The viability of space DCs seem to depend entirely on the existence of higher latency, lower throughput, single card to single server workloads, which IMO isn't the primary class of workload today (and if one holds the belief that frontier models will maintain their dominant position, this will remain rack/row/facility level, which isn't viable for orbital facilities to handle as currently designed).
If that workload does expand in usage, and the economics of orbital DCs make sense (with high launch costs, zero ability to do maintenance, and weight penalties for shielding and radiators), then surely ocean based data centers (e.g. Panthalassa) make more sense (lower 'launch' costs, high but existant latency on maintenance, no weight penalties), and if those make sense then surely the distributed land based facilities (e.g. Rune Energy) or home based facilities (e.g. Span XFRA) make even more sense (easier maintenance, low/no permitting overhead because they're covered by existing permits, reasonable quality terrestrial fiber internet).
I don't think it's quite as crazy as you make it sound.
Per unit area, the radiators dissipate a similar amount of energy as the solar panels do, so they'll be a similar size. That's doable.
The upside is there's no land and power generation or grid connection issue; you get solar power 24/7 and the latency to earth is just a few ms.
It would also be possible to fly the individual satellites close together so that you can do high bandwidth optical links between them, i.e. you're not limited to single rack workloads.
A single GB200 rack costs about $3M, transporting its 1.4 tons on a Falcon 9 would cost a similar amount, around $4-5M.
That's got to be too expensive for now, but it's also not in a range that's totally off.
The idea is essentially a bet on land and power continuing to be a big issue and launch costs getting a lot cheaper, which they very well might with Starship. The rest is engineering challenges that are probably solvable.
Low Earth orbit is 90 to 120 minutes, depending on altitude. Even if the satellite is currently close to you, it will soon be on the other side of the world.
Assuming that orbital data centers make sense, it will be an interesting challenge to figure out how to best mitigate that. Short tasks that require little data will be easy, but everything else needs rethinking. Can you tolerate variable latency? Should you duplicate the data across many satellites? Or should the data constantly migrate from satellite to satellite?
That’s at least 20 ms of round trip time, realistically much more. Especially since it’s not geostationary so the distance to the closest ground station will in most cases be much longer than straight down. I would be surprised if the average ping is below 100 ms.
Latency isn't really the issue here (lots of inference serving happening across oceans, which is probably 100-150ms RT), but bandwidth ends up becoming an issue for large scale transfers (mostly model weights, not the actual inference tokens/queries).
Why are you guys rehashing the exact calculations that were made in the article? The article is all about these calculations, and they give ranges of values for each cost to see the the possible price points for everything in order to make it worthwhile.
They take all of the bits you guys are talking about into account in the article. You might disagree with some of the numbers, I guess, but that doesn't seem to be what you guys are talking about.
Finding land/power/space to do a _single_ rack is not hard as ~every office building in America has 200kW of extra power and HVAC water at ~25C. Plus it only costs $4-5k to stick that $5M rack in the back of a truck to drive it there. No permits, easy maintenance, stable network connectivity.
Why use a satellite to do what your IT closet can already do?
And putting things in the ocean is only marginally less idiotic due primarily to sea water being corrosive, wet, and full of creatures that tend to want to grow all over anything.
Microsoft did attempt ocean racks in a sealed container , and it sounded like the 3 year test ended successfully but was not renewed.
I presume the cost of setup, teardown, lead-time, low power density when scaled up, and lack of convenient access for mid-life repairs made it not worth the trouble
Yes, but that apparently proves the point of the poster above you.
The moderate improvements in cooling we're not worth the extra complexity.
So it just didn't make sense to do outside of a single media stunt project.
Data centers in space don't even get the cooling benefit.
The whole idea just doesn't survive modest consideration. Data centers aren't particularly dangerous. There's no forcing function to put them away from where people live. The best location for a data center is going to be a decent sized urban center with existing utilities (power and network) and acceptable land costs.
There are nearly forty of them in Atlanta, people walk by them every day with no idea what they are.
This is the sort of thing that highlights why space is such a dumb place to put a DC.
Everything about storing a DC underwater is easier than it is in space (except corrosion).
It's easier to access, cheaper, easier to cool, easier to get enough power, easier to update.
Heck, doing a second story data-center resolves a ton of issues with today's data-centers (while being a bit more expensive due to additional structural support requirements).
There are so many options for terrestrial DCs to overcome whatever problem space might supposedly solve, and they are all cheaper than the space solution.
Spacecraft in LEO deal with corrosion from - among other causes - atomic oxygen atoms up there, which are eager to cause oxidation in almost anything they come in contact with. The rate of this can be greater than on Earth's surface, though I'm not sure how oceans compare for corrosion.
Overall it's probably true that it's less of an issue than with sea water, but it's not a nonissue either.
how is power easier? solar power in orbit seems like a no-brainer - no losses due to the atmosphere / weather, and with the right orbit you can get sun all the time. Was the ocean DC tide-powered or something? how does that compare?
It looks like ISS can generate 240KW peak from 2500m2 of solar panels, and runs at half of that nominally [0].
A single B200 can consume up to 1200W. A single GB300 system needs 14.5kW. So, you can only have ~15 (incl. satellite's energy needs) GB300 with ISS' solar panels, which is absurd. See [1] for the size of the panels we're talking about.
People dream about Starlink-style close flying systems to aggregate them to a single data center. With these power requirements, it's more like an AI hallucination. If we can build way more efficient chips or run smaller models on modest hardware, etc. We may go there, but needing ISS sized solar panels to run frontier models in space?
Also, don't forget the radiators. You also need those to pump that heat out of these systems.
In space, you have to deploy all the power production you want to consume in the same package that's consuming that power.
On earth, you can build a power plant miles from the DC. And you can use all sorts of different sources of power, tidal, wind, nuclear, solar, fossil fuels. All options. And like the DC, if there's a problem with any of those sources, you can go fix it or replace it with different power sources. If there's a problem with the solar panels in space, your DC is dead.
I think if we manufactured most of the infrastructure and maybe the chips or solar panels in space it might actually make sense. Nobody is talking about that though, I guess Elon is saving the we will make the solar panels in space for a down quarter.
I would not expect that to happen any sooner than 50 years from now. If a flight to the Moon on a Starship really requires a dozen refuelling flights, then we need an even bigger rocket (because the larger a rocket is, the more efficient it gets). Then we need to develop and build automated and teleoperated lunar factories, maybe ship the electronics from Earth for final assembly on the Moon, establish lunar mining and metalworking to build the structures, motors, actuators, solar panels, radiators, find raw materials and process them into heat exchanger fluids and propellants, and launch the satellites at least to Lunar orbit so they can self propel to their destination orbit where latency is not two seconds.
Even with an intelligence explosion guiding the development of all these technologies, it’d take many decades.
How would it make sense? Have you seen how things are manufactured? Why would you want to make something harder and more expensive than it has to be? It's delusional to think that these things will work better in space/on mars than on earth.
Some metallurgy on the Moon might be easier - at least the atmosphere doesn’t want to react with the materials. For other processes, you actually want some of the atmosphere to do exactly that.
We’ll need to develop base industries from scratch for these environments, but, eventually, we’ll get there. The downside is that it’ll take at least multiple decades, if not more than a century.
The article makes several big assumptions that are questionable (experimental cooling can scale, permit issues on land will only become worse, etc). But more importantly, their analysis of the risks of placing a data center in orbit is completely off. They mention that it costs tens of millions to launch a missile to destroy a satellite, and it might be cheaper to simply replace it (based on launch costs). But one strike can hit many targets, those targets may contain hardware costing multiple times the value of their launch, and you will be out of compute while you scramble to launch a replacement. We can also assume that placing big fat valuable targets in space would have many actors wondering how to make it cheaper to destroy them.
For those reasons alone, and because compute will become more and more strategic for defense, I find it very hard to believe that data centers in space will represent even 5% of all compute in the next 10 years. I think a higher proportion will rather be underground or in bunkers than in space.
Space is like the deep ocean, you never really know what's lurking in the shadows. A well placed pebble, a pellet of paint or a beam of microwaves could ruin a fragile satellite without being none the wiser.
If you want a safer datacenter you put it underground and you can safely assume that if it's hit by a bunker buster you have much much bigger problems.
Orbital data centers look more viable than we expected going in and the case rests primarily on launch cost. The engineering challenges seem significant but none appear intractable. Commercial off-the-shelf solar, passive radiators, and standard error-correcting codes seem sufficient for a baseline architecture. We expect limited demos and shipments of compute to orbit in the coming years but our best estimate is that orbital data centers are unlikely to represent a meaningful share of compute before 2030. They might scale quickly thereafter however, and are more likely to be bottlenecked by chip production rates rather than by launch cadence. A generally important dynamic is that terrestrial data centers likely face headwinds that worsen as compute scales up while orbital data centers face constraints that relax with scale.
Anyone who thinks we’ll have data centres in space before we have a civilisation changing step up in surface to orbit space drives and materials science is an idiot.
How the hell does anyone think a rocket is ever going to compete with a prime mover towing two trailers, or an ocean going vessel.
All of this sounds a lot like how ULA, ESA thought reusable rockets were a pipe dream. These are engineering challenges, not physical laws that need to be broken.
yeah, but:
Data centers, fundamentally, are a solved problem. The rest is just cost balancing - and space looks like complete rubbish when it comes to all kinds of costs, including and especially when it comes to maintenance.
This sounds like someone who concludes that FTL is possible because “once people thought that flight was impossible!”
But we have and had ample evidence of flight in nature, we just didn’t understand how it worked. A similar search across the observable universe reveals no sign of FTL behavior.
Sometimes when you understand the science a comparison to some time when people didn’t isn’t as moving or compelling as you seem to imagine.
And all those conclusions take is cost to space to reduce 15 times; new cooling tech that isn't used anywhere to work at scale but only in space (if ground datacenters can use it too, their cost decrease); that ops cost line to not do anything useful so it can be just dropped; those 10mm radiation shields to not weight anything, because they didn't enter the costs; and that ground datacenters to use more expensive tech than they use today (just not improving is not enough).
We will have space datacenters some day. But I'm not holding my breath.
The more I read HN the more I realize that I am more of an empiricist than a rationalist.
So many people here are opining as if they are correct. I have a large amount of doubt and I like the fact that the world is structured so that people can try things out and see if it works.
If the world was run by the naysayers in this thread, almost nothing would ever get done.
I have doubt but I don't believe in the idea of something like this being strictly impossible.
This is turning out just like everything else. We establish a certain conclusion that is preferred and then walk backward into whatever evidence supports that conclusion.
In this case, AI needs to fail for certain groups. It is an apparent loss of identity. The notion of it going to space where it's a lot harder to get at (protest/regulate/etc) must really bother these people. I don't need datacenters in orbit to succeed. Whether they do or do not doesn't bother me very much. The concept is interesting but not essential. The inverse is clearly not true with the counterparties. I really don't understand why you would go on the internet and be a total shitlord regarding how orbital datacenters are impossible and fuck you for even suggesting they might be.
Oh my god, it'd be way easier to, I don't know, build datacenters near Tashkent, Uzbekistan, where [0] is located. Here's your almost unlimited solar power, cheap land, and still better latency and easier cooling.
> it'd be way easier to, I don't know, build datacenters near Tashkent, Uzbekistan, where [0] is located.
To build? Yes, it would be easy. But to not get those datacenters annexed by Russia would be pretty hard. Russia experiences shortages of hardware right now, I'm sure Putin would be happy if we placed some ready-made datacenters nearby so he could grab them.
Even put launch cost aside and assuming the best case for starship, most of the whitepapers I've seen seem to hand-wave the weight, cost, and complexity of cooling. You need large panels with special coating, pumps, and pipes - and that will get hit with micro-meteors so you need redundancy.
Land based or even sea wave-based datacenters are all much more feasible.
I don‘t think both sides are equal though. The vibes on one side is the default assumption that this is a dumb idea and is never going to work. The vibes on the other side goes against traditional wisdom but it will somehow work anyway.
Vibing with the default assumption is cheap, and it should be cheap. While we have no reason to believe otherwise there is no reason to go looking for a lot of evidence for why we are right. Everybody believes so, so why bother convincing us right? Vibes are enough until proven otherwise.
Vibing against the default assumption is expensive, and it should be expensive. Extraordinary claims require extraordinary evidence and all that. Your job is to prove that you are right, that the traditional wisdom is wrong, and that everybody has to change their mind. Vibing in this direction is not nearly enough. In fact, the vibes in this direction are worth nothing.
The only use case for datacenter in space is to have high communication speed with satellites that have low compute. Though that does beg the question of why not just add more compute to those satellites going forward.
Perhaps companies should consider putting their data centers underground (out of sight and out of mind). Space seems incredibly expensive and risky in comparison.
Oh I'm sorry, are you not even thinking of the lesser American ground worm? We're going to need a 15 year weevil analysis just to even think about digging a one metre borehole to test whether the worms will be affected.
Outside cost of launch I have understood that scaling it to sufficient power envelope with needed cooling is biggest step which most likely makes it unreasonable at scale. That is being able to use enough power to get reasonable amount of compute. Scale it small and low power enough and it is pretty simple. But at that point you might not be talking about amount of compute you expect.
Especially fun when you consider how often gigawatts are talked about with AI. And what that would mean just for cooling as in area... Or then dividing that to sufficient number of individual satellites...
Wouldn't it be cheaper to build a nuclear plant to power data centers than to launch the data centers into space? Nuclear is expensive, but rockets, orbital infrastructure, and giant cooling radiators aren't exactly budget options.
That’s just not true. For the same cost as a nuclear plant of some rated power, you can build far more solar, wind, and batteries.
The pro-nuclear propaganda is suspiciously associated with people with a history of also wanting to build nuclear weapons. The Liberal Party of Australia is an example — they seem very interested in overpriced nuclear power and obtaining nuclear powered subs capable of launching nuclear tipped missiles.
For this article, they put their thumb on the scale. It's just nonsense conclusions from bad assumptions.
A 100kW Earth based solar array and a 500kWh grid scale battery that will last ~20 years is already a tiny fraction of the cost of even one Starlink satellite. It doesn't matter even if it was free to put in orbit, as only lasts five years once it's up there.
Their radiator assumptions are also laughable. 13x-28x better than ISS with carbon wizardry entirely on paper. TRL level what exactly?
Google's paper on their Suncatcher project (where they recently put their TPU in orbit and received a response from a prompt) is pretty interesting to read:
Outside of the physical, technical, and maintenance challenges of putting data centers in space, I simply don’t want more trash littering the night sky.
The discussion of cooling in this piece is more convincing than I would have expected, but there remains a glaring social problem:
Should we consider orbit to be outside of terrestrial tax jurisdiction?
The majority of the cost of putting these data centres into orbit would be borne by terrestrial jurisdictions. The places that built institutions, funded education, provide medical care, etc. for the people who built these things. The environmental impact of building these data centres and launching them will be paid by those living on Earth's surface.
On Earth, we often fail to recoup many of these costs because of out-sourcing, but at least some other jurisdiction benefits. If a poor country is lifted up by the benefits of out-sourcing, at least some good has been done. If we let billionaires simply blast off and evade taxes, that's a net loss to the societies that made this possible with no benefit to anyone but the billionaires. That's fundamentally wrong.
Space may belong to no one nation, but neither is it true that space belongs to no one at all. The space around Earth belongs to all of humanity. Orbital data centres should pay terrestrial taxes. It's not clear where those taxes should go, but we need to hammer out the laws to make sure they are paid.
What happens to all the rare earth materials once the space DC have been decommissioned? Isn't it a horrible waste of the already hard to come-by resources to send it to a disposable trash heap in space, all in the name of forever chasing profits?
All these wild plans are pretty funny given hyperscalers won't even build "in the middle of nowhere" in the US but rather they put their sites in the edge of exist industrial areas with the aim to repurpose every bit of industrial infrastructure plus aiming to get all the incentives these small town (corruptly imo) offer industries for relocation.
I mean, the US could have a rational process of building new energy infrastructure (which should be only solar, wind and nuclear imo) but the actual process is opportunistic and haphazard. So it's not strange to see the NIMBYs biting them. It's like "well if you hadn't been a d*k, this might not be happening to you..."
How feasible would that be? Don't we have to bring down some satellites from the orbit at some point? How would that work if space datacenter has a limited time use?
Isn’t “space grade” a result of high space launch costs? If it becomes cheap enough to launch hardware into space, it doesn’t need to last as long to meet ROI.
For people selling the anti-sabotage / anti-datacenter-terrorism angle: the math doesn't work.
The energy needed to achieve orbital altitude is vastly less than the energy needed to reach orbit. About 1kg of rocket fuel will carry a small explosive and some tiny steel spheres to LEO altitude. The projectile will not be in orbit; it won't have escape velocity -- in fact it will briefly hit zero velocity (wrt the earth) at orbital altitude and then fall back to earth.
The very large cloud of shrapnel created will encompass LEO orbital height for several seconds. If an orbiting datacenter passes through the cloud during those seconds, the shrapnel will be moving at orbital velocity wrt the datacenter. In other words, just one piece will take it out.
With relatively simple rocketry (and, these days, an abliterated LLM to aid you) a dedicated luddite would have several opportunities per day to hit one of a constellation of satellites. If they wanted one very specific satellite they'd have one or two chances a year. The targeting and steering is not hard -- the orbital paths are publicly known and easily observable; you can use RF time-of-flight to implement exceptionally precise position/altitude/velocity measurements from the rocket (no need for GPS -- use your own transmitters on the ground). This is not hard. And it will only get easier as AI advances.
*I strongly discourage anybody reading this from doing the above.*
I'm posting it in order to show people that anti-sabotage is absolutely not a selling point for orbital datacenters. If anything, they are far more vulnerable to luddite attacks. The people channelling billions into this tech know this too. So their motive is something else. I don't know what, but it isn't this. The math doesn't work.
Even if they aren't financially viable, I think they're attractive because companies want to keep certain kinds of data/processing on everyone/everything someplace where there's a lower chance of government regulation (gdpr/etc).
If this is the case, I wouldn't be surprised to see them eventually support these kinds of regulations because it would enhance their moat (now earth orbit in addition to large capex for equipment and a sufficiently large user base).
In that case, international water is far easier. There already are server farms on old oil rigs.
But, as these things go, regulation isn't that much about where you put the computers, but about where you register your company, and where your customers are living.
It's easier, but it's riskier too. Someone can come along a take your stuff a lot easier at sea than they can in orbit. And weather is always tricky, although launches aren't easy either.
In terms of regulation, it depends on the specifics. Sometimes it depends on where the customer lives, but even if it does, when a platform has no business operations in a region and there aren't international treaties between the consumer country and where they're operating from to resolve that sort of thing, the most the regulating country can do is block them.
It seems to be in certain cases. For example Meta is subject to GDPR because they have EU headquarters, and were subject to fines because they transferred EU data to US servers.
If they have headquarters outside the EU, and only use data they've already acquired/processed to make money off of EU consumers from outside the EU, GDPR shouldn't be an issue for them in the same way.
And GDPR makes it harder for their competitors to catch up because it's harder for those competitors to get data.
Probably, but not in the same way. I'm guessing they wouldn't have been fined for external data transfers if their servers were located outside the EU.
It could be that the EU would block them if they wanted to host fb/etc... including ads, and didn't have a presence in the EU, but I'm not sure if that's the case.
That's not accurate. You can build/do/etc things in some places and not in others because of differences in local law/regulation.
Where you put something, like a data center, can be relevant. If Meta didn't have a business presence in the EU, and operated their data centers outside of the EU, they wouldn't have been subject to GDPR in the same way they were.
I couldn't find a single instance of "upgrade", "recover", or "recycle" in there.
What happens if a major advancement renders all the GPUs obsolete? How can they upgrade the hardware and resell the old hardware?
Are they just going to vaporize all the gear and precious metals?
IDK, the whole thing seems silly at the current state of space logistics.. For these and many more reasons.
It's popular mechanics/science levels of practical fantasy, but somehow now those stories end up swirling around the Internet and never flush. Would have been like having 5 newspaper articles about space elevators every month for years.
I fail to see the economic case. You will need very large solar arrays and radiation hardened components, both of which make it uneconomical compared to a data center on Earth. This is another of Musk's genius ideas like the tunnels that were supposed to be carrying us under cities by now.
The problem is that experiments into this by Google have concluded that radiation hardening is unnecessary. HBM memory far surpassed their expectations for survival, exceeding 5 years in their simulated proton beam on earth. Localized shielding for consumer equipment is more than enough, especially combined with ECC.
Solar power and power dissipation are similarly solvable. The only bottleneck currently is launch costs which is why it’s expected around 2030 is when these DCs will explode
> Solar power and power dissipation are similarly solvable
So it is not outright stupid, just has a few extra costs.
So again, what's the economical benefit? Because you surely have no power and dissipation problem down on Earth, so why add a whole rocket launch into the equation? Because big fire goes brr?
The PV at least can potentially be much less robust than on Earth, since it doesn't have to contend with wind and weather. CdTe thin films, for example, can be just 1 micron thick (the absorption length of solar photons in that material is ~0.1 microns). So potentially the material cost of space PV can be much less than the same peak power PV on Earth.
The bet that they are making is that, in the near-future, energy on Earth will be such an issue (especially if you want to scale) that DCs on Earth will cost more than components + launch costs
Why would energy be an issue? The Sun is shining, wind is blowing, the ocean has waves, nuclear is nuclear. All of these will just get more efficient in using the same constant resource.
It's difficult to imagine any sort of compute in space given the state of the art rad hardened processor is about as fast as a computer Apple sold in 1998.
You don’t have to imagine. The starlink constellation is a working and profitable massive deployment of orbital compute. (It doesn’t use rad hardened hardware as far as I can tell)
Starlink satellites have such a low orbit that they experience non-negligible atmospheric drag, which is why they have to be replaced at such a rapid rate. All proposals for orbital datacenters propose placing them in sun-synchronous orbits for maximum solar exposure, which puts them right at the edge of the Van Allen Belt at best, meaning they'd be exposed to dramatically more radiation than a Starlink satellite.
Musk's "genius idea" there was to (pardon the pun) derail high speed rail and other similar investments so he could sell more cars.
What are Hyperloop/Boring actually doing now? It doesn't take that much investment to human-drive a 10-20 Teslas up and down the Vegas Strip in a straight tunnel at normal speeds.
> and becomes cheaper than any current terrestrial energy source at around $50/kg
Heck for $50/kg I could see it, too. Now with Falcon 9 or Heavy it's on the order of thousands, isn't it? Just need a measly two orders of magnitude improvement and then we can start talking.
Does gauging this based on Falcon 9 launch costs make sense now? SpaceX has already deployed two payloads of satellites using Starship. Falcon 9 is a deprecated system because SpaceX intends to use Starship for future payload missions.
We can't yet know the ultimate cost of Starship payloads. All we have is SpaceX projections of $10-100/kg. Maybe that is realized, maybe not. Given the scale of the LA launch site they're building, they intend to have the launch cadence to deliver lower launch costs than anything achieved with Falcon 9 or any other launch system ever created.
Underlying this seems to be the assumption that those pesky governments, pushed around by public opinion, will keep on blocking or delaying terrestrial data centres, but space-based ones will be fine.
The idea that launch rates could increase to the rates required to launch gigawatts of computing into space, without attracting a huge backlash, is fantasy. And such a backlash would be entirely justified - the only reason why the environmental effects of space launch aren't a focus of environmentalists is because there aren't many of them.
Just skipping to the Conclusions section, I find this article to be wildly biased toward space data centers. Some of this reasoning is ridiculous, insane.
For one thing, I think there's a huge assumption that we just can't make energy cheap on earth (Cost per GW). Solar/battery costs are plummeting, nuclear energy is still a great option (see: the quantity of in-progress Chinese nuclear power plant builds in progress, I think around 30 projects under construction right now), and we might just figure out fusion reactors one day soon.
"Favors: Earth today; Space later" - seems way too confident.
Land Use: Does this article just not realize how large the earth is? The earth will never run out of space for data centers. You can even put them in skyscrapers if you want.
Long-run scaling: How does this row make a lick of sense? This is a lot like saying that if I build my house in space then I could have 100 parking spots in my driveway. The problem is, I don't need 100 parking spaces in my driveway. Like, yeah, we know that there's infinite room to put stuff in space.
Carbon / local environmental burden: Another insane take on this subject. Are we really going to hand-wave over burning a shitload of rocket fuel to put stuff in space? "Favorable vs gas-backed terrestrial buildout" why are we comparing against a gas-backed buildout?
No we will not. The reason is simple: Space is not in the cloud, it is above. Anything that's not in the cloud is bad.
Silliness aside, has anyone come up with a cooling system that would work well for a friggin data center in space? Or is it still all just a wishful thinking game, ignoring physics?
The plan as I understand it is to pack it tightly with TPUs with liquid cooling out to radiators and flying a tight cluster of these next to each other using high bandwidth laser links. That lets you balance the amount of thermal radiation you need to emit vs the amount of TPUs in a single satellite. Thus it’s not a DC in a single satellite but more like a rack and then a bundle of racks flying beside each other.
Or you know, just have tpus with liquid cooling in huge racks, with optical cables or whatever. We can put a shitton of them into a warehouse sized space, and we have basically infinite energy to power and cool it, with physical connection to the internet for ultra fast speeds!
The problem is that building the physical DCs is a cost problem not just in terms of energy availability (it’s not infinite) but also politics and legal regulations. Ignoring that is why it makes the space idea seem silly but it’s a real one
> Silliness aside, has anyone come up with a cooling system that would work well for a friggin data center in space? Or is it still all just a wishful thinking game, ignoring physics
That cooling is already perfectly viable.
However at the temperatures chips operate at, because stuff on the ground is getting convective cooling as well as radiative, a radiator designed for LEO will reject slightly more heat on the ground than it would in orbit.
It is absolutely a wishful thinking game, ignoring physics. It's a bet made by investors who don't know or care about the science, and are looking for a modal outcome of a failed company purchased by SpaceX or by an even dumber company looking to compete with SpaceX.
I recommend reading up on the results of Google’s project suncatcher. IMHO the results have provided a roadmap to space DCs with the only blocker being launch cost which are targeted to become economically viable in 2030s; sooner for spacex because they can lower their margins to zero or even negative for the launch itself.
Whenever I asked someone to dig down into "against the laws of physics" arguments, they turned out to instead rest on a pile of tacit and pessimistic technological assumptions.
Perhaps you could explain in detail why you think the laws of physics have to be violated for space data centers to work.
As a Linux user, I can already build such a system quite trivially by installing Debian on a Raspberry Pi, attaching it to a weather balloon, and setting up a cron job to maintain orbit. The cooling is handled by opening a window. From Windows or Mac, the balloon could be accessed through built-in software.
So if launch costs drop 15x and radiator technology improves 12 to 30x and chip lose is improved 2x and there are absolutely no innovations that make terrestrial deployment cheaper, it's totally slightly cheaper!
I wouldn't put it past him to launch 20,000 of them before moving on to the next thing.
Or two. It's a broad range, but with his personality and prior history both are possible before he announces whatever the next pivot is.
Just, not a million. There's better things to do with the capacity to put a million up there, and better places to put a million (last I checked) 120 kW compute racks and associated PV systems.
In space solar energy is available all the time. That is not the case in those locations, and the storage needs to get around that are the worst on Earth.
You don't have to pack your energy generator into the same package where your compute would go, for laughably little energy. You can have a single cable go between a huge power plant and the data center.
Even if data centres in space become cost-effective, what's the point? The only non-contrived answer that I can think of is to avoid NIMBYism, but pushback against data centres can't be that bad.
The only place where a space data center makes sense is in Mars orbit to reduce the amount of mass that needs to be landed. Every thing else is a smoke and mirrors game to justify rolling a bunch of failed businesses into SpaceX.
Even then. If you're building a million satellites, they only need to be about 42 meters long for you to make them into a contiguous ring in LEO. Smaller (22.5m) or fewer (537,000) for the same in a 200 km orbit around Mars.
Do that, and you can quickly bootstrap an orbital ring and make rockets obsolete from surface to orbit of either body.
In the case of Mars, the planet-wide dust storms make
it a good idea to locate your PV off-planet, and an orbital ring is also very convenient for that.
Because it's held up by being in orbit (strictly: a bit faster than that), you can by the same process as a maglev hang a stationary-with-respect-to-ground cable off it; and because the cable is only as long as your orbit is high (~200 km or whatever) rather than the 36,000 km you'd need for a full space elevator, we don't need to invent any supermaterials for it.
Structually, it barely matters. The physics works even if each element isn't joined contiguously to the next, it's just that if you do make it contiguous you can also use it for power delivery and worry less about elements drifting out of alignment.
What specifically do you disagree with from the article to say its Smoke and Mirrors? I think their analysis is reasonable and that it will reduce to a launch cost problem.
Giving it a quick look they're doing the same thing as everyone else and handwaving away really difficult engineering problems with some back of the envelope math that only looks at one or two aspects.
Like their section on collision avoidance isn't any more sophisticated than "Starlink has a great record so far" when we're talking about a very different scale of how much mass and surface area has to be moved around in a coordinated way.
This is in effect an editorial from two postgrads, not an actual detailed workup. I don't feel like spending time going into any more effort based on that.
Saying it is equivalent to launch costs is hand waving. Their "median" scenario would involve Starship working at the lower bound of their operational goal for it.
The real main purpose of putting data centers in space is not economic, and it’s not to guard against threats from major powers who could easily destroy them. It’s so that the ultra rich can have data centers that the rest of us cannot destroy. I make no argument about whether we should be thinking of doing so or not, but it is interesting to think of the implications is it not? And it certainly fits the personalities involved and their public statements.
The viability of space DCs seem to depend entirely on the existence of higher latency, lower throughput, single card to single server workloads, which IMO isn't the primary class of workload today (and if one holds the belief that frontier models will maintain their dominant position, this will remain rack/row/facility level, which isn't viable for orbital facilities to handle as currently designed).
If that workload does expand in usage, and the economics of orbital DCs make sense (with high launch costs, zero ability to do maintenance, and weight penalties for shielding and radiators), then surely ocean based data centers (e.g. Panthalassa) make more sense (lower 'launch' costs, high but existant latency on maintenance, no weight penalties), and if those make sense then surely the distributed land based facilities (e.g. Rune Energy) or home based facilities (e.g. Span XFRA) make even more sense (easier maintenance, low/no permitting overhead because they're covered by existing permits, reasonable quality terrestrial fiber internet).
I don't think it's quite as crazy as you make it sound.
Per unit area, the radiators dissipate a similar amount of energy as the solar panels do, so they'll be a similar size. That's doable.
The upside is there's no land and power generation or grid connection issue; you get solar power 24/7 and the latency to earth is just a few ms.
It would also be possible to fly the individual satellites close together so that you can do high bandwidth optical links between them, i.e. you're not limited to single rack workloads.
A single GB200 rack costs about $3M, transporting its 1.4 tons on a Falcon 9 would cost a similar amount, around $4-5M.
That's got to be too expensive for now, but it's also not in a range that's totally off.
The idea is essentially a bet on land and power continuing to be a big issue and launch costs getting a lot cheaper, which they very well might with Starship. The rest is engineering challenges that are probably solvable.
>you get solar power 24/7 and the latency to earth is just a few ms.
Not at the same time. To have few-ms latency you need to be in/below LEO, but then earth blocks the sun half the time.
It can be done in a dawn-dusk, sun synchronous orbit, less than 1000 km up.
Low Earth orbit is 90 to 120 minutes, depending on altitude. Even if the satellite is currently close to you, it will soon be on the other side of the world.
Assuming that orbital data centers make sense, it will be an interesting challenge to figure out how to best mitigate that. Short tasks that require little data will be easy, but everything else needs rethinking. Can you tolerate variable latency? Should you duplicate the data across many satellites? Or should the data constantly migrate from satellite to satellite?
That’s at least 20 ms of round trip time, realistically much more. Especially since it’s not geostationary so the distance to the closest ground station will in most cases be much longer than straight down. I would be surprised if the average ping is below 100 ms.
Probably still fine for most applications though.
Latency isn't really the issue here (lots of inference serving happening across oceans, which is probably 100-150ms RT), but bandwidth ends up becoming an issue for large scale transfers (mostly model weights, not the actual inference tokens/queries).
With laser terminals you’ll get a few gigabits per second. That’s possible both between sats and to the ground station. Probably not cheap though.
Why are you ignoring all of the costs of actually being in orbit here?
A normal satellite runs something similar to 1U worth of computer and costs roughly what you quoted in total...
Why are you guys rehashing the exact calculations that were made in the article? The article is all about these calculations, and they give ranges of values for each cost to see the the possible price points for everything in order to make it worthwhile.
They take all of the bits you guys are talking about into account in the article. You might disagree with some of the numbers, I guess, but that doesn't seem to be what you guys are talking about.
Finding land/power/space to do a _single_ rack is not hard as ~every office building in America has 200kW of extra power and HVAC water at ~25C. Plus it only costs $4-5k to stick that $5M rack in the back of a truck to drive it there. No permits, easy maintenance, stable network connectivity.
Why use a satellite to do what your IT closet can already do?
People are fudging orders of magnitude because it sounds cool and they are stupid.
Starmind is already one rack.
I don't think we need that many words. It's an obviously preposterously stupid idea that has been promoted only to defraud investors.
Unfortunately most people even here don’t have a working knowledge of the relevant tech, and seem to lack a similar grasp of the thermodynamics.
A non-trivial number still think that the hard part of colonizing Mars is the rocket to get there. What can you do?
Correct.
And putting things in the ocean is only marginally less idiotic due primarily to sea water being corrosive, wet, and full of creatures that tend to want to grow all over anything.
Microsoft did attempt ocean racks in a sealed container , and it sounded like the 3 year test ended successfully but was not renewed.
I presume the cost of setup, teardown, lead-time, low power density when scaled up, and lack of convenient access for mid-life repairs made it not worth the trouble
https://www.microsoft.com/en-us/research/project/natick/
Yes, but that apparently proves the point of the poster above you.
The moderate improvements in cooling we're not worth the extra complexity.
So it just didn't make sense to do outside of a single media stunt project.
Data centers in space don't even get the cooling benefit.
The whole idea just doesn't survive modest consideration. Data centers aren't particularly dangerous. There's no forcing function to put them away from where people live. The best location for a data center is going to be a decent sized urban center with existing utilities (power and network) and acceptable land costs.
There are nearly forty of them in Atlanta, people walk by them every day with no idea what they are.
This is the sort of thing that highlights why space is such a dumb place to put a DC.
Everything about storing a DC underwater is easier than it is in space (except corrosion).
It's easier to access, cheaper, easier to cool, easier to get enough power, easier to update.
Heck, doing a second story data-center resolves a ton of issues with today's data-centers (while being a bit more expensive due to additional structural support requirements).
There are so many options for terrestrial DCs to overcome whatever problem space might supposedly solve, and they are all cheaper than the space solution.
> except corrosion
Spacecraft in LEO deal with corrosion from - among other causes - atomic oxygen atoms up there, which are eager to cause oxidation in almost anything they come in contact with. The rate of this can be greater than on Earth's surface, though I'm not sure how oceans compare for corrosion.
Overall it's probably true that it's less of an issue than with sea water, but it's not a nonissue either.
https://en.wikipedia.org/wiki/Corrosion_in_space
how is power easier? solar power in orbit seems like a no-brainer - no losses due to the atmosphere / weather, and with the right orbit you can get sun all the time. Was the ocean DC tide-powered or something? how does that compare?
It looks like ISS can generate 240KW peak from 2500m2 of solar panels, and runs at half of that nominally [0].
A single B200 can consume up to 1200W. A single GB300 system needs 14.5kW. So, you can only have ~15 (incl. satellite's energy needs) GB300 with ISS' solar panels, which is absurd. See [1] for the size of the panels we're talking about.
People dream about Starlink-style close flying systems to aggregate them to a single data center. With these power requirements, it's more like an AI hallucination. If we can build way more efficient chips or run smaller models on modest hardware, etc. We may go there, but needing ISS sized solar panels to run frontier models in space?
Also, don't forget the radiators. You also need those to pump that heat out of these systems.
The outlook doesn't look good at this point.
[0]: https://en.wikipedia.org/wiki/Solar_panels_on_spacecraft#Spa...
[1]: https://www.sgmk.edu.pl/wp-content/uploads/2025/10/iss.png
In space, you have to deploy all the power production you want to consume in the same package that's consuming that power.
On earth, you can build a power plant miles from the DC. And you can use all sorts of different sources of power, tidal, wind, nuclear, solar, fossil fuels. All options. And like the DC, if there's a problem with any of those sources, you can go fix it or replace it with different power sources. If there's a problem with the solar panels in space, your DC is dead.
That's both easier and a lot more flexible.
In Microsoft’s case, they ran a cable, which is much simpler to do than putting solar panels in orbit.
How is Panthalassa doing? Anyone know?
That being said. If I were an AI, I would lobby for datacenters in space as hard as I can to put the off switch comfortably out of reach of humans.
The benefit is that they will fall down in 5-10 years.
That's an automatic off switch.
That's an eternity for an exponentially growing intelligence.
Exponentially growing on the same hardware? That's... optimistic. Not, IMO, realistic.
I think if we manufactured most of the infrastructure and maybe the chips or solar panels in space it might actually make sense. Nobody is talking about that though, I guess Elon is saving the we will make the solar panels in space for a down quarter.
I would not expect that to happen any sooner than 50 years from now. If a flight to the Moon on a Starship really requires a dozen refuelling flights, then we need an even bigger rocket (because the larger a rocket is, the more efficient it gets). Then we need to develop and build automated and teleoperated lunar factories, maybe ship the electronics from Earth for final assembly on the Moon, establish lunar mining and metalworking to build the structures, motors, actuators, solar panels, radiators, find raw materials and process them into heat exchanger fluids and propellants, and launch the satellites at least to Lunar orbit so they can self propel to their destination orbit where latency is not two seconds.
Even with an intelligence explosion guiding the development of all these technologies, it’d take many decades.
How would it make sense? Have you seen how things are manufactured? Why would you want to make something harder and more expensive than it has to be? It's delusional to think that these things will work better in space/on mars than on earth.
Some metallurgy on the Moon might be easier - at least the atmosphere doesn’t want to react with the materials. For other processes, you actually want some of the atmosphere to do exactly that.
We’ll need to develop base industries from scratch for these environments, but, eventually, we’ll get there. The downside is that it’ll take at least multiple decades, if not more than a century.
The article makes several big assumptions that are questionable (experimental cooling can scale, permit issues on land will only become worse, etc). But more importantly, their analysis of the risks of placing a data center in orbit is completely off. They mention that it costs tens of millions to launch a missile to destroy a satellite, and it might be cheaper to simply replace it (based on launch costs). But one strike can hit many targets, those targets may contain hardware costing multiple times the value of their launch, and you will be out of compute while you scramble to launch a replacement. We can also assume that placing big fat valuable targets in space would have many actors wondering how to make it cheaper to destroy them.
For those reasons alone, and because compute will become more and more strategic for defense, I find it very hard to believe that data centers in space will represent even 5% of all compute in the next 10 years. I think a higher proportion will rather be underground or in bunkers than in space.
Space is like the deep ocean, you never really know what's lurking in the shadows. A well placed pebble, a pellet of paint or a beam of microwaves could ruin a fragile satellite without being none the wiser.
If you want a safer datacenter you put it underground and you can safely assume that if it's hit by a bunker buster you have much much bigger problems.
> We can also assume that placing big fat valuable targets in space would have many actors wondering how to make it cheaper to destroy them.
Yup. See True Anomaly, among others.
you just need one missile, maybe 2-3.
space debris will ruin the rest.
Defense compute can move to bunkers, but that is probably small percentage of overall need
From the article:
Conclusions
Orbital data centers look more viable than we expected going in and the case rests primarily on launch cost. The engineering challenges seem significant but none appear intractable. Commercial off-the-shelf solar, passive radiators, and standard error-correcting codes seem sufficient for a baseline architecture. We expect limited demos and shipments of compute to orbit in the coming years but our best estimate is that orbital data centers are unlikely to represent a meaningful share of compute before 2030. They might scale quickly thereafter however, and are more likely to be bottlenecked by chip production rates rather than by launch cadence. A generally important dynamic is that terrestrial data centers likely face headwinds that worsen as compute scales up while orbital data centers face constraints that relax with scale.
This is complete bullshit.
Anyone who thinks we’ll have data centres in space before we have a civilisation changing step up in surface to orbit space drives and materials science is an idiot.
How the hell does anyone think a rocket is ever going to compete with a prime mover towing two trailers, or an ocean going vessel.
We’re not exactly short of land.
And no, space is not cold.
All of this sounds a lot like how ULA, ESA thought reusable rockets were a pipe dream. These are engineering challenges, not physical laws that need to be broken.
yeah, but: Data centers, fundamentally, are a solved problem. The rest is just cost balancing - and space looks like complete rubbish when it comes to all kinds of costs, including and especially when it comes to maintenance.
This sounds like someone who concludes that FTL is possible because “once people thought that flight was impossible!”
But we have and had ample evidence of flight in nature, we just didn’t understand how it worked. A similar search across the observable universe reveals no sign of FTL behavior.
Sometimes when you understand the science a comparison to some time when people didn’t isn’t as moving or compelling as you seem to imagine.
And all those conclusions take is cost to space to reduce 15 times; new cooling tech that isn't used anywhere to work at scale but only in space (if ground datacenters can use it too, their cost decrease); that ops cost line to not do anything useful so it can be just dropped; those 10mm radiation shields to not weight anything, because they didn't enter the costs; and that ground datacenters to use more expensive tech than they use today (just not improving is not enough).
We will have space datacenters some day. But I'm not holding my breath.
We can just repurpose Deep Space 9, the Cardassians aren’t using it and nobody really cares about Bajor.
The more I read HN the more I realize that I am more of an empiricist than a rationalist.
So many people here are opining as if they are correct. I have a large amount of doubt and I like the fact that the world is structured so that people can try things out and see if it works.
If the world was run by the naysayers in this thread, almost nothing would ever get done.
I have doubt but I don't believe in the idea of something like this being strictly impossible.
This is turning out just like everything else. We establish a certain conclusion that is preferred and then walk backward into whatever evidence supports that conclusion.
In this case, AI needs to fail for certain groups. It is an apparent loss of identity. The notion of it going to space where it's a lot harder to get at (protest/regulate/etc) must really bother these people. I don't need datacenters in orbit to succeed. Whether they do or do not doesn't bother me very much. The concept is interesting but not essential. The inverse is clearly not true with the counterparties. I really don't understand why you would go on the internet and be a total shitlord regarding how orbital datacenters are impossible and fuck you for even suggesting they might be.
Oh my god, it'd be way easier to, I don't know, build datacenters near Tashkent, Uzbekistan, where [0] is located. Here's your almost unlimited solar power, cheap land, and still better latency and easier cooling.
[0] https://en.wikipedia.org/wiki/Solar_furnace_of_Uzbekistan
> it'd be way easier to, I don't know, build datacenters near Tashkent, Uzbekistan, where [0] is located.
To build? Yes, it would be easy. But to not get those datacenters annexed by Russia would be pretty hard. Russia experiences shortages of hardware right now, I'm sure Putin would be happy if we placed some ready-made datacenters nearby so he could grab them.
This premise always struck me as a solution in search of a problem, which when implemented introduces even more problems.
Even put launch cost aside and assuming the best case for starship, most of the whitepapers I've seen seem to hand-wave the weight, cost, and complexity of cooling. You need large panels with special coating, pumps, and pipes - and that will get hit with micro-meteors so you need redundancy.
Land based or even sea wave-based datacenters are all much more feasible.
Remarkably few comments are engaging with the actual assumptions, arguments, and calculations in the article. Beware vibes, in both directions!
I don‘t think both sides are equal though. The vibes on one side is the default assumption that this is a dumb idea and is never going to work. The vibes on the other side goes against traditional wisdom but it will somehow work anyway.
Vibing with the default assumption is cheap, and it should be cheap. While we have no reason to believe otherwise there is no reason to go looking for a lot of evidence for why we are right. Everybody believes so, so why bother convincing us right? Vibes are enough until proven otherwise.
Vibing against the default assumption is expensive, and it should be expensive. Extraordinary claims require extraordinary evidence and all that. Your job is to prove that you are right, that the traditional wisdom is wrong, and that everybody has to change their mind. Vibing in this direction is not nearly enough. In fact, the vibes in this direction are worth nothing.
i heard someone say the same thing about thranos even after it was looking clear that they were shady
You can put computers in space, but can you really call it a data centre at the scale that it can be done at?
Surely a single AWS/GCP/Azure/etc DC on earth is so big we won't be getting close to a similar amount of compute in space for the foreseeable?
The economic case also needs to be explained. What are people going to do with a DC in space?
The only use case for datacenter in space is to have high communication speed with satellites that have low compute. Though that does beg the question of why not just add more compute to those satellites going forward.
Perhaps companies should consider putting their data centers underground (out of sight and out of mind). Space seems incredibly expensive and risky in comparison.
I’m thinking semi-mobile datacenters near active volcanos powered by the heat of the lava/magna flows would be “cool”
Or maybe we should just stop trying to centralize compute in the first place...
Oh I'm sorry, are you not even thinking of the lesser American ground worm? We're going to need a 15 year weevil analysis just to even think about digging a one metre borehole to test whether the worms will be affected.
Outside cost of launch I have understood that scaling it to sufficient power envelope with needed cooling is biggest step which most likely makes it unreasonable at scale. That is being able to use enough power to get reasonable amount of compute. Scale it small and low power enough and it is pretty simple. But at that point you might not be talking about amount of compute you expect.
Especially fun when you consider how often gigawatts are talked about with AI. And what that would mean just for cooling as in area... Or then dividing that to sufficient number of individual satellites...
Wouldn't it be cheaper to build a nuclear plant to power data centers than to launch the data centers into space? Nuclear is expensive, but rockets, orbital infrastructure, and giant cooling radiators aren't exactly budget options.
Aalo Atomic is working with datacenter maker Crusoe.
There isn’t really any other option than nuclear.
Solar & wind can’t power industry.
That’s just not true. For the same cost as a nuclear plant of some rated power, you can build far more solar, wind, and batteries.
The pro-nuclear propaganda is suspiciously associated with people with a history of also wanting to build nuclear weapons. The Liberal Party of Australia is an example — they seem very interested in overpriced nuclear power and obtaining nuclear powered subs capable of launching nuclear tipped missiles.
Cost comparison: https://chatgpt.com/share/6af8ce92-bd10-83ec-88be-1f61051091...
Well named account. I assume it's a parody account.
Nuclear needs to get down to a capex of maybe $3/W to have a chance. This is a very big ask.
For this article, they put their thumb on the scale. It's just nonsense conclusions from bad assumptions.
A 100kW Earth based solar array and a 500kWh grid scale battery that will last ~20 years is already a tiny fraction of the cost of even one Starlink satellite. It doesn't matter even if it was free to put in orbit, as only lasts five years once it's up there.
Their radiator assumptions are also laughable. 13x-28x better than ISS with carbon wizardry entirely on paper. TRL level what exactly?
Google's paper on their Suncatcher project (where they recently put their TPU in orbit and received a response from a prompt) is pretty interesting to read:
https://www.cell.com/joule/fulltext/S2542-4351(26)00362-4
Their blogpost is more "marketing-ish" but the videos are worth a watch too:
https://blog.google/innovation-and-ai/models-and-research/go...
We barely even have functioning trains or planes.
There's a lot of long shot assumptions here, but I feel the biggest is $250/kg launch costs (let alone $100 or $50).
The long shot assumptions are only really applied in one direction as well, never really in favour of earth DCs.
Outside of the physical, technical, and maintenance challenges of putting data centers in space, I simply don’t want more trash littering the night sky.
https://youtu.be/yZpbJJLk1ug
The discussion of cooling in this piece is more convincing than I would have expected, but there remains a glaring social problem:
Should we consider orbit to be outside of terrestrial tax jurisdiction?
The majority of the cost of putting these data centres into orbit would be borne by terrestrial jurisdictions. The places that built institutions, funded education, provide medical care, etc. for the people who built these things. The environmental impact of building these data centres and launching them will be paid by those living on Earth's surface.
On Earth, we often fail to recoup many of these costs because of out-sourcing, but at least some other jurisdiction benefits. If a poor country is lifted up by the benefits of out-sourcing, at least some good has been done. If we let billionaires simply blast off and evade taxes, that's a net loss to the societies that made this possible with no benefit to anyone but the billionaires. That's fundamentally wrong.
Space may belong to no one nation, but neither is it true that space belongs to no one at all. The space around Earth belongs to all of humanity. Orbital data centres should pay terrestrial taxes. It's not clear where those taxes should go, but we need to hammer out the laws to make sure they are paid.
You can tax the owners on the profits. (Or the revenue, absent profit, if you want to be punitive.)
What happens to all the rare earth materials once the space DC have been decommissioned? Isn't it a horrible waste of the already hard to come-by resources to send it to a disposable trash heap in space, all in the name of forever chasing profits?
What rare earth materials? Are you referring to Shellenberger's BS that PV uses rare earths?
Wouldn't it be far cheaper to put DCs in offshore windmill parks? You have the power, cooling water abundant, fewer "NIMBY" than onshore DCs?
Is this already done?
All these wild plans are pretty funny given hyperscalers won't even build "in the middle of nowhere" in the US but rather they put their sites in the edge of exist industrial areas with the aim to repurpose every bit of industrial infrastructure plus aiming to get all the incentives these small town (corruptly imo) offer industries for relocation.
I mean, the US could have a rational process of building new energy infrastructure (which should be only solar, wind and nuclear imo) but the actual process is opportunistic and haphazard. So it's not strange to see the NIMBYs biting them. It's like "well if you hadn't been a d*k, this might not be happening to you..."
How feasible would that be? Don't we have to bring down some satellites from the orbit at some point? How would that work if space datacenter has a limited time use?
It's really hard to get electronic "space grade" there is a lot more radiation there.
Also I wonder how long the ODC needs to operate to offset the energy spend on getting it in orbit.
Isn’t “space grade” a result of high space launch costs? If it becomes cheap enough to launch hardware into space, it doesn’t need to last as long to meet ROI.
Underground or in skyscrapers make more sense
For people selling the anti-sabotage / anti-datacenter-terrorism angle: the math doesn't work.
The energy needed to achieve orbital altitude is vastly less than the energy needed to reach orbit. About 1kg of rocket fuel will carry a small explosive and some tiny steel spheres to LEO altitude. The projectile will not be in orbit; it won't have escape velocity -- in fact it will briefly hit zero velocity (wrt the earth) at orbital altitude and then fall back to earth.
The very large cloud of shrapnel created will encompass LEO orbital height for several seconds. If an orbiting datacenter passes through the cloud during those seconds, the shrapnel will be moving at orbital velocity wrt the datacenter. In other words, just one piece will take it out.
With relatively simple rocketry (and, these days, an abliterated LLM to aid you) a dedicated luddite would have several opportunities per day to hit one of a constellation of satellites. If they wanted one very specific satellite they'd have one or two chances a year. The targeting and steering is not hard -- the orbital paths are publicly known and easily observable; you can use RF time-of-flight to implement exceptionally precise position/altitude/velocity measurements from the rocket (no need for GPS -- use your own transmitters on the ground). This is not hard. And it will only get easier as AI advances.
*I strongly discourage anybody reading this from doing the above.*
I'm posting it in order to show people that anti-sabotage is absolutely not a selling point for orbital datacenters. If anything, they are far more vulnerable to luddite attacks. The people channelling billions into this tech know this too. So their motive is something else. I don't know what, but it isn't this. The math doesn't work.
No
Even if they aren't financially viable, I think they're attractive because companies want to keep certain kinds of data/processing on everyone/everything someplace where there's a lower chance of government regulation (gdpr/etc).
If this is the case, I wouldn't be surprised to see them eventually support these kinds of regulations because it would enhance their moat (now earth orbit in addition to large capex for equipment and a sufficiently large user base).
In that case, international water is far easier. There already are server farms on old oil rigs.
But, as these things go, regulation isn't that much about where you put the computers, but about where you register your company, and where your customers are living.
It's easier, but it's riskier too. Someone can come along a take your stuff a lot easier at sea than they can in orbit. And weather is always tricky, although launches aren't easy either.
In terms of regulation, it depends on the specifics. Sometimes it depends on where the customer lives, but even if it does, when a platform has no business operations in a region and there aren't international treaties between the consumer country and where they're operating from to resolve that sort of thing, the most the regulating country can do is block them.
GDPR does not care about where your servers are located.
It seems to be in certain cases. For example Meta is subject to GDPR because they have EU headquarters, and were subject to fines because they transferred EU data to US servers.
If they have headquarters outside the EU, and only use data they've already acquired/processed to make money off of EU consumers from outside the EU, GDPR shouldn't be an issue for them in the same way.
And GDPR makes it harder for their competitors to catch up because it's harder for those competitors to get data.
Aren't they subject to GDPR because they do business in the EU with EU citizens? I don't see how that would change.
Probably, but not in the same way. I'm guessing they wouldn't have been fined for external data transfers if their servers were located outside the EU.
It could be that the EU would block them if they wanted to host fb/etc... including ads, and didn't have a presence in the EU, but I'm not sure if that's the case.
There are only three ways to deal with regulations you don’t like: political, legal, and force.
Where you put something is completely irrelevant.
That's not accurate. You can build/do/etc things in some places and not in others because of differences in local law/regulation.
Where you put something, like a data center, can be relevant. If Meta didn't have a business presence in the EU, and operated their data centers outside of the EU, they wouldn't have been subject to GDPR in the same way they were.
Earth has massive amounts of mostly unused land that is comparatively close enough by speed of light in fibre.
vs Space:
* Can drive a truck out there
* No limit on weight or dimensions (within reason)
* No giant explodey thing transporting it to into where it needs to be
* No constant falling towards the earth that needs to be averted with thrusters
* If building in the middle of nowhere there is also lots of space for solar panels
* Heat can be rejected into the atmosphere/earth/nearby bodies of water etc instead of radiation into the vacuum of space
* De-comissioning it doesn't also involve slamming a DC worth of mass into the atmosphere and burning up otherwise recyclable materials.
* You don't look like a doofus for doing bad math.
Drive through Nevada outside of Vegas and you’ll see land everywhere. Yet, solar plants still build out slowly. It’s more than just land availability.
I couldn't find a single instance of "upgrade", "recover", or "recycle" in there.
What happens if a major advancement renders all the GPUs obsolete? How can they upgrade the hardware and resell the old hardware?
Are they just going to vaporize all the gear and precious metals?
IDK, the whole thing seems silly at the current state of space logistics.. For these and many more reasons.
It's popular mechanics/science levels of practical fantasy, but somehow now those stories end up swirling around the Internet and never flush. Would have been like having 5 newspaper articles about space elevators every month for years.
I fail to see the economic case. You will need very large solar arrays and radiation hardened components, both of which make it uneconomical compared to a data center on Earth. This is another of Musk's genius ideas like the tunnels that were supposed to be carrying us under cities by now.
The problem is that experiments into this by Google have concluded that radiation hardening is unnecessary. HBM memory far surpassed their expectations for survival, exceeding 5 years in their simulated proton beam on earth. Localized shielding for consumer equipment is more than enough, especially combined with ECC.
Solar power and power dissipation are similarly solvable. The only bottleneck currently is launch costs which is why it’s expected around 2030 is when these DCs will explode
> Solar power and power dissipation are similarly solvable
So it is not outright stupid, just has a few extra costs.
So again, what's the economical benefit? Because you surely have no power and dissipation problem down on Earth, so why add a whole rocket launch into the equation? Because big fire goes brr?
The PV at least can potentially be much less robust than on Earth, since it doesn't have to contend with wind and weather. CdTe thin films, for example, can be just 1 micron thick (the absorption length of solar photons in that material is ~0.1 microns). So potentially the material cost of space PV can be much less than the same peak power PV on Earth.
The bet that they are making is that, in the near-future, energy on Earth will be such an issue (especially if you want to scale) that DCs on Earth will cost more than components + launch costs
Why would energy be an issue? The Sun is shining, wind is blowing, the ocean has waves, nuclear is nuclear. All of these will just get more efficient in using the same constant resource.
It's difficult to imagine any sort of compute in space given the state of the art rad hardened processor is about as fast as a computer Apple sold in 1998.
You don’t have to imagine. The starlink constellation is a working and profitable massive deployment of orbital compute. (It doesn’t use rad hardened hardware as far as I can tell)
Starlink is impressive but calling it compute is a stretch.
It's not that far off from a residential router with a bigger antenna.
Starlink satellites have such a low orbit that they experience non-negligible atmospheric drag, which is why they have to be replaced at such a rapid rate. All proposals for orbital datacenters propose placing them in sun-synchronous orbits for maximum solar exposure, which puts them right at the edge of the Van Allen Belt at best, meaning they'd be exposed to dramatically more radiation than a Starlink satellite.
Musk's "genius idea" there was to (pardon the pun) derail high speed rail and other similar investments so he could sell more cars.
What are Hyperloop/Boring actually doing now? It doesn't take that much investment to human-drive a 10-20 Teslas up and down the Vegas Strip in a straight tunnel at normal speeds.
> and becomes cheaper than any current terrestrial energy source at around $50/kg
Heck for $50/kg I could see it, too. Now with Falcon 9 or Heavy it's on the order of thousands, isn't it? Just need a measly two orders of magnitude improvement and then we can start talking.
Does gauging this based on Falcon 9 launch costs make sense now? SpaceX has already deployed two payloads of satellites using Starship. Falcon 9 is a deprecated system because SpaceX intends to use Starship for future payload missions.
We can't yet know the ultimate cost of Starship payloads. All we have is SpaceX projections of $10-100/kg. Maybe that is realized, maybe not. Given the scale of the LA launch site they're building, they intend to have the launch cadence to deliver lower launch costs than anything achieved with Falcon 9 or any other launch system ever created.
Underlying this seems to be the assumption that those pesky governments, pushed around by public opinion, will keep on blocking or delaying terrestrial data centres, but space-based ones will be fine.
The idea that launch rates could increase to the rates required to launch gigawatts of computing into space, without attracting a huge backlash, is fantasy. And such a backlash would be entirely justified - the only reason why the environmental effects of space launch aren't a focus of environmentalists is because there aren't many of them.
Just skipping to the Conclusions section, I find this article to be wildly biased toward space data centers. Some of this reasoning is ridiculous, insane.
For one thing, I think there's a huge assumption that we just can't make energy cheap on earth (Cost per GW). Solar/battery costs are plummeting, nuclear energy is still a great option (see: the quantity of in-progress Chinese nuclear power plant builds in progress, I think around 30 projects under construction right now), and we might just figure out fusion reactors one day soon.
"Favors: Earth today; Space later" - seems way too confident.
Land Use: Does this article just not realize how large the earth is? The earth will never run out of space for data centers. You can even put them in skyscrapers if you want.
Long-run scaling: How does this row make a lick of sense? This is a lot like saying that if I build my house in space then I could have 100 parking spots in my driveway. The problem is, I don't need 100 parking spaces in my driveway. Like, yeah, we know that there's infinite room to put stuff in space.
Carbon / local environmental burden: Another insane take on this subject. Are we really going to hand-wave over burning a shitload of rocket fuel to put stuff in space? "Favorable vs gas-backed terrestrial buildout" why are we comparing against a gas-backed buildout?
Do we need to? Do we want to? Who is "we" here?
No we will not. The reason is simple: Space is not in the cloud, it is above. Anything that's not in the cloud is bad.
Silliness aside, has anyone come up with a cooling system that would work well for a friggin data center in space? Or is it still all just a wishful thinking game, ignoring physics?
The plan as I understand it is to pack it tightly with TPUs with liquid cooling out to radiators and flying a tight cluster of these next to each other using high bandwidth laser links. That lets you balance the amount of thermal radiation you need to emit vs the amount of TPUs in a single satellite. Thus it’s not a DC in a single satellite but more like a rack and then a bundle of racks flying beside each other.
Or you know, just have tpus with liquid cooling in huge racks, with optical cables or whatever. We can put a shitton of them into a warehouse sized space, and we have basically infinite energy to power and cool it, with physical connection to the internet for ultra fast speeds!
(Sorry, the snark is not directed at you)
The problem is that building the physical DCs is a cost problem not just in terms of energy availability (it’s not infinite) but also politics and legal regulations. Ignoring that is why it makes the space idea seem silly but it’s a real one
> Silliness aside, has anyone come up with a cooling system that would work well for a friggin data center in space? Or is it still all just a wishful thinking game, ignoring physics
That cooling is already perfectly viable.
However at the temperatures chips operate at, because stuff on the ground is getting convective cooling as well as radiative, a radiator designed for LEO will reject slightly more heat on the ground than it would in orbit.
It is absolutely a wishful thinking game, ignoring physics. It's a bet made by investors who don't know or care about the science, and are looking for a modal outcome of a failed company purchased by SpaceX or by an even dumber company looking to compete with SpaceX.
I recommend reading up on the results of Google’s project suncatcher. IMHO the results have provided a roadmap to space DCs with the only blocker being launch cost which are targeted to become economically viable in 2030s; sooner for spacex because they can lower their margins to zero or even negative for the launch itself.
Whenever I asked someone to dig down into "against the laws of physics" arguments, they turned out to instead rest on a pile of tacit and pessimistic technological assumptions.
Perhaps you could explain in detail why you think the laws of physics have to be violated for space data centers to work.
it's the equivalent of PMs asking "can we just make our own llm to save costs"
It will definitely happen and when it happens, HN will deny it happened, just like self-driving cars.
Solar wind and magnetic storm will fry the data center into space junk …
I have a few qualms with this:
As a Linux user, I can already build such a system quite trivially by installing Debian on a Raspberry Pi, attaching it to a weather balloon, and setting up a cron job to maintain orbit. The cooling is handled by opening a window. From Windows or Mac, the balloon could be accessed through built-in software.
One cool possibility is superconductance with a number of materials.
The catches:
o Shielding from the sun's light and radiant heat.
o And efficiently radiating the system heat into space (since there is no air to convect or conduct the heat away).
Simple answer, no.
Long answer, https://www.dia.mil/Portals/110/Images/News/Military_Powers_...
So if launch costs drop 15x and radiator technology improves 12 to 30x and chip lose is improved 2x and there are absolutely no innovations that make terrestrial deployment cheaper, it's totally slightly cheaper!
Yes, for military reasons.
I thought this was one of these joke websites that just say "No."
I mean, Elon will probably shoot some GPUs up there and call it a success, but I'm not sure that should count as "data centers in space."
I wouldn't put it past him to launch 20,000 of them before moving on to the next thing.
Or two. It's a broad range, but with his personality and prior history both are possible before he announces whatever the next pivot is.
Just, not a million. There's better things to do with the capacity to put a million up there, and better places to put a million (last I checked) 120 kW compute racks and associated PV systems.
If we don’t put data centers in the North Pole or in Antarctica, I really doubt that we put data centers in space.
In space solar energy is available all the time. That is not the case in those locations, and the storage needs to get around that are the worst on Earth.
Wind, waves, nuclear.
You don't have to pack your energy generator into the same package where your compute would go, for laughably little energy. You can have a single cable go between a huge power plant and the data center.
A huge cable to the North Pole or to Antarctica would be way too expensive.
Also... nuclear? Please don't be silly.
Bro the North Pole doesn’t have land FYI
Hey the geomagnetic pole is probably currently somewhere over Ellesmere Island, though has been drifting away.
neither does space
Even if data centres in space become cost-effective, what's the point? The only non-contrived answer that I can think of is to avoid NIMBYism, but pushback against data centres can't be that bad.
Only if launch costs fall to 10% of current costs.
IMO, they need to fall below the land cost for the same computing power somewhere in a desert.
The only place where a space data center makes sense is in Mars orbit to reduce the amount of mass that needs to be landed. Every thing else is a smoke and mirrors game to justify rolling a bunch of failed businesses into SpaceX.
Even then. If you're building a million satellites, they only need to be about 42 meters long for you to make them into a contiguous ring in LEO. Smaller (22.5m) or fewer (537,000) for the same in a 200 km orbit around Mars.
Do that, and you can quickly bootstrap an orbital ring and make rockets obsolete from surface to orbit of either body.
In the case of Mars, the planet-wide dust storms make it a good idea to locate your PV off-planet, and an orbital ring is also very convenient for that.
Maybe I missed something obvious, but how does an orbital ring make rockets obsolete for surface to orbit trips?
Because it's held up by being in orbit (strictly: a bit faster than that), you can by the same process as a maglev hang a stationary-with-respect-to-ground cable off it; and because the cable is only as long as your orbit is high (~200 km or whatever) rather than the 36,000 km you'd need for a full space elevator, we don't need to invent any supermaterials for it.
https://en.wikipedia.org/wiki/Orbital_ring
What would you like to make this orbital ring out of?
Structually, it barely matters. The physics works even if each element isn't joined contiguously to the next, it's just that if you do make it contiguous you can also use it for power delivery and worry less about elements drifting out of alignment.
https://en.wikipedia.org/wiki/Orbital_ring
Pangolin scales?
Goodness no, all subspecies are listed on Wikipedia as having various kinds of "Threatened" conservation statuses.
What specifically do you disagree with from the article to say its Smoke and Mirrors? I think their analysis is reasonable and that it will reduce to a launch cost problem.
Giving it a quick look they're doing the same thing as everyone else and handwaving away really difficult engineering problems with some back of the envelope math that only looks at one or two aspects.
Like their section on collision avoidance isn't any more sophisticated than "Starlink has a great record so far" when we're talking about a very different scale of how much mass and surface area has to be moved around in a coordinated way.
This is in effect an editorial from two postgrads, not an actual detailed workup. I don't feel like spending time going into any more effort based on that.
Saying it is equivalent to launch costs is hand waving. Their "median" scenario would involve Starship working at the lower bound of their operational goal for it.
"Failed businesses" is some insane cope.
The impulse to discredit someone's "art" because of their politics has always stuck me as incredibly juvenile.
It's as if people can't imagine competence - even in some narrow way - in someone who doesn't share all their exact beliefs.
It seems especially silly because literally every great man and woman of history shares only a tiny fraction of our modern beliefs.
Just so you know, this is what cope looks like.
Sure bud. The richest man alive is a huge business failure and I'm just coping about it.
The real main purpose of putting data centers in space is not economic, and it’s not to guard against threats from major powers who could easily destroy them. It’s so that the ultra rich can have data centers that the rest of us cannot destroy. I make no argument about whether we should be thinking of doing so or not, but it is interesting to think of the implications is it not? And it certainly fits the personalities involved and their public statements.
Since when have the public destroyed data centers?
Hopefully soon.
Definitely need to see a Microsoft data centre in Uranus.
Sorry. Not sorry.
No.
We kinda already have IMO.
Starcloud-1 is already in orbit, and many other plans by them and other firms are already well into development.
And Lonestar has a backup server on the moon. Seriously.