You operate them like Microsoft's submerged data center project: you don't do maintenance, whatever fails fails. You start with enough redundancy in critical components like power and networking and accept that compute resources will slowly decrease as nodes fail
No operational needs is obviously ... simplified. You still need to manage downlink capacity, station keeping, collision avoidance, etc. But for a large constellation the per-satellite cost of that would be pretty small.
How do you make a small fortune? Start with a big one.
The thing being called obvious here is that the maintenance you have to do on earth is vastly cheaper than the overspeccing you need to do in space (otherwise we would overspec on earth). That's before even considering the harsh radiation environment and the incredible cost to put even a single pound into low earth orbit.
If you think the primary source of electricity is solar (which clearly Musk does), then space increases the amount of compute per solar cell by ~5x, and eliminates the relatively large battery required for 24/7 operation. The thermal radiators and radiation effects are manageable.
The basic idea of putting compute in space to avoid inefficient power beaming goes back to NASA in the 60s, but the problem was always the high cost to orbit. Clearly Musk expects Starship will change that.
ISS cooling is 16KW dissipation. So like 16 H200. Now imagine you want to cool 100k instead of 16.
And all this before we talk about radiation, connectivity (good luck with 100gbps rack-to-rack we have on earth), and what have you.
—
Sometimes I think all this space datacenters talk is just a PR to hush those sad folks that happen to live near the (future) datacenter: “don’t worry, it’s temporary”
> ROSA is 20 percent lighter (with a mass of 325 kg (717 lb))[3] and one-fourth the volume of rigid panel arrays with the same performance.
And that’s not the current cutting edge in solar panels either. A company can take more risks with technology choices and iterate faster (get current state-of-the-art solar to be usable in space).
The bet they’re making is on their own engineering progress, like they did with rockets, not on sticking together pieces used on the ISS today.
How much maintenance do you need? Lets say you have hardware whose useful lifespan due to obsolescence is 5 years, and in 4, the satellite will crash into the atmosphere anyways.
Let's say given component failure rates, you can expect for 20% of the GPUs to fail in that time. I'd say that's acceptable.
A lot. As someone that has been responsible for trainings with up to 10K GPUs, things fail all the time. By all the time I don't mean every few weeks, I mean daily.
From disk failings, to GPU overheating, to infiniband optical connectors not being correctly fastened and disconnecting randomly, we have to send people to manually fix/debug things in the datacenter all the time.
If one GPU fails, you essentially lose the entire node (so 8 GPUs), so if your strategy is to just turn off whatever fails forever and not deal with it, it's gonna get very expensive very fast.
And thats in an environment where temperature is very well controlled and where you don't have to put your entire cluster through 4 Gs and insane vibrations during take off.
Note how Musk cleverly doesn't claim that not doing maintenance drives down costs.
Nothing in there is a lie, but any substance is at best implied. Yes, 1,000,000 tons/year * 100kW/ton is 100GW. Yes, there would be no maintenance and negligible operational cost. Yes, there is some path to launching 1TW/year (whether that path is realistic isn't mentioned, neither what a realistic timeline would be). And then without providing any rationale Elon states his estimate that the cheapest way to do AI compute will be in space in a couple years. Elon is famously bad at estimating, so we can also assume that this is his honest belief. That makes a chain of obviously true statements (or close to true, in the case of operating costs), but none of them actually tell us that this will be cheap or economically attractive. And all of them are complete non-sequiturs.
If you ramp up the economies of scale to make those things - radiation protection and cost per pound - the calculus changes. It's supposed to synergize with Starship, and immediately take advantage of the reduced cost per pound.
If the cost per pound, power, regulatory burden, networking, and radiation shielding can be gamed out, as well as the thousand other technically difficult and probably expensive problems that can crop up, they have to sum to less than the effective cost of running that same datacenter here on earth. It's interesting that it doesn't play into Jevon's paradox the way it might otherwise - there's a reduction in power consumption planetside, if compute gets moved to space, but no equivalent expansion since the resource isn't transferable.
I think some sort of space junk recycling would be necessary, especially at the terawatt scale being proposed - at some point vaporizing a bunch of arbitrary high temperature chemistry in the upper atmosphere isn't likely to be conducive to human well-being. Copper and aluminum and gold and so on are also probably worth recovering over allowing to be vaporized. With that much infrastructure in space, you start looking at recycling, manufacturing, collection in order to do cost reductions, so maybe part of the intent is to push into off-planet manufacturing and resource logistics?
The whole thing's fascinating - if it works, that's a lot of compute. If it doesn't work, that's a lot of very expensive compute and shooting stars.
Or, just saying, be critical of ideas and think them through, and take in what experts say about it, and determine for yourself what's up. If a bunch of people who usually seem to know what they're talking about think there's a legitimate shot at something you, as a fellow armchair analyst, think is completely impractical, it makes sense to go and see if maybe they know something you don't.
In this case, it's all about Starship ramping up to such a scale that the cost per pound to orbit drops sufficiently for everything else to make sense - from the people who think the numbers can work, that means somewhere between $20 and $80 per pound, currently at $1300-1400 per pound with Falcon 9. Starship at scale would have to enable at least 2 full orders of magnitude decrease in price to make space compute viable.
If Starship realistically gets into the $90/lb or lower range, space compute makes sense; things like shielding and the rest become pragmatic engineering problems that can be solved. If the cost goes above $100 or so, it doesn't matter how the rest of the considerations play out, you're launching at a loss. That still might warrant government, military, and research applications for space based datacenters, especially in developing the practical engineering, but Starship needs to work, and there needs to be a ton of them for the datacenter-in-space idea to work out.
Or, just saying, we should eat babies because they are abundant and full of healthy nutrition for adult humans. [1]
Just because an idea has some factors in its favor (Space-based datacenter: 100% uptime solar, no permitting problems [2]) doesn't mean it isn't ridiculous on its face. We're in an AI bubble, with silly money flowing like crazy and looking for something, anything to invest it. That, and circular investments to keep the bubble going. Unfortunately this gives validation to stupid ideas, it's one of the hallmarks of bubbles. We've seen this before.
The only things that space-based anything have advantages on are long-distance communication and observation, neither of which datacenters benefit from.
The simple fact is that anything that can be done in a space-based datacenter can be done cheaper on Earth.
The idea here is that the economics of launch are changing with Starship such that the "incredible cost" and "overspeccing" of space will become much less relevant. There's a world where, because the cost per kg is so low, a data center satellite's compute payload is just the same hardware you'd put in a terrestrial rack, and the satellite bus itself is mass-produced to not-particularly-challenging specs. And they don't have to last 30 years, just 4-ish, when the computer is ready for retirement anyway.
Will that come to be? I'm skeptical, especially within the next several years. Starship would have to perform perfectly, and a lot of other assumptions hold, to make this make sense. But that's the idea.
My point is even if it were free to put things in space and radiation did not need mitigation, you're still paying a lot to have hardware that can't be maintained. If it were cheaper we wouldn't be doing online maintenance on Earth. Name a single datacenter on the rocky surface of the Earth that is opting to not have maintenance.
Unless I missed something the Microsoft underwater data center was basically a publicity stunt.
Anyone who thinks it makes sense to blast data centers into space has never seen how big and heavy they are, or thought about their immense power consumption, much less the challenge of radiating away that much waste heat into space.
Well the thing is that it seemed to have been successful beyond all expectations despite being that? They had fewer failures due to the controlled atmosphere, great cooling that took no extra power, and low latency due to being close to offshore backbones. And I presume you don't really need to pay for the land you're using cause it's not really on land. Can one buy water?
Space is pretty ridicolous, but underwater might genuinely be a good fit in certain areas.
Hot saltwater is the worst substance on earth, excepting, maybe, hydrofluoric acid. You really don't want to cool things with ocean water over an extended period of time. And filtering/purifying it takes vast amounts of power (e.g. reverse osmosis).
I was listening to a Darknet Diaries episode where Maxie Reynolds seems to make it work: https://subseacloud.com/ I don't know how profitable they are, and I doubt this is scalable enough, but it can work as a business.
An 8 GPU B200 cluster goes for about $500k right now. You'd need to put thousands of those into space to mimic a ground-based data center. And the launch costs are best case around 10x the cost of the cluster itself.
Letting them burn up in the atmosphere every time there's an issue does not sound sustainable.
Are launch costs really 10x!? Could I get a source for that?
In the back on my head this all seemed astronomically far-fetched, but 5.5 million to get 8 GPUs in space... wild. That isn't even a single TB of VRAM.
Are you maybe factoring in the cost to powering them in space in that 5 million?
The Falcon Heavy is $97 million per launch for 64000 kg to LEO, about $1,500 per kg. Starship is gonna be a factor 10 or if you believe Elon a factor 100 cheaper. A single NVidia system is ~140kg. So a single flight can have 350 of them + 14000kg for the system to power it. Right now 97 million to get it into space seems like a weird premium.
Maybe with Starship the premium is less extreme? $10 million per 350 NVidia systems seems already within margins, and $1M would definitely put it in the range of being a rounding error.
But that's only the Elon style "first principles" calculation. When reality hits it's going to be an engineering nightmare on the scale of nuclear power plants. I wouldn't be surprised if they'd spend a billion just figuring out how to get a datacenter operational in space. And you can build a lot of datacenters on earth for a billion.
If you ask me, this is Elon scamming investors for his own personal goals, which is just the principle of having AI be in space. When AI is in space, there's a chance human derived intelligence will survive an extinction event on earth. That's one of the core motivations of Elon.
A Falcon Heavy takes about 63 tons to LEO, at a cost of about $1,500 per kg.
A server with 4x H200s and some RAM and CPU costs about $200k, and weighs about 60kg, with all the cooling gear and thick metal. As is, it would cost $90k to get to LEO, half of the cost of the hardware itself.
I suppose that an orbit-ready server is going to cost more, and weigh less.
The water that serves as the coolant will weigh a lot though, but it can double as a radiation shield, and partly as reaction mass for orbital correction and deorbiting.
Power draw is max 10.2 kW but average draw would be 60-70% of that. let's call it 6kW.
It is possible to obtain orbits that get 24/7 sunlight - but that is not simple. And my understanding is it's more expensive to maintain those orbits than it would be to have stored battery power for shadow periods.
Average blackout period is 30-45 minutes. So you'd need at least 6 kWh of storage to avoid draining the batteries to 0. But battery degradation is a thing. So 6 kWh is probably the absolute floor. That's in the range of 50-70 kg for off-the-shelf batteries.
You'd need at least double the solar panel capacity of the battery capacity, because solar panels degrade over time and will need to charge the batteries in addition to powering the gpu's. 12 kW solar panels would be the absolute floor. A panel system of that size is 600-800 kg.
These are conservative estimates I think. And I haven't factored in the weight of radiators, heat and radiation shielding, thermal loops, or anything else that a cluster in space might need. And the weight is already over 785 kg.
Using the $1,500 per kg, we're approaching $1.2 million.
Again, this is a conservative estimate and without accounting for most of the weight (radiators) because I'm too lazy to finish the napkin math.
I think we're on the same page. Lifting the actual computing devices would be not that expensive, compared to lifting a lot of other related mass, principally the cooling systems, and the solar panels.
The solar panels used in space are really lightweight, about 2 kg / m² [1], it's like ten times lighter weight than terrestrial panels. Still they need load-bearing scaffolding, and electrical conductors to actually collect the hundreds of kilowatts.
You mean you operate them like Microsoft's failed submerged data center project [1]. When pointing at validating past examples you are generally supposed to point at successes.
> By 2024, Project Natick had been inactive for several years, though it was referenced in media as though it was ongoing. That year, Microsoft confirmed that the project was inactive and that it had no servers underwater.
I wouldn't exactly call this a success, for that matter.
To me, failed, implies some sort of real failure, not just, "eh, won't make us enough money" a la Google/business since forever/the exec who's pet project it was moved on/had babies/was fired for unrelated reasons/some other human thing unrelated to the technical proposition.
If, like, sea-water entered and corroded the system and it blew up and ate babies, and caused Godzilla, that would be a failure. It just being not quite interesting enough to go after seems... I mean I guess it is, but on a "meh" level.
"eh, won't make us enough money" would make space data centres a failure too. The whole point is to do compute more cheaply than a traditional data centre.
But if we’re going down that line of thinking then it’s a poor comparison. I could open a data centre on the ground and use the same principle of zero maintenance, and it would be way cheaper and way more powerful.
This will totally work since we have an unlimited amount of rare earth elements we can just ship off into space never to see again. Infinite raw materials + infinite power equals infinite AI!!!
My understanding was that access to very large body of cold water was a core feature for the project. The water was to be used for cooling relatively efficiently or cheaply.
My point was that they are both quite hostile environments for different reasons. In the same way space has abundant power supply, subsea has an abundant heat sink.
No operational needs is obviously ... simplified. You still need to manage downlink capacity, station keeping, collision avoidance, etc. But for a large constellation the per-satellite cost of that would be pretty small.