutions, it turns into a cost problem. And if that cost is too high (for whatever arbitrary threshold you use for that) it becomes an optimization problem.
This whole thread reads like a lot of "but ... but ... but ...". It all boils down to people assuming things about what is too much or too hard. And it's all meaningless unless you actually bother to articulate those assumptions. What exactly is too hard here? What would it take to address those issues? What would the cost be? Put some numbers on it. There are also all sorts of assumptions about what is valuable and what isn't. You can't say something is too hard or too costly without making assertions about what is worth paying for and what isn't.
The answers are going to be boring. We need X amounts of giga tons launched to orbit at Y amount of dollars. OK great. What happens if launch cost drops by 1 or 2 orders of magnitude? What happens if the amount of mass needed drops because of some engineering innovation? Massively dropping launch cost is roughly what SpaceX is proposing to do with Star Ship. Is it still "too hard"? You can't have that debate until you put numbers on your assertions.
There's a bit of back of the envelope math involved here but we're roughly talking about a million satellites. In the order of ~2.5 million tonnes of mass (at 2.5 ton per satellite). Tens of thousands of Star Ship launches basically. It's definitely a big project. We're talking about 1-2 order magnitude increase of the scale of operations for SpaceX going from lower hundreds to thousands of launches per year spread over maybe 10-15 years to work up to a million satellites.
I'm more worried about what all that mass is going to do when it burns up in the atmosphere / drops in the oceans. At that scale it's no longer just a drop in the ocean.
The average temperature of deep space is approximately -270.45°C or 2.73 Kelvin), which is just above absolute zero. This baseline temperature is set by the Cosmic Microwave Background (CMB) radiatio...
Which is absolute nonsense, because vacuum has no temperature.
gemini says that the NVIDIA DGX H100 is 130kg and takes 11kW.
It says space-based radiators in the 100kW range are approx 15kg per kW. And space-based solar panels are approx 1kg per kW.
So let's says we're talking about 1 system that bundles 9 DGX H100's. That's 1.2T for the computing system, 1.5T for the radiator, 100kg for the solar panels, and let's say 2T for the propulsion, propellant, guidance, and all the other spacecraft stuff. That's a total of about 5T, and the radiator is just about 20% of the mass budget.
The power radiated is proportional to the 4th power of the temperature, so they would be incentivized to develop a heat exchanger with a high temperature working fluid.
And nobody ever calls them out on it.
Today's data centres are optimised for reliability, redundancy, density, repairability, connectivity and latency. Most of advertised savings come not from placing the data centre in space, but the fact that advocates have argued away the need for absolutely everything that modern data centres are designed to supply, except for the compute.
If they can really build a space data centre satellite for as cheap as they claim, why launch it? Just drive it out into the middle of the desert and dump it there. It can access the internet via starlink, and already has solar panels for power and radiators for cooling. IMO, If it can cool itself in direct sunlight in space, it can cool itself in the desert.
The main thing that space gains you over setting up the same satellite in the desert is ~23 hours of power, vs the ~12 hours of power on the ground. And you suddenly gain the ability to repair the satellite. The cost of the launch would have to be extremely cheap before the extra 11ish hours of runtime per day outweighed the cost of a launch; Just build twice as many "ground satellites".
And that's with a space optimised design. We can gain even more cost savings by designing proper distributed datacenter elements. You don't need lightweight materials, just use steel. You can get rid of the large radiators and become more reliant on air cooling. You can built each element bigger, because you don't have to fit the rocket dimensions. You could even add a wind turbine, so your daily runtime isn't dependant on daylight hours. Might even be worth getting rid of solar and optimising for wind power instead.
An actual ground optimised design should be able to deliver the same functionality as the space data centre, for much cheaper costs. And it's this ground optimised distributed design that space data centres should be compared to, not today's datacenter which are hyper-optimised for pre-AI use cases.
-------------------
Space data centres are nothing more than a cool Sci-Fi solution looking for a problem. There have been mumblings for years, but they were never viable (even bitcoin mining was a bit too latency sensitive). Space data centre advocates have been handed a massive win with this recent AI boom, it's the perfect problem for their favourite solution to solve.
But because it's a solution looking for a problem, they are completely blind to other solutions that might be an even better fit.
Not to go all Ian Malcolm, but half this comment section is spending so much time wondering if we could build a space data center, without stopping to ask if it made any goddamn sense whatsoever to do so.
Elon has already built tons of data centers here on earth. He knows how to build them quickly. People even build them in tents these days.
If you want to radiate away the heat, you are either limited by the Stefan-Boltzmann equation which requires extraordinarily large radiators at any reasonable operating temperature, or have to develop a "super-Planckian" radiator technology, something which while it may be theoretically possible doesn't seem to actually exist yet as a practical technology.
The only other plausible technology I can think of would be to use evaporative or sublimation-based cooling, but that would consume vast quantities of mass in the process, every bit of which would have to be delivered to space first.
Has anyone seen any published work that suggests it is actually anywhere near economically feasible to dissipate megawatts of power in space, using either these or any other technology?
The moon has:
- Some water
- Some materials that can be used to manufacture crude things (like heat sinks?)
- a ton of area to brute force the heat sink problem
- a surface to burry the data centers under to solve the radiation problem
- close enough to earth that remote controlled semi-automated robots work
I think this would only work if some powerful entity wanted to commit to a hyper-scale effort.
I suspect this is really the fundamental idea behind this whole plan.
Almost any reason why the moon is better than in orbit is a point for putting it on earth.
i think the moon likely does contain vast mineral deposits though. when europeans first started exploring australia they found mineral anomalies that havent existed in europe since the bronze age.
the Pilbara mining region is very cool. it contains something like 25% of the iron ore on earth, and it is mostly mined using 100% remote controlled robots and a custom built 1000 mile rail network that runs 200-300 wagon trains, mostly fully automated. it is the closest thing to factorio in real life. 760,100 tonnes a year of iron ore mined out and shipped to China.
I too don't think it's currently a sensible solution. But the author completely unable to make a proper case. For instance, just to refute that one claim, there are many reasons to do it in space even at an cost.
Space-based data centers provide an off-world backup that is immune to Earth-specific disasters like earthquakes, floods, fires, or grid collapses. Servers in orbit are physically isolated from terrestrial threats, making them safe from riots, local warfare, or physical break-ins.
Moving infrastructure to space solves local community disputes by removing the strain on residential power grids and freeing up land for housing or nature. Space data centers do not deplete Earth’s freshwater supply for cooling, unlike terrestrial centers which consume billions of gallons annually.
Solar panels in orbit can access high-intensity sunlight 24 hours a day without interference from clouds, night, or the atmosphere.
Data stored in space can exist outside of national borders, protecting it from seizure, censorship, or the legal jurisdiction of unstable governments. Data transmission can be faster in space because light travels roughly 30% faster in a vacuum than it does through fiber optic cables.
Processing data directly in orbit is necessary for satellites and future space stations to avoid the delay and cost of beaming raw data back to Earth
To that end, a small data center space isn’t about unit-economics, it’s a bigger mission. So the question we should consider is what can we put into space the further that mission. Can we put a meaningful sum of human knowledge out there for preservation? It sounds like “yes,” even if we can’t train ChatGPT models out there yet.
The whole time I was there it was a mental game of trying to steel man the contradictory or incoherent stuff, using my brain power to try and rewrite things to make sense.
After some years, I woke up and realized that’s what I was doing, and even if I could do it in my mind, that didn’t make the source material rational.
Heres hoping you have a similar moment.
Datacenters in space have a lifespan measured in years. Single-digit years. Communicating with such an installation requires relatively advanced technology. In an extinction level crisis, there will be extremely little chance of finding someone with the equipment, expertise, and power to download bulk data. And don't forget that you have less than a decade to access this data before the constellation either fails or deorbits.
Meanwhile people who actually care about preserving knowledge in a doomsday crisis have created film reels containing a dump of GitHub and enough preamble that civilizations in the far future can reconstruct an x86 machine from scratch. These are buried under glaciers on earth.
We've also launched (something like) a microfilm dump of knowledge to the moon which can be recovered and read manually any time within the next several hundred or thousand years.
Datacenters in space don't solve any of the problems posed because they simply will not last long enough.
High performance chips are made for the shielded atmosphere. Imagine the cost launching all the extra shielding that you don't need on earth.
It is beyond stupid. Comical levels. I can't believe people are trying to find any justification.
Making a dent into making humans a multiplanetary species requires making a lot of companion species as well; the task requires much more elementary stuff (relative to the mission), at the ground level, than Musk is demonstrating to do (at technical, entrepreneurial and political level).
This is a con, from the start. It just worked so far so some people fall for it.
(Yes, I know what steel manning is)
If the AI data-center used only 10MW then each could have two redundant SMR's assuming the cooling challenges have been worked out but then we could have nuclear reactor disposal and collision issues.
[1] https://hackaday.com/2024/02/05/starlinks-inter-satellite-la... (and this is two years ago!) [2] https://resources.nvidia.com/en-us-accelerated-networking-re...
And some of us are reading these things and trying to be polite.
But at some point patience runs thin and the only response that breaks through the irrationality is some variation of "what if unicorns and centaurs had teamed up with Sauron?"
The limit of the ratio of useful:useless "what if's" approaches zero.
I also remember, roughly 10 years ago, people saying that the amount of effort to discredit bullshit is wildly out of whack. Which makes bullshit basically asymmetric warfare.
So here we are, in this thread, actually spending time attempting to discredit bullshit.
I've heard stories that over a decade ago teams inside hyperscalars had calculated that running completely cryogenically cooled data centers would be vastly cheaper than what we do now due to savings on resistive losses and the cost of eliminating waste heat. You don't have to get rid of heat that you don't generate in the first place.
The issue is that at the moment there are very few IC components and processes that have been engineered to run at cryogenic temperatures. Replicating the entirety of the existing data center stack for cryogenic temps is nowhere near reality.
That said, once you have cryogenic superconducting integrated circuits you could colocate your data centers and your propellant/oxidizer depots. Not exactly "data centers off in deep space" since propoxd tend to be the highest traffic areas.
take an h100 for example. it will need something like 1kW to operate. that's less than 4 square meters of solar panel
at 70C, a reasonable temp for H100, a 4 square meter radiator can emit north of 2kW of energy into deep space
seems to me like a 2x2x2 cube could house an H100 in space
perhaps I'm missing something?
Apart of that, I do agree that space data centers are probably just a marketing stunt at this point, although some things could obviously be done to increase their chances, like more lightweight designs on GPUs, something that was never a big topic before.
Asside from the other excellent comments on power consumption, cooling and radiation. One point I didn't see being made in the comments much is maintenance costs.
Now I don't find myself in the facility of a data center often in daily life, however I do know that medium to big data centers require 24/7 hardware replacement. I believe this is what those 5 guys with the bikes and scooters are doing in every data center. That would be very difficult, near impossible in space (with the current space fairing infrastructure).
press x to doubt
> on 21 February 2008, the US Navy destroyed USA-193 in Operation Burnt Frost, using a ship-fired RIM-161 Standard Missile 3 about 247 km (153 mi) above the Pacific Ocean.
1. Inference
2. Training
Inference just might be doable in space because it is "embarrassingly parallel" and can be deployed as a swarm of thousands of satellites, each carrying the equivalent of a single compute node with 8x GPUs. The inputs and outputs are just text, which is low bandwidth. The model parameters only need to be uploaded a few times a year, if that. Not much storage is required , just a bit of flash for the model, caching, logging, and the like. This is very similar to a Starlink satellites, just with bigger solar panels and some additional radiative cooling. Realistically, a spacecraft like this would use inference-optimised chips, not power-hungry general purpose NVIDIA GPUs, LPDDR5 instead of HBM, etc...Training is a whole other ballgame. It is parallelisable, sure, but only through heroic efforts involving fantastically expensive network switches with petabits of aggregated bandwidth. It also needs more general-purpose GPUs, access to petabytes of data, etc. The name of the game here is to bring a hundred thousand or more GPUs into close proximity and connect them with a terabit or more per GPU to exchange data. This cannot be put into orbit with any near-future technologies! It would be a giant satellite with square kilometers of solar and cooling panels. It would certainly get hit sooner or later by space debris, not to mention the hazard it poses to other satellites.
The problem with putting inference-only into space is that training still needs to go somewhere, and current AI data centres are pulling double-duty: they're usable for both training and inference, or any mix of the two. The greatest challenge is that a training bleeding edge model needs the biggest possible clusters (approaching a million GPUs!) in one place, and that is the problem -- few places in the world can provide the ~gigawatt of power to light up something that big. Again, the problem here is that training workloads can't be spread out.
Space solves the "wrong" problem! We can distribute inference to thousands of datacentre locations here on Earth, each needs just hundreds of kilowatts. That's no problem.
It's the giaaaant clusters everyone is trying to build that are the problem.
I suspect that Musk wants to build space data centers in order to mitigate political and societal problems, which may yet prove more intractable than cooling in space.
The current expansion of terrestrial data centers has already caused a huge backlash. Their adversaries may well try to regulate them out of existence, at least locally. If an important jurisdiction like California or Germany subjects building of new data centers to regulations similar to building, say, a new nuclear power station, they will achieve a de-facto stop on further development there even without banning them outright.
Space, while not entirely lawless, is much harder to regulate this way. Local authorities have no power over it, nor do governments of nations without space capabilities. Big authorities of big nations (the FAA etc.) do, but they will likely be more friendly to already established launch businesses like SpaceX, not least because of the geopolitical dimensions of having a vibrant space sector.
From Musk's POV, this may be worth the additional cost and technical troubles.
https://www.pbs.org/newshour/world/pentagon-embraces-musks-g...
Data centers in space make absolute sense when you want as close to real time analysis on all sorts of information. Would you rather have it make the round trip, via satellite to the states? Or are you going to build these things on the ground near a battlefield?
Musk is selling a vision for a MASSIVE government contract to provide a service that no one else could hope to achieve. This is one of those projects where he can run up the budget and operating costs like Boeing, Northrup etc, because it has massive military applications.
Is this all an effort to utilize more efficient solar panels? Are solar panels really the limiting factor for data centres?
The self-driving car worked too well. Tesla is promising that for over a decade now, and still can't deliver. They came much closer to the goal, but are still very far away from it. Shareholders don't seem to care.
This might also be a new vehicle to mask any space warfare technology deployments.
BUT the fact that we are even arguing about whether or not we should be putting data centers into space is so incredibly absurd to someone who watched the Challenger explode and assumed that space wouldn't be ventured into again in my lifetime.
People don't realize how much the priors have changed. Take a minute to appreciate that. We are living in a world where people are debating if it makes sense to spend a bazillion dollars to put a hard disk into orbit.
I wonder if the Klingons are good at cyber warfare.
https://climatecosmos.com/sustainability/how-close-are-we-to...
However one flaw in this critique is that is only looks at the cost of ground-based solar panels and not their overall scalability. That is, manufacturing cost is far from the only factor. There is also the need for real estate in areas with good sun exposure that also have sufficient fresh water supply for cleaning.
When we really consider the challenges of deploying orders of magnitude more terrestrial solar, it really requires a more detailed and specific critique of the orbital vision. Positive includes near continuous solar exposure (in certain orbits) and no water requirements.
Much has been said of cooling but remember, there is a lot of literal space between the satellites for radiative cooling fins. It is envisioned they would network via optical links, and each mini satellite would be roughly on the order of a desktop GPU (not a whole data center rack). The vision is predicated on leveraging a ton of space for lots of mini satellites on the order of a Dell desktop tower. The terrestrial areas that are really cold are also not that great for solar exposure.
Personally I don't know how it will play out but the core concern I have about making these kinds of absolutist predictions is they make weak assumptions about the sustainable scalability of terrestrial power. And that is definitely the case here in that it only looks at the manufacturing cost of solar.
Engineering is always a question of tradeoffs.
Launch costs are dropping, and we’re still using inefficient rockets. Space elevators & space trains, among others, can drop this much more, the launch costs are still dropping, even using rockets, maybe we’ll never get to elevators & trains the costs will drop so low!
Radiation shielding is not required for VLEO or LEO, and phenomenally more capable aerospace processors are near - hi Microchip Inc! There are many other radiation solutions coming, no doubt with nuclear power.
Satellites can be upgraded at scale, though for many things, it does not make $ sense to upgrade them, but fuel , reaction wheels, solar panels, among other things do make $ sense to replace.
Latency was technically solved in 1995 & 2001 with the first laser comms missions NASDA’s ETS-VI kiku-6 and ESA’s Artemis , and Laser crossbars for comms are common. A full laser TDRS no RF is not yet extant but soon. Earth to deepspace was just demonstrated by ESA.
Cooling can be significantly improved due to lower launch costs, heat piping, RTGs, TEGs, and thermoradiative cells, not to mention sunside solar and darkside inline radiators
Furthermore, it is very likely that as neuromorphics with superior SWaP emerge, we could see very different models of space based computation.
Economic tradeoffs should drive many of these decisions as I’m not discussing the other applications of datacenter in space
You're saying they're going to steal the night? We'll see the sun in the day, radiative cooling for surveillance AI in the time formerly known as night?
I'll confess that the numbers aren't nearly as bad as I'd thought. Apparently, you can dissipate 1MW at 100°C with a 17m diameter sphere at night. So it's like the size of a small house. It doesn't even glow. On the other hand, you need a lot of temperature differential to move the heat out fast enough, which means your TPUs are going to be hellishly hot.
Though you'd probably only run it when it's in the sun and radiate in other directions, so you don't have to store the power in heavy batteries. You need a 56m diameter disk of solar panels to provide 1MW, don't forget that.
(All figures were vibe calculated with Claude and are unchecked.)
If you're looking at this and saying "Lol, no, we don't need data centers in space just to power more GPT sycophancy and some health insurance company's RAG workflows." Then you're right, so just move on from that usage and consider the things we'd like to do in space, and especially the things we're already doing but want to do more of.
I know, my gut was telling me this is ridiculous, premature at best even through the lens of expanding space industry. But then with respect to the need for data centers-- had I even thought about it a year ago, my gut would have laughed if someone had said "Buy Wester Digital, HDD's are a growth stock".
* They assume 1 satalite = 1 GPU. This is quite funny, actually. A single H200 floating in space with a solar panel and an antenna. In reality, a satellite would pack as many chips as the heat/power allows. A Starlink-sized satellite should be able to hold 40 or so chips. There's no reason why a larger satellitte couldn't hold, say, 1024. * They mention training, but sampling is what makes sense here. Training is a different beast, and requires high reliability, high bandwidth, low latency, and a lot of IO. Space would not be ideal for this. I'd expect training to remain terrestrial and just do sampling in space. (FWIW, sampling will be most of the compute allocation). * Also, no one upgrades GPUs in datacenters, they just add new nodes and leave the old ones there. Google still has their P100 nodes running. Not being able to fix them, though, is a significant concern.
Yeah, space is cold, but also you don't have access to large thermal masses with which to exchange heat, so the fact that space is cold does not help much.
In space the only option for cooling is large radiators. If you had a data center in space you'd need enormous radiators -- much larger than the data center itself.
(On Earth we can exchange heat with the environment, and the environment includes convection and the water cycle and ultimately can expel excess heat via high altitude condensation of water vapor where most of the heat released escapes to space. As well clouds can both block insolation as well as keep heat below trapped, but altogether between high altitude condensation and blocking insolation this is the mechanism by which Earth keeps its temperature as a random walk around the average that we enjoy.)
If you assume that these people aren't completely stupid, then there is some reason why they want this workload running at great physical distance from all the people down on Earth. It's probably not to protect people on Earth. After all they'll happily deorbit satellites and other junk from orbit and let it rain down on us. And they will happily destroy the environment with all those rocket launches too. Therefore it must be to protect the workload from us.
What is a workload that is something that people would probably want to destroy, and which would also provide enough value to offset the expense to launch and run in space? The only thing that might make sense is a military AI platform. Think something that observes Earth, launches missiles, and controls terrestrial drone armies remotely, with relatively low latency.
It gets built and launched thanks to endless military budget, and once it is up there, running such an AI from space means that effectively the only people who can take it out are nation state level foes who can launch rockets into low earth orbit. And this thing is a satellite, probably part of a network that is watching the Earth all the time. Start building something that looks like a rocket launch site, and the AI will see, then you get hit by a missile or taken out by a drone first before you get a chance to attack the platform.
It sounds like sci-fi, but in the future, if we let it happen, there could absolutely be nearly invulnerable autonomous AI platforms in space overseeing everything, and making decisions, and issuing commands. Of course there could still be a massive solar flare event, or a Kessler syndrome event that releases us all from AI enforced servitude. Anyway, it's a not so fun thought experiment, and let's hope this stays sci-fi, so we can just enjoy a fun Hollywood film about this rather than experiencing it firsthand.
The biggest problem folks had was even with equipment with 99.9% reliability something breaks every day due the huge raw number of devices involved. And most network equipment is not any where close to being radiation hardened.
I had some fun with it with Bezo’s fist bumping folks because SpaceX was cleaning BlueOrigin clock.
I talked to one of their lawyers and didn’t hear anything afterwards. I left AWS and a couple of years later Amazon announced AWS ground station. I wonder how much my paper contributed to green lighting that project.
Then he talked about datacenters in space and this is something I have some appreciation for, and I immediately knew he couldnt have done much Physics, and sure enough, I was right.
There are "experts" out there who basically have no idea what they are talking about, "it is absolute zero in space in the shadow!", as though radiative cooling is that effective.
And that's not even talking about part failures. How do we replace failed parts in space? This is a scam, but everybody is afraid to openly challenge eloquent "experts" who are confidently wrong.
https://www.nextbigfuture.com/2022/02/spacex-reusable-rocket... -- looks like target price for Starship launch would be $3--$5m according to the author.
Wouldn't the /kg price to SpaceX be:
3000000/100000 = $30/kg -- 5000000/100000 = $50/kg?
If they recover everything and produce fuel at scale, wouldn't it drop the cost even more.
What many people quote here are commercial rates, I think. SpaceX won't pay those prices.
Can someone check my math
Maybe there's a very latency sensitive need to send realtime targeting information to a tomahawk missile in flight? But it's also too bandwidth, compute or cost intensive to send a firehose of raw spy-satellite data to a disposable one-way attack munition?
The data centers in space are actually for the spy satellites to use. That's all I got for practical applications.
Author made a fatal mistake. By flying enough hardware in space, you can simply blot out the sun and steal their solar capacity. Drink their milkshake with a long straw!
I think it’s all farce and technically unsound, but I also think that grok-5-elononly is a helluva drug. It’s really got him ready to rally investors behind “spreading the light of consciousness to the universe”. Oh to see the chat logs of their (Elon and his machine girlfriend)’s machinations.
PS. I think the authors argument that millions of satellites might run into each other is silly. There are like 1.5 B cars.
Listen, I totally agree, the tech makes absolutely no sense. It does not. But the fact that someone is willing to spend money on figuring this out is pretty good. The worst thing is going to happen, we'll have a cheaper space travel. And let the guys to have the first hit at it, wasting money on an enormous amount of research needed.
Ain't my money being spent.
As long as we don't have to use Russian rockets to send the US payload to the orbit, I'm cool with it.
But more abstractly, it's our resources that are being allocated. The planet as a unit is deciding where to put it's effort. Apparently we're not very good at this
The answer to that is that coordination problems are really hard. Much harder even than what are currently unsolved engineering problems. In fact, SpaceX can only launch from California because they have DOD coverage for their launches. Otherwise the California Coastal Commission et al. would have blocked them entirely. Perhaps the innovation for affordable space Internet is combining it with mixed-use technology.
The truth is that in America today self-driving cars (regulated by a state board run by bureaucrats) are easier to build than trains (regulated by every property owner on the train route). Mark Zuckerberg tried to spend some money evaluating a train across the Bay and had to give up. But Robotaxi service is live in San Francisco.
So if there is an angle that makes sense to me it's that they anticipate engineering challenges beatable in a way where regulatory challenges are not.
I also checked out your blog and got 2 interesting articles in 2 tries. If you have some personal favourites and listing them is not a bother, I'd be happy to read them.
Even if it's more expensive, Spacex will be able to deploy hardware when no one else can because all the gas turbines and existing power plants have been exhausted and the lead time to build new ones is 5+ years out because of the bureaucratic overhead.
Tesla's valuation has been nuts for a while. The music was going to stop playing at some point, so something something robotaxis, something something androids, something something AI. Keep the investors duped while you can move money around and leverage it to stay relevant.
Cars are out, social media as well (especially X), but Space is still in, and even more so AI. So let's move the cost center with world domination potential (AI) over to the one company that's making money and has a still has a cash-out potential via an IPO.
I'm just so tired of it all. I actually think 'boutique' businesses (companies that generate real value to real users and are profitable now) are the only thing that can save our economy medium-term, but investors and the government are having none of it. And the result is that these bait-and-switch scams will continue.
Nvidia GPUs, particularly H100s, have a failure rate orders of magnitudes higher than traditional CPU-only hardware. I myself have accidentally melted an H100 during a large-scale training run.
While it’s trivial to replace a broken GPU in the ground, it seems to be infeasible in space during the life of the satellite. Getting a human or robot “fixer” spacecraft to it would likely cost more than the $30K GPU itself.
I'm no expert on solar but I thought there was some upper limit on how much power ground-based solar panels can generate per area based on how much energy gets through the atmosphere all the way to ground - and that panel efficiency was approaching that limit.
However, I don't doubt ground-based panels can continue to improve in cost and other metrics and thus exert competitive pressure on space-based solutions.
First mover advantage, and all.
And hardware that is happy in high-radiation environments is not going to be fast.
As an alleged human, I'd like to preserve my option to interfere.
While technically not impossible, the space data center vision appears primarily designed to support SpaceX’s anticipated mid-2026 IPO and justify a $1.5 trillion valuation rather than solve near-term compute constraints.
You cannot do two things at the same time:
1. Make electronics small. 2. Make electronics radiation resistant.
It’s a numbers game. The denser things are packed, the higher the probability that a random high energy particle or ray will damage something.
20 years ago NASA was buying old chips, because those were less susceptible, modern “radiation hardened by design” chips are better than those, but still slower than those for planetary use.
I'm taking the parts of this write-up I don't have expertise with a grain of salt after seeig this.
Kessler cascades are real. Particularly at high altitudes. They're less of a problem in LEO. And in no case can they "[cripple] our access to space." (At current technology levels. To cripple access to space you need to vaporise material fractions of the Earth's crust into orbit.)
The sentence you mention was indeed a give away, but there are many others. Worst case scenario, nothing works and Elon burns a bunch of money, part of which goes into jobs and research. Best case scenario, we actually move away from technologies from the 50's and end up with daily, cheap earth-to-low-orbit (ideally something better than that - how about the moon?), no more whining about energy costs, and laser communication IRL. That's just the obvious stuff.
Being "realistic" and "having a budget" is what companies like Google do. That's all good, but we have enough of those already.
Well, maybe "higher", but not really high.
The lower the altitude, the larger the odds of making one, in a quadratic fashion. But also the lower the altitude, the less time it will last.
There is some space where it lasts basically forever but is small enough for it to happen. It's higher than LEO, and way lower than things like GEO.
Given the solar constant 1361 W/m^2, you can calculate the temperature range based on the emissivity and absorptivity. With the right shape and “color”, the equilibrium temperature can be cooler than most people thought.
I suppose that a space data center powered 100% by solar is no different than this iron ball in principle.
Also why talk about training not inference? That needs data centers too and could be what they're intending to do.
So this post is clearly not an effort to objectively work out the feasibility but just a biased list of excuses to support the author's unsubstantiated opinion.
Taking a creative step back, perhaps datacenters in space support something with Mars?
As much as that might not seem realistic, I also have to counterbalance it with operationalizing and commercializing SpaceX, Starlink and Tesla relatively quickly when so much stays at the R&D stage for so long.
I’d even bet that when they do IPO, there will be ZERO mention of “space data centres” in the prospectus!
The author forgot to add that this is only true from the perspective of their own bias.
To someone else it might make a lot of sense, e.g. someone who expects militant resistance to the "data centers" from the general public or some other actor that is highly unlikely to achieve space capabilities.
what am I missing here?
They make no sense otherwise.
The only other thing I can think of is the whole thing is just a scheme to get investment and they’re never going to actually go through with it.
At this point I kind of think the former is more likely.
then anything that drives money towards your work on that goal is worth pursuing particularly if you think time is short.
But heat radiation rates are proportional to temperature to the 4th power!!!!
That is a magical law. The quality of heat pumps used to concentrate heat, will drive the economics and structure of heat dissipation.
Seldom do we get constraints that favorable to work with.
Besides this, it's concerning how much stuff we're sending to space. One day we'll have to start worrying about satellite parts falling on us.
These companies wanted to merge for financial reasons and the invented reason is nonsensical. We shouldn't even give the nonsensical reason the benefit of trying to make sense of it.
Reusable rockets make no sense.
Autonomous cars make no sense.
Data centers in space make no sense. <--- You are here.
Humanoid robots make no sense.
Solar roof tiles: makes no sense
Lot's of tiny tunnels under cities: makes no sense
Performance of the new roadster: makes no sense
All four of the above were likely scams. Musk is not beyond running a scam.
- Reusable rockets have been a thing since the Space Shuttle in 1981, building on a 1969 plan for reusable space vehicles. - https://en.wikipedia.org/wiki/Space_Shuttle
- Autonomous cars: human chauffeurs and taxis (and trains) so we can be moved around without doing the driving, go back to the first cars. We haven't had the technology to build them (and arguably don't and won't have until we get near AGI).
- Data centers in space ... ???? Bueller? Bueller?
- Humanoid robots were seen in Fritz Lang's Metropolis film in 1927, they've always made sense. What doesn't make sense is lying about having built humanoid robots and then having to admit they were being remote controlled, cough Tesla.
"Just change the law" ok sure we'll get right on it.
My take-away is - SpaceX is still an extremely good stock to hold. However, the stupid money will buy the stock at IPO on the promise of space datacentres.
When SpaceX inevitably u-turns on this plan and the stock plummets temporarily, THAT will be a good time to buy in.
What they are trying to do is an a very ambitious engineering challenge in several highly integrated domains, from spaceships to robotics, gpu, server design , ai. Typical stakes are high margins are high.
Project also pushes boundaries of what human can do the same way starlink did. 20 years ago starlink scale was also an "impossible" thing. Is it possible now to push one gpu and serve it from space? Yes, can you do it at scale? Big question.
Obstacles :
Price per kg to orbit. They aim to go from $150 to $10 per kg. Can they deliver? Big question, but having something that demands so many starship, like GPU heavy tasks, will help them achieve this goal. Benefits? No rent, no cooling issues, cheap sun-powered electricity 24 hours a day, unlike anywhere else on Earth.
Jurisdiction. Servers can't be shut down or taken away by police, etc.
Cooling. Yes, it's a vacuum, but with $10 per kg, you can deliver pipes and coolant, and since you don't have space constraints, you can build these pipes with a robot, making that datacenter extremely cheap.
3. Labor. If a robot can do primitive tasks combined with design that is fully remote, you labor costs goes almost to 0.
the output: with $10 per pg, delivering solar panels, coolant + robot for some urgent fixes. Robots even with current technology can swap a harddrive for example, especially if hardware is built to be mantained by robots not human. You don't need large construction, you invest into design, that once assembled , cost you maybe comparable amount of hardware to deliver. After that it runs for free!
The economy if this obviously beating anything exised on earth.
Putting data centers in space keeps them out of reach of humans with crowbars and hammers, which may have been a vulnerability for those robots Tesla is building.
The website insists that you let it record your voice in order to show you the dangers of AI. Is it trolling the visitor? https://civai.org/talk
Does not feel like a vibe.
This is while they try to find a solution to earn money with it.
Where is the tech?
- have very non-deterministic latency
- are located outside of a country that can protect you (ie China could disrupt your space data center)
- have to pay millions of dollars to swap out hardware
Modulo some efficiency losses, most of the electricity it generates is leaving the satellite. Contrast with a datacenter, where most of the energy is spent heating up the chips, and the rest is spent moving the heat away from those chips.
Space is cold but has little mass. Either heat can radiated or transfered. To transfer heat, mass which easily absorbs heat is required. The moon might be suitable for that.
We're living in the Age of Distraction… amusing ourselves to death (as usual).
The regulatory framework is getting more and more difficult for data centers.
The options are move to countries with less of an uphill regulatory burden (UAE?), but this comes with other issues.
Space it is.
You do this when the most fragile part in the system fails. Solar panels good for 25 years but the SSDs burn out after 2? Incinerate the lot!
This kind of thinking is late capitalist brain rot. This kind of waste should be a crime.
Anyone planning expenditures as large as a modern data center thinks about all kinds of risks (earthquakes, climate, power, etc), and so perhaps there is a premium for GPUs that are out of the reach of your median angry unemployed guy.
(yes, this is nuts, but I can easily imagine some fever-dream pitch meeting where Musk is talking about it)
This statement is actually completely false. The bottleneck is not cost of building data centers, but the energy accessible on the planet. How much we're willing to pay for increasing that is currently very unclear, but it's far more than the current cost of building a terrestrial data center.
As Jensen Huang famously touted their performance per watt, saying "Our customers won't buy our competitor's chips even if they were free."
But general purpose compute no
No? You'd only need one with lots of gpus on the ship at the same time
entirely out of jurisdiction, where it is prohibitively expensive to travel, and impractical for any physical seizure.
you dont need to compute, just store it and P2P amongst satellites.
essentially an orbital NAS.
1) Kessler syndrome is a contingency.
2) This is a logistics issue, not a physical impossibility.
3) Those are different tradeoffs (solar in space). There is not really an argument there.
All in all this is extremely weak reasoning, which is quite the contrast with the definitive title.
I throw this to the "nerds need to feel smarter than Elon" pile of articles. :)
"So here's what I did. I built a simple model that reduces the debate to one parameter: cost per watt of usable power for compute. The infographic below lets you change the assumptions directly. If you disagree with the inputs, great. Move the sliders. But at least we'll be arguing over numbers that map to reality.
The model is deliberately boring. No secret sauce. Just publicly available numbers and first-principles physics: solar flux, cell efficiency, radiator performance, launch cost, hardware mass, and a terrestrial benchmark that represents the real alternative: a tilt-wall datacenter sitting on top of cheap power. "
"Here's the headline result: it's not obviously stupid, and it's not a sure thing. It's actually more reasonable than my intuition thought! If you run the numbers honestly, the physics doesn't immediately kill it, but the economics are savage. It only gets within striking distance under aggressive assumptions, and the list of organizations positioned to even try that is basically one."
To me it looks like the next Musk’s grift. Remember Mars? Have you heard about it recently? He threw it to the Internet and everyone got excited for a minute. Then nerds did quick math and it didn’t make any sense. And so everyone forgot about Mars. This is the same. Hype everyone up for a week or two to inflate stock before the merger/purchase/IPO/whatever. That is all.
Datacenters in space is ambiguous enough to mean on lunar soil which provides plenty of heat dissipation using geothermal heat pumps.
Similarly mass to orbit is also less problematic if silicon factories (including the refineries) are built on lunar soil as well.
https://pluralistic.net/2024/05/17/fake-it-until-you-dont-ma...
Seems like a pretty obvious "no" to me. Loudoun County is a much better choice, just to pick one alternative. Antarctica is an awfully inhospitable place and running a data center there would be a nightmare.
And yet it's way better than space. It's much easier to get to. Cooling is about a thousand times easier. The radiation environment is much more forgiving.
This whole concept is baffling to me.
(Incidentally, a similar thought experiment is useful when talking about colonizing Mars. Think about colonizing the south pole. Mars is a harsher environment in just about every way, so take the difficulties of colonizing the south pole and multiply them.)
1) Water scarcity and energy scarcity here on earth
2) It will drive down launch costs and promotes investment in orbital facilities and launch capabilities.
those two reasons alone are enough.
Current satellites get around 150W/kg from solar panels. Cost of launching 1kg to space is ~$2000, so we're at $13.3(3)/Watt, that just power, let's assume that cooling will cost us same per kg, the same amount need to be dissipated so let's round it to $27
One NVidia GB200 rack is ~120kW. To just power it, you need to send $3 240 000 worth of payload into space. Then you need to send additional $3 106 000 (rack of them is 1553kg) worth of servers. Plus some extra for piping. We're already at $6.3 mil a pop for just hauling it up to orbit, with no cost of solar cells included
I'd imagine comparable hardware for just some solar + batteries on ground is around $200k. I dunno where the repeated 5x cost number comes from. I suspect whoever pushed it was just lying
No company has ever made an investment in something that ended up being more expensive than calculated, or so expensive it bankrupted them.
I mean, I still remember promises of $1000-per-kg for space launches, and how e.g. Gigafactory will produce half of the world battery supply, and other non-scientific fiction peddled by Musk. Remember when SpaceX suggested in 2019 that the US Army could use its Starship rockets to transport troops and supplies across the planet in minutes? I do. By the way, have they finished testing Starship yet, is it ready?
- Data centres need a lot of power = giant vast solar panels
- Data centres need a lot of cooling. That's some almighty heatsinks you're going need
- They will need to be radiation-hardened to avoid memory corruption = even more mass
- The hardware will be redundant in like 2 years tops and will need replacing to stay competitive
- Data centres are about 100x bigger (not including solar panels and heat sinks) than the biggest thing we've ever put in space
Tesla is losing market share (and rank increasingly poorly against alternatives), his robots are gonna fail, this datacentre ambition needs to break the laws of physics, grok/twitter is a fake news pedo-loving cesspit that's gonna be regulated into oblivion. Its only down from here on out.
Hey! It can be de-orbited onto the location of your choosing. I bet you can sell this service to the DoD!
Barring that, you can sell it on the global market to the highest bidder.
This is BS, everyone knows that this is BS, but because this is Elon, there are still people who don't call out the BS.
It might be distraction, he might be delusional, he might be asking his investors to stop asking for profit by giving them shares from SpaceX, but this is not him discovering new physics.
How's that full-self driving promised for decades working out?
How's the destruction of USAID working out (oh you wouldn't know a million dead now)
We already have a data-center in space, sort-of, here's how many radiation panels it has to deploy for just the heat produced from the small number of low-power computers
* https://i.sstatic.net/cpIBo.jpg
(ISS, all those white panels are thermal heat radiators)
If SpaceX, by being a company serving the federal government are covered by a law that would make its offices (on Earth, duh) a protected area ... then could they by some law-bending make that protection also encompass the data centres that contain the AI-generated CSAM and training data, in order to protect them from being raided by state law enforcement?
That does not have to sound reasonable to us ... only to Musk.
Adding a global UHVDC grid to even out dips in local PV performance due to cloud cover and the diurnal cycle on spaceship earth seems to be magnitudes cheaper and scaleable than this loony pitch.
The only thing making this hard is requiring supranational collaboration.
If the nodes are spinning around the earth at orbital velocities, then all the benefits of physical locality are thrown out the window.
I thought that was actually quite interesting/practical, because if there is a problem, you can just bury the problem.
not like tmi/fukushima/chernobyl
It never fails here. Ya'll are soooo determined to assume ulterior motivations. He has always been direct about his motivations, whether it's about politics or business decisions.
To suspect his whole plan for datacenters in space is a ruse or something to drive up some stock price (or whatever), is... just ridiculous. You really think they person who leads SpaceX (more satellites than the rest of the world combined) and Xai (with a competitive frontier model in 2 years starting from scratch), and Tesla (more inference compute than any company on Earth) really knows less than you about the math and physics of this idea??
His immediate focus for Starship has shifted from Mars colonisation to exponential expansion of intelligence, as he thinks it's the best path to extend consciousness. Ridiculous or not, he believes that, and that's the motivation.
Is that possible in our lifetime? I'd be optimistic about that. Can SpaceX pull that off? Space what? ...
GPT-4o mini: The term for financial moves where new investors are continually recruited to pay off previous ones is often referred to as a "Ponzi scheme." Another similar term is "pyramid scheme," where returns are paid to earlier investors from the contributions of newer investors, but with a structure that typically requires participants to recruit others to earn returns. Both schemes are unsustainable and illegal.
as always: imho. (!)
we already had this topic before, an example for another good article regarding physical arguments against this idea would be:
"Datacenters in space are a terrible, horrible, no good idea" ~ late 2025
* https://taranis.ie/datacenters-in-space-are-a-terrible-horri...
TL;DR: It's not going to work.
idk ... maybe elon has something else in mind with this merger!?
cheers,
a..z
Just do the basic thermal heat transfer math.
What does "serious" mean
What matters is that investors and shareholders love to hear about future space data centers.
Obligatory /s.
You can’t read a 300 word article in 1930 and know that a formula 1 engine could not be built.
Disagree there are bunch of scenarios where Data Centers in space make sense. Like nuclear annihilation and having vaults across the globe to communicate and get back lost information because ground data centers would be wiped out by EMP from blasts.
(If you can't xcancel it yourself your hacker card is revoked.)
As most engineers realize right away, it is not going to be profitable to operate a regular datacenter in space, per the article (and I agree), so something else is going on here. Almost all the discussion is about feasibility, which is not by itself going to explain the situation.
It is clearly somewhat feasible to build Starlink level infrastructure and operate it profitably. I would posit that the narrative is a funding vehicle for a more conservative, incremental objective.
The very fact that the infrastructure is in space places the datacenter on the legal and geopolitical high ground. It's hard to raid servers if they are in orbit. It's hard to disable, audit, or arm-wrestle into submission. It doesn't have to have the scale we've come to expect in 2026 to be useful. And it's for inference, not training, of course. Useful levels of inference is computationally cheap. There are implications with the financial system as well.
In combination with PLTR technology, what I see is another intelligent and strategic move by Musk to enable and be part of hegemony. He is a central player not making decisions in isolation. They are playing a game with different rules, and therefore different unit economics.
Put those three together and maybe it’s possible to push physics to its limits. Faster networking, maybe 4x-5x capacity per unit compared to earth. Servicing is a pain, might be cheaper to just replace the hardware when a node goes bad.
But it mainly makes sense to those who have the capability and can do it cheaply (compared to the rest). There’s only one company that I can think of and that is SpaceX. They are closing in on (or passed) 8,000 satellites. Vertical integration means their cost-base will always be less than any competitor.
This is false, it's hard to cool things in space. Space (vacuum) is a very good insulator.
3 are ways to cool things (lose energy):
- Conduction
- Convection
- Radiation
In space, only radiation works, and it's the least efficient of those 3 options.Do you know the cost of sending up a payload of them?
Do you know how much $$ you need to extract from those payloads to make the cost of sending them up make sense?
Do you know how much they've lied about Starlink revenue and subscription counts?