HN Reader

What really worries me is that I keep hearing "cooling is cheap and easy in space!" in a lot of these conversations, and it couldn't be farther from the truth. Cooling is _really_ hard and can't use efficient (i.e. advection-based air or water cooling) approaches and are limited to dramatically less efficient radiative cooling. It doesn't matter that space is cold because cooling is damned hard in a vacuum.

The article makes this point, but it's relatively far in and I felt it was worth making again.

With that said, my employer now appears to be in this business, so I guess if there's money there, we can build the satellites. (Note: opinions my own) I just don't see how it makes sense from a practical technical perspective.

Space is a much harder place to run datacenters.

> This is all to say that the current discourse is increasingly bothering me due to the lack of rigor; people are using back-of-the-envelope math, doing a terrible job of it, and only confirming whatever conclusion they already want. Calculating radiation and the cost of goods is not difficult. Run the numbers.

> References: Gemini, Gemini, ChatGPT, ChatGPT, Gemini, ChatGPT, Gemini, ChatGPT, Grok, Gemini (There are sub-references from these services in the GitHub.)

I think, if you're going to make statements like this - especially from a position of expertise, you should be personally verifying the numbers and citing their sources directly. What good is asking the reader to trust an AI on your behalf? They should trust you.

(To be clear, I suspect the conclusions drawn are still correct.)

Will these space-based data centers run on rad-hard silicon (which is dog slow compared to anything on Earth) or just silently accept wrong results, hardware lockups and permanent failure due to the harsh space environment? Will they cool that hardware with special über-expensive high-temperature Peltiers that heat the radiators up to visible incandescence so that the heat can be shed with any efficiency? There's zillions of those issues. The whole idea is just bonkers.

The analyses here miss the economic realities of building datacenters. "Just use land", "just use nuclear", "just use water". All of this is contested. A system of lawsuits and regulations turns negative externalities (even ones you aren't convinced of!) into costs you can weigh against. So, like hydrogen vs. RP-1, it's not enough to pick a handful of physical performance metrics. It has to win holistically.

If you can produce any kind of economically productive compute node and add it to (for example) the Starlink network, and launch on a reusable vehicle, you carry on installing them as fast as you can build them.

So, the move is to turn the problem of contested land use into a manufacturing problem.

This is not so easy to pin down on a spreadsheet, and will be decided at the level of the business unit. If SpaceX can put a GPU/TPU on the grid more economically than the other guy, then it doesn't matter if they have ammonia in the pipes instead of water.

Grab your popcorn.

This is an interesting analysis, and I like the sliders that let you instantly show the impacts of system trades.

The one glaring hole that I see is the challenge of moving the data to/from the datacenter while it's on orbit. Bandwidth to/from space isn't free. FCC/ITU licenses are required, transmitters/receiviers/modems/DSP/antennas all add to SWAP (size, weight, and power). Ground-stations are needed to move the data up/down, but those have recently become a commodity too. Still, they're not free. (see: https://aws.amazon.com/ground-station)

There is also the added latency between earth-based users and space-based datacenters, which may be a deal breaker for some applications.

Another issue I don't see covered are the significant differences between space-based hardware and terrestrial hardware. The space stuff needs to be radiation tolerant, and that usually makes it a lot slower and a lot more expensive than the terrestrial stuff, all other things being equal.

In the end, space-based datacenters are highly impractical even if you assume that Starship can put anything into orbit very cheaply.

Does anybody actually work with H100s and the like? Their failure rate is so high, I dont understand why anybody will even consider it feasible to put the machines in orbit or even the sea. By my ballpark estimate, if you have 800 H100s, after 6 months, about 100 would be overheating or throttling, and a few will disappear and one or two will crash the machine with load.

Here's some math on how affordable that abundant LEO solar energy is:

First you have to pay energy to get to LEO

A Starship Launch costs[0] 51.75 TJ of energy in terms of its methane fuel.

It will be able to take a payload of 150 tonnes or 331,000 pounds[1].

How many computers is that?

One online estimate says a computer weights 80 lbs or 35 kg.

So 150000 kg / 35 kg/computer = approximately 4285 computers that we can launch into orbit per Starship.

51.75TJ / 4285 computers = approximately 12.08 GJ per computer to place it in orbit.

Let's say each computer is a H200 and consumes 700 watts continuously. How long would it need to run in orbit before it used as much energy for computation as it took to launch it?

12.08 GJ / 700 W = 12,080,000,000 J / 700 J/s = approximately 17,257,143 seconds.

Or about 6.5 months to break even on energy.

That sounds pretty good, except my estimate for the weight of each compute unit and associated power system & cooling etc. are probably underestimates by one or two orders of magnitude. In which case you'd be looking at 5 to 50 years to break even on energy, by which time the chips are obsolete and need to be replaced anyway.

[0] https://space.stackexchange.com/questions/66480/how-much-ene... [1] https://en.wikipedia.org/wiki/SpaceX_Starship#Description

I am struggling with a why for this (other than “huh cool, that will get investors”). All the jurisdiction and regulation arguments and the “we could get the costs down” seem to meet the objection of “for the same investment we could do just as well or better on the ground”.

The one that does not is the physics of the whole thing. I struggle to work out how exactly but being slightly time dilated compared to the ground does not seem like a win, but being able to gather data from opposite sides of the planet slightly faster than cables does seem like a potential win. Most stock exchanges make a significant chunk of their revenues renting out data space, so it seems a possibility.

But either way it seems very niche.

François Chollet https://x.com/fchollet/status/1999982683708150014 :

"Datacenters in space" make for a catchy narrative and an interesting demo, but the math simply doesn't work.

When considering factors like launch cost, maintenance complexity, and the cost of high-bandwidth communications (latency included), there is no realistic set of economic and engineering assumptions under which orbiting datacenters become cost-competitive with simply building conventional nuclear-powered (or renewable energy-powered) datacenters on the ground.

In fact we're off by 50-100x. Dramatic launch cost reductions still won't make it work. And of course if you invest a lot in specific lines of tech to make it work you then have to consider that the same can also be invested in better ground-based nuclear, bringing the cost of power down for everyone.

I love the sliders, but note that the numbers on this site literally came from ChatGPT, so there is plenty of room for disagreement.

Seems like according to this analysis it all hinges on launch cost and satellite cost. This site's default for Starship launch cost is $500/kg, but SpaceX is targeting much lower than that, more like $100/kg and eventually optimistically $10/kg (the slider doesn't even go that low). At $100/kg (and assuming all the other assumptions made on the site hold) then you break even on cost vs. terrestrial if you can make the satellites for $7/watt (excluding GPUs, as the whole analysis does).

Scott Manley had a video about this last year.

https://www.youtube.com/watch?v=d-YcVLq98Ew

The short of it is that cooling is likely the biggest problem, given you will need to pump the heat to the backside and radiate it away, and the amount of mass you will need to dedicate to cooling works against deployments and increases the cost per unit significantly. Not to mention, the idea of these huge deployments runs into potential space debris issues.

Whenever one of these ventures actually manages to launch a proof of concept, I think we'll be able to quickly discern if there is actually a near-future here.

> That "why" is almost missing from the public conversation. People jump straight to hardware and hand-wave the business case, as if the economics are self-evident. They aren't.

But then he never answers that fundamental question, and jumps straight to the hardware and power and cost? What problems are orbital data centers trying to solve? What optimizations are they intended to deliver? Are these optimizations beneficial to everyone who uses a data centers, or just operators or users of orbiting satellite constellations?

> But the knock-on effects are why this keeps pulling at people. If you can industrialize power and operations in orbit at meaningful scale, you're not just running GPUs. You're building a new kind of infrastructure that makes it easier for humans to keep spreading out. Compute is just one of the first excuses to pay for the scaffolding.

This seems to be the closest we get to a “Why”, but it doesn’t make much sense. A constellation of 40,000 satellites with GPUs “infrastructure that makes it easier for humans to keep spreading out”? How?

> The target I care about is simple: can you make space-based, commodity compute cost-competitive with the cheapest terrestrial alternative? That's the whole claim. … Can you deliver useful watts and reject the waste heat at a price that beats a boring Crusoe-style tilt-wall datacenter tied into a 200–500 MW substation?

Isn’t the answer clearly “No”? The default settings of his model — which I assume he considers optimal — tell us that power for orbital data enters will cost 3.5X terrestrial ones. And that only SpaceX has the vertical integration to do even attempt to do this. So again, where is the competitive advantage?

Also, I don’t understand why he’s including satellite construction and launch costs for a 40,000 satellite constellations in his analysis, if he’s assuming SpaceX as he claims. Wouldn’t SpaceX simply implement these compute capabilities in the next gen of Starlink, so which would reduce costs significantly.

> It might not be rational. But it might be physically possible.

But isn’t that precisely what everyone has been saying? I don’t think the question has been whether orbital data centers are possible, it’s been whether they are rational. And that centers foremost h the unanswered question, Why is this a good idea?

OP here. Happy to answer any questions, appreciate the boost.

When Starcloud put together that whitepaper the first thing I looked at was the launch costs[1]. It references a $5M cost to launch, which right away made absolutely no sense to me. Just a cursory search shows launch costs are around $50M per launch, if not more.

It's great that this site drills down even further to demonstrate that there is absolutely no point at which the launch costs ever make this economical or viable, so I really don't understand what people are doing.

Especially because this site was harping for years about the cost of launches and putting things in to orbit, the whole reason why SpaceX got started and has grown as it has. As soon as that became an inconvenient number, we now just make things up (Just pretend that launch costs are 10% of what they actually are to get people to invest?).

[1]: https://starcloudinc.github.io/wp.pdf

Did a similar back-of-the-napkin and got 5x $ / MW of orbital vs. terrestrial. This article's analysis is ~3.4x.

I do wonder, at what factor of orbital to terrestrial cost factor it becomes worthwhile.

The greater the terrestrial lead time, red tape, permitting, regulations on Earth, the higher the orbital-to-terrestrial factor that's acceptable.

A lights-out automated production line pumping out GPU satellites into a daily Starship launch feels "cleaner" from an end-to-end automation perspective vs years long land acquisition, planning and environment approvals, construction.

More expensive, for sure, but feels way more copy-paste the factory, "linearly scalable" than physical construction.

tbh this feels lot like people throwing Drake equation around. You put in whatever random numbers together and you can get any result you want.

Aside from the economics, the question is why do it in orbit vs on land (or sea)?

What are the regulatory/legal gains? Lack of jurisdiction means open slather?

What are the national security gains? Redundancy and resiliency by each satellite being a "micro-compute" connected by high speed laser links? So more resilient to attack?

Why do it at all?

Great article, with dollars and numbers as it should be.

The proponents of orbital wunderbars may have different calculations and there's only one way to resolve the differences - make them pay for their space-centers themselves, making sure no public money is spent on the project which is also required to carry zero default risk and pay regular taxes without the right to claim deductions or write-offs for any losses. All this is necessary to avoid the usual corporate tax games.

I can sketch a contract to that end in like 10 minutes. Everything else is a waste of time.

Economics for who? The builder of the data center and plethora of all contractors and sub-contractors would see great economics though. Even the sponsor/owner of the data center might see economics work out, if you consider the reputational gain (why did we land on Moon? what's the economics?), experience gained and considering the burn of someone else's money. The money of mega companies that go into this kind of "monuments" is not exactly theirs.

One of the main reasons for putting "compute" and "storage" in space is that it is out of reach for the general public and would allow for more stability in exceptionally intense tyranny.

It is much easier to blow things up on land than in space, and the 'negative externalities' simpler to make assumptions about.

The value of this to the people who would be in charge of this "compute" and "storage" is likely much larger than the difference in energy cost.

AFAIK compute heavy datacenter in space don't work. But if you already have a vast fleet of laser connected LEO satellites throwing some efficient SSDs into them can make a lot of sense. A large portion of the traffic is fairly static, e.g. video based content or even model weights. Caching that will save you ground to space side of the transmission. This will let you put more user beams on satellites and use less ground stations.

this does not makes sense from dollars pov to me, I ran a back of napkin session with claude & gemini on this and the short of it is you need a magical weightless radiator for cooling and even then it wont work because the launch costs need to be sub $100 before this can be feasible. this does not even factors the 5y amortization and LEO orbit drag correction.

It then occurred to me that they (all major AI companies) know all of these facts but still pushing for it so there must be another reason. Then I recalled the offhand statement from the openAI lady about govt backstop for infra, which was strongly opposed by public and AI czar. this might be be a backdoor way of injecting that backstop capital in terms of subsidies now for results in 5 years or so. and needless to say after pilot programs those will fail spectacularly.

I'm not really interested in the problems that can come with orbital compute. We've seen them listed ad nauseam.

Have we seen any benefits to orbital computing by launching a cluster of raspberry pis to LEO? Surely this isn't an impossible task to test out on a smaller scale?

I’d love to see these variables fitted to learning curves. That would give you a forecast for when, if learning continues as predicted, the economics could be competitive. (If it doesn’t, you need a new paradigm first.)

On the ground you don’t have to fling tons of sensitive electronics to orbital speeds, you can have someone pop by to check whether everything is ok if you get an alert, you have a thick atmosphere cutting down solar radiation, you can cool everything relatively cheaply, and you can run fiber and power cables directly into the building. In space ou have to design super robust and self maintaining machinery. It’s a cool marketing stunt, but I don’t get the economics behind doing it for real.

I suspect that this orbital data centers isn't entirely about dollars (No doubt dollars are important).

I suspect it is about the regulatory environment. The regulatory environment on data centers is moving quickly. Data centers used to be considered a small portion of the economy and thus benign and not worth extorting/controlling. This seems to be changing, rapidly.

Given that data centers only exchange information with their consumers they are a natural candidate for using orbit as a way to escape regulators.

Further, people are likely betting that regulators will take considerable time to adjust since space is multinational.

Very doable with Photonic compute in orbit: Way less waste heat (huge when you can only radiate it away), lower power draw per FLOP (smaller solar arrays), and photons don't give a damn about cosmic radiation causing bit flips.

I am reminded of how space exploration has come largely before deep ocean exploration, seems like a human bias.

Putting data centers under water makes way way more sense than into space.

This is very well done! I love including all the sliders so that we can play with the (reasonable looking) assumptions the author has made. Like the author, I share their surprise that the result did not come out even more in favor of terrestrial GPUs.

Really interesting article. Curious to see the ChatGPT link attached

But when I click on it, I get this error.

Failed to load shared conversation. Request is not allowed. Please try again later. (403, 9aebe525df75165e-BLR)

Hilarious how they could easily validate this by PoC. Even putting a small RasPi cluster in orbit. Do that and make a profit. Then ask for money.

Terrestrial may be cheaper but it can be burned down by peasant mobs. To become an immortal god you must remove all meat-based legacy threats.

I guess I'd assume that the premise driving this would be that there will eventually be enough business in space that it's necessary for space-centric use, and that terrestrial use is just a fringe benefit or loss leader or something.

But oddly this doesn't seem to be how the concept is typically framed.

My second level curiosity is how much cheaper/competitive it'd be if we had space elevators.

The 'Satellite Size' slider does not seem to impact the cost, also not in relation to changing the other settings?

I sometimes wonder if there are people out there who just read too much Neuromancer, and they think they can construct their own Tessier-Ashpool orbital dynasty.

I suppose there are several other Oligarchs In Space stories and movies since then, but if the point of the space station is to host AI, that narrows it down a bit.

Or perhaps it's performative, designed to spook gullible politicians into changing laws to "keep" businesses that were never actually going to go somewhere else anyway.

For someone who is "increasingly bothered by the lack of rigor in the current discourse", the author sure has no qualms about using LLM outputs as primary sources. This is an immediate red flag that discredits the entire article.

Cons of orbital data centers:

- Ludicrously expensive to setup

- Need radiation-hardened silicon

- Ludicrously expensive maintenance requiring highly specialized operators (a.k.a astronauts)

- High risk of losing the entire equipment to a rocket failure (not infrequent even for modern launch vehicles)

- Supplying enough electrical power would be extremely difficult

- Cooling would be extremely difficult

- Geosynchronous orbits have at least 200ms of communication latency

- Lower orbits means the data center would not stay in place and require complicated tracking antennae and/or a communication mesh a la Starlink, again increasing latency and complexity

Pros of orbital data centers:

- ??????

...why are we doing this again?

do it on mars then... that would have the added benefit of heating the planet so we could live on it. It seems so obvious if you think about it. Someone transfer eleventy trillion dollars to Elon Musk so we can get started.

I realize terrestrial data centers have environmental risks, but are the risks greater for an orbital data center? I would think space debris, solar flares, or a bad actor satellite with a laser could do a lot of damage. Good luck repairing the orbital data center.

it would only make sense if land availability for DCs is 0 or truly expensive…

Someone might be foreseeing an scenario like that? Are satellite launchers behind this hype?

> ...we should be actively goading more billionaires into spending on irrational, high-variance projects that might actually advance civilization. I feel genuine secondhand embarrassment watching people torch their fortunes on yachts and status cosplay. No one cares about your Loro Piana.

I 100% agree with this. There are ~2,600 billionaires in the world and we should encourage all of them to spend their money. Even buying a superyacht is a benefit to the economy. But the best billionaires, like Bill Gates and Elon Musk, are actually trying to advance the tech tree.

We are honestly lucky that Musk is wired funny. Any normal human being would retire and hang out on the beach with supermodels after all the abuse he has taken. But he takes it all as a personal challenge and doubles down. That is both his worst quality and his best.

As an expert in this field, I can assure that cost is not the most important factor. Demand is.

For someone engaging in a lot of fun, sci fi utopian thinking, he still falls prey to libertarian thought:

"I'll go one step further and say the quiet part out loud: we should be actively goading more billionaires into spending on irrational, high-variance projects that might actually advance civilization. I feel genuine secondhand embarrassment watching people torch their fortunes on yachts and status cosplay. No one cares about your Loro Piana. If you've built an empire, the best possible use of it is to burn its capital like a torch and light up a corner of the future. Fund the ugly middle. Pay for the iteration loops. Build the cathedrals. This is how we advance civilization."

That can be done easily (and has been done many times in the past! And in the present, elsewhere in the world outside the US!) by TAXING the billionaires and using that money for government funded research programs such as DARPA, NSF, national space programs that are actually ambitious and risk taking and held to timelines.

Americans need to get over this idea that billionaires are gods that we must pray to and instead see them as just normal citizens who need to be taxed way more.

I have no idea what this Grok assisted article is trying to say. But the data-center-in-space hype is irrational. It hand-waves cooling and bit flip errors. It does not explain why we need chat bots in space (we don't need them on earth either).

It is a nice talking point for the U.S. Saudi Investment Forum. The Saudis apparently buy anything:

https://xcancel.com/elonmusk/status/2000603814249079165#m

The only benefit as I perceive it re: orbital data center hardware is regulatory avoidance. Think...DDOS machines that can't be shut off; or financial hosting services for unsavory individuals. However, it's very expensive by all metrics (including those talked about in the article), and frankly, these satellites are sitting ducks for the hunter killer satellites the various space powers have, if they actually wanted to do something about these hypothtical data centers and the problems they would cause.

This is AI slop in a pretty dress. It's fascinating that space-based datacenters are such a catastrophically bad investment that even limp apologia like this can, at best, argue that maybe it's not quite as bad as you think, as long as you still manage to ignore half the costs because you're a loser who delegated your thinking to a chatbot.

the cost of having very fast up/down syncrounous data/compute,comunications to any random spot on the planet, at any time, in a very hard to interupt or detect manner, is, what it is

and can not be compared to anything else

so the conclusion is orbital datacenter is a very bad idea ?

There was already an article of how this thing is just ain’t gonna happen:

https://taranis.ie/datacenters-in-space-are-a-terrible-horri...

—

If I were to guess, my first bet would be grand PR damage control for all the Mexicans, Irish, and what have you as in “don’t worry, we’ll soon be in space and out of your backyard” (no, they won’t).

https://www.nytimes.com/2025/10/20/technology/ai-data-center...

Right now all my favourite pirate sites are based in Russia but moving them to extrateritorial space would be genius!

Americans couldn't shoot at it in fear of igniting a space war with China.