> Some parts of the moon are permanently shadowed and therefore extremely cold, as low as -173 °C. This means that no energy or water would need to be expended to cool the data center.
That doesn't sound right to me. If there's no air, then only black body radiation can be used to cool the data center. That means a massive radiator, a lot larger than a heat-to-air radiator+fan used on earth.
This is as big a scam or waste as those solar mirror people. Throwing a USB stick out the lunar rover window and calling it a data center. Data centers usually do stuff other than sit in the dust.
Feel free to invest though, perhaps if you feel good about discarding hard drives on the moon I could interest you in space mirrors and for a low low price I'll lease you the spot where your harddrive lands for 100 years.
> Throwing a USB stick out the lunar rover window and calling it a data center. Data centers usually do stuff other than sit in the dust.
Okay, so here me out: the ultimate cold storage for someone like Iron Mountain. You'd have to understand that you'll need a minimum of 96 hours for retrieval time, and it's gonna be expensive to get that retrieval rocket there and back. Or, build a big dish and send the data via satellite signals.
If you want your data to survive a global nuclear armageddon or Chicxulub impactor, a flash drive on the moon isn't a bad place to put it.
You know what they say, 3-2-1 backup: three copies of your data, stored on two different types of media, with one copy kept offsite to guard against common-cause data loss. Depending on the cause, off-site could be a difficult place to go!
Let's tale it at face value that a hard drive on the moon does not entropy. Let's also take it as given that the data is so important that you would pay millions of dollars to place it on the moon and millions more to get it back when needed in the event that all your other forms of backup failed due to some global catastrophe.
1) At the moment, we can't even get living people off the international space station let alone land on the moon and take off from the same spot twice.
2) If a space based proof of concept was practical, why would we not store our hard drive on the ISS. It is looking for some excuse to remain in operation and we can already come and go from it on a semi-regular basis?
3) If there was a global catastrophe to the extent that only moon based archives remained, then how are we going to go get them? This crisis destroys all data archives but preserves our space program?
4) Once we did get the drive back, what exactly might we do with it considering all other forms of data storage were destroyed?
5) If the data on the drive was so valuable that we were willing to pay millions of dollars for the chance that after Armageddon we could still get it back... Then why would the Chinese not just wait for us to place the drive then go get it themselves? Surely you would never encrypt it as the key would be just as vulnerable to loss as the data.
> 1) At the moment, we can't even get living people off the international space station let alone land on the moon and take off from the same spot twice.
This is absolute garbage. It's not even close to being true. Since the astronauts arrived on the Boeing craft, SpaceX has delivered and retrieved other astronauts. They are not still on the ISS because there's no ability to bring humans back, but because of a scheduling logistics situation.
Continuing to push this scheduling snafu as being unable is just nonsense, and you are as well for pushing it.
> 2) If a space based proof of concept was practical, why would we not store our hard drive on the ISS. It is looking for some excuse to remain in operation and we can already come and go from it on a semi-regular basis?
There's only so much space on the ISS. Also, it's being decommissioned soon, so unless some company wants to take it over as a business--which NASA is open to yet no takers--they've contracted SpaceX to de-orbit the station.
That's because of the incompetence of the manufacturer of the craft used, not humanity's inability to make this happen. So this isn't even close to being a fair comparison.
Are you suggesting that we "simply" build an entire space transit system so people can come and go to the moon on short notice in order to support hard drive archiving?
> They've got it all figured out, you just don't understand.
I'd love to know how they plan to offset the cost of moving every single nut and bolt to the moon with... Cheaper cooling?
I mean, Microsoft experimented with sinking data centers under the ocean. That's certainly cheaper and more performant than shooting a rocket to the moon. That experiment ended. Why?
Iirc a big issue they had was the fact it's nearly impossible to service the interior once you sink it down. That would basically get an order of magnitude worse with it on the dark side of the moon.
A 1 meter square heat exchanger in a vacuum at 20C will emit about 1 kilowatt at -173C. So about as much as a small space heater per small panel. So a 1 megawatt datacenter would need about 300,000m^2 or 0.3 km^2 of surface area to cool it.
But geothermal cooling would be great on the moon too. Run a pipe 2 meters under the lunar surface and it is -21C.
I think the whole idea though is to make a low wattage space-stead so you can store copies of Moana out of reach of Disney cease and desist letters.
> But geothermal cooling would be great on the moon too. Run a pipe 2 meters under the lunar surface and it is -21C.
Isn't the moon geologically dead though - no water or geological movements?
I worry this would just result in the ground absorbing the waste heat and eventually becoming too warm to effectively cool anything. Especially because the ground itself would eventually still be limited by the rate of radiative cooling into space, right?
You have to worry about changing the ground temperature even on earth FYI. When designing district heating/cooling systems with borehole fields, one of the things that you check for is to make sure that you don’t inject too much heat (or extract too much) seasonally - ideally it’s roughly balanced so any drift year over year is small.
Obviously things like the diffusivity (so conductivity, mass, density etc) of the ground matter a lot, as does the rate of heat exchange at the surface for it to reject (or absorb) heat to the environment.
Looked up some papers, and seemingly super low compared to what I would have initially guessed - probably because it’s porous/fluffy/sharp dust with lots of small voids/less compacted I’m guessing. Like, orders of magnitude less than the ground on earth. Not my area of expertise though and was just cursorily skimming papers for values. Specific heat cap and density seem like what you would expect for any rocky materials.
Geological activity is usually a problem for geothermal cooling. Conductivity of lunar subsurface is your main problem. Generally what makes geothermal really effective is the monstrous thermal mass of solid and liquid material compared to air as well as it's much higher conductivity. While that is true on Earth, on the Moon things are very different.
The problem isn't so much geological activity or lack thereof, as the nature of lunar regolith. Lunar regolith has a conductivity of 0.004W/mK. That is lower than aerogel! So unless the subsurface has a much higher conductivity, using subsurface cooling would be doomed.
Edit: Lunar Regolith is only the first 4-5 meters of the lunar surface.
Their point is sure, you locally dump the heat. Where does it go then? There's not ground water to act as a sink so you're stuck with basically a big regolith insulator, and there's barely any atmosphere, so you're back to the black body of the ground with extra steps and a large local sink. That sink isn't infinite when you're talking even house scales, much less thermal scales of a large datacenter. Cooling works on earth because the atmosphere moves the heat.
The regolith is 0.004W/m2K. Less than aerogel. Less than an open vacuum! Ouch!
Although the regolith is only 4-5 meters thick, so you could probably just go under it and see what the subsurface is like with regards to conductivity.
> But geothermal cooling would be great on the moon too. Run a pipe 2 meters under the lunar surface and it is -21C.
It won't stay -21C for very long, if you pump heat into it.
Really, radiative cooling is your only longer term option.
Btw, you can make your radiative cooling a lot more efficient than you gave in your example, if you run it at a higher temperature. Radiated power grows with the fourth power of (absolute) temperature. So, run your chips at something closer to eg 100C and you radiate more than 2.5x as much power.
> It won't stay -21C for very long, if you pump heat into it.
You are right, and also it would require digging which is a lot harder than laying out panels on the surface. Back of the napkin it's a tossup depending on the conductivity of lunar sub-surface material and how much pipe you lay. Just like on Earth.
> Btw, you can make your radiative cooling a lot more efficient than you gave in your example
This is true too, heat pumps could even get higher radiator temps than 100C if you like.
Heat pumps are typically 200-400% efficient. That is, they move 2-4x as much heat as they produce to move that heat. Although that is usually radiating energy to a fairly warm atmosphere. If they are pumping into an extremely cold environment you will get insane efficiencies since you are 'going with the flow' with regards to entropy increase.
Putting a data center on the moon is not the real business.
The real business is the fake contracting companies the founders own who will hoover up all the contracts, do a lot of on-paper contracting making the founders very wealthy before the fake lunar business goes bankrupt.
No, I’m saying it’s stupid to put this on the moon. Every time someone says, “just do this” there are complexities and costs that they don’t account for. Where will you get the immense energy needed to melt supercooled ice?! It’s ridiculous that I’m even arguing, it’s such a stupid idea.
> No, I’m saying it’s stupid to put this on the moon.
I agree with that. Just not for the heating/cooling reasons.
> Where will you get the immense energy needed to melt supercooled ice?!
The problem was too much heat. Now it's too little heat. You can't have both. The energy comes from the server farm. The coolant comes from the ice. You melt your first ice with the heat from the server farm (or an initial solar field).
The ice isn’t in one place, it’s spread around in the regolith at varying densities. How do you collect and melt it? Sure you have heat and nothing to do with it but your ice isn’t all in a neat pile. My point is that you can go on and on like this and come up with a thousand ways the project will fail before it’s even started. Anyways, I’m done arguing, we both agree it’s stupid but just for different reasons.
You could probably use significantly less coolant if you're using heat pipes. The coolant is mainly gaseous and only a small mass remains liquid during the cycle
There is a certain amount of energy you need to move from the chips to the atmosphere or to the geology, and a liquid coolant vs a gaseous coolant will make a BIG difference in how much of that heat you can move. I don't know why I am arguing, putting a data center on the moon is dumb as hell.
Rock is so well known as a thermal insulator that we build buildings from it (brick, concrete, stone blocks). I don't think it's going to help much with your heat transfer problem.
I read somewhere else that regolith is spiny and porous, and with the leading theory of the moon being that it was the result of some massive object hitting Earth a long time ago, and then the ejected material coalescing in orbit and cooling, it makes sense to me that it would be more of an insulator than a conductor. I only know a bit of materials science from what I learned in nuclear power school back in the day, but for a conductor you want to have, generally, no air voids. I recall learning how bubbles in the steel for the plant could not only lead to failures, but also lead to thermal stress points because heat would build up on the "thinner" cross section of metal when there were pores all over (heat transfer being dependent a lot on the cross sectional width, IIRC). Anyways, it's unlikely they'll actually build anything on the moon anytime soon, especially a "data center".
There's a serious lack of terraforming here on Earth.
Oh, we're overpopulated? Buy some single-family homes, buy out the government, and upzone it into apartments. You're still cheaper than space flight and with a shorter commute to work than Rapture
> If there's no air, then only black body radiation can be used to cool the data center.
to put it into numbers - at 80C (353K) 1m2 radiates 880Watts
I think though that instead of the Moon we'll be putting data centers into orbit - for 1KWt GPU we'll need 5m2 solar panels and 1m2 radiator - all together under 10kg, ie. $1000 at Starship prices while the GPU itself is $20K+ .
The kicker here is that the Starship launch price is cheaper than installing solar on the ground ( $2K/KWt and higher)
I wonder if this info came from the Intuitive Machines or from article editor?
I can't find anything related to cooling on Intuitive Machines website. BTW, the website looks like investor bait, not a real company that has a future.
"Amit Verma, a professor of electrical engineering at Texas A&M University Kingsville who is not affiliated with the project, says there may be technical advantages to hosting data on the moon as well. Some parts of the moon are permanently shadowed and therefore extremely cold, as low as -173 °C. This means that no energy or water would need to be expended to cool the data center. And the electrical components will perform more efficiently."
I'm guessing Verma only thought about the electrical aspects, and simply didn't think about the different atmospheric conditions (i.e. not having one) as that's outside of the conditions an electrical engineer typically deals with. I can see how someone can make such a "oops, didn't think of that" mistake when a journalist asks for a comment.
"According to the United Nations’ 1967 outer space treaty, space and the moon are “not subject to national appropriation by claim of sovereignty,” and as such poses a loophole for data sovereignty laws."
That will probably work about as well as the proposal to put a data center on Sealand [1]. Or Cryptoland. Or Satoshi Island. Or Blueseed.[2]
Or the Space Kingdom of Asgardia, which launched a successful satellite with some data storage in 2017.[3] That lasted until 2022, when the satellite re-entered.
Whether countries have jurisdiction over the moon or not is irrelevant, because countries do have jurisdiction over the person or corporation that will run the lunar data center.
> [...] because countries do have jurisdiction over the person or corporation that will run the lunar data center.
The people who originally sent the thing up, can give up their access.
As long as you don't have to physically touch the thing again, you can use some clever cryptography, so that no one is technically running it.
That's easiest, if you just let no one have any privileged access. But you can use (public key) cryptography or similar to give some anonymous people on earth access. Or, for peek publicity value, the data centre can give access to whoever holds a specific bitcoin.
You can really trade it as a token on some blockchain, and the data centre itself can read the chain (as long as someone sends it to them, but anyone can do that).
The data centre will do the bidding of whoever has the private key that's associated with that specific token on the blockchain.
But the whole blockchain bit isn't even necessary: some anonymous person on earth can hold the private key. When they want to pass on the rights to the datacentre, they send it a command (signed with their own private key) to obey some other key in the future.
Governments can't figure out who holds the relevant key. (At least not in theory. In practice, they can try to do classic spy craft and meta data analysis etc.)
So even though the owner of the key might sit in Germany or the US or Switzerland, those countries don't know that.
It might have a better chance that those ocean-based proposals.
There are a lot of countries with navies that could just happen to decide to conduct a live fire training exercise in the general area as your ocean facility and just happen to have an accident that takes you out, with enough plausible deniability that they probably would not get in any serious trouble over it.
There are a much fewer number of countries that could take out a lunar facility and I don't think any could do it in a way that has any plausible deniability.
> There are a lot of countries with navies that could just happen to decide to conduct a live fire training exercise in the general area as your ocean facility and just happen to have an accident that takes you out, with enough plausible deniability that they probably would not get in any serious trouble over it.
Why would countries not already do the same thing to commit high seas piracy today?
> There are a much fewer number of countries that could take out a lunar facility and I don't think any could do it in a way that has any plausible deniability.
Sure, but if the goal is to get around copyright law, i think the usa would be very happy to do it. No plausable deniability needed.
"1967 outer space treaty, space and the moon are “not subject to national appropriation by claim of sovereignty"
because no country on earth can utilize full space venture (yet), do you think this treaty will hold if lets say US decide that they would colonize the moon for a start because no one would be able to do anything
Nope you're wrong. Lonestar has this all figured out. They even have a landing page and dramatic promo video, which is a clear sign of how much due diligence they put in after receiving their oversubscribed $5M seed investment.
Middle English: from Old French lunatique, from late Latin lunaticus, from Latin luna ‘moon’ (from the belief that changes of the moon caused intermittent insanity).
Very weird pitch. There are datacenters in nuclear bunkers and all the major providers let you do redundant copies across continents.
So what exactly is the threat model here? Astroid pulverises earth but moon somehow stays ok?
>it’s impossible to accommodate all potential customers in any one location, except in outer space
Extra territorial jurisdiction is a thing so don't think outer space treaty on sovereignty will help. Could just as well put it on the antarctic and that works better on cooling too.
It's cool as an experiment ofc but doesn't seem to make any sense.
Exactly. It doesn't solve data sovereignty in a practical way for anything other than backups, and even that could be debated. At the end of the day, everyone still wants low-latency access to their data.
> Over 100 countries worldwide have laws that restrict where certain data can be processed and stored, often to within that country itself. As a data center provider, it’s impossible to accommodate all potential customers in any one location, except in outer space. According to the United Nations’ 1967 outer space treaty, space and the moon are “not subject to national appropriation by claim of sovereignty,” and as such poses a loophole for data sovereignty laws. An American satellite is under American law, but it can carry a black box inside it that’s under British law, or any other country’s. A moon-based data center can host as many separate black boxes as needed, to accommodate all of its diverse customers. Governments seem particularly interested in this prospect.
Setting feasibility aside, seems strictly like this is privacy for the entities that deserve it the least. Privacy is for the individual and their "owned" data, not for governments to craft digital black sites.
We can barely tolerate the latency and throughput limitations between Europe and the US for data center processing... and now we're okay with data being wirelessly transmitted (read: super slow) from Earth to the Moon?
Building something on the moon would be cool, but a data center? Unlikely.
$ units
Currency exchange rates from FloatRates (USD base) on 2022-12-14
3753 units, 113 prefixes, 120 nonlinear units
You have: (2 * 238854 mi) / c
You want: ms
* 2564.4291
/ 0.00038995034
Over 2.5 seconds round trip. And just imagine the on-call shifts. "Sorry honey, PDU failed. I'll be back four days after the next-next launch window opens."
The moon doesn't make a ton of sense, but Antarctica sure does. It's cold, it's dry, and if you load it up with GPUs, the latency is not a huge problem because you'd mostly be using it for training runs. And it's still on Earth, so you can connect it with an undersea cable.
Biggest issue would be power. Not sure what the geothermal situation is there, but given that they get most of their power from diesel, it's probably not great. You could build a big solar array, but then you can only use it for 1/3 of the year.
Btw, instead of Antarctica, you could put your data centre in Iceland (with more or less the same pros, but fewer cons). And: people are actually doing that!
Iceland has a great geothermal power story, but the air is not nearly as cold. The reason Antarctica is good is because you would just blow outside air into the datacenter, no chillers required.
Yes, Iceland is a bit warmer than Antarctica. But cooling is still relatively easy: Iceland only gets to about 14C in summer days. Round it up, and say it's definitely below 30C.
If you run your computers at about 60C, that's still plenty of difference.
You can also look into dumping your heat into water. Either ocean or freshwater.
It approximately doesn't matter in the longer run and on a global scale whether you release your heat into the ground, atmosphere or sea. It's all conducted around sooner or later anyway.
I'm surprised that, somehow, the article barely even touches upon the question of bandwidth. Terrestrial data centers often have hundreds of gigabits per second of transit - how much bandwidth would be available to this lunar data center, to what endpoint, and how reliable will that be? (Would one of these permanently shadowed areas of the moon even have line-of-sight to the Earth's surface, or would it have to be relayed through a satellite?)
The transit is even worse than that, right? The moon orbits the earth so at different times of day it is over different parts of the earth's surface. So the route is not only constantly changing, but sometimes you have to go around the entire earth (either via satellite relay or fiber) and then cover the distance from the earth to the moon on top of that.
It's hard to imagine any scenario where this proposal really makes sense.
If you have a problem that requires and hour or a day to compute, then spending fifteen minutes for data transfer up and down (particularly in the face of lowered costs) is often a profitable trade. Movie studio render farms are a classic example of such compute jobs. Weather or geological resource prediction could be another. There are many such high-compute jobs in practice.
Not only are those jobs relatively rare, but they also require a lot of hardware and a lot of power. Both of those are going to be in short supply. (Solar power requires sunlight, and the target locations are deliberately in shadow.)
> “When you place data centers in environments that are already very, very cold...the performance actually also improves significantly,” Verma says. “Because when you go down in temperature, things like electrical resistance also go down.”
Oh, yes yes yes. Until it stops working entirely, that is. Some resistance is part of the design. Ask any LN2 overclocker.
Any former Googlers here from long ago also mildly surprised this wasn't an announcement from google?
Circa the "glory days" in 2007, "Google Moonbase" was the (90%? jokingly) go-to answer internally for the next crazy, ambitious project the company should work on. "Google Moonbase" was also the go-to answer when friends or family would ask what cool thing you're working on. Especially if you were working on actually secret projects like Chrome or Android at the time.
At current prices, I think the moon is about as expensive as getting into Fort Knox by force.
Probably about as survivable, too.
Starship, if they solve the remaining issues before politics catches up with them, could solve the first problem.
I've seen some interesting ideas for contact-free drilling that might help with the second, but for now they're experimental* — we've got a lot of things in space tech that need R&D spending, which is a great opportunity on a forum like Hacker News, but does mean dreamers like me need to wait.
>Lonestar’s CEO Christopher Stott says it is to protect sensitive data from Earthly hazards.
If storing your data into 2 or 3 datacenters spread across the planet isn't safe enough from disaster, it's not clear that the moon will be any better since after a global disaster that destroys all copies of the data, it's likely that there will be no one left on earth that still needs or wants that data.
Yes. Meteorites are being systematically collected to get impact statistics in areas that are tectonically undisturbed. They seem equally good places: Antarctica, Atacama desert, Greenland. In terms of bandwidth and maintainance they seem preferable to the moon. See the Arctic Code Vault in Svalbard.
If we're talking about the surface of the Earth (i.e. no clever "In the core" stuff) then... everywhere. And by surface lets say that we're talking about everything from the seafloor to the upper atmosphere.
It would probably be cheaper to put it in the caldera of an active volcano than on the Moon. You could certainly get a bathysphere full of archival tape to the bottom of the Challenger Deep for less than you could get the same on the Moon.
oil rig would be perfect, position it over one of the cold currents than make up the global "conveyor belt" heat transport system, and enjoy limitless water @3~4°c to cool any sized data cluster.....position it just right, and the rig can just tap into a natural gas resevoir and use that to power the data cluster, pumps, stuff, whatever, stick some giant wind turbines out there, paint them yellow, make everybody happy
Why not just in orbit around the earth? It could be placed on the opposite side of the sun so it's always in the dark if they think that's a compelling reason.
Not only a "data center". Nokia has LTE 4G connectivity for moon local data onboard the lander/hopper Grace.
It's supposed to be able to travel/hop up to 100m high with a range of 25km from Athena to dive into ditches with shadow areas. Quite an interesting mission.
Why don't they build a datacenter below sea water on earth? There is so much water to cool the system and I bet they can arrange some sort of one directional water flow with no energy expenditure.
Solar panels can provide the electricity for the datacenter.
Those will be installed in the nearest shore (sea water destroys even solar panels)
aside from all the practical arguments listed out in other comments, i simply don't support corporations building in space. space should only be multi-national projects, similar to the way the ISS is managed
I can think of a bunch of great reasons to be wary about this, but at the same time it feels inevitable, doesn't it? Biggest reason I don't like it is the fact a big rock lobbed at earth from the moon is one of the lowest cost of entry WMD programs imaginable.
Seems like it's only a matter of time before someone deploys their capital to another space rock, convinces people to go with them, and then declares sovereignty of that rock. What comes with that is all the same stories history tells over and over, just again, on a different rock.
I'll take your word for it. I lose intuition for how far away the moon is and how much delta - V it takes to get back to earth.
I do think the moon has an advantage from a standpoint that the material is already there. You have to "shoot up hill" to stop the gun from getting loaded.
protect sensitive data by putting it in a high radiation environment, at high risk of micrometeorites, and where getting rid of tons of generated heat is hilariously impractical.
heat transfer is going to be a bitch. on the moon, there's no atmosphere, so no convection. as there is no erosion on the moon, regolith is very spiky and has very little contact between adjacent particles, so it's shit at conduction.
Yes! And run there all autonomous ad+slop+BS-generators on one hand , and all the autonomous personal-electric-monks-that-are-to-be-irradiated-with-that on the other hand - with minimal latency and maximal throughput.
That doesn't sound right to me. If there's no air, then only black body radiation can be used to cool the data center. That means a massive radiator, a lot larger than a heat-to-air radiator+fan used on earth.