Jeff Bezos throws his hat in the ring for an orbital data center megaconstellation, too

"Space-based data centers will be a complement to terrestrial infrastructure."

See full article...

 

[AdrianS]

Is this article intended as rage bait? Posit a patently unnecessary and ruinously expensive idea. Then offer zero critique of that proposal. I expect more from ars than this.

The arguments that the companies make themselves on their own websites are ridiculous. There is no data center land crisis, nor energy crisis that is best solved by putting mass compute into orbit. Leave aside the dubious financials. Wtaf

We've had quite a few uncritical "press release as news" space articles recently.
It is what it is.

 

[aenoiclurker]

Let's stop and consider your 144 kW per rack of electrical power. If we're in sun-synchronous orbit at the terminus, we can expose our solar panels 100% of the time, but they're only going to be about 30% efficient at converting solar power into electricity.

[...]

So please, feel free to crap on the idea of datacenters in space for any number of compute or financial reasons. But let's retire the idea that cooling in space is a technology bottleneck.

The ISS EEATCS ammonia-loop system can radiate a maximum 70kW of waste heat. It weighs 14 metric tons, which a quick google suggest will cost around $100 million to launch into orbit. So more than $200 million for a single rack of processors just for the cooling. Now add all the stuff you actually need to do any useful work.

I will indeed crap on the idea of datacenters in space because an impossibly expensive solution that makes zero economic sense isn't going to happen, even if theoretically you could engineer around the cooling problem.

Edit: Shit. I didn't account for a failure, so double everything for redundancy. You're looking at a billion dollars in overheads per rack of capacity, at a time that AI companies are haemorrhaging money. It's a fantasy.

 

[@inemesitaffia]

Someone call me when any of these rich idiots have any idea how they are going to cool megawatts of computers in the vacuum of space.

Lines like these are about stock price manipulation.

Stock in a privately singly held company.

 

[aenoiclurker]

This thorough, factual comment is very welcome. Contrast with people here complaining about the insane space data center gold rush while at the same time skim-reading the content and comments they're replying to in their own desperate rush to get their comments onto the first page.

I see your comment got more downvotes than upvotes, but no counterarguments. Therefore all the negative vote count indicates is the degree to which the downvoters lost the argument.

The OP ignores the reality in which any cooling system exists. If it's so expensive as to be commercially dead on arrival then pointing out what we already know doesn't add to the argument. The ISS exists and everybody isn't dead so we already know that it's possible to cool satellites in space.

If cooling was the only blocker I'd have more patience with this kind of BS, but Microsoft tried putting a datacenter under water to test if cooling was easier. It was. A hugely successful experiment in cooling, according to MS, but commercially non-viable and never repeated because the problems of connectivity and access for maintenance made the whole thing a non-starter.

I'm sure putting your delicate electronics in multiple satellites in <checks notes> fricking space will make project-killer issues with connectivity and access go away.

 

[JoHBE]

At the current rate that literally EVERYTHING seems to be imploding Down Here, by the time these fantasy constellations are up there, they will be everything that is left.

The disconnect is astonishing.

 

[JoHBE]

This line really should not be so matter of fact. It is an extremely speculative statement. It's not clear how this would improve, let alone enable, AI. It also doesn't really say what kind of AI it would instantiate, as current LLM's certainly don't need to be in space. It's not even clear in-space processing is that good of an idea in general. Indeed, there's a lot to argue against it.

The level of discourse and thinking and seriousness , over the last year and a half, has been set to an historically low level by the White House. And it drags everything with it. As long as consequences are sufficiently delayed, a plausible relation to reality is entirely unnecessary.

 

[Doomlord_uk]

I also don't see

So the alleged data center in space uses a really long USB cable to communicate with a single point on the ground? Or do smoke signals work in space?

OK so if the ground stations are not in the US, what does the FCC care then?

 

[VoterFrog]

Can we please, please, please, retire concerns about cooling in space? For all the problems the idea of datacenters on-orbit have, cooling is not one of them. Or, at least, the math actually closes. This is at least the third time I've walked through this exercise, but perhaps this time it will stick.

Let's stop and consider your 144 kW per rack of electrical power. If we're in sun-synchronous orbit at the terminus, we can expose our solar panels 100% of the time, but they're only going to be about 30% efficient at converting solar power into electricity.

For every square meter of solar panel pointed perfectly perpendicular to the sun, there's 1361 W/m^2 of solar power of which 400 W/m^2 will turn into useful electrical power. That means that for every 144 kW rack you describe above, you need to have 352 m^2 of solar panel to power it. That's a square 18.7 m on a side. Are you starting to see why the cooling issue closes?

The important thing to realize is that for all the massive real estate I'm going to have to allow for solar panels also provides a massive amount of real estate for radiators. And since the solar panels are always pointed at the sun, the back side is always pointed at empty space.

At this point, we have to make assumptions about surface temperatures, etc. but there are a few points to realize. If the radiators and the solar panels are the same temperature (given that the former is on the back side of the latter), then we have to consider that both sides of the solar panel / radiator are actually acting as radiators. Sure, in the direction of the sun the solar panels are certainly receiving a net amount of energy. However, for most of the hemisphere where a photon can try to emit, it's empty space too. If you're going to use heat pumps or the like to increase the radiator temperature, then you can mostly ignore the radiation from the solar panel side and have to consider just the area of the radiator (especially since you'll need more power to drive the heat pumps and therefore more solar panels).

So, for grins, let's consider a double-sided system (front and back) that has to reject (1361 W/m^2) / 2 . Use your favorite online blackbody calculator to determine the surface temperature of an ideal surface that can reject the amount of power that the sun puts on the panels. In the end, you'll find that you need 331 K or a 58 deg. C skin temperature. That's hot to touch, but certainly not too hot to run integrated chips or solar panels.

If we distribute your 144 kW of compute in that 352 m^2 area, there won't even need to be hot spots. If you're being smart and having a separable compute node that connects to cooling and power, you'll probably want a liquid loop of some sort to draw the heat away. You'll probably need 10 deg. C higher temperatures on your GPUs than the final 58 deg. C mark, but that's still very reasonable.

So please, feel free to crap on the idea of datacenters in space for any number of compute or financial reasons. But let's retire the idea that cooling in space is a technology bottleneck.

Is anybody making the argument that cooling is physically impossible or are you attacking a straw man? The cooling loop you glossed over makes the server rack the largest non-human-carrying payload ever launched into space (unclassified, at least). How is that not a bottleneck to wanting to launch thousands of them into space? These guys will essentially be in business launching cooling equipment into space... With a little bit of data center attached.

 

[Argent Claim]

This thorough, factual comment is very welcome. Contrast with people here complaining about the insane space data center gold rush while at the same time skim-reading the content and comments they're replying to in their own desperate rush to get their comments onto the first page.

While it is indeed a "factual comment", it also has little to do with the one it was replying to nor the comment that that comment was replying to.

The problem is not cooling as an engineering issue but cooling as a cost issue. Yes, the cost of keeping a radiator running in space is effectively free (at least if you're not launching repair or maintainance missions to it), but the cost of launching that equipment in the first place is not free and constellation-type data centers would literally be required to replace their cooling equipment (like everything else) on a regular basis to destructive re-entry of the entire platform on which they are mounted.

Radiators are more massive and mechanically complicated than Solar panels of the same surface are, and the former will cut into the amount of computing power that can be delivered per launch while both are additional marginal production costs of sorts that proponents of orbital data centers have left totally unaccounted for in their cost estimates.

Also (and this is amusingly not the first time I've seen it for this topic) the whole proposal of adding radiators behind photovoltaics for, "real estate" is absolutely pointless. It's space, not a dense urban environment or claustrophobic subterranean bunker. There is no need to constrain deployed volume in such a way; just mount the radiator arrays somewhere else.

I see your comment got more downvotes than upvotes, but no counterarguments. Therefore all the negative vote count indicates is the degree to which the downvoters lost the argument.

Unless you can read minds, all down voting means is that people clicked the downvote button. Imagining the motivations of why people downvoted the comment is a waste of time.

 

[Wickwick]

The ISS EEATCS ammonia-loop system can radiate a maximum 70kW of waste heat. It weighs 14 metric tons, which a quick google suggest will cost around $100 million to launch into orbit. So more than $200 million for a single rack of processors just for the cooling. Now add all the stuff you actually need to do any useful work.

I will indeed crap on the idea of datacenters in space because an impossibly expensive solution that makes zero economic sense isn't going to happen, even if theoretically you could engineer around the cooling problem.

Edit: Shit. I didn't account for a failure, so double everything for redundancy. You're looking at a billion dollars in overheads per rack of capacity, at a time that AI companies are haemorrhaging money. It's a fantasy.

Hmm, that seems to fall into an economic argument. The last paragraph of my post almost invites that criticism:

So please, feel free to crap on the idea of datacenters in space for any number of compute or financial reasons. But let's retire the idea that cooling in space is a technology bottleneck.

 

[Wickwick]

Is anybody making the argument that cooling is physically impossible or are you attacking a straw man? The cooling loop you glossed over makes the server rack the largest non-human-carrying payload ever launched into space (unclassified, at least). How is that not a bottleneck to wanting to launch thousands of them into space? These guys will essentially be in business launching cooling equipment into space... With a little bit of data center attached.

The post I quoted was the third exchange of a discussion about the possibility of cooling. So yeah, not a strawman.

 

[fenris_uy]

Is anybody making the argument that cooling is physically impossible or are you attacking a straw man? The cooling loop you glossed over makes the server rack the largest non-human-carrying payload ever launched into space (unclassified, at least). How is that not a bottleneck to wanting to launch thousands of them into space? These guys will essentially be in business launching cooling equipment into space... With a little bit of data center attached.

https://meincmagazine.com/space/2026/...lation-too/?comments=1&post=44319961#comments

So, you decided to crap in a comment before reading the thread of comments back to see if your comment was right?

Commenters have been saying that you can't cool it. Not that it's too expensive, but that the engineering doesn't adds up.

When Musk talks about this, you can assume that he is ignoring the engineers and accountants in SpaceX saying that this is too expensive because he wants to prop up SpaceX value for the IPO.

But as far as we know, BO isn't accepting outside investors, so if Bezos is also talking about this, maybe the engineers/accountants found out that if they manage to hit certain numbers then the system works.

Wickwick numbers were about putting a rack per sat, maybe that's not needed. If you are consuming less than 30kW you don't even need a separate radiator if your sat is the same size as a Starlink V2.

So now cost are down. Now you need to know some internals about the company running public LLM models. How much compute are they using while training vs how much compute are they using while doing inference. How much power does a prompt uses in their different models? How long does it takes from query to response? Etc.

With that set of information you can conclude if you are going to make money with this or not.

Also, unlike Starlink, you don't need to populate that much of your shells to start making money with the service. If you could use a communication constellation as your communication backbone, you can provide service from the first sat that you launch.

Does it makes economical sense?, I don't know. To me not even building datacenters on Earth makes economical sense for AI given that all of them keep losing money selling their services, so I'm probably not the one to ask about this.

 

[Wickwick]

https://meincmagazine.com/space/2026/...lation-too/?comments=1&post=44319961#comments

So, you decided to crap in a comment before reading the thread of comments back to see if your comment was right?

Commenters have been saying that you can't cool it. Not that it's too expensive, but that the engineering doesn't adds up.

When Musk talks about this, you can assume that he is ignoring the engineers and accountants in SpaceX saying that this is too expensive because he wants to prop up SpaceX value for the IPO.

But as far as we know, BO isn't accepting outside investors, so if Bezos is also talking about this, maybe the engineers/accountants found out that if they manage to hit certain numbers then the system works.

Wickwick numbers were about putting a rack per sat, maybe that's not needed. If you are consuming less than 30kW you don't even need a separate radiator if your sat is the same size as a Starlink V2.

So now cost are down. Now you need to know some internals about the company running public LLM models. How much compute are they using while training vs how much compute are they using while doing inference. How much power does a prompt uses in their different models? How long does it takes from query to response? Etc.

With that set of information you can conclude if you are going to make money with this or not.

Also, unlike Starlink, you don't need to populate that much of your shells to start making money with the service. If you could use a communication constellation as your communication backbone, you can provide service from the first sat that you launch.

Does it makes economical sense?, I don't know. To me not even building datacenters on Earth makes economical sense for AI given that all of them keep losing money selling their services, so I'm probably not the one to ask about this.

I will add that the 2D approximation is the limit for cooling. Once you have a massive structure with solar cells on one side and radiators on the other you have reached the minimal limit of cooling. Any sort of 3D structure (such as arrays of satellites flying in clusters) will have better cooling than my numbers. As such, my approximation is the most conservative back-of-the-envelope calculation one can make.

 

[fenris_uy]

About the economics.

https://www.tomshardware.com/tech-i...gas-plant-equivalent-to-nine-nuclear-reactors

10GW of compute needs $33B just to build the power plants needed. And an additional $4B in power lines.

As I said in another comment, if you get 80kW of compute per sat, BO is proposing building a 4GW datacenter, so for a comparable Earth bound datacenter, that would be budget of $14B just in power. Then you need to add how much nat gas is that 10GW in power going to cost each year.

 

[Statistical]

I also don't see

OK so if the ground stations are not in the US, what does the FCC care then?

It cares because to get a launch license from the US it needs FCC approval.

However yes if the company operating the sats is not in the US, the launch vehicle company is not a US company, the launch site is not in the US and none of the ground stations are in the US then yes the FCC wouldn't be involved.

 

[aenoiclurker]

Hmm, that seems to fall into an economic argument. The last paragraph of my post almost invites that criticism:

Which I explicitly addressed. Economically, this is pure fucking bunkum. Simply Not Going To Happen.

You may get one outrageously expensive POC launched so Musk/Bezos can declare they have the biggest dick and then they'll never mention it again.

Edit: You should have spotted I had made a basic error in my maths. You don't need double the cooling capacity of the ISS - that 70kW is already allowing for failure of a cooling loop. You actually need four times the capacity of the ISS to cool a constant 144kW - one rack - of GPU's. Good fucking luck making a profit with that kind of overhead.

 

[AlexisR200X]

This line really should not be so matter of fact. It is an extremely speculative statement. It's not clear how this would improve, let alone enable, AI. It also doesn't really say what kind of AI it would instantiate, as current LLM's certainly don't need to be in space. It's not even clear in-space processing is that good of an idea in general. Indeed, there's a lot to argue against it.

Its all a scam to lure in investor money. Data centers rely on low latency and require ongoing upgrades and repairs. How do yo do that in space at a competitive price, in a timely fashion? You don't. All they are doing at that point is cluttering the orbits with soon to be obsolete hardware that costs way more and is. Meanwhile traditional datacenters are kept upgraded for a fraction of the cost. Anyone investing in it will do so chasing the AI in space buzzword but in reality they will be signing up for a rather blatant rug pull.

 

[aenoiclurker]

Commenters have been saying that you can't cool it. Not that it's too expensive, but that the engineering doesn't adds up.

Bullshit. Please link to the comment where someone has claimed that it's impossible as opposed to stupidly expensive. I went back - no such comment exists.

 

[mannyjupiter]

Spending billions launching power hungry AI chips into orbit is a waste. It would be better to spend that money improving the compute to energy use efficiency. Perhaps moving away from waste heat generating electronic chips to full optical processing.

 

[McTurkey]

You don't have to put entire racks in a single satellite--but then you create other problems. Like needing 100,000+ satellites to do the same compute as a single terrestrial data center (depending on the size of "data center' you want to compare to since they're all different).

What makes you think the size of the constellation is inherently a problem?

 

[fenris_uy]

Bullshit. Please link to the comment where someone has claimed that it's impossible as opposed to stupidly expensive. I went back - no such comment exists.

I have a link to that exact comment in the comment that you are responding to. It's not about how you are going to pay for the cooling, it's about how they are going to do it.

Someone call me when any of these rich idiots have any idea how they are going to cool megawatts of computers in the vacuum of space.

How they are going to do it? Either from surface radiation of the sat itself, or separate radiators if the sat uses too much power.

 

[aenoiclurker]

What makes you think the size of the constellation is inherently a problem?

When you have to launch multiple times the photovoltaic and cooling capacity of the ISS to service one rack of GPU's how do you figure that the equivalent of 100,000 of the largest satellite humanity has put into orbit is even remotely possible?

Edit: and with GPUs failing after around five years (probably MUCH earlier in a high-radiation environment) you're going to have to constantly replace them. You will have to launch thousands of them a year - multiple ISS's sent into orbit per day.

Seriously: what the fuck are you people smoking?

 

[NetMage]

There is no need to constrain deployed volume in such a way; just mount the radiator arrays somewhere else.

Where else on a satellite in sun terminator orbit will have enough shade?

 

[aenoiclurker]

I have a link to that exact comment in the comment that you are responding to. It's not about how you are going to pay for the cooling, it's about how they are going to do it.

I saw it - it doesn't say what you claim.

how they are going to cool megawatts of computers in the vacuum of space.

Not that's it's impossible, but how are they going to do it? It's a fair question, given the scale of the problem.

Either from surface radiation of the sat itself, or separate radiators if the sat uses too much power.

Already addressed above using the example of the ISS. Try reading.

 

[NetMage]

Is anybody making the argument that cooling is physically impossible or are you attacking a straw man?

Plenty of comments have said that if you had read them.

The cooling loop you glossed over makes the server rack the largest non-human-carrying payload ever launched into space (unclassified, at least).

So what?

How is that not a bottleneck to wanting to launch thousands of them into space?

Starship and New Glenn 9x4 are going to be capable of huge payloads at low cost.

These guys will essentially be in business launching cooling equipment into space... With a little bit of data center attached.

How is that different from AST SpaceMobile, a company that sends IP huge antennas with a tiny bit of radio attached? And why does it matter?

 

[fenris_uy]

I saw it - it doesn't say what you claim.


Not that's it's impossible, but how are they going to do it? It's a fair question, given the scale of the problem.



Already addressed above using the example of the ISS. Try reading.

The ISS example is about cost, not how. It claims what it says in the text that I pasted, how are they going to do it. The how is already answered, how much it's going to cost, it's another different question, that wickwick says that it's the true question that BO and SpaceX need to answer. Not the how.

How much, and how, are not the same question.

 

[NetMage]

No, much of the hardware literally dies within 5 years. The mean time between failures for an Nvidia GPU used for AI training is a couple years. Less if it's highly utilized.

Why do you think they will be using space based data centers for training. They will want to run models trained on Earth where the data and bandwidth is located, and just run them in the space data centers.

But this makes me think the data center satellites could be built into two pieces that autonomously join in orbit. One piece is the compute and the other the communication, power and cooling. After five years you de-orbit the compute portion and launch a new compute portion to attach for another five years.

 

[aenoiclurker]

Plenty of comments have said that if you had read them.

The ISS example is about cost, not how. It claims what it says in the text that I pasted, how are they going to do it. The how is already answered, how much it's going to cost, it's another different question, that wickwick says that it's the true question that BO and SpaceX need to answer. Not the how.

How much, and how, are not the same question.

Bullshit. If the "how" makes it impossible to operate commercially then the answer may as well be "can't be done".

FFS: if the answer is "spend a billion dollars to cool a single racks worth of GPUs" then you do not have a viable solution to putting datacenters into orbit. Stop smoking crack.

 

[fenris_uy]

Bullshit. If the "how" makes it impossible to operate commercially then the answer may as well be "can't be done".

FFS: if the answer is "spend a billion dollars to cool a single racks worth of GPUs" then you do not have a viable solution to putting datacenters into orbit. Stop smoking crack.

Operating commercially viable and can't be done aren't the same thing. SLS isn't commercially viable, it's launching. Vulcan isn't commercially viable (ULA is going to end losing money on it), they are launching. Starliner isn't commercially viable, they are still trying to launch.

I'm not saying that it makes economic sense, wickwick isn't saying that it does economic sense. I'm saying that the engineering is possible, so people should stop arguing about the engineering, and start only questioning the how are you going to make money.

Also the 144kW per sat isn't a known requirement for the constellation. So it shouldn't be used to argue about the economics. From Scott Manley video, a 20kW Sartlink sat doesn't needs any radiator, it can radiate the heat just fine from the size of it. So, maybe a 40kW sat twice as big can do the same, I don't know, but if you could, now you don't need to launch 2 to 4 ISS sized radiators per sat. There is probably a size under 144kW in which you only need to launch 1 ISS sized radiator per sat and get enough work done in the sat that it can do actual work. Do you make money out of it? Depends on how much do you charge for that actual work. Right now, OpenAI et al aren't making money even with normal datacenters.

 

[aenoiclurker]

Operating commercially viable and can't be done aren't the same thing.

In the context of Jeff Bezos promising to launch a mega-constellation of satellites providing a datacentre in space, yes it fucking is.

Nobody is going to put tens, let alone hundreds of ISS-sized satellites into space when the same functionality can be provided for a tiny fraction of the cost on the ground. You may be starry eyed at the idea but it's still fantasy fucking bullshit.

Grow up.

 

[aenoiclurker]

Also the 144kW per sat isn't a known requirement for the constellation. So it shouldn't be used to argue about the economics.

That's the base requirement for a single rack of H200 GPU's. What the fuck do you think they're going to put up there? An abacus?

So, maybe a 40kW sat twice as big can do the same, I don't know, but if you could, now you don't need to launch 2 to 4 ISS sized radiators per sat. There is probably a size under 144kW in which you only need to launch 1 ISS sized radiator per sat and get enough work done in the sat that it can do actual work.

You only need to launch one half-billion dollar ISS-sized cooling system per satellite? Oh well then - that's obviously as easy as falling out of bed. After all, we've already launched dozens of things that size, right? Wait - what? Only one? And nobody is prepared to pay to keep it aloft?

Can you even hear yourself?

 

[didhestealtheraisins]

The future sucks.

This would solve the water problem for data centers. How does that suck?

 

[didhestealtheraisins]

What's the play here? I don't understand what the benefit of a data center in the sky is over a data center in Stuttgart, Arkansas.

Or is this just a ploy for investor dollars and to justify Bezos' space ambitions?

Don’t have to find water to cool everything. Don’t have to pay for the real estate.

 

[NetMage]

This would solve the water problem for data centers. How does that suck?

There is no water problem for data centers. If they had to, they would just recycle the water but it is more economical not to at this point.

 

[Statistical]

What makes you think the size of the constellation is inherently a problem?

Because it is. AI workloads rely on high speed datalinks BETWEEN servers. 800 Gbps is common right now but the trend is towards higher bandwidths.

 

[Wickwick]

Where else on a satellite in sun terminator orbit will have enough shade?

Since you asked...

Your large-scale structure could be in the shape of a "T." The top of the "T" is the solar panels and the stem is your radiators. You probably need to put reflectors on the side pointed at the earth, but the other side can be as long as you want. So you could have 100X the area for radiators as solar panels. It's expensive, but possible.

 

[Wickwick]

Bullshit. If the "how" makes it impossible to operate commercially then the answer may as well be "can't be done".

FFS: if the answer is "spend a billion dollars to cool a single racks worth of GPUs" then you do not have a viable solution to putting datacenters into orbit. Stop smoking crack.

If it's an economic argument, why harp on the radiators? They're going to be one of the cheaper things to send up - certainly cheaper than the solar panels. Cheaper than the rest of the structural elements or the station-keeping, etc.

The thermal issue is a dead one. In numerous threads it's the first technical (not economic) objection and it's indicative of someone that's doing engineering by association rather than by numbers.

 

[Snark218]

If it's an economic argument, why harp on the radiators? They're going to be one of the cheaper things to send up - certainly cheaper than the solar panels. Cheaper than the rest of the structural elements or the station-keeping, etc.

The thermal issue is a dead one. In numerous threads it's the first technical (not economic) objection and it's indicative of someone that's doing engineering by association rather than by numbers.

All questions of feasibility are both economic and technical questions. The two are inextricable. It’s not worth discussing one or the other, as if the people proposing these systems can afford to ignore either practicability or costs.

 

[JustUsul]

All questions of feasibility are both economic and technical questions. The two are inextricable. It’s not worth discussing one or the other, as if the people proposing these systems can afford to ignore either practicability or costs.

I would argue it is worth discussing where the technical challenge actually lie, because it lends legitimacy to argument. It's important to understand the underlying points in why space-based data centers are a dumb idea in the first place. When someone says "this is a non-starter because of the inherent thermodynamics of cooling a data center only through radiative transfer, modulo some stupidly large surface area" ignores the fact that the power generated by a typical data center first requires massive amounts of solar panels. And if you've solved that problem, then you've also solved the surface area requirements needed for your radiators. If I'm understanding the logic correctly, it should be conceivable for a space data center that is powered solely by solar panels to radiate away the heat that it generates (ie., you can't generate more heat than you've already collected from the sun, which means technically you can radiate that away).

I feel like people have lost the plot when they're yelling at Wickwick for simply pointing this out (while also suggesting there are a whole slew of other reasons to be critical of this whole endeavor).

 

[Wickwick]

I would argue it is worth discussing where the technical challenge actually lie, because it lends legitimacy to argument. It's important to understand the underlying points in why space-based data centers are a dumb idea in the first place. When someone says "this is a non-starter because of the inherent thermodynamics of cooling a data center only through radiative transfer, modulo some stupidly large surface area" ignores the fact that the power generated by a typical data center first requires massive amounts of solar panels. And if you've solved that problem, then you've also solved the surface area requirements needed for your radiators. If I'm understanding the logic correctly, it should be conceivable for a space data center that is powered solely by solar panels to radiate away the heat that it generates (ie., you can't generate more heat than you've already collected from the sun, which means technically you can radiate that away).

I feel like people have lost the plot when they're yelling at Wickwick for simply pointing this out (while also suggesting there are a whole slew of other reasons to be critical of this whole endeavor).

That's a reasonable summary.

If first-order thermal analysis suggested you need 4x the area of the solar panels for radiation or you needed a skin temperature of 400 deg. C then that would be a reason to question the efficacy of data centers in space regardless of the economics of how they get up there or the value of the service they're providing. But a skin temperature of 60 or 70 deg. C seems like a reasonable starting point to ground the rest of the discussion.

In my mind, the economics don't close, but that's for today's world. But that's a strong function that today's AI datacenters are losing money rather than making it. Should that change and should the demand for compute scale at a strong exponential pace, the cost basis could change. And it's important to both be aware of changing markets and to be in a position to take advantage of however things shake out.

Mind you, the economics of compute wouldn't have closed in 1960 for everyone to carry a computer in their pocket. But today, several of the most valuable companies in the world exist almost entirely because of ubiquitous compute. I certainly don't have the expertise of foresight to declare that space-based datacenters will never make economic sense. But I can at least bracket known physics and help describe the external factors that must be in-play to make them a reality. And along those lines, the cooling seems to be one of the smaller concerns.