I am just honestly confused. There seem to be no logical reason for it, except that it would raise the stock value of rocket producers like SpaceX and Blue Origin.
I am just honestly confused. There seem to be no logical reason for it, except that it would raise the stock value of rocket producers like SpaceX and Blue Origin.
Radiators work in space. The area required for solar arrays easily accommodates cooling. And there’s the issue that you actually need to heat electronics in space for them to function properly.
So really the issue is heating them and not cooling them. But yeah overall you would need heat pumps.
It’s funny how a guy with no engineering experience whatsoever thinks they know more than people putting fucking rockets into space.
You can even do fun things with TEG. Utilizing the difference in temperature between two surfaces to generate energy.
And the hotter your radiator the better it radiates. So yes heat pumps would move heat, which is funny we call it heat cuz it would actually be relatively cold compared to the radiators (potentially), and your radiator operates at T^4 scaling. You get that thing(radiator) as hot as possible that would be the goal.
Really it’s more of a Goldilocks problem. Or basically just a heat pump problem you have to be able to pipe your heat perhaps to the processor perhaps away from the processor depending the Tj(thermal junction.)
That is just a load of bs. Heat pumps do not operate efficiently at high temperatures, and they only add more heat to the circuit. You cannot generate energy from the temperature difference, because then you would slow heat transfer significantly while generating an insignificant amount of power.
It has already been calculated, and radiators are a really big issue for “space datacenters”. Even with our current cooling technology advancements. Server hardware needs to be kept at a cool temperature (ideally not higher than 50°C), and most importantly, it needs to be constant in spite of very variable hardware load. That is really difficult to manage in space, and an issue that needs to be considered even for our non datacenter satellites.
But even if cooling was trivial, space datacenters make absolutely no sense.
For starters, it wouldn’t be possible to put any server in the market up there. Chips are very vulnerable to the radiation in space, which is why space CPUs are stupidly expensive and way slower than earth CPUs. They need to use less efficient litographies, with very special shielding that isn’t commonly available. The sheer compute and memory density of our current servers would be impossible to properly shield, even if we ignore the cost. ECC RAM and ZFS pools for drives would very quickly stop protecting your server from errors, since the bit flips would be too common, and your computations would often be incorrect.
If you ignore both cooling and space radiation, you still have the issue of accesibility. Datacenters have people always coming and going, because hardware needs maintenance. Sometimes you want to upgrade the server, change the connections, add or remove hardware, and of course, you need to replace the parts that break over time. Hardware becomes old really quickly, especially for the compute intensive workloads that are planned for those “space datacenters”.
Datacenters are kept fast and competitive through constant upgrading of the parts. Compute intensive servers also churn through parts very quickly, because even at 50-60°C temps, keeping hardware at 100% utilization isn’t the best. This is further exhacerbated by temperature fluctuations, which can prematurely kill the silicon.
Like someone else mentioned in another comment, Microsoft already tried dumping containers with servers into the ocean, and putting wind turbines on top. That’s free cooling, very stable temperatures, and free energy. Even if it was a success in every measure, Microsoft dropped it because it wasn’t accessible enough, which made it unusable for any real usecase.
The truth is, datacenters are a market where cost is very important, and solutions are already very competitive. Putting a datacenter in orbit would be stupidly costly, even if you could fix all the issues mentioned. Free electricity does not offset those costs, and that’s without taking into account the cost of the rockets for putting the datacenters in orbit.
TL;DR: There are so many issues and extra costs, even when ignoring cooling, that make datacenters in space a stupid thing to even consider.
And before you say I’m not an engineer and I don’t know what I’m talking about, server infrastructure is my field. I’ve worked with datacenters, maintained server stacks and I myself have worked on those cost calculations. I know what I’m talking about.
Thanks for saving me the writeup. Probably would have taken me some research to find peer reviewed sources again but this is exactly what I have found in the past.
Plus the engineers they rave about are the same ones incapable of building a new giant rocket with a launch capacity to orbit over the one banana threshold so that made me chuckle.
https://youtu.be/FlQYU3m1e80
Grab some popcorn and learn something.
Opens link, sees Scott Manley and realize we ain’t in peer reviewed territory by a long shot.
This is a troll, they used chatGPT (which they called “Chad GPT”) to reply to me before. They don’t know shit, but it’s funny seeing them think they do. Read their user description, it’s so funny.
You can’t even understand the Stefan-Boltzmann equation. Gotcha. Have a blissful day.
I think your problem is that you are thinking of typical datacenters with commodity hardware orbiting the planet. Which is obviously Ralph-levels of ridiculous… Mmm paste!
Anywho, for entertainment I’ll respond.
Specialized heat pumps can hit several hundred degrees. Recall T^4(Stefan boltzmann law)? 120C for a perfect blackbody can balance the incoming solar flux at 1 AU. Realistically, a coefficient of emissivity(epsilon) of 0.7 would need a radiator at ~160°. This is trivial. There is no heat generated in the circuit unless there’s a direct energy loss creating it (“friction”). It’s conserved and it doesn’t matter. So what?
Semiconductors don’t have to run under 50° or stay “cool” nor is there need for any sort of constant temperature. I have never heard anything so misguided. That’s not even true for terrestrial data centers. There is an extremely wide range of operational temperatures for hardened semiconductors built for purpose for orbital deployment. And since that’s what we’re talking about maybe that’s what you should think about? Moving on…
Take xAI/SpaceX, they plan to modularize them into 100-150 kW satellites that’ll mesh network. So if you get 7,000 to 10,000 of them you then have a ‘gigawatt datacenter.’ Not one giant monolithic multiple square kilometer datacenter complex (which seems to be what you’re thinking.) oof, moving on…
Accessibility? Deorbit and launch upgraded replacements, lol. Which addresses quite a few points… Just like starlink. What the fuck is the story about Microsoft? So what?! How is it even relevant? I think you’ve been on Earth too long. Moving on!
Technically, you would have custom Asics built for purpose… You may have heard of terafab? At least that’s the plan for Elon muskrat! You can trade-off shielding mass, radiation hardening, redundancy, error correction, orbital altitude, component selection, replacement cadence, and so forth. You know, specifically design it for the environment. Does that make sense to you? Are you groking it yet, sunny? ;)
Honestly, I’m totally skeptical about the economics, but tech bros have a shit ton of money to throw at the problem. Do I think we should do it? No, but when has that ever stopped anybody with more money than sense?
You don’t know what you’re talking about. Heat pump circuits don’t generate extra heat? Semiconductors don’t need to stay cool? They don’t need constant temperature? You don’t need accesibility? Dismissing Microsoft’s story because it happened on earth? Quoting Musk’s numbers without questioning if they’re even possible? You think ASICs will fix everything?
Yes, heat pumps generate heat.
Yes, semiconductors need to stay cool and have constant temperature.
Yes, you don’t need to use standard hardware, but then, why spend 100x more on hardware to have 40x less performance? It makes no sense. You can only use appropriate hardware on earth.
Yes, you need accessibility. Are you gonna deorbit and burn a whole “100-150 kW” satellite just because you needed to change a cable on a patch panel? Microsoft’s story demonstrates how even with free power and cooling, accessibility breaks the deal because it’s that important. And Microsoft “only” had to refloat a container and open a hatch. They didn’t have to burn the whole thing.
Additionally, burning metals in the atmosphere is incredibly harmful for the ozone layer. So there’s that.
Even if you built that “gigawatt datacenter” in space, that’s a meaningless term. It only makes some sort of sense on earth, where we know the performance/watt of common hardware. The term is a sensationalized thing, and the correct measure is performance, like flops. A gigawatt datacenter in space, with hardened hardware, will be way slower than a 50 megawatt datacenter on earth.
You say you could have custom ASICs built for the purpose? While some massive companies have made their own hardware before, look at nvidia. The only reason they’re at the position they’re in is because they have designs for AI accelerators. Only nvidia can design those fast chips, and even if they’re really expensive, most AI datacenters (and even compute intensive non AI datacenters) are full of those. Nobody can build anything, ASIC or not, that beats nvidia.
ASICs are also a really bad choice for a datacenter, because they’re expensive, have to be custom built for a task, and make your super expensive datacenter useless for everything else. Even changing the training method of the AI will make your ASICs useless.
And no, you can’t just “build a chip for the environment” and expect to not have to cool it to the same strict levels, or shield it, or have error correction. Semiconductors are not magic, they work in a certain way. And of the litographic processes announced for the so called “terafab”, none of them can be radiation hardened. They’re too small.
I don’t know why I even bothered to write this, and I don’t think I’ll bother to write the next reply when you inevitably quote more of elon’s delusions at me. You clearly have never been close to a datacenter, and you don’t understand what you’re talking about. You believed musk when he said it totally made sense, and your whole argument makes no sense at any level.
I didn’t quote Elon at all. Why you so upset?
You’re arguing against a version of the idea that nobody actually proposed. And you’re still doing it.
ELI5:
A computer uses electricity. Almost all of that electricity eventually becomes heat. In space, you don’t cool the computer by blowing air over it. You conduct the heat to a radiator, and the radiator emits infrared radiation into space. That’s literally what spacecraft have been doing for decades.
A heat pump doesn’t magically destroy heat, either. It moves heat from one place to another while consuming additional power. If you have 100 kW of computers and a heat pump consumes 10 kW, you ultimately have roughly 110 kW that has to leave the spacecraft. Nobody is disputing that.
The important question is whether 100-150 kW can be radiated into space. It can. At roughly 400 K, an idealized high-emissivity radiator emits about 1.3 kW/m². So 150 kW requires on the order of 115 m² of radiating surface(they use double-sided radiators so it’s actually half that), before accounting for engineering margins, Earth IR, view factors, etc. At higher radiator temperatures the required area gets dramatically smaller. This isn’t speculative physics.
“Semiconductors need to stay cool” is also not equivalent to “semiconductors must be operated at terrestrial datacenter temperatures.” Semiconductor reliability generally improves at lower temperature, but the allowable junction temperature is much higher than room temperature. Different electronics can also be designed for different temperature ranges. The spacecraft thermal system’s job is to keep the junctions within their specified limits, not to maintain the entire satellite at 21°C because that’s what a conventional server room does.
Radiation is a much more legitimate engineering problem. But “the chips must therefore be old, slow rad-hard CPUs” doesn’t follow. There are several ways to address radiation: shielding, redundancy, ECC, watchdogs, scrubbing, fault-tolerant architectures, selective hardening, shorter mission life, and using commercial silicon where the risk is acceptable. Starlink already demonstrates that SpaceX is willing to use this general philosophy rather than designing every component like a 1980s NASA deep-space probe.
And ASICs aren’t supposed to “fix everything.” They’re useful because AI/datacenter workloads are already extremely amenable to specialized accelerators. An ASIC doesn’t have to be a single-purpose chip that becomes worthless if you change one training algorithm. Modern accelerators contain programmable processors, memory systems, matrix engines, interconnects, etc. Even Nvidia’s own products are effectively highly specialized compute architectures rather than magical general-purpose computers.
“Nvidia is the only company capable of making fast AI chips” is an assertion, not an engineering law. Google has TPUs. Amazon has Trainium/Inferentia. Microsoft has Maia. Meta has MTIA. Jim Keller is behind tenstorrent. Various other companies have developed AI accelerators. The existence of Nvidia does not establish that nobody else can design an accelerator.
Likewise, “100× more expensive and 40× slower” is a number you invented unless you have an actual orbital hardware architecture and benchmark to support it. If the claim is that orbital compute is uneconomical, then compare actual quantities: $/FLOP, FLOP/W, useful compute per kilogram, launch cost/kg, radiator mass per kW, solar-array mass per kW, radiation-induced error rate, expected lifetime, and maintenance/replacement cost. That’s an engineering argument. “It can’t possibly work because terrestrial datacenters don’t do it” isn’t. And the Microsoft underwater-datacenter experiment doesn’t demonstrate that orbital datacenters are impossible. It demonstrated that maintenance accessibility has value. That’s obvious. An orbital data center would have to be designed around that constraint, just as satellites, submarines, aircraft, offshore platforms, and nuclear reactors are.
Finally, the “gigawatt” point is fair in one narrow sense: watts alone don’t tell you useful computational performance. You need performance/W, performance/kg, cost/FLOP, etc. But that doesn’t make gigawatts meaningless. Power is one of the fundamental constraints on a compute system. If somebody puts 1 GW of useful compute power in orbit, the relevant question is exactly what performance they get from it. So yes, there are serious engineering questions about orbital AI datacenters. Radiation, thermal design, mass, launch cadence, degradation, optical communications, power generation, and maintenance are all real problems.
But “there are engineering problems” and “therefore the concept violates physics” are two very different conclusions.
The former is engineering.
The latter is just internet physics.
You just got fucking schooled by Chad GPT. Have a blissful day.
I said I wouldn’t answer, but I just can’t help it. I’m literally laughing at this reply.
Yeah, I guess that explains why everything is so misinformed and incorrect.
I’m not even gonna bother answering any of the points, as just 5 minutes of research is enough to disprove them.
Though, I am gonna share my favourite part of your rant:
Yeah, I guess the Application Specific Integrated Circuit doesn’t have to be application specific after all.
I’ll also share this other gem with you. You wrote it yourself! I just made a few changes to it to make it more accurate.
“It’s funny how a guy with no engineering experience whatsoever thinks they know more than people designing server systems and computing solutions, who have experience working with datacenters.”
Yeah I get it.
It’s pretty fucking hilarious when an AI can appear more intelligent than what appears to be a person but I don’t even know if you’re a person either so you just might be a more stupid AI than Chad GPT.