Hey there, folks who know more science than me. I was listening to a recent daily Economist podcast earlier today and there was a claim that in the very near future that data centres in space may make sense. Central to the rationale was that ‘space is cold’, which would help with the waste heat produced by data centres. I thought that (based largely on reading a bit of sci fi) getting rid of waste heat in space was a significant problem, making such a proposal a non-starter. Can you explain if I am missing something here??
Comments
Not knowing the context of the claim, I have to say this is one of the stupidest things I have heard proposed. At our present level of technology and space-based infrastructure, I have trouble imagining anywhere on Earth which is habitable by humans for any stretch of time that would make a worse place to put a data center than orbit. Literally the only possible upside I could think of off the top of my head would I guess be cheap solar power?
So, I guess the answer to the question in your post is no, not unless we both are, and whatever it is we’d both be missing would have to be pretty massive to outweigh the cons.
Sure deep space itself can be “cold” but if were talking any Earth orbits, temps swing wildly between -250F and +250F depending on the “day” or “night” side.
You are correct that a vacuum doesnt transfer heat well at all, which is what you need to cool hot things down. Data centers in orbit are a no go until some major breakthrough in radiators happens, otherwise youre looking at football+ size radiators to do the job.
Space based data centers have two huge hurdles, cooling and power generation, same as on the ground but several factors more expensive in a vacuum. A space based data center would need a fission reactor and a radiative cooling technology that just doesnt exist yet.
Saying “‘space is cold” while somewhat true, is the wrong way to think about it. Space is empty, and empty doesn’t have a temperature, hot or cold. As humans, we would simply perceive this “emptiness” as “cold”, but we know “cold” doesn’t exist.
You are correct; waste heat is an issue in space, and the proposal is dead on arrival.
Space isn’t cold. The term doesn’t really make sense in a vacuum (or near vacuum if you want to be pedantic). Instead, vacuum is a perfect insulator.
The only method by which heat can transfer in space is radiation. There aren’t any molecules to convect heat away, and you’re not touching anything you can conduct heat to.
Data centers in space make sense for only one reason: basically free power with lots of solar panels. LOTS of solar panels. For every other aspect of data center requirements space is kind of terrible. And given the power requirements of an average data center, I don’t know that even solar is going to cut it. Not without much bigger panels than you’d expect. (or you move your data satellite closer to the sun for more power that way.)
Heating/cooling, maintenance, upgrades, latency, all of these would be much harder problems for a datacenter in space.
Space has way, way fewer particles floating around than our atmosphere, so the usual transmission of heat by particles bumping into you doesn’t happen much at all. The extremely low pressure of that environment also causes liquids to boil and evaporate (Like how water boils at a lower temp in Colorado due to the lower air pressure).
Evaporation is an endothermic process, as to “leap” from liquid to gas, water molecules have to absorb a certain amount of energy, stealing it from the surface they’re on. So most things that have lots of liquid water will freeze in space due to that rapid evaporation.
Heat is mainly transmitted in space by light. If a metal sheet was floating in space, it would get pretty hot as the sun shone on it. If it was instead in the shadow of a planet, it would cool–albeit very slowly. Since very few particles are bumping into it, the only other way heat energy can escape is through radiation. The sheet will glow in the infrared, like we all do when you see infrared camera footage. This is a much slower process than being in an atmosphere, though, so it’d take quite a while. Eventually, as long as the sheet is not in direct sunlight, it would cool to extremely low temperatures, roughly -270C or ~3 Kelvin.
In a sense, space is neither cold nor warm, because there’s not enough matter around to have a temperature. Yes, getting rid of waste heat in space is an issue, because you need matter to carry the heat away. They deal with it by building heat sinks out of heat-conducting material, to direct the heat to where it can radiate away, but that’s not as efficient as, say, a nice refreshing breeze.
Now, if they tried building data centers under the ocean, that would make a lot of other things harder, but it would probably be great for cooling.
Other people might chime in with actual maths, but because no one has yet I’ll just hand-wave: the problem is that the concept of temperature as we usually experience it kinda breaks down in a vacuum. Usually you can think of it as the average kinetic energy of all the particles in a volume. With essentially no particles the vacuum of space is “cold” – but there are also no nearby atoms to conduct any heat away from something that is warm, so it’s difficult to shed any extra heat you do have. You’re reduced to radiating heat away as infra-red from big heatsink-radiators, which is less effective than using air or water to carry the heat into the environment like you would do on Earth. Sometimes The Economist should stick to economics.
Space is cold by comparison to earth temps. However, heat needs a medium to dissipate. So using that temperature differential for cooling is kinda a non starter.
Also space is full of sources of incoming radiation (like the sun) that would contribute to heating as well, thus exacerbating the cooling problem.
Yes, heat needs to radiate away TO something – air, water, an object. Space means relatively little matter (there is matter, just significantly less dense than you would find in our atmosphere). Space also means heat absorbed from the sun. Cooling electronics in space is challenging.
Spacecrafts will usually have radiators for just that with large surface areas. When the radiators can’t be used, sometimes they’ll move heat into liquid stores (e.g. freon) to radiate off later.
Space may be cold (on average), but it’s not thermally conductive. You would die of oxygen starvation long before you would freeze to death.
space is a vacuum, or very close to it, so there is nothing to transfer heat to. it would be like putting a server in thermos. they would have to build massive radiator fins that passively output infrared radiation, which I have a hard time believing would be able to keep up with the output of a server farm. they would also have to shield it from the sun since, while space itself is cold (or rather has almost nothing in it to carry energy), radiation from the sun will quickly heat up a solid surface. and of course they would have to figure out how to transmit all the data at a speed that justifies the cost. but other than that it sounds like a great idea.
The radiative heat flux equation is `q = σ * ε * A * T^4` . If you are a satellite in space trying to cool via a heat exchanger exposed to cosmic background radiation, the net heat flux is `q = σ * A * (ε * T_sattelite^4 – T_cmb^4 ) ` , In this equation, `A` is the surface area of your heat exchanger, the bigger it is the more heat you can shed. T_cmb is about 2.726 Kelvin, so yes, space is very cold. T_satelite is your spacecraft or datacenter temperature, so something like 400 Kelvin.
This is neat and all, but the problem is the sigma term, the Stefan-Boltzmann constant which sits at 5.67 x 10^-8 W/m2/K4. This very small constant is why cooling via radiative transfer is so slow, even though there is a large temperature difference between the spacecraft and the cosmic background.
From my calculation, a radiator with an emissivity of 0.8 (emissivity of carbon) can shed 1161 W for every square meter.
Whoever that economist was, they should stick to economics for sure. It’s hard to think of a worse place for a data center than Earth orbit, for many reasons.
You can’t run fiber optic cable to it
Datacenters need a constant supply of relatively heavy replacement hardware.
Even a relatively low orbit would lead to unacceptable latency because of the distance the signal has to travel.
And as you pointed out, waste heat is an issue. The vacuum of space in fact makes it harder to cool large scale infrastructure, not easier.
“Cold” is determined by the average molecular motion of a region. There are two general ways this can happen: 1- lots of atoms in the region, but all moving very slowly, or 2- very few atoms in the region, but moving at any speed.
Space is of the second kind. And having such low density, it’s incredibly inefficient at transferring heat. Because of this, space is actually terrible for cooling high-energy electronics.
I read this article earlier today.
Classically speaking there are three forms of heat transfer: conduction, convection, and radiation. In space there is effectively no conduction or convection outside an object, because there are very few particles to conduct or carry heat.
Radiation is fairly insignificant as a form of net heat transfer in many scenarios on earth. This is largely because you absorb about as much radiative heat from around you as you emit, and because conduction and convection can be powerful. However, objects on the ground still generally emit a significant amount of radiative heat straight up into space (or rather up into the atmosphere, where much of it is absorbed by greenhouse gases).
In space, heating and cooling are different from on the ground. For an object like a satellite, incoming radiation from the sun and outgoing radiation from the object are not hindered by an atmosphere. Thus solar radiation is very potent and an object emits net radiative heat in nearly every other direction. If an object is moving relative to heat sources and sinks in such a way that its exposure changes rapidly, this can be a major problem in one direction or another, and may be what the sci fi is talking about.
Starcloud, the company mentioned in The Economist, intends to take advantage of both the sun exposure (via solar panels) and the radiative cooling while managing its orbits such that the amount of sun exposure and cooling is predictable. In this case, or so they claim, it does indeed make sense that the satellites will be able to maintain a favorable thermal equilibrium.
Space isn’t really cold the way you’d normally think of it. Cold is the sensation of heat transferring from one body into another. Space is a vacuum so there isn’t anything for that heat to leave into. People freeze in space because all the water in them boils off in the vacuum and that takes their heat with it. I don’t know how that would help data centers. even worse, when they’d be exposed to the sun, they’d be heated up directly without any atmosphere and would reach hundreds of degrees.
Things in space will eventually reach thermal equilibrium, where they radiate away the same amount of energy as they are producing and/or receiving.
If the object in question is a rock somewhere in the outer solar system, then it will not be producing its own heat, and it will be receiving very little from the sun, so in the few billion years since the Solar System has formed, it can reach that equilibrium at a very low temperature.
A data center in orbit around Earth, however, will be producing its own heat, and it will be close enough to the Sun that it will be receiving a significant amount of energy from sunlight, and these two points together will dramatically raise its equilibrium temperature well above that of the rock in the previous example.
It is possible to keep something like that cool, but it would be much easier and cheaper to do so on Earth, where you can take advantage of having an atmosphere to dump excess heat into.
Space is cold in a sense that there’s almost nothing there so there is no heat transfer through conduction.
But it works both ways – you’re not getting heat but you can get rid of heat only through radiation.
If you’d keep the servers on an orbit that is always shaded from the sun by the earth it could work, otherwise, well, there’s a reason that space stuff is white or silver. And just look at the size of radiators on iss.
Space is cold because all objects radiate their energy away, including heat energy, and without something to provide more energy (like radiation from the sun), objects eventually radiate all their energy and freeze.
Radiation does not require a medium, which is why things still freeze in the vacuum of space. However, the rate of radiation depends on the object, and can be quite slow, so a heatsink in space would not be effective at all.
If you’re thinking about water, remember that water (and other liquids) freeze rapidly in space because the drop in pressure lowers their boiling point to zero, causing them to boil immediately. This resulting gas cloud will then freeze, as the individual molecules will radiate their energy simultaneously. What you’re left with is a cloud of ice particles. Thus, this freezing of water didn’t occur because “space is cold,” but because individual molecules don’t have much energy to radiate after so much was used in the boiling process.
That depends a lot of what “is cold” means to you. In the sense of “you lose a lot more thermal energy to radiation than you gain”, most of it is very cold. In the sense of the technical definition of temperature (speed of individual atoms) of the interstellar medium (the tiny handful of hydrogen atoms that still zoom around in the emptiness), it is actually very hot, like thousands of degrees, but they don’t really matter since the pressure (amount of atoms total) is so small that almost nothing of that “warmth” transfers to larger things floating around in it.
In terms of easy ways to discharge heat, you’re correct, space may be cold but if you make your own heat in space you’ll have a much harder time getting rid of it than if you had, say, a nearby flowing river. (I assume that the economist probably didn’t really know what they were talking about, or maybe they meant “on an object” like an asteroid rather than floating freely which may increase your effective radiation surface dramatically — but the distance probably makes that impractical too.)
Space is “cold” but its hard to transfer heat to. The only real method is radiation, which is not a good transfer method unless you’re REALLY hot (like melting your datacenter hot). There is a reason we use things like liquid cooled devices. Conduction and convection are way better at transferrring heat
Cold is the relative absence of heat. Heat could be though of as just the average kinetic energy of molecules.
In space those freepy moving space molecules aren’t moving very fast so when they run into a faster vibrating molecule in a heat sink on the side of some scifi ship, they absorb some of that energy and bounce off at a faster speed than they ran into the ship with. That molecule absorbed some heat!
Unfortunately the number of molecules in space running into that heat sink are few and far between. And they can only absorb a tiny tiny tiny amount of heat in their shot collision… So while space is cold, it’s also thermally isolating.
Now collisions between molecules is only one form of heat transfer. Radiation is another. If you keep the space station in the shadow of a planet so it isn’t warmed by the sun, and you use a heat sink that is good at shedding energy via black body radiation (think your oven/stove heat element glowing when warm), then it could reject heat to space as light while absorbing very little from the other radiating bodies near by.
Really all you need is for net heat emitted to space to be greater than heat absorbed from space in order to cool your space station.
A lot of wrong answers here. Space is cold, about 2.7 K, even in vacuum. Its just that you equilibriate with that temperature via radiation, which can be slow. However, you can speed up the process of radiation using a large surface area. A simple calculation via Stefan-Boltzmann law suggests that a 1 m^2 radiator could keep a 400W power source at a steady 300K (80F). A large data center would need maybe 30,000 m^2. It’s big, but not impossibly big. The savings is that the cooling would be completely passive, not requiring any additional power for cooling. But the cost of building such a large radiator in space would probably cancel such savings.
Temperature is a property of matter, space is literally just space, for our intents and purposes it contains no matter.
Because space contains no matter, it doesn’t have a temperature. Things in space which are made of matter do, however, have a temperature.
Because there is no matter, objects retain their heat very well, as they can only radiate heat through thermal radiation.
As for the data center idea, I think it’s mostly nonsense.
Even if the data center is shielded from the sun, it will produce waste heat of its own and it’ll heat up even if there is no external source of heat. All that heat would have to be expelled through radiative cooling which isn’t really a fast way of doing things. The ISS has massive arrays to cool a relatively small volume.
Radiation is a significant problem in space and it can mess with electronics by flipping bits in computers.
It’s incredibly expensive to launch things into space let alone a whole data center. Building the thing is its own headache, not to mention maintaining it.
Unless the data center is right above you in a low orbit, there would be more lag because the data has to travel farther than it would on Earth.
Basically, it’s just way easier to just do these things on Earth and I don’t see any advantages whatsoever.
I think the person who theorized this was way out of their depth.
Space is cold in the sense that the average temperature is low. But it doesn’t absorb heat the way something like ice water does. The only way things lose heat in space is via evaporation and radiation, which is slow compared to convection and conduction.
You’d die from hypothermia in the sea in a matter of minutes, but assuming something else doesn’t kill you, it’d take hours to succumb to hypothermia in a complete vacuum.
Depends where you are in space. Outside of the solar system and away from any stars, yeah it would be something like -300C. However, if you’re still in our solar system then the side facing the sun would be around 300C while the side facing away from the sun would be around -300C. So you would be freezing and burning simultaneously.
Your average person is probably familiar with the fact that “vacuum insulated” bottles exist, and that they are very good at preventing the transfer of heat.
The vacuum of space is the ultimate insulated layer and a space station is essentially the inside of the best insulated bottle in the universe. It is extraordinarily difficult for heat to escape.
Space is a great insulator, eliminating both convection and conduction modes of heat transfer. It’s like you are in a vacuum thermos!
You can only lose heat through radiation, but unfortunately it’s governed by the Stefan-Boltzmann constant. That constant is really small (5.67×10^-8 W/m²•K^4), so you need to either have a big enough surface area, or be hot enough to have meaningful heat loss to offset it.
Yes, space is intensely cold – around 3 Kelvin based on the equilibrium temperature reached by objects well shaded from the sun.
But space is also very, very empty, so you can really only lose heat via radiation, which isn’t very fast unless you have huge radiators.
So no, it’s crappy place for data centers, or anything else that needs to shed a lot of heat. And you will not flash-freeze if exposed to it, though give your body a few…days(?) after you die to cool down, and it will freeze solid enough.
The temperature is most useful not for cooling things off, but keeping them cold. E.g. the JWST took many days to cool down to its ultra-low operating temperature… but then it just happily stays there with no more effort than a good sun-shield.
That has to be one of the worst ideas in recent memory. One of the biggest challenges for space craft in general is cooling because the only way to get rid of heat is through radiation.
Also, the hardware of data centers fails constantly, so you’d need a crew of engineers, and regular shipments of heavy replacements.
Then there’s the issue of communication. The fastest satellite link I could find is about 100 Gbit/s, and that’s experimental. About 200 Mbit/s are more typical. The former might just be enough for a small data center, but absolutely not for AWS scale…
There are plenty of cold places on Earth. There’s zero benefit for putting a data center in space aside from hyping up gullible investors, so I expect Elon Musk to announce it within the year.
“Space is cold” is more or less true. If you dropped an object halfway between here and Andromeda, in intergalactic space, with no star nearby to warm it up, that object would slowly cool down. It would eventually reach a very low temperature, near absolute zero.
However it would cool down very slowly. Space is quite insulating. If, instead of an object, we dropped you there naked, after imbuing you with the power to not quickly die in space, you wouldn’t feel cold at all.
Cause you produce your own heat. And space would not leech your heat away faster than you produce it.
With two exceptions. You’d get cold eyes and mouth. Those places are wet. Liquid water isn’t stable without atmospheric pressure. Your tears and saliva would start evaporating. And that is a chemical process that sucks heat. So you’d get cold there. But with a pressurized helmet, and the rest of your body exposed (ok, and a pair of pressurized britches, Superman was onto something after all), you’d be fine as pine.
Yes, space is cold, but using space as a way to vent off excess heat is an amazingly awful idea.
Think about what it’s like being in 50 degree water. Pretty cold, right? Cold enough to be dangerous.
Now think about 50 degree weather. A little chilly, but kids play in that sort of weather all the time.
Water sucks the heat from you faster than air of the same temperature because water is denser than air. It’s about 800 times denser! So it’s much better at transferring heat.
Air is about 10,000,000,000,000,000,000,000,000 denser than space (space still has matter in it – it’s not a perfect vacuum. It’s just really close). Space is very cold but it’s just awful at transferring heat. Getting rid of heat was a major issue for space travel, and it’s really not as simple as just pushing it into space.
As cold as space is, the lack of atmosphere means you’re likely burning to death from direct sunlight rather than freezing to death. Your data center is probably gaining heat from being in space, not losing it.
space is cold, technically, inasmuch as the interplanetary medium, as nondense at it is, has a temperature.
that said, you are right that its heat-sucking-capacity (so to speak) is extremely poor, despite the technically-low-temperature. as said, vacuum is one of the worst possible ways to transfer heat. it’s like a yeti cup, which is vacuum insulated, only a yeti vacuum is orders of magnitude crappier than actual space vacuum. so a millilmeter thick layer of space is like a meter of yeti insulation (or however many orders of magnitude it is).
space is cold, technically, but a goddamned terrible heatsink.
so short version, you’re absolutely correct
Lets look at Moon. At lunar day surface temperature us 120C, at lunar nightsurface temperature is – 120C (pretty cold to be there). How come?
All objects emits infraread radiation. This way they loose energy. If they emit more than they receive – object temperature reduces.
So if you poke hand in space in shadow for a minute (you need some way to prevent depressurisation) it will not became cold. After few hours you will be freased.
Now lets talk about why unprotected human exposed to soace will freeze almost instantly. This is whole different process tied to pressure. Thermodynamics stats: when oreasure decreases – temperature decreases and vise versa. Sudden exposure to space have a tremendous preasure decrease causing very fast cooling
Really depends on on where you are. For instance if you’re in earth orbit, one side of your spacesuit could boil water and the other side would freeze the CO2 in your breath to dry ice, but as you mentioned actual heat transfer is tricky.
That said, dealing with heat is one of the smaller engineering challenges with putting a data center in space and it would be a HUGE problem.
Everyone is addressing orbit, but you didn’t clarify. Were they talking about the moon? If not, they were completely pulling out of their ass. If they did mean moon, they should have been talking about using the moon as a giant heatsink, not space itself. And even then, the engineering to keep cosmic radiation from damaging the data and shielding the sunlight while it is facing the sun and relays to get the data to Earth while it is facing away from us would probably make the economics a flop.
It isn’t a firm 100% bad idea though, you just need to benefit from the other aspects of such a data center. The most obvious is preserving knowledge in the event of Armageddon, but there can be other scenarios where the moon is more desirable, such as a read-only server without physical access for immutable records or an exceptionally hard to destroy datacenter for military secrets.
Of course this starts departing from reality to Sci-Fi, but at least the discussion would be coherent.