How are they solving the heat dissipation issues?
Both of those are expensive as hell, by the way.
Cooling via radiation follows Stefan–Boltzmann: P = εσAT⁴. Let's assume a good surface (emissivity ~0.9) at 300 K (27 °C) at 400 W per square meter per side. A flat panel radiating from both faces into deep space gets 800 W/m, not including the losses from, say, the Sun, or from IR coming off the Earth. Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and minimum 1,200–1,500 m² of radiator.
In case ya didn't know - 1 MW is tiny from a present-day-datacenter perspective. It's like 8 racks. So we're talking orbital megastructures here, many many many square kilometers, and this is with all the best case assumptions, and magic radiator panels that never see the sun, or the earth, or the moon.
This is just the basic numbers here, by the way. There's a garbage truck full of other unsolvable problems if you poke your head in there.
Aside from the "Avoid Regulations" aspect, and the "Everything That Burns Deorbiting is Depreciation" aka "The Starlink Trick", I'm not sure what the hell the draw is.
one is marketing.
the other is that you could make tiny datacenters and flood the sky with them. in effect, not datacenters at all, but some kind of dataswarm coordinating at literal lightspeed via lasers.
they'd still be wildly expensive to deploy, and probably litter the orbit zone with fast-moving debris.
You want to go smaller and go one rack only sure, it's still hundreds of square meters. Check the size of current orbital structure for a point of reference, you can't dwarf those and call it a "dataswarm of tiny datacenters flooding the sky".
For that you might need more advanced stuff like liquid droplet radiators (https://en.wikipedia.org/wiki/Liquid_droplet_radiator), heat sinks & pulsed operation. Still, it should be eventually doable.
As for space data centers - I think the main issue is the complete lack of in space infrastructure for resource mining, processing and manufacturing & maintenance. It is kinda like building your first practical steam locomotive & the deciding to build directly an airliner. No suitable materils, experience, work force, material sources, etc.
We eventually went from locomotives to airliner, in an incremental manner & expanding the supporting infrastructure to support the ever more ambitious projects.
Method 1: same as how u2 planes dumped their data: air drop physical media containing data and catch it in the air.
Method 2: laser based emission to specific detectors.
Method 3: baseball style communication: station is under observation and manipulates in some way to serve as a signalling language.
Method 4: numbers station
Method 5: bill yourself as an isp and have some coded syntax that can be supplied in plain sight with the rest of isp traffic.
Training models takes weeks, are you really worried about a couple hours?
> Can be jammed by drone with laserpointer
Ok, you're just joking
Why do you think GPS is so easy to spoof?
Actually I just googled (clauded?) to learn more about space lasers and found a paper describing interrupting quantum key distribution (tamper evident but not jam proof!) with a 1kw laser on the ground pointed at the satellite! So it really does not take many photons to fuck up an encrypted signal.
“ Vulnerability of Satellite Quantum Key Distribution to Disruption from Ground-Based Lasers”
Definitely not signals from low earth orbit.
never? I doubt that.
Technology will improve over time. Eventually I bet it will become cheaper.
Have you tried building in the U.S.? Why do you think it's so expensive to build in the U.S.? It's due to regulation and red tape.
Thermodynamics says no today, no tomorrow, and no 100 years from now. That's not ever going to change.
The old Chicago School boogeymen of unions, regulations, and red tape are effectors, but they're a pimple on the back of the world-mutating effects of seven-some decades of USD Seigniorage.
I imagine we're going to be living through the reverse of this Seigniorage experiment quite soon.
2) The International Space Station has used a dual-loop ammonia/water-based heat pump to cool the station temperatures. It's been in place for several decades. Heat pumps are a proven technology.
Other satellites have also used heat pumps, such as SES-17 in geostationary orbit https://www.esa.int/Applications/Connectivity_and_Secure_Com...
If we want the heat pump's cold end at about 40–65°C, then for each 1MW of GPU heat, we need another 1MW of heat pump power. Now you need 2MW of solar power.
Good news is that the radiator at 227C (500K) can emit about 5× more heat per square meter than at 57C (330K)
Looks like some experimental pumps within this region have CoP around 30%: https://www.sciencedirect.com/science/article/abs/pii/S03605...
So you'll need a lot of additional energy to run the pumps. Which will require additional radiator area.
That's thumping the Carnot limit: [[T_cold / (T_hot − T_cold)]].
2.5, while rejecting at 500 K, cold side's at least 357 K (eeehhhhhhh 84 °C) . . . and that's an absolutely perfect Carnot machine. At 50% Carnot -- a pretty good heat pump, real world performance is 40-60 -- cold side's at 417 K (144 °C). 417k, feeding your GPU coolant loops.
With those numbers, ideal carnot would be 500/(500-357) = 3.5.
Multi-stage could potentially get you to a COP of 2 or so. So 0.5MW.
Depends on which heat you want
Heat adding to hot side: COP_heat = Q_hot / W = T_hot / (T_hot − T_cold).
Heat leaving the cold side: COP_cool = Q_cold / W = T_cold / (T_hot − T_cold).
Another one (more common in the day to day, for me at least): heat-engine efficiency, η = 1 - T_cold / T_hot. Cycle forward to make work from heat.
Luckily, there are almost no rock in space.
Though I do wonder if it would be possible to have some kind of internal heat pump driven by electrical power to juice up the temperature of the radiators to increase the power being radiated away? E.g., run a heat pump to increase the temperature of a working fluid and then run high temperature radiators? I think it would work and I don't immediately see that it would violate the laws of thermodynamics? (this is ignoring all practically, I'm sure the engineering would be devilishly hard, although if you're already shooting for the moon you might as well throw in some artificial gravity to boot, it's not like the robots get motion sickness)
We need > 2x more solar panels than we need radiators. Doesn't this imply radiation isn't really the limiting factor here?
Also also, how do you get the heat out there, to the panel? You are talking a lot of plumbing, and a lot of working fluid. As mentioned, ISS's two external cooling loops hold about 540 kg of ammonia combined to dump 70 kW of heat, so at the megawatt scale thousands and tens of thousands of kg of working fluid. I know, new heat exchanger, lots of people have brought up "Magic Heat Exchanger Technology X", but no one's given me a cite to at least lay down the fundamentals on how these things manage to bang all over the Carnot numbers.
I'm no downer on space solar. It's probably the long term solution for humanity's energy needs - an equatorial band on the Moon, or even Mercury, would give us more power than we could ever reasonably use. But presenting it as the fix for the data center problem . . . I'm sorry, it sounds like people chasing other people's money. Which is what got us into this mess in the first place.
(This follows because a physical object can't absorb more light than a perfect blackbody spanning its convex hull. A perfect blackbody by definition absorbs 100% of incident light, which is a hard upper bound. Any line incident on an object is also incident on its convex hull).
(Consider an isothermal object that emits more power than a blackbody in the shape of its hull at the same temperature. If you were to place that object in a closed system at thermal equilibrium, the interior of an insulated emissive sphere—combining assumptions, it would emit more power than it absorbs, in violation of the 2nd law. Starting from an isothermal system, the object would grow colder, and the enclosing container hotter).
At any rate, 1 MW for a single satellite is fine. Just launch several thousand of those and you get to real numbers.
Also, there's no need to talk about "magic" radiators. You orient them so they're at a knife edge to both the Sun and Earth. This is not difficult (the Moon is irrelevant).
They run fine for a short while, but not nearly as long.
Heat accelerates all aging processes. It's how they artificially age chips in order to calculate MTBF.
“ We’re working on a number of different approaches for this, including a combination of heat pipes and radiators to cool the chips. So far, our team has tested the technology in a thermal vacuum chamber that simulates both the thermal and vacuum environment in space. We’ll see how our new TPU cooling system works in space and refine our designs as we learn more.”
TPU: 100,000+ watts/square-meter
Radiator: ~300 watts/square-meter
> The biggest challenge was how to cool the A.I. chips, which heat up when they perform calculations and process information. Fans, which typically help dissipate the heat, do not work in space. So the Silicon Valley company instead developed a cooling system that uses layers of conductive material to expel the heat into space.
> The bottom layer is made up of Google’s A.I. chips, which sit on a green motherboard. The next layer consists of “thermal interface material,” a pale green putty that comes in sheets like Fruit Roll-Ups and connects the chips to layers of aluminum and copper, radiating heat away from the motherboard. Finally, there is a radiator panel, which projects heat into space.
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
Is it at least a 1:1 usage / cooling cycle?
You don't need civilian scale compute in space, but you absolutely can put miliary application up there and get filthy rich, like Elon, who doesn't care if it ever works, because he's getting paid as taxi service to shoot stuff up.
Sounds very probable: gives a plausible reasons for sending a lot of infra up, absolutely doesn't have to be profitable or even effective, puts the focus on AI instead of the MIC, and is in line with the kind of contracts Big Tech runs after.
To me this model explains a lot of why so many big companies seem to be investing into what every expert I've heard says goes against basic physics. The only thing missing for it to be more than an interesting idea is why China and the EU are not fighting this.
It's Golden Dome when Google does it.
I guess it's something along the same line when Jeff Bezos does it. He does indeed plan it.
What is China's motive? why do they also need to pretend they want space data centers? [0], [1]
How about Europe's? [2]
[0] - https://www.tomshardware.com/tech-industry/space/china-puts-...
[1] - https://www.reuters.com/science/china-vows-develop-space-tou...
(I guess you could try to capture the radiative photons via the photoelectric effect?? but I don't think it works.)
"Scientists" being detractors and everyone going online to comment about how something is impossible while the people doing the thing just do it then all the detractors quietly shrug it off until the next time is so consistent it feels like Groundhog Day. I'd really like to learn about why so many people get off by saying something that other people are working on is impossible.
I guess you are doing the opposite?
> "Scientists" being detractors and everyone going online to comment about how something is impossible while the people doing the thing just do it
Let's make a chart for how many times this has worked vs failed.
You see, this is the difference between bullshit and not. I can claim "we'll do timetravel soon !", but unless I share some "things" (whatever they are) that might sustain that affirmation, this is just bullshit
Does it mean that no solution exists ? No. Just that today none are known (as far as I can tell).
Your desire to call out bullshit is stronger than your ability to GSD.
People are working on problems I cannot manage (who could have known) and they cannot say anything about that but we have to trust them to bring results out of thin air ? Is this politics .. ?
If your personal accusations can be boiled down to : "shame ! you are not gullible!", then please know that, here in France, not being gullible is a compliment.
The only valid argument against data centers in space is the economic cost. If the per megatoken price for space datacenters is higher than terrestrial, then this won't work.
But the price for terrestrial datacenters keeps going up and technology keeps dropping the price of space-based.
If you are trying to generate a profit, every extra kg puts you further in the red. It still costs over $1000/kg of mass to get into orbit.
But if it's an economic argument, then you need to do the actual math. How big is the radiator? How low could the price to orbit go? How much can you charge per million tokens in 2030?
The reason this is a fake argument is because the validity depends on the math, and nobody advancing the "you can't cool stuff in space" argument is actually doing the math.
Sure, things would be different if the cost was $1/kg, but short of somebody building a space elevator that's just not going to happen.
But are you sure that $100/kg is not competitive? I don't remember all the math, but even their initial AI1 design[1] would throw a lot of profit, if you can sell at the price they offered to Anthropic.
I suspect the price of manufacturing the satellite, plus chips, is the dominating factor, not necessarily the launch costs.
Why wouldn;t you? he delivered on almost everything he promised, just the timeline was a lot later.
we are talking sci-fi things here, these will take time but can be done.
would you rather we bury our heads in the sand and never innovated beyond basics?
A 1GW datacenter with chips operating at 100C (which is probably doable) will need a radiator that has a surface of one square _kilometer_, and this is with all the favorable assumptions. Realistically you'll need about 2x of that.
If you want your DCs to be on a 1000km orbit (for reasonable ping times), you'll be able to _resolve_ these satellites with a naked eye!
Sorry. But this idea is fundamentally unworkable.
Deploy 4,000 and you're at 1 GW. That's 160 launches.
BTW: SpaceX has already manufactured and launched 10,000 Starlink satellites and Falcon 9 launches about 150 times per year. None of this seems unworkable.
Look at the numbers. 1kW of nuclear power capacity on Earth is around $2000, and that's 24/7 guaranteed power. So a 250kW cluster needs $500000 to cover its power demand with near 100% reliability and with some ongoing cost.
And if we're OK with some interruptions, then we can use solar+wind at around $100000 and with essentially no ongoing cost. If we assume the absolutely best projected launch cost of $100 per kg (vs the current one of ~$800), that's just 1 ton of material in space!
So you're off by 2-3 orders of magnitude in cost. And this kind of "it's unprofitable" is actually a fundamental issue.
Now your argument is that it's unprofitable. Here's the calculator: https://andrewmccalip.com/space-datacenters
Specifically, if SpaceX can get the price of a satellite down to $8 per watt (about $2 million USD) then it will compete with terrestrial.
I just don't understand how you can be so certain that they can't do that. I'm not certain that they can, but being certain that it's impossible seems completely evidence-free.
So no, the calculations show that space is NOT feasible unless you want to do that for nefarious reasons: evading regulations, using AI for criminal enterprises, military use, that sort of thing.
Only these applications have the profit margin that even comes close to justifying it.
If you want math then https://andrewmccalip.com/space-datacenters exists. The numbers are grim for orbital DC. Even if you drag the launch cost slider all the way to $1/kg (by the way this is literally sci-fi, per ChatGPT air freight of semiconductors from Taiwan to Ohio costs $9/kg and ocean/train freight costs a bit under $1/kg for a reasonable shipment so good luck with $1/kg to LEO this century) it is still more than twice as expensive as terrestrial DCs.
That sounds absolutely possible. But in any event, we're now arguing a different thing.
The only thing I'm arguing is that cooling is a solved problem. I don't know if SpaceX will ever get the price down so that it is economical. But I'm convinced that there is no universe in which they hit their foreheads and say, "Oh no, we forgot about cooling!"
> No additional mass for liquid cooling loop infrastructure; likely needed but not included
> Thermal: only solar array area used as radiator; no dedicated radiator mass assumed
In hardware and mfg. solvable vs. solved is a big difference. And I too believe that SpaceX's engineers know about radiator panels. But the more cynical interpretation is that whatever the SpaceX engineers think about the technical merits, they are not being asked for that. They are just being asked for a pretext that justifies the xAI acquisition. Elon is also discussing lunar satellite factories that launch the satellites via railgun. Now, is this physically impossible? No, that isn't physically impossible either and I will seriously defend the physical possibility of this. It's not going to happen though.
And you could spend all the engineering costs on building some seriously efficient terrestrial DCs, but somehow all these analyses start with "assume that launch and satellite technology advances manyfold and terrestrial DCs stagnate or become less efficient, then if you squint the two numbers get kinda close."
This is why that calculator, even under extremely optimistic assumptions for orbital, and even if you assume launch is zero, still cannot make it competitive with terrestrial.
> If you assume Starlink satellite $/W
Why would you assume that, given that high power output is not a design goal of Starlink?
ETA: power is absolutely important for Starlink. SNR is very important for shannon capacity and satellite systems are often limited in this respect.
Anyway, my point is that Starlink is not primarily designed to harvest as much solar power as possible the way data center sats are. I'm sure they're not making it inefficient on purpose, but what you're suggesting is like using the cost/W of a solar powered traffic camera to estimate the cost/W of a solar farm. Yes, they both use solar power, but they have entirely different design goals.
Cost is a different matter of course, but it seems incongruous to me that after being confronted with the fact that your power density assumptions were off by a factor of 2x you would continue to insist ChatGPT's weird assumptions about manufacturing costs scaling linearly with mass are accurate. As was already pointed out upthread, even if all the calculator's other assumptions are correct they can turn a profit if they get the per satellite cost (excluding the cost of the GPUs) below $2 million. I think the only big question mark here is launch cost.
So... not easy? I don't suppose we're at a point where you could reasonably send a large-enough radiator for a multi-gigawatt cluster into space?
This is all a weird speedrun or race. If something we should be working on setting up resource mining from the Moon & asteroids, materials processing on the Lunar surface & in orbit, simple manufacturing in space, etc.
Instead some people think we can jump straight to a computronium Dyson swarm. :P
You don't need to send a multi-gigawatt satellite to space. You just need to launch a few thousand 250 kW satellites. That's not against the laws of physics.
If you know all that out to 2040 then you must be a time traveler. Please try to fix our timeline rather than wasting time on HN.
>> The current plan is for 250 kW peak with 160 m^2 radiators. It will weigh about 4 tons so you can pack 25 on one flight of Starship.
>> Deploy 4,000 and you're at 1 GW. That's 160 launches.
Convective radiation does not happen in space and this challenge is far more significant than your comment implies. Rather than "a dumb hunk of metal", radiators for spacecraft are often made of ceramics and carbon laminates with higher IR emissivity than convective radiators made of simple metals.
From the article you're commenting on:
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
The Thermal Control section on NASA's Small Spacecraft documentation center is quite an interesting read for this subject: https://www.nasa.gov/smallsat-institute/sst-soa/thermal-cont...
ML Infrastructure comes with some pros (larger emissive footprint) and cons (exponentially larger TDP) compared to the concerns there, but if you aren't familiar with the challenges of heat dissipation in space, please give it a read.
There's also a pretty interesting pop-sci article on cooling the Webb telescope, since it needs to be especially cold for its purpose. Not directly related, but may give insight into both challenges and solutions as well as well. https://science.nasa.gov/mission/webb/science-overview/scien...
The equation is:
A ~ (1000 P) / (2 e k T^4)
A is the radiator area in square meters
P is the power in kilowatts
e is emissivity (usually 0.9)
k is the constant 5.67e-8
Both Google and NASA are worried about this for a reason, if you think they are wrong, you should offer your assistance to them, rather than debating me.
If your argument against space datacenters is "radiator materials are too expensive" then I just think that's not a very good argument.
Moreover, you can easily calculate how big of a radiator you need for a given power level and temperature. You can use the Stefan–Boltzmann law:
A ~ (1000 P) / (2 e k T^4)
A is the radiator area in square meters
P is the power in kilowatts
e is emissivity (usually 0.9)
k is the constant 5.67e-8
For SpaceX's 175 kW satellites, they will need ~170 square meters of radiators.
None of this is impossible or even difficult to calculate. That's why I think this is the laziest argument against space data centers. There are so many other more reasonable arguments (like whether they will be economically competitive) but people love to latch on to this one for some reason.
That's like a quarter of a rack of modern AI hardware. Modern AI datacenters are now in the multi-GW range.
Now the argument is, what, you can't launch that many satellites?
How can anyone be certain of any of those numbers without (a) knowing how the technology will evolve, and (b) doing the math?
I'm just astounded that people can have such confidence.
They are showing that space-based DCs only make sense for criminal enterprises. Which probably IS what's going on here.
The 15-minute figure is what they currently designed for because they have mass constraints. If they wanted a different figure (like forever) they could do it with a larger radiator.
As you said, the physics is solved! We know exactly how much heat a surface radiates in space--there is literally an equation for it. We know how to cool stuff in space. Will the price be low enough to make a profit? That's the real question. But stop worrying about cooling in space.
If you have something useful to contribute regarding how to actually reasonably put up enough mass to make this remotely a feasible idea then please contribute. We have the technology to prevent global warming, and doing that is far easier than this, and yet even that is apparently not feasible for humanity.
What's unfeasible about that? SpaceX has already launched 10,000 Starlink satellites. Falcon 9 launches 150 times per year.
People use that argument because it takes zero thought to make and significant effort to refute.
"Space is cold" is classic zero effort to make, significant effort to refute stuff: proponents imply complete nonsense about vacuums being optimal for cooling and anyone who understands why this is a lie to gull retail investors ends up getting bogged down in "sure, but Boltzmann equations proving it's prohibitively expensive doesn't prove it isn't possible"...