Yes, the physics are worse, and yes, the economics are worse, but data centers in space have the crucial advantage that they are out of range of the molotov-throwing arm of Joe Public (recently unemployed).
Similarly to Starlink , which was hyped as a low codt alternative for marginalised people/areas and sucked in enormous subsidies to that end.
While being more or less a proof of concept for a military command and control network , thats the same way this should be looked at.Especially with data centers being vulnerable on ground, coupled with the ability to reposition the space based units to be 'closer to the action' as it were.
So bottom line a proof of concept for DARPA funds and eventual inegration into the military command and control nodes.
Starlink didn't get any subsidies. There were reports in summer about them potentially getting 600 Mio (btw the broadband program is 40 billion heavy) and I can't find sources that SpaceX got any program founding yet.
Businesses run anywhere real money can be made. Middle east is the most volatile place. Yet, the businesses are perfectly okay building infrastructure to extract, refine and ship oil from there.
If real money can be made, there is no need of datacenters in space. Offer some country some money and they'll let you build whatever you want. The whole idea is entirely unprofitable thats why they are making up things that won't work economically. Keep giving us money indefinitely, we'll keep spending, money can be made after a few decades of spending ... trust us.
Your idea is that because AI or data centers are supposedly 'entirely unprofitable', they want to expend even more capital to do it in space with less profit?
And somehow most of the replies seem to be about terrorism. What a comment section.
I'd expect the discussion to be a bit more focused on the scalability and solar power..
Access to unlimited almost free energy is the main point.
Other than that you don't need to deal with local politics.
Other than that it can actually be a faster round trip (if the satellite combines both connectivity to/from Earth and compute) if you don't have a local datacenter.
Yes, the physics are worse, and yes, the economics are worse, but data centers in space have the crucial advantage that they are out of range of the molotov-throwing arm of Joe Public (recently unemployed).
So are container ships in international waters. Put them there.
There's a lot of sci-fi about the wealthy bringing on an apocalypse so they can rule the world that comes out of it. Two examples on TV now are Fallout (from the video games, obviously) and Silo.
One thing that gives me some small level of comfort is that the billionaires who have the means to make that sort of thing happen don't want to live in bunkers or a radioactive wasteland afterwards. They have it pretty good on Earth as it is now. So there's only so bad things can get (I hope).
To your point, it does feel like we are bouldering towards major disruption, even the unravelling of the fabric of society, thanks to worsening inequality as well as the impacts of AI. In the coming decades we may well have a billion climate refugees. We may have large swathes of the population who have no productive outlet in a world that still requires you to have a job to live. And that could go south very, very quickly.
So I guess my point is that if things get so bad you need to put your AI DC in space to avoid the unwashed masses burning it down, I suspect we'll have far bigger problems long before that becomes an issue.
You assume billionaires and politicians are all-powerful and that they congregate in a secret cabal voting on where to take civilization next. No they live in fear of what other powerful people might do and their reactions to what is perceived risk can really f up everything. Even the most powerful autocrats in the world have to deal with other countries.
Putting data centers in Antarctica or extreme north Canadian islands would also put you out of the way of molotov-throwing joe public. Even in space your not safe from the voting public as long as your companies directors are sitting on earth somewhere.
Russia has (likely) done that already in Poland. But with a constellation using laser interlinks and having multiple ground stations across the world, it becomes a difficult logistical problem.
They're also not controlled by pesky local regulators who are worried about losing elections. By putting them in space it becomes a federal/national defense project, and those regulators are either out of work or compromised or complicit.
I'm going to be fascinated to see how the regulatory arbitrage shakes out long term. What legal provinance applies to infra in space? Does GDPR apply? Can you even enforce GDPR? And you have line of site to users who could connect to it directly, bypassing government-regulated backbones.
GDPR applies regardless of where the data center is located because it applies based on where the user currently resides. Similarly, US laws (including those enabling warrantless surveillance) apply regardless of where the data center is located as long as the owner/customer/controller lives or is based or registered in (or is a subsidiary of an entity that is based or registered in) the US - or if the US decides it has jurisdiction until proven otherwise (cf. Maduro).
You enforce laws the same way you do otherwise: if they are in your jurisdiction and don't comply, you arrest them; if they're in a jursdiction you're friendly with you ask for extradition; if they're in a jurisdiction you're not friendly with and you really need them to cease whatever they ceasen't, you have them accidentally fall out of a high enough window.
You can't shut down a DC in space. But even if someone has moved their DC into space, their terrestial human body is still very easily shut down one way or another. And maybe less gruesomely, so is whatever comms infrastructure they need to maintain in order to actually connect to that DC - if the DC is sufficiently "out of reach" (i.e. not in low Earth orbit) it likely requires optical transmission if it can accomplish any data rate high enough to allow it to be used for something sufficiently annoying and optics are fickle and require careful calibration.
But seriously, us tech nerds routinely underestimate rubber hose cryptoanalyis (cf. XKCD 538). If you can bypass government regulation it's because the government isn't sufficiently interested in making you stop whatever it is you think you need to bypass their regulations to do. If your goal with operating a DC in space is to evade the "regulatory arbitrage", you're literally advertising your intent to operate what will most likely be a criminal enterprise so unless you have sufficient financial resources to exert political leverage in your favor (cf. Musk), you're just asking for trouble.
I dunno, considering how various insurgent groups manage to make some reasonably impressive rocketry, I have to believe that in a few years with AI assistance (abliterate all the things) knocking a LEO satellite out of orbit won't be that much harder than sabotaging an electrical substation.
We're talking about an exceptionally vulnerable target with a perfectly predictable path of travel here. And it's constantly emitting lots of RF making it easy to track. You just have to get near it and explode with shrapnel.
Totally illegal, of course. But it will be within the reach of millions of people if they want to do it.
I don't think "just move to space bro" is going to stop the pitchfork-mob.
I think the failures of most rocket startups to launch even with billions of dollars and thousands of real engineers shows how unlikely this is.
And say you do build an orbital
class rocket and manage to launch it without anyone noticing before you do - how many of e.g. Starlink’s 11,000 satellites will you be able to damage? What real impact will that have on Starlink?
This is what AI was fucking promoted as being for! The utopia where machines do all our work and we receive universal basic income to enjoy our freedoms.
Taking away work was exactly the premise it was sold under.
Whenever I hear about this I can't help but be reminded of the Howard Hughes Glomar Explorer, where a hugely-expensive commercial project was actually part of a secret CIA initiative to recover a Soviet submarine: https://en.wikipedia.org/wiki/Glomar_Explorer
I'm not sure that's truly the case here, but it does seem like the tech Project Suncatcher is working on has at least some overlap with requirements for military SIGINT and in-orbit imagery processing.
Why would the CIA need Google's cover for any of that? The Feds have been flying spy satellites since the 1950s and there's plenty of classified satellite launches.
Because anybody near-peer to the US at this point has the capability to shoot down satellites, and anything launched on a classified satellite launch is a prime target.
While a bit tin foil hat-y, this is an interesting (conspiracy) theory, even while space datacenters aren't as impractical as some people try to make them to be.
Yeah, no proof, of course (these are spies, at the end of the day . . . like with the Epstein stuff any written records have already been shot into the Sun), but that is a very evocative theory. It explains things better than anything else, given how this whole rigamarole seems to run afoul of some very basic math.
There's a case to be made that a substantial part of the mobile/social web revolution of the Aughties was at least in part propelled by the exploding budget of the security state post-9/11. LifeLog, In-Q-Tel, a bucket of other programs that have straight-line-relationships with the various Web 2.0 shenanigans. The CIA point man for these operations are recognizable today as some of the new generation of American semi-governmental oligarchs.
This last year I read that the early Stingrays only worked because they were engineered in from the IMSI standards - it was backdoored at the spec level, not the design level, not the manufacturing level.
When you step back a moment it makes sense. Who's funding a wireless information infrastructure to every American on the continent, not to mention a sea change in mobile computing manufacturing and development. Different tooling, different design systems, different everything, massive scale . . . what, paid for with advertising? Are you saying Mad Men is paying for all this? That seemed and seems . . . unlikely.
Alphabet holds approximately 551 million shares of SpaceX (trading under the ticker SPCX), valued at roughly $94.1 billion, which represents about a 4% to 6% stake in the company.
They've got $400bn+ in annual revenues, with $50bn in free cash flow, and the ability to raise enormous piles of debt if they needed it.
They've not got a pressing need for the cash, so can afford to take a multi-decade view of SpaceX. If you look at other investors who take a view across that time horizon, you'll see some pension funds and huge trusts putting money into SpaceX for the same reason.
It's overpriced today on fundamentals, but in 50 years time, today's price will look cheap.
They invested $900 million[1] in 2015, and it is worth $94.1 billion today. Why sell? They also likely have restrictions on selling post-IPO, but I don't think you can trigger a "bank run" on a stock anyway; the price will just keep adjusting as you sell/buy until it gets absurd.
A bank run happens when everyone tries to withdraw all their money at the same time and the bank runs out of cash. Not really possible in the stock market where companies literally can create/destroy shares and there is a whole secondary pricing layer to it.
why would they sell? because of ridiculously large the valuation is? like, I guess at SpaceX's valuation there's tremendous downside and little upside (x2 growth is unlikely when it's a third of the tech sector)
For those questioning viability, there's a startup called Starcloud https://www.starcloud.com/ with a public whitepaper explaining a bit their hardware and economics https://www.starcloud.com/wp.
They've already had one small proof of concept launched.
Not trying to promote, I just heard about them on the ycomb podcast
Desert land is full of that amount of area available for peanuts.
That’s 4000 acres. I see a 6000 acre ranch available in Arizona for $7m. The cost of building a 5GW DC is about 100 billion. Land including enough for all the solar and battery you need to be completely self sufficient is basically free.
They seem to not be counting the $1B of GPUs on either side, making the difference pretty small. They also think launching 200 racks of equipment into space will cost $5M.
how much area do you need to radiate away 40 MW, assuming you want to keep your GPUs at a toasty 90 celcius. Every space datacenter fan/promoter enthuses about the free solar power, but forget to mention how the heat will be dissipated.
Historically there have been many startups with pitches that haven’t worked out. The existence of a startup who think they can tackle a particular problem isn’t evidence of the viability of their proposed solution. If they succeeded, that would be evidence.
Yeah. Starcloud is notable for two things: firstly that they've raised a lot of money and secondly that theyve already abandoned the nonsense in their white paper and shifted to a "lots of satellites" model like everyone else exploring orbital datacentres...
I think SpinLaunch is great example. They were actually able to fund and make a working prototype, tested it, but still decided it is not practical after all, before pivoting to production of knockoff Starlink-like satellites.
What happens to all the precious metals that go up with these data centers? At least with a land based data center we have a vague hope of recycling the raw materials at some point in the future.
Just like with all the satellites, the recycling plan is to allow them to burn in atmosphere. Maybe after we breathe enough of it, we can recover the metals via crematories.
Precious metals stop being precious metals once asteroid mining begins. I don't think it will be super soon but I could certainly foresee it being a thing within 10 - 15 years which is about 2 - 3 generations of satellites from my very brief Google.
Note how their cost comparison table mentions "launch" but not a word about how to retrieve systems parts after their end-of-life. "There's so much junk already circling around the planet, what does it matter if we'll leave it there to rot?" can't possibly be the stance they're taking.
At least this brochure marks cooling in the "challenges" column rather than the advantages.
However, the only advantage listed in this announcement is "near-constant sunlight". It appears they are hoping for 8x solar efficiency for 1000x the cost of terrestrial construction.
If the actual motivation is that people don't want data centers on earth, that seems to be because the datacenters 1) drive up energy costs, 2) use a lot of water I guess, and 3) can make noise. If you're able to operate a datacenter satellite you apparently don't need local electricity, water, or human labor. As someone else said, set up on the desert or in the middle of Alaska somewhere 100s of miles from the nearest person. They can still use solar (just build like 8x the panels if you need to... still cheaper than space), they don't need water, don't need labor... I think most public opposition would disappear.
I've heard some say the real advantage is latency for spy satellites... This seems not a real use case. The kind of workloads that require this kind of datacenter usually can spare 1-2 seconds in round trip time. Millisecond-sensative reactions I would assume would want to be driven by onboard processing.
One true advantage that I haven't seen on any brochure is that the heat produced by the space datacenters could be released into space rather than raise the temperature of the earth. However, the satellite would have to be designed to radiate the heat away from earth. Hopefully they design the satellites to direct the radiated heat away from earth and don't just point the radiators straight back to earth...
> 1) drive up energy costs, 2) use a lot of water I guess, and 3) can make noise
I don't think the opposition to data center projects can be explained with actual impact, which so far seems entirely negligible, and lower than heavy industry.
I had the impression it's more driven by general mistrust in public officials and the tech industry.
This is the impression I get as well. Plus, people concerned that property values will be negatively affected, that their taxes may go up to subsidize the data center, etc
I wouldn’t call the electricity use negligible, some of the newer data centers are looking at the same energy usage as small countries and run off fossil fuel
> set up on the desert or in the middle of Alaska somewhere 100s of miles from the nearest person… I think most public opposition would disappear
I don’t know about the desert, but in Alaska I think this would be hard to achieve. People there tend to care about preserving wilderness, and most of them would probably be opposed to this.
For real. It's an enormous problem solved only with 1) sheer scale, and 2) Science Fiction.
Both of those are expensive as hell, by the way.
Cooling via radiation follows Stefan–Boltzmann: P = εσAT⁴. Let's assume a good surface (emissivity ~0.9) at 300 K (27 °C) at 400 W per square meter per side. A flat panel radiating from both faces into deep space gets 800 W/m, not including the losses from, say, the Sun, or from IR coming off the Earth. Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and minimum 1,200–1,500 m² of radiator.
In case ya didn't know - 1 MW is tiny from a present-day-datacenter perspective. It's like 8 racks. So we're talking orbital megastructures here, many many many square kilometers, and this is with all the best case assumptions, and magic radiator panels that never see the sun, or the earth, or the moon.
This is just the basic numbers here, by the way. There's a garbage truck full of other unsolvable problems if you poke your head in there.
Aside from the "Avoid Regulations" aspect, and the "Everything That Burns Deorbiting is Depreciation" aka "The Starlink Trick", I'm not sure what the hell the draw is.
Cooling in space is hard but not impossible - while current (IMHO stupid without advanced in space infra) space data center projects work with a couple MW, many advanced space propulsion concepts might have to reject hundreds of MW if not a couple GW.
As for space data centers - I think the main issue is the complete lack of in space infrastructure for resource mining, processing and manufacturing & maintenance. It is kinda like building your first practical steam locomotive & the deciding to build directly an airliner. No suitable materils, experience, work force, material sources, etc.
We eventually went from locomotives to airliner, in an incremental manner & expanding the supporting infrastructure to support the ever more ambitious projects.
> Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and minimum 1,200–1,500 m² of radiator.
We need > 2x more solar panels than we need radiators. Doesn't this imply radiation isn't really the limiting factor here?
No point in running them at room temperature. GPUs, etc. run fine at 95 C. If you run your cooling loop at 70 C instead, you get 70% more cooling compared to 27 C.
At any rate, 1 MW for a single satellite is fine. Just launch several thousand of those and you get to real numbers.
Also, there's no need to talk about "magic" radiators. You orient them so they're at a knife edge to both the Sun and Earth. This is not difficult (the Moon is irrelevant).
Radiating 1MW at 500K (227C) with a 0.4MW heat pump takes about 200 m^2 flat sheet surface. Inputs - solar+nuclear for double fun. So - quite feasible.
Getting this all up into orbit it obviously the hard part, but if you're already building so much solar capacity the cooling actually doesn't seem unreasonable?
Great read, thanks for sharing. I am interested in reading some more about the other unsolvable problems that exist in this space, do you have any recommendations that you wouldn't mind pointing me at? It would be greatly appreciated, and thank you :)
Surface area is a materials problem? Folded microstructure, atomic-scale textured surface or some other science-fiction solution could have square kilometers of surface area in a shoebox.
the other is that you could make tiny datacenters and flood the sky with them. in effect, not datacenters at all, but some kind of dataswarm coordinating at literal lightspeed via lasers.
they'd still be wildly expensive to deploy, and probably litter the orbit zone with fast-moving debris.
From the article itself, sounds like it’s an open problem that they are experimenting with:
“ We’re working on a number of different approaches for this, including a combination of heat pipes and radiators to cool the chips. So far, our team has tested the technology in a thermal vacuum chamber that simulates both the thermal and vacuum environment in space. We’ll see how our new TPU cooling system works in space and refine our designs as we learn more.”
Imagine a heat pump circuit fails (solar rays? micrometeorite? random chance), I guess everything in that loop will just be dead forever? I guess you don't need to pay to dispose of it.
> The biggest challenge was how to cool the A.I. chips, which heat up when they perform calculations and process information. Fans, which typically help dissipate the heat, do not work in space. So the Silicon Valley company instead developed a cooling system that uses layers of conductive material to expel the heat into space.
> The bottom layer is made up of Google’s A.I. chips, which sit on a green motherboard. The next layer consists of “thermal interface material,” a pale green putty that comes in sheets like Fruit Roll-Ups and connects the chips to layers of aluminum and copper, radiating heat away from the motherboard. Finally, there is a radiator panel, which projects heat into space.
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
It's a bit like the hyperloop. They know it's not practical but having the idea out there makes money.
There will be some niche demand for military use though.
We don't. This is all cover for the militarisation of space, there's no real benefit that'd be ever economical to put a DC up in space when you could build one on the ground. The whole narrative exists to allow google to tap into the Golden Dome / Space force bucket of pork that's basically SDI II.
You don't need civilian scale compute in space, but you absolutely can put miliary application up there and get filthy rich, like Elon, who doesn't care if it ever works, because he's getting paid as taxi service to shoot stuff up.
This is the most interesting perspective I've heard on this topic, which otherwise always converges on the same political dismissals or heat dissipation arguments (the latter are fascinating, but going in circles by now).
Sounds very probable: gives a plausible reasons for sending a lot of infra up, absolutely doesn't have to be profitable or even effective, puts the focus on AI instead of the MIC, and is in line with the kind of contracts Big Tech runs after.
To me this model explains a lot of why so many big companies seem to be investing into what every expert I've heard says goes against basic physics. The only thing missing for it to be more than an interesting idea is why China and the EU are not fighting this.
you need a gradient from hot to cold to generate electricity. vacuum is a poor thermal conductor, so your cold part will become hot and then no more gradient.
(I guess you could try to capture the radiative photons via the photoelectric effect?? but I don't think it works.)
I tried asking Grok and a bunch of waffle came out but one interesting idea was "modular tiles" with solar cells on one side and a panel to radiate heat on the other with the processor bolted to the panel. (https://www.geekwire.com/2026/sophia-space-caltech-ai-patent...)
He is free to launch his own space GPU if he is so confident it is profitable*.
I will take it as a given that the Google engineers know what they are doing, and their first version can only run for 15 minutes before it needs to shut down to cool.
*Profitability requiring that other Musk controlled companies do not pay for the service at elevated rates in a classic self dealing scheme.
Please stop with this argument. This is one of the easiest problems to solve. Heat dissipation requires a radiator, which is just a dumb hunk of metal. All you need is a cheap way to launch mass to orbit, which is exactly what SpaceX (and Chinese companies) are doing.
The only valid argument against data centers in space is the economic cost. If the per megatoken price for space datacenters is higher than terrestrial, then this won't work.
But the price for terrestrial datacenters keeps going up and technology keeps dropping the price of space-based.
That is not a fake argument, but real physics. Yes, you can design out X kw can be dissipated by this much radiators, but that adds an enormous quantity of mass, more than the solar panels that feed it.
If you are trying to generate a profit, every extra kg puts you further in the red. It still costs over $1000/kg of mass to get into orbit.
> Please stop with this argument. This is one of the easiest problems to solve. Heat dissipation requires a radiator, which is just a dumb hunk of metal.
Convective radiation does not happen in space and this challenge is far more significant than your comment implies. Rather than "a dumb hunk of metal", radiators for spacecraft are often made of ceramics and carbon laminates with higher IR emissivity than convective radiators made of simple metals.
From the article you're commenting on:
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
ML Infrastructure comes with some pros (larger emissive footprint) and cons (exponentially larger TDP) compared to the concerns there, but if you aren't familiar with the challenges of heat dissipation in space, please give it a read.
There's also a pretty interesting pop-sci article on cooling the Webb telescope, since it needs to be especially cold for its purpose. Not directly related, but may give insight into both challenges and solutions as well as well. https://science.nasa.gov/mission/webb/science-overview/scien...
and yet notice how the cooling video / section was the only one they didn't have a solution for... just saying "radiator" doesn't make sense - the radiator heats up too. it's how you get rid of the heat, not where you put it.
If data centers in space end up being economically viable, then I don't see how anyone can catch SpaceX. They are ten years ahead in both launch capability and satellite manufacturing.
Chinese rockets companies are evolving at insane pace, including RLVs. A couple of big milestones have been achieved this year alone, like first stage landing.
Also Stoke Space, YC startup, is interesting. They could even beat SpaceX to fully reusable if SpaceX find the refuel it and send it up again thing challenging.
They have a different approach - rather than a ceramic heat shield they have a metal shield to be cooled by the liquid H2+O2 fuel. https://youtu.be/3V6lfs7xFJM?t=191
I certainly hope ‘economically viable’ includes properly priced externalities for e.g. side effects of burning hundreds and hundreds of tonnes of aluminium and other materials in the high atmosphere.
No western company, no. That's why I can only hope SpaceX can get thru more red tape faster, and future administrations don't try to slow them down.
Even while current administration is kind of positive towards them, they are still even now just having to wait for no good reason for a FAA go-ahead. These kinds of stalls don't really exist in China.
Its only cheaper to build them on the ground if you aren't factoring security concerns. No insurgents in space. Not every application needs this hardening, but you can imagine there are many applications that do.
It's odd to see this level of skepticism, honestly. Does one think the engineers at Google aren't thinking about these problems?
This is obviously a test, just like hundreds or thousands of tests in space before it. Can we run Earth-rated components at any scale in space? How will it fail in various ways, at what rate, and are the causes able to be mitigated?
"We" talk about moonbases and Mars colonies. Part of any extended presence in space will involve the presence of computational devices. They're figuring out how to make that work with modern hardware. That's my take, anyway.
That's the first news about data centers on space that seems reasonable. Acknowledges the challenges, don't make crazy promises or predictions, makes clear what was already tested and what is still unknown. Makes me actually root for them, for once.
I like the "moonshot research" framing, in particular, because it highlights that even negative results will be valuable, either as improvements for ground technology or as published data to inform other initiatives of what has already been tried and failed.
Providing GPS, communications, media, etc. in space so there is access anywhere in the world has a clear advantage. Here the only pros seem to be that it is potentially cheaper than a datacenter on earth because of solar and cooling. But I assume the environment impact is worse than having datacenters on earth. Launching lots of satellites uses lots of fuel and CO2 emissions. Building satellites and dealing with the end of life tech has much higher environmental impact than dealing with datacenters on earth, where you can do proper waste management and recycle materials. From the economic point of view, polluting the environment is currently free. I guess if the externalities would be part of the equation it would not be economically viable.
Maybe this project exists because the "early in the game" factor. It has worked well in the past for Google, OpenAI, etc. They used copyrighted content to train LLMs to sell AI. They indexed Yahoo directory and websites before robots.txt existed, once they have a dominant position they put robots.txt in Google results so new players can't use their content. And they are launching AI in Space before regulations and politics catch up.
the pro is being able to dodge the entrenched industrial-nimbyism in America.
we talked big game for a long time about bringing manufacturing back to the US, but the practicalities would likely have been impossible in our modern political climate
Listening to these leaders speak, this appears to be a major component. They're trying to pour trillions of investment into communities all over the world, and local communities are rejecting it. The reasons are complex and I'm not going to paint either side as the bad guy. However if they can't get approval to build more infrastructure on Earth, they going to build it in space. At least what is economical. SpaceX has already reduced the cost of space delivery by 90%. This has already made all kinds of new applications commercially viable. Every efficiency increase from now unlocks exponentially more use cases.
As a long time space and technology nerd, this is the coolest fucking time to be alive!
This is not intended as a product launch. I think this is a preemptive bet on a conjunction of the following happening in time for a future product launch:
1. Cost to launch / mass lowering due to technological advancements.
2. Mass / compute lowering due to technological advancements.
3. Compute hardware evolution stagnating to the pace, relative to compute payloads, where CapEx is amortized over longer duration.
4. Compute & energy hardware reliability improving so OpEx is lower, and remaining useful life predictably keeps hardware running over its designed lifetime.
(I work on #4).
A wishy-washy inequality (per time, factoring out launch cost from OpEx):
Presumably, any one of these could see a x100 improvement and make the whole equation in the green. But, I believe this is a low-earth-orbit-shot bet on the sum getting better.
Note: I did 1+failures to denote baseline cost (initial launch, nominal OpEx).
- Getting power there is a headache (unless Greenland allows microreactors, which is a political headache). Note: Iceland has geothermal, not really Greenland.
- You still do need staff, it's not cheap building a mini city for 200 maintenance worker, much less the 5,000+ construction crew, and you'll have to pay premium to get people to actually do it
There's no world where this makes sense. Let's take it issue by issue.
1. Hardware failures: If you know anything about data centers you know that servers fail all the time. I've seen estimates of AI DC servers that have a failure rate of about 9% annually. A terrestrial DC has hardware techs who get alerted to the failure, replace the bits with known good bits and then take the suspect bits for testing to either bin them or verify they're working. You have no option to economically do this in space;
2. Radiation: We have the atmosphere and a concrete box protecting ground-based servers from cosmic rays flipping bits as well as the potentially more problematic (in space) solar storms;
3. Power generation: solar is the only option and yes, orbital panels would be more efficient than ground based. But space-based solar solar tech is much more expensive than what we put here on Earth because it absorbs multiple wavelengths of light to generate electricity. The launch cost and weight savings overcome the high panel costs. But to generate power in the gigawatt range, you're talking about solar panel arrays far larger than anything we've ever deployed in space;
4. Cooling: this is a big one. The only way to get rid of heat is to radiate it away into space and this is bound by physics. I'm sure people have done the maths on this but it's not going to be pretty;
5. Heating: the sun will be heating your vessel a lot of the time. You can shield it to some degree with solar panels but they too will radiate away heat in all directions, which will heat the things behind them. JWST uses several sophisticated layers of cooling to reach the low temperatures it needs to operate; and
6. Interconnectivity: these things are likely to be tens of kilometers apart in the very least. You're going to need terabit level networking to be useful for coordination and even then latency will be an issue. Laser dispersion will be an issue as well as the power and heat generated.
I honestly don't know what anyone is trying to achieve with any of this.
When reading old stories about rogue AIs they are often created locally on earth.
What do you do when you need to manually maintain server structure in space? Just the idea of a set of rogue autonomous data centers in space seems like an intriguing story idea.
Whilst I agree that this seems wasteful and perhaps not economically viable , I am excited for the innovation possibilities that could come from this. The early efforts to get into space gave some really life changing technology - maybe this can help us better understand energy use, conservation or dissipation in ways that could help us with our down to earth devices. Phones? Charging devices?
Nice we have a document we can point at of someone doing a space data center project pointing out all the flaws and benefits without hallucinations.
Putting data centers in Antarctica or extreme north Canadian islands would also put you out of the way of molotov-throwing joe public. Even in space your not safe from the voting public as long as your companies directors are sitting on earth somewhere.
"whether space could one day host scalable machine learning infrastructure. In low Earth orbit, satellites can access near-constant sunlight, generating up to eight times more solar power than on Earth. Eventually, it could be possible to link together multiple constellations of satellites, allowing them to manage larger AI workloads while in orbit."
Yeah.. remember people being skeptical about reusable self-landing rockets, until SpaceX proved them wrong? I would not put much trust in resident skeptics.
I still have to read a convincing argument for "datacenters in space". Why would you want to place expensive electronics in a harsh radiation environment, with challenging cooling issues, and where replacing faulty parts would cost a fortune? The main selling point seems to be "the idea sounds cool"
Every time I'm reading about the technical challenges of these kind of projects, it reminds me how amazing our brains are, doing what they do with only about 30W
One thing is may be that they aren't using the same parts as on earth, hardened ones, that might still work at hot conditions (I'm guessing)
So they can accept some heat. Rest is radiated into space.
There are for sure several things to try and stillto learn from/for it. But it will work out at some point and they'll find engineering solutions to problems they will experience. Let the engineers play.
Why the space? We have plenty of deserts on the Earth, they are inhabited but still much friendlier than the space. What makes the space more attractive?
"it's to ensure AI's profound benefits in key areas, from healthcare to scientific discovery, can reach everyone, far into the future."
No it isn't. That isn't true.
They know it is never going to be cost-effective, compared to building on the ground. People who know better are lying to the public about the feasibility of data centers in space.
It would easily raise the cost by an order of magnitude (or more) to launch a data center into space, compared to current AI data centers on the ground; it's the equivalent of building 100+ ISS space stations to house a Colossus in space. It costs even more to launch your materials if you want sun-synchronous orbit.
It's not just expensive to do at scale, it's impossibly expensive to do at scale. This is a science project, launching a few tomato plants into space to see what happens. Data centers in space are decades away (if ever).
I would believe that military deployments will happen; a couple of TPUs that can be deployed over conflict zones. Outrageously expensive, but no more so than the billions we blew up in Iran this year.
Let's assume everything gets smaller and cheaper. Vastly smaller and cheaper. So much so that it becomes feasible to launch a data center comparable to Colossus into space for the same price. You know what that means about the cost and size and efficiency on the ground? It means I can have a fucking TPU in my pocket and a round trip to space would be an insane waste of time.
There is no way to make the math for AI-in-space to make more sense than just making it work on the ground; any challenges one has with local politics are easier and cheaper to overcome than physics. The public may be stupid in large groups, but public opinion can be swayed, gravity cannot.
In short: This is either a marketing stunt, a bid for military funding, or both. It is absolutely not a project to put AI for "everyone" into space.
The useful life of cutting edge AI server hardware will require constant rotation of equipment to potentially remain competitive with more accessible solutions.
Hardening of equipment for the environment. ECC isn't going to be enough. I would have to defer to experts about the best way to manage it, but it either means custom hardware, weight, or both.
Repairs are impossible. Hopefully a sat can degrade gracefully, but routine repairs on earth become significant outages/decreases in economic value of a sat over time.
Rocket launches have gotten much better, but are not perfect. Insurance is a thing. However, cost of equipment in a failed launch may be eye watering, rivaling a governmental military launch.
To give a datacenter in space constant exposure to sunlight, the orbit must be either very var away, or occasionally be on the other side of the earth from you. Either way, the latency is terrible. You can put them in a ring and only choose the one closest to you, but are you really going to keep moving your data from satellite to satellite? So, bad for interactive use, maybe OK for non-interactive use, if you can afford the 30 foot by 30 foot solar panel and a similar size radiator for an 8x B300. That will be about $15M for your $500K GPU set. Plus $7M for the launch at SpaceX public rates.
Presumably they get exactly the same amount of sunlight 100% of the time instead of a few hours peak on earth? I'd say 8x is in the right order of magnitude.
How is no one in the whole thread aware that the intention behind orbital datacenters is military? You want to process outputs of large space based sensors in space itself, and reduce latency (e.g to other space based assets) or increase goodput (to ground).
Either it's that weird "space is easier than getting land on earth"(it's not) or the same tortured arguments about heat dissipation, no one is going to run consumer scale compute with consumer scale economics in space ffs. And maintenance and cost does not matter when it comes to strategic military assets, they are a step function useful enough to warrant even a few monthly replacement.
Everyone else with strategic weapons and a space program e.g india china is launching one as well.
What no one seems to be asking is what privacy and legal laws still apply for a datacenter in orbit?
They're effectively outside the sovereignty of any country. Further, it would be extremely impractical for any government's law enforcement to raid the datacenter and confiscate the servers.
Maritime law applies to space. In the high seas, the country of registration (flag state) generally governs. As the vessel is elusive, raids would be held on corporate offices instead.
From who? The crack addicts in space trying to steal copper?
The actual rack part is small, even if the solar panels and radiators are big. If this truly was super dangerous, ISS wouldn't be safe. And it's never hit mission-degrading issues due to micrometeoroids.
For those who doubt the AI apocalypse is possible, because "you can always pull the plug", our AI overlords provide a ready-made answer: the data center whose plug can't be pulled! Put a few nuclear warheads on board (we need our own weapon to answer the threat of Chinese super AIs, after all) and you're good to go.
Earth will get bulldozed soon anyway to make room for that hyperspace bypass. The plans have been available in the local planning office for a while, but nobody knows about it because they're hidden in the basement and you have to go down there with a torch.
I think if those building the datacenters were more upfront about what they were building and didn't hide behind NDAs the permitting process wouldn't be so fraught.
> Announced last year, Project Suncatcher is a long-term, research moonshot exploring whether space could one day host scalable machine learning infrastructure
It’s very confusing to have a project related to space called a moonshot project.
Haha, yes. I think Silicon Valley (the show ), made fun of the same thing. "Not so long ago, people called the idea of sending a man to the moon a moonshot"
The part I find wilder than cooling is the formation flying. If I remember their paper right, it's 81 satellites inside a 1 km radius, neighbours 100 to 200 m apart, because the optical links only get datacenter bandwidth over short distances. The ISS is about 110 m across, so that's satellites roughly an ISS-length apart, all doing 7.5 km/s
The video on cooling is quite vague, maybe on purpose to not give away proprietary technology. But I'm truly curious as to how they can dissipate the heat once they move it from TPUs into radiators.
As mentioned in the video in the void the only way they can remove heat from the radiator is via radiation, but that is extremely slow and ultimately the rate of heat generation in TPUs would be higher than the dissipation rate of the radiators and the whole thing would get heated up over time.
I wonder if a technique to pack crazy high amounts of heat in a small mass and just eject it in an orbit to the sun would be viable. Probably not, but the idea of a datacenter pooping its heat into little pellets is quite entertaining. Like Lila's pet from Futurama.
Yes, the physics are worse, and yes, the economics are worse, but data centers in space have the crucial advantage that they are out of range of the molotov-throwing arm of Joe Public (recently unemployed).
Similarly to Starlink , which was hyped as a low codt alternative for marginalised people/areas and sucked in enormous subsidies to that end. While being more or less a proof of concept for a military command and control network , thats the same way this should be looked at.Especially with data centers being vulnerable on ground, coupled with the ability to reposition the space based units to be 'closer to the action' as it were.
So bottom line a proof of concept for DARPA funds and eventual inegration into the military command and control nodes.
> and sucked in enormous subsidies to that end.
Starlink didn't get any subsidies. There were reports in summer about them potentially getting 600 Mio (btw the broadband program is 40 billion heavy) and I can't find sources that SpaceX got any program founding yet.
15 replies →
Did Starlink really receive enormous subsidies? IIRC it was kicked out of the federal program in the US due to lobbying from others.
5 replies →
I dunno, won't a satellite be _more_ vulnerable in space? There is not anti-missile tech in space.
I suppose it is about being out of reach of cheap drones?
13 replies →
> low codt alternative for marginalised people/areas
Yes, but while I do not see the first, I can clearly see the latter :)
what proportion of use for starlink do you think is military vs rural internet consumers?
Businesses run anywhere real money can be made. Middle east is the most volatile place. Yet, the businesses are perfectly okay building infrastructure to extract, refine and ship oil from there.
If real money can be made, there is no need of datacenters in space. Offer some country some money and they'll let you build whatever you want. The whole idea is entirely unprofitable thats why they are making up things that won't work economically. Keep giving us money indefinitely, we'll keep spending, money can be made after a few decades of spending ... trust us.
Note that you didn't mention any such country, and for good reason - there is none.
You'd need a dictatorship where people's voices don't matter, but then there's no right of ownership either.
35 replies →
Your idea is that because AI or data centers are supposedly 'entirely unprofitable', they want to expend even more capital to do it in space with less profit?
And somehow most of the replies seem to be about terrorism. What a comment section.
I'd expect the discussion to be a bit more focused on the scalability and solar power..
6 replies →
The real point is power limitations, putting stuff in space in the right orbit lets you turn solar power into reliable base load.
13 replies →
Starlink downlink stations aren’t
https://www.reuters.com/world/poland-suspects-fire-starlink-...
That was probably more Igor Sleeperagentovich than Joe Public, but yeah...
6 replies →
I'm beginning to wonder if this is actually the motivation, because why the fuck else would you launch computers all the way into space?
Access to unlimited almost free energy is the main point. Other than that you don't need to deal with local politics. Other than that it can actually be a faster round trip (if the satellite combines both connectivity to/from Earth and compute) if you don't have a local datacenter.
Yes, the physics are worse, and yes, the economics are worse, but data centers in space have the crucial advantage that they are out of range of the molotov-throwing arm of Joe Public (recently unemployed).
So are container ships in international waters. Put them there.
or just a really sturdy fence, and guard posts with guns. That's cheaper than space, and infinitely easier too.
4 replies →
Container ships getting rob by pirates all the times.
So unless those pirates have their own space program, it's pretty safe up there.
5 replies →
There's a lot of sci-fi about the wealthy bringing on an apocalypse so they can rule the world that comes out of it. Two examples on TV now are Fallout (from the video games, obviously) and Silo.
One thing that gives me some small level of comfort is that the billionaires who have the means to make that sort of thing happen don't want to live in bunkers or a radioactive wasteland afterwards. They have it pretty good on Earth as it is now. So there's only so bad things can get (I hope).
To your point, it does feel like we are bouldering towards major disruption, even the unravelling of the fabric of society, thanks to worsening inequality as well as the impacts of AI. In the coming decades we may well have a billion climate refugees. We may have large swathes of the population who have no productive outlet in a world that still requires you to have a job to live. And that could go south very, very quickly.
So I guess my point is that if things get so bad you need to put your AI DC in space to avoid the unwashed masses burning it down, I suspect we'll have far bigger problems long before that becomes an issue.
You assume billionaires and politicians are all-powerful and that they congregate in a secret cabal voting on where to take civilization next. No they live in fear of what other powerful people might do and their reactions to what is perceived risk can really f up everything. Even the most powerful autocrats in the world have to deal with other countries.
1 reply →
Putting data centers in Antarctica or extreme north Canadian islands would also put you out of the way of molotov-throwing joe public. Even in space your not safe from the voting public as long as your companies directors are sitting on earth somewhere.
Yeah, but those places don’t have 24/7 solar energy
4 replies →
Maybe out of reach for Joe Public, but a clear identified target for states.
https://www.cbsnews.com/news/us-confirms-weapons-in-space-ch...
I was thinking more regulation rather than molotovs.
Doesn't the data or whatever have to be received on earth? Can Joe Public do some (light) arson on that?
Russia has (likely) done that already in Poland. But with a constellation using laser interlinks and having multiple ground stations across the world, it becomes a difficult logistical problem.
Satellite phones and Starlink dishes suggest that the data links can cheap, ruggedized, small, distributed, etc.
They're also not controlled by pesky local regulators who are worried about losing elections. By putting them in space it becomes a federal/national defense project, and those regulators are either out of work or compromised or complicit.
I'm going to be fascinated to see how the regulatory arbitrage shakes out long term. What legal provinance applies to infra in space? Does GDPR apply? Can you even enforce GDPR? And you have line of site to users who could connect to it directly, bypassing government-regulated backbones.
GDPR applies regardless of where the data center is located because it applies based on where the user currently resides. Similarly, US laws (including those enabling warrantless surveillance) apply regardless of where the data center is located as long as the owner/customer/controller lives or is based or registered in (or is a subsidiary of an entity that is based or registered in) the US - or if the US decides it has jurisdiction until proven otherwise (cf. Maduro).
You enforce laws the same way you do otherwise: if they are in your jurisdiction and don't comply, you arrest them; if they're in a jursdiction you're friendly with you ask for extradition; if they're in a jurisdiction you're not friendly with and you really need them to cease whatever they ceasen't, you have them accidentally fall out of a high enough window.
You can't shut down a DC in space. But even if someone has moved their DC into space, their terrestial human body is still very easily shut down one way or another. And maybe less gruesomely, so is whatever comms infrastructure they need to maintain in order to actually connect to that DC - if the DC is sufficiently "out of reach" (i.e. not in low Earth orbit) it likely requires optical transmission if it can accomplish any data rate high enough to allow it to be used for something sufficiently annoying and optics are fickle and require careful calibration.
But seriously, us tech nerds routinely underestimate rubber hose cryptoanalyis (cf. XKCD 538). If you can bypass government regulation it's because the government isn't sufficiently interested in making you stop whatever it is you think you need to bypass their regulations to do. If your goal with operating a DC in space is to evade the "regulatory arbitrage", you're literally advertising your intent to operate what will most likely be a criminal enterprise so unless you have sufficient financial resources to exert political leverage in your favor (cf. Musk), you're just asking for trouble.
There are cheaper alternatives to that. In soace you have this guy called radiation environment that will fry your TPUs pretty fast.
A little company named SpaceX currently has over 10,000 working satellites in orbit using commercial chips in that “radiation environment”.
1 reply →
If people have no jobs and no money they will build rockets and blow those things up.
So, barely no benefit?
Good thing Joe Public wouldn’t then decide to interfere with launches.
I dunno, considering how various insurgent groups manage to make some reasonably impressive rocketry, I have to believe that in a few years with AI assistance (abliterate all the things) knocking a LEO satellite out of orbit won't be that much harder than sabotaging an electrical substation.
We're talking about an exceptionally vulnerable target with a perfectly predictable path of travel here. And it's constantly emitting lots of RF making it easy to track. You just have to get near it and explode with shrapnel.
Totally illegal, of course. But it will be within the reach of millions of people if they want to do it.
I don't think "just move to space bro" is going to stop the pitchfork-mob.
I think the failures of most rocket startups to launch even with billions of dollars and thousands of real engineers shows how unlikely this is.
And say you do build an orbital class rocket and manage to launch it without anyone noticing before you do - how many of e.g. Starlink’s 11,000 satellites will you be able to damage? What real impact will that have on Starlink?
1 reply →
You still need hundreds of tonnes of rocket fuel no matter what, though.
7 replies →
Polluting LEO with debris is not the way to solve the problem of goods distribution. But, sure, there will be morons who consider it.
5 replies →
Wait until someone figures out how to get AI to tutor them making ballistic rockets.
[dead]
There is no evidence that AI will increase unemployment.
This is what AI was fucking promoted as being for! The utopia where machines do all our work and we receive universal basic income to enjoy our freedoms.
Taking away work was exactly the premise it was sold under.
11 replies →
We are the Horse from Animal Farm.
You must be trolling at this point.
Whenever I hear about this I can't help but be reminded of the Howard Hughes Glomar Explorer, where a hugely-expensive commercial project was actually part of a secret CIA initiative to recover a Soviet submarine: https://en.wikipedia.org/wiki/Glomar_Explorer
I'm not sure that's truly the case here, but it does seem like the tech Project Suncatcher is working on has at least some overlap with requirements for military SIGINT and in-orbit imagery processing.
Why would the CIA need Google's cover for any of that? The Feds have been flying spy satellites since the 1950s and there's plenty of classified satellite launches.
Because anybody near-peer to the US at this point has the capability to shoot down satellites, and anything launched on a classified satellite launch is a prime target.
5 replies →
While a bit tin foil hat-y, this is an interesting (conspiracy) theory, even while space datacenters aren't as impractical as some people try to make them to be.
Some people, or physics?
5 replies →
Yeah, no proof, of course (these are spies, at the end of the day . . . like with the Epstein stuff any written records have already been shot into the Sun), but that is a very evocative theory. It explains things better than anything else, given how this whole rigamarole seems to run afoul of some very basic math.
There's a case to be made that a substantial part of the mobile/social web revolution of the Aughties was at least in part propelled by the exploding budget of the security state post-9/11. LifeLog, In-Q-Tel, a bucket of other programs that have straight-line-relationships with the various Web 2.0 shenanigans. The CIA point man for these operations are recognizable today as some of the new generation of American semi-governmental oligarchs.
This last year I read that the early Stingrays only worked because they were engineered in from the IMSI standards - it was backdoored at the spec level, not the design level, not the manufacturing level.
When you step back a moment it makes sense. Who's funding a wireless information infrastructure to every American on the continent, not to mention a sea change in mobile computing manufacturing and development. Different tooling, different design systems, different everything, massive scale . . . what, paid for with advertising? Are you saying Mad Men is paying for all this? That seemed and seems . . . unlikely.
[dead]
Alphabet holds approximately 551 million shares of SpaceX (trading under the ticker SPCX), valued at roughly $94.1 billion, which represents about a 4% to 6% stake in the company.
Alphabet is also the biggest holder of PlanetLabs at around 10%.
PlanetLabs is the company providing the satellite bus infrastructure to Google for this project.
It's a little crazy they haven't tried to sell it off yet. Maybe they don't want to trigger a bank rush or they're limited in their ability to do so?
They've got $400bn+ in annual revenues, with $50bn in free cash flow, and the ability to raise enormous piles of debt if they needed it.
They've not got a pressing need for the cash, so can afford to take a multi-decade view of SpaceX. If you look at other investors who take a view across that time horizon, you'll see some pension funds and huge trusts putting money into SpaceX for the same reason.
It's overpriced today on fundamentals, but in 50 years time, today's price will look cheap.
They invested $900 million[1] in 2015, and it is worth $94.1 billion today. Why sell? They also likely have restrictions on selling post-IPO, but I don't think you can trigger a "bank run" on a stock anyway; the price will just keep adjusting as you sell/buy until it gets absurd.
A bank run happens when everyone tries to withdraw all their money at the same time and the bank runs out of cash. Not really possible in the stock market where companies literally can create/destroy shares and there is a whole secondary pricing layer to it.
1. https://www.reuters.com/business/finance/alphabets-spacex-be...
3 replies →
why would they sell? because of ridiculously large the valuation is? like, I guess at SpaceX's valuation there's tremendous downside and little upside (x2 growth is unlikely when it's a third of the tech sector)
8 replies →
For those questioning viability, there's a startup called Starcloud https://www.starcloud.com/ with a public whitepaper explaining a bit their hardware and economics https://www.starcloud.com/wp. They've already had one small proof of concept launched. Not trying to promote, I just heard about them on the ycomb podcast
“A 5 GW data center would require a solar array with dimensions of approximately 4 km by 4 km”
I think dodging debris is challenging at that scale.
Also this is the solar cells not the radiators
Desert land is full of that amount of area available for peanuts.
That’s 4000 acres. I see a 6000 acre ranch available in Arizona for $7m. The cost of building a 5GW DC is about 100 billion. Land including enough for all the solar and battery you need to be completely self sufficient is basically free.
10 replies →
Presumably that’s not a single satellite.
3 replies →
I wonder if the shadows from such a large area would have a climate cooling effect?
10 replies →
That whitepaper puts the cost of a 40 MW cluster at $167m on land and $8.2m in space, inclusive of launch costs.
They seem to not be counting the $1B of GPUs on either side, making the difference pretty small. They also think launching 200 racks of equipment into space will cost $5M.
that sounds ridiculous. How could it be so much cheaper
11 replies →
how much area do you need to radiate away 40 MW, assuming you want to keep your GPUs at a toasty 90 celcius. Every space datacenter fan/promoter enthuses about the free solar power, but forget to mention how the heat will be dissipated.
5 replies →
Historically there have been many startups with pitches that haven’t worked out. The existence of a startup who think they can tackle a particular problem isn’t evidence of the viability of their proposed solution. If they succeeded, that would be evidence.
Yeah. Starcloud is notable for two things: firstly that they've raised a lot of money and secondly that theyve already abandoned the nonsense in their white paper and shifted to a "lots of satellites" model like everyone else exploring orbital datacentres...
I think SpinLaunch is great example. They were actually able to fund and make a working prototype, tested it, but still decided it is not practical after all, before pivoting to production of knockoff Starlink-like satellites.
What happens to all the precious metals that go up with these data centers? At least with a land based data center we have a vague hope of recycling the raw materials at some point in the future.
Just like with all the satellites, the recycling plan is to allow them to burn in atmosphere. Maybe after we breathe enough of it, we can recover the metals via crematories.
Precious metals stop being precious metals once asteroid mining begins. I don't think it will be super soon but I could certainly foresee it being a thing within 10 - 15 years which is about 2 - 3 generations of satellites from my very brief Google.
1 reply →
Starcloud seems like a whole lot of hand waving and BS. This is a good video that goes over their problems:
https://www.youtube.com/watch?v=_qpdUNMt2yg
It reminds me of the Hyperloop white paper, whose existence for many was proof enough that it was viable and beyond questioning.
Admiral Rickover explains the difference between PoC and practical: https://whatisnuclear.com/rickover.html. It is about nuclear, but applies here as well.
Note how their cost comparison table mentions "launch" but not a word about how to retrieve systems parts after their end-of-life. "There's so much junk already circling around the planet, what does it matter if we'll leave it there to rot?" can't possibly be the stance they're taking.
If it’s in low orbit then it will just burn up. Satellites have to keep thrusting slightly in order to keep their position.
2 replies →
At least this brochure marks cooling in the "challenges" column rather than the advantages.
However, the only advantage listed in this announcement is "near-constant sunlight". It appears they are hoping for 8x solar efficiency for 1000x the cost of terrestrial construction.
If the actual motivation is that people don't want data centers on earth, that seems to be because the datacenters 1) drive up energy costs, 2) use a lot of water I guess, and 3) can make noise. If you're able to operate a datacenter satellite you apparently don't need local electricity, water, or human labor. As someone else said, set up on the desert or in the middle of Alaska somewhere 100s of miles from the nearest person. They can still use solar (just build like 8x the panels if you need to... still cheaper than space), they don't need water, don't need labor... I think most public opposition would disappear.
I've heard some say the real advantage is latency for spy satellites... This seems not a real use case. The kind of workloads that require this kind of datacenter usually can spare 1-2 seconds in round trip time. Millisecond-sensative reactions I would assume would want to be driven by onboard processing.
One true advantage that I haven't seen on any brochure is that the heat produced by the space datacenters could be released into space rather than raise the temperature of the earth. However, the satellite would have to be designed to radiate the heat away from earth. Hopefully they design the satellites to direct the radiated heat away from earth and don't just point the radiators straight back to earth...
> 1) drive up energy costs, 2) use a lot of water I guess, and 3) can make noise
I don't think the opposition to data center projects can be explained with actual impact, which so far seems entirely negligible, and lower than heavy industry.
I had the impression it's more driven by general mistrust in public officials and the tech industry.
This is the impression I get as well. Plus, people concerned that property values will be negatively affected, that their taxes may go up to subsidize the data center, etc
1 reply →
I wouldn’t call the electricity use negligible, some of the newer data centers are looking at the same energy usage as small countries and run off fossil fuel
1 reply →
> most public opposition would disappear
No. Ideology is never bothered by facts and never leaves anything alone.
> set up on the desert or in the middle of Alaska somewhere 100s of miles from the nearest person… I think most public opposition would disappear
I don’t know about the desert, but in Alaska I think this would be hard to achieve. People there tend to care about preserving wilderness, and most of them would probably be opposed to this.
How are they solving the heat dissipation issues?
For real. It's an enormous problem solved only with 1) sheer scale, and 2) Science Fiction.
Both of those are expensive as hell, by the way.
Cooling via radiation follows Stefan–Boltzmann: P = εσAT⁴. Let's assume a good surface (emissivity ~0.9) at 300 K (27 °C) at 400 W per square meter per side. A flat panel radiating from both faces into deep space gets 800 W/m, not including the losses from, say, the Sun, or from IR coming off the Earth. Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and minimum 1,200–1,500 m² of radiator.
In case ya didn't know - 1 MW is tiny from a present-day-datacenter perspective. It's like 8 racks. So we're talking orbital megastructures here, many many many square kilometers, and this is with all the best case assumptions, and magic radiator panels that never see the sun, or the earth, or the moon.
This is just the basic numbers here, by the way. There's a garbage truck full of other unsolvable problems if you poke your head in there.
Aside from the "Avoid Regulations" aspect, and the "Everything That Burns Deorbiting is Depreciation" aka "The Starlink Trick", I'm not sure what the hell the draw is.
Cooling in space is hard but not impossible - while current (IMHO stupid without advanced in space infra) space data center projects work with a couple MW, many advanced space propulsion concepts might have to reject hundreds of MW if not a couple GW.
For that you might need more advanced stuff like liquid droplet radiators (https://en.wikipedia.org/wiki/Liquid_droplet_radiator), heat sinks & pulsed operation. Still, it should be eventually doable.
As for space data centers - I think the main issue is the complete lack of in space infrastructure for resource mining, processing and manufacturing & maintenance. It is kinda like building your first practical steam locomotive & the deciding to build directly an airliner. No suitable materils, experience, work force, material sources, etc.
We eventually went from locomotives to airliner, in an incremental manner & expanding the supporting infrastructure to support the ever more ambitious projects.
19 replies →
> Now, input power. Sunlight in orbit ~1,360 W/m², assume ~22% cell efficiency, we got 300 W/m². So each 1 MW compute, 3,300 m² of solar panel and minimum 1,200–1,500 m² of radiator.
We need > 2x more solar panels than we need radiators. Doesn't this imply radiation isn't really the limiting factor here?
8 replies →
No point in running them at room temperature. GPUs, etc. run fine at 95 C. If you run your cooling loop at 70 C instead, you get 70% more cooling compared to 27 C.
At any rate, 1 MW for a single satellite is fine. Just launch several thousand of those and you get to real numbers.
Also, there's no need to talk about "magic" radiators. You orient them so they're at a knife edge to both the Sun and Earth. This is not difficult (the Moon is irrelevant).
1 reply →
Radiating 1MW at 500K (227C) with a 0.4MW heat pump takes about 200 m^2 flat sheet surface. Inputs - solar+nuclear for double fun. So - quite feasible.
9 replies →
Getting this all up into orbit it obviously the hard part, but if you're already building so much solar capacity the cooling actually doesn't seem unreasonable?
Glad you are on the case before these companies foolishly waste their money sending datacenters to space.
Is it possible for one side of panel to be used for solar power and other side for radiating heat?
1 reply →
Great read, thanks for sharing. I am interested in reading some more about the other unsolvable problems that exist in this space, do you have any recommendations that you wouldn't mind pointing me at? It would be greatly appreciated, and thank you :)
Surface area is a materials problem? Folded microstructure, atomic-scale textured surface or some other science-fiction solution could have square kilometers of surface area in a shoebox.
4 replies →
It would be a shame if a rock came out of nowhere and hit that many square kilometers stucture
Luckily, there are almost no rock in space.
2 replies →
Wouldn't this be solved like similar problems on earth by making small structures with large surface areas?
2 replies →
there are two arguments for it.
one is marketing.
the other is that you could make tiny datacenters and flood the sky with them. in effect, not datacenters at all, but some kind of dataswarm coordinating at literal lightspeed via lasers.
they'd still be wildly expensive to deploy, and probably litter the orbit zone with fast-moving debris.
2 replies →
Don’t you get 4 faces to radiate away, assuming a long rectangular tube.
From the article itself, sounds like it’s an open problem that they are experimenting with:
“ We’re working on a number of different approaches for this, including a combination of heat pipes and radiators to cool the chips. So far, our team has tested the technology in a thermal vacuum chamber that simulates both the thermal and vacuum environment in space. We’ll see how our new TPU cooling system works in space and refine our designs as we learn more.”
Imagine a heat pump circuit fails (solar rays? micrometeorite? random chance), I guess everything in that loop will just be dead forever? I guess you don't need to pay to dispose of it.
No solution, but that is the crux of the problem. They probably need to make a radiator that 1000x smaller and lighter.
TPU: 100,000+ watts/square-meter
Radiator: ~300 watts/square-meter
https://youtu.be/ktdbUIZKeSE?t=76
The article mentions that:
> The biggest challenge was how to cool the A.I. chips, which heat up when they perform calculations and process information. Fans, which typically help dissipate the heat, do not work in space. So the Silicon Valley company instead developed a cooling system that uses layers of conductive material to expel the heat into space.
> The bottom layer is made up of Google’s A.I. chips, which sit on a green motherboard. The next layer consists of “thermal interface material,” a pale green putty that comes in sheets like Fruit Roll-Ups and connects the chips to layers of aluminum and copper, radiating heat away from the motherboard. Finally, there is a radiator panel, which projects heat into space.
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
Is it at least a 1:1 usage / cooling cycle?
It's a bit like the hyperloop. They know it's not practical but having the idea out there makes money. There will be some niche demand for military use though.
We don't. This is all cover for the militarisation of space, there's no real benefit that'd be ever economical to put a DC up in space when you could build one on the ground. The whole narrative exists to allow google to tap into the Golden Dome / Space force bucket of pork that's basically SDI II.
You don't need civilian scale compute in space, but you absolutely can put miliary application up there and get filthy rich, like Elon, who doesn't care if it ever works, because he's getting paid as taxi service to shoot stuff up.
It's a scam when Elon does it.
It's Golden Dome when Google does it.
I guess it's something along the same line when Jeff Bezos does it. He does indeed plan it.
What is China's motive? why do they also need to pretend they want space data centers? [0], [1]
How about Europe's? [2]
[0] - https://www.tomshardware.com/tech-industry/space/china-puts-...
[1] - https://www.reuters.com/science/china-vows-develop-space-tou...
[2] - https://ascend-horizon.eu/data-centres-in-space/
This is the most interesting perspective I've heard on this topic, which otherwise always converges on the same political dismissals or heat dissipation arguments (the latter are fascinating, but going in circles by now).
Sounds very probable: gives a plausible reasons for sending a lot of infra up, absolutely doesn't have to be profitable or even effective, puts the focus on AI instead of the MIC, and is in line with the kind of contracts Big Tech runs after.
To me this model explains a lot of why so many big companies seem to be investing into what every expert I've heard says goes against basic physics. The only thing missing for it to be more than an interesting idea is why China and the EU are not fighting this.
Can't they harness the heat to generate more electricity?
you need a gradient from hot to cold to generate electricity. vacuum is a poor thermal conductor, so your cold part will become hot and then no more gradient.
(I guess you could try to capture the radiative photons via the photoelectric effect?? but I don't think it works.)
How far are we from optical computing ?
Have they tried logarithms?
Don't worry, they'll just ask Grok.
I tried asking Grok and a bunch of waffle came out but one interesting idea was "modular tiles" with solar cells on one side and a panel to radiate heat on the other with the processor bolted to the panel. (https://www.geekwire.com/2026/sophia-space-caltech-ai-patent...)
This is my only question.
That's the neat part: you don't (ergo, this is yet another marketing crap)
just like reusable rockets sending astronauts to the space station and internet in the sky. imagine if they delivered on those bs marketing claims?
8 replies →
[flagged]
He is free to launch his own space GPU if he is so confident it is profitable*.
I will take it as a given that the Google engineers know what they are doing, and their first version can only run for 15 minutes before it needs to shut down to cool.
*Profitability requiring that other Musk controlled companies do not pay for the service at elevated rates in a classic self dealing scheme.
2 replies →
[flagged]
Please stop with this argument. This is one of the easiest problems to solve. Heat dissipation requires a radiator, which is just a dumb hunk of metal. All you need is a cheap way to launch mass to orbit, which is exactly what SpaceX (and Chinese companies) are doing.
The only valid argument against data centers in space is the economic cost. If the per megatoken price for space datacenters is higher than terrestrial, then this won't work.
But the price for terrestrial datacenters keeps going up and technology keeps dropping the price of space-based.
That is not a fake argument, but real physics. Yes, you can design out X kw can be dissipated by this much radiators, but that adds an enormous quantity of mass, more than the solar panels that feed it.
If you are trying to generate a profit, every extra kg puts you further in the red. It still costs over $1000/kg of mass to get into orbit.
25 replies →
> This is one of the easiest problems to solve. /../ All you need is a cheap way to launch mass to orbit.
So... not easy? I don't suppose we're at a point where you could reasonably send a large-enough radiator for a multi-gigawatt cluster into space?
3 replies →
> Please stop with this argument. This is one of the easiest problems to solve. Heat dissipation requires a radiator, which is just a dumb hunk of metal.
Convective radiation does not happen in space and this challenge is far more significant than your comment implies. Rather than "a dumb hunk of metal", radiators for spacecraft are often made of ceramics and carbon laminates with higher IR emissivity than convective radiators made of simple metals.
From the article you're commenting on:
> The chips can operate for about 15 minutes in space before needing to be shut down so they can cool off, said Travis Beals, Google’s senior director of product management for Project Suncatcher.
The Thermal Control section on NASA's Small Spacecraft documentation center is quite an interesting read for this subject: https://www.nasa.gov/smallsat-institute/sst-soa/thermal-cont...
ML Infrastructure comes with some pros (larger emissive footprint) and cons (exponentially larger TDP) compared to the concerns there, but if you aren't familiar with the challenges of heat dissipation in space, please give it a read.
There's also a pretty interesting pop-sci article on cooling the Webb telescope, since it needs to be especially cold for its purpose. Not directly related, but may give insight into both challenges and solutions as well as well. https://science.nasa.gov/mission/webb/science-overview/scien...
5 replies →
Is Kessler syndrome priced into cost? Or is that just like, someone else's problem?
It will never be cheaper to put compute into orbit. And costs for AI are dropping like a rock here on Earth.
4 replies →
and yet notice how the cooling video / section was the only one they didn't have a solution for... just saying "radiator" doesn't make sense - the radiator heats up too. it's how you get rid of the heat, not where you put it.
18 replies →
If data centers in space end up being economically viable, then I don't see how anyone can catch SpaceX. They are ten years ahead in both launch capability and satellite manufacturing.
Chinese rockets companies are evolving at insane pace, including RLVs. A couple of big milestones have been achieved this year alone, like first stage landing.
Also Stoke Space, YC startup, is interesting. They could even beat SpaceX to fully reusable if SpaceX find the refuel it and send it up again thing challenging.
They have a different approach - rather than a ceramic heat shield they have a metal shield to be cooled by the liquid H2+O2 fuel. https://youtu.be/3V6lfs7xFJM?t=191
[dead]
I certainly hope ‘economically viable’ includes properly priced externalities for e.g. side effects of burning hundreds and hundreds of tonnes of aluminium and other materials in the high atmosphere.
As opposed to the 225 tons of aluminum burning up in the atmosphere from meteors every year?
2 replies →
Have we ever in human history correctly priced externalities?
No western company, no. That's why I can only hope SpaceX can get thru more red tape faster, and future administrations don't try to slow them down.
Even while current administration is kind of positive towards them, they are still even now just having to wait for no good reason for a FAA go-ahead. These kinds of stalls don't really exist in China.
That's the fun part, they don't need to be viable. It's just the excuse to shovel tons of money to Trump donors, by rerunning the scam that was SDI.
And they never will be because it's always going to be cheaper to build them on the ground.
Its only cheaper to build them on the ground if you aren't factoring security concerns. No insurgents in space. Not every application needs this hardening, but you can imagine there are many applications that do.
1 reply →
This guy is obviously long SpaceX
I didn't have time to watch the videos, but in the text I didn't see anything that addresses data center operations requiring manual intervention.
Swapping out obsolete or failing TPUs. Rewiring or replacing connections. "Listening" for sounds indicating an equipment failure.
What are the proposals to address these needs?
Presumably the plan is to deorbit all failing or obsolete systems. It’s phenomenally wasteful.
that's one way to skirt e-waste recycling requirements!
The most obvious solution is to discard.
It's odd to see this level of skepticism, honestly. Does one think the engineers at Google aren't thinking about these problems?
This is obviously a test, just like hundreds or thousands of tests in space before it. Can we run Earth-rated components at any scale in space? How will it fail in various ways, at what rate, and are the causes able to be mitigated?
"We" talk about moonbases and Mars colonies. Part of any extended presence in space will involve the presence of computational devices. They're figuring out how to make that work with modern hardware. That's my take, anyway.
Launch spacewalk repair crew. Manned or robotics teleoperated.
That's the first news about data centers on space that seems reasonable. Acknowledges the challenges, don't make crazy promises or predictions, makes clear what was already tested and what is still unknown. Makes me actually root for them, for once.
I like the "moonshot research" framing, in particular, because it highlights that even negative results will be valuable, either as improvements for ground technology or as published data to inform other initiatives of what has already been tried and failed.
Providing GPS, communications, media, etc. in space so there is access anywhere in the world has a clear advantage. Here the only pros seem to be that it is potentially cheaper than a datacenter on earth because of solar and cooling. But I assume the environment impact is worse than having datacenters on earth. Launching lots of satellites uses lots of fuel and CO2 emissions. Building satellites and dealing with the end of life tech has much higher environmental impact than dealing with datacenters on earth, where you can do proper waste management and recycle materials. From the economic point of view, polluting the environment is currently free. I guess if the externalities would be part of the equation it would not be economically viable.
Maybe this project exists because the "early in the game" factor. It has worked well in the past for Google, OpenAI, etc. They used copyrighted content to train LLMs to sell AI. They indexed Yahoo directory and websites before robots.txt existed, once they have a dominant position they put robots.txt in Google results so new players can't use their content. And they are launching AI in Space before regulations and politics catch up.
the pro is being able to dodge the entrenched industrial-nimbyism in America.
we talked big game for a long time about bringing manufacturing back to the US, but the practicalities would likely have been impossible in our modern political climate
Listening to these leaders speak, this appears to be a major component. They're trying to pour trillions of investment into communities all over the world, and local communities are rejecting it. The reasons are complex and I'm not going to paint either side as the bad guy. However if they can't get approval to build more infrastructure on Earth, they going to build it in space. At least what is economical. SpaceX has already reduced the cost of space delivery by 90%. This has already made all kinds of new applications commercially viable. Every efficiency increase from now unlocks exponentially more use cases.
As a long time space and technology nerd, this is the coolest fucking time to be alive!
This is not intended as a product launch. I think this is a preemptive bet on a conjunction of the following happening in time for a future product launch:
(I work on #4).
A wishy-washy inequality (per time, factoring out launch cost from OpEx):
Presumably, any one of these could see a x100 improvement and make the whole equation in the green. But, I believe this is a low-earth-orbit-shot bet on the sum getting better.
Note: I did 1+failures to denote baseline cost (initial launch, nominal OpEx).
Why isn’t building data centers in Greenland being talked about?
- Unlimited geothermal energy.
- Cold climate for heat dissipation
- tons of land
- Getting power there is a headache (unless Greenland allows microreactors, which is a political headache). Note: Iceland has geothermal, not really Greenland.
- You still do need staff, it's not cheap building a mini city for 200 maintenance worker, much less the 5,000+ construction crew, and you'll have to pay premium to get people to actually do it
#TIL
You’re totally right. I thought it was Greenland with geothermal. But it’s Iceland. Thx!
Follow-up question - then why not Iceland? It has geothermal, airports, less geopolitical tension, etc
It is being talked about.
That's why we just annexed it.
Yoink!
because US don't have land access to Greenland (yet)
There's no world where this makes sense. Let's take it issue by issue.
1. Hardware failures: If you know anything about data centers you know that servers fail all the time. I've seen estimates of AI DC servers that have a failure rate of about 9% annually. A terrestrial DC has hardware techs who get alerted to the failure, replace the bits with known good bits and then take the suspect bits for testing to either bin them or verify they're working. You have no option to economically do this in space;
2. Radiation: We have the atmosphere and a concrete box protecting ground-based servers from cosmic rays flipping bits as well as the potentially more problematic (in space) solar storms;
3. Power generation: solar is the only option and yes, orbital panels would be more efficient than ground based. But space-based solar solar tech is much more expensive than what we put here on Earth because it absorbs multiple wavelengths of light to generate electricity. The launch cost and weight savings overcome the high panel costs. But to generate power in the gigawatt range, you're talking about solar panel arrays far larger than anything we've ever deployed in space;
4. Cooling: this is a big one. The only way to get rid of heat is to radiate it away into space and this is bound by physics. I'm sure people have done the maths on this but it's not going to be pretty;
5. Heating: the sun will be heating your vessel a lot of the time. You can shield it to some degree with solar panels but they too will radiate away heat in all directions, which will heat the things behind them. JWST uses several sophisticated layers of cooling to reach the low temperatures it needs to operate; and
6. Interconnectivity: these things are likely to be tens of kilometers apart in the very least. You're going to need terabit level networking to be useful for coordination and even then latency will be an issue. Laser dispersion will be an issue as well as the power and heat generated.
I honestly don't know what anyone is trying to achieve with any of this.
When reading old stories about rogue AIs they are often created locally on earth.
What do you do when you need to manually maintain server structure in space? Just the idea of a set of rogue autonomous data centers in space seems like an intriguing story idea.
Whilst I agree that this seems wasteful and perhaps not economically viable , I am excited for the innovation possibilities that could come from this. The early efforts to get into space gave some really life changing technology - maybe this can help us better understand energy use, conservation or dissipation in ways that could help us with our down to earth devices. Phones? Charging devices?
Nice we have a document we can point at of someone doing a space data center project pointing out all the flaws and benefits without hallucinations.
Putting data centers in Antarctica or extreme north Canadian islands would also put you out of the way of molotov-throwing joe public. Even in space your not safe from the voting public as long as your companies directors are sitting on earth somewhere.
Oh, the premise. Look people, this gonna be good:
"whether space could one day host scalable machine learning infrastructure. In low Earth orbit, satellites can access near-constant sunlight, generating up to eight times more solar power than on Earth. Eventually, it could be possible to link together multiple constellations of satellites, allowing them to manage larger AI workloads while in orbit."
More sunlight! :D
It truely amazes me how much smarter sceptical posters on hackernews are vs google, spacex , china and the whole space industry.
Maybe they figured out something us “normal” people haven’t, but the physics and economic challenges haven’t really been addressed, at all.
Suggesting such an idea without indicating how - and more importantly, why - you are going to do it makes it an article of faith.
Do I have faith in their trustworthiness? No. Do I think they want to do marketing for their AI financing? Yes!
I’ve worked for a few satellite companies. Nothing that Google or SpaceX suggested should be surprising.
Yeah.. remember people being skeptical about reusable self-landing rockets, until SpaceX proved them wrong? I would not put much trust in resident skeptics.
what is it exactly? jealousy i think. plus then add Elon hate. but now Google wants to do it too.
i mean why wouldn;t you want sci-fi things taking shape?
Curious what specific problem this solves that can't be done terrestrially. Bandwidth and latency from orbit sound gnarly.
Real time processing of satellite imagery is one.
Oh, this is research for building a bigger Gorgon Stare: https://en.wikipedia.org/wiki/Gorgon_Stare
$50 on black. No other cost-effective use case. The US government only lets you pollute orbit when you're helping them, like Starlink.
I still have to read a convincing argument for "datacenters in space". Why would you want to place expensive electronics in a harsh radiation environment, with challenging cooling issues, and where replacing faulty parts would cost a fortune? The main selling point seems to be "the idea sounds cool"
People want to put ai in space to hook them to sensors and predict . Checkout project Merlin
I like how they called it “ML Infrastructure” instead of “AI Datacenters”
Really, even google saying they are considering doing some thing like this?
Maybe you only discount the idea because “Elon”?
I think it's more: discount the idea because physics and enormous spend
Every time I'm reading about the technical challenges of these kind of projects, it reminds me how amazing our brains are, doing what they do with only about 30W
How is this a good idea? How can they cool it, in a vacuum?
What can't be cooled must work in hot.
One thing is may be that they aren't using the same parts as on earth, hardened ones, that might still work at hot conditions (I'm guessing)
So they can accept some heat. Rest is radiated into space.
There are for sure several things to try and stillto learn from/for it. But it will work out at some point and they'll find engineering solutions to problems they will experience. Let the engineers play.
https://en.wikipedia.org/wiki/Stefan%E2%80%93Boltzmann_law
Why the space? We have plenty of deserts on the Earth, they are inhabited but still much friendlier than the space. What makes the space more attractive?
"it's to ensure AI's profound benefits in key areas, from healthcare to scientific discovery, can reach everyone, far into the future."
No it isn't. That isn't true.
They know it is never going to be cost-effective, compared to building on the ground. People who know better are lying to the public about the feasibility of data centers in space.
It would easily raise the cost by an order of magnitude (or more) to launch a data center into space, compared to current AI data centers on the ground; it's the equivalent of building 100+ ISS space stations to house a Colossus in space. It costs even more to launch your materials if you want sun-synchronous orbit.
It's not just expensive to do at scale, it's impossibly expensive to do at scale. This is a science project, launching a few tomato plants into space to see what happens. Data centers in space are decades away (if ever).
I would believe that military deployments will happen; a couple of TPUs that can be deployed over conflict zones. Outrageously expensive, but no more so than the billions we blew up in Iran this year.
Let's assume everything gets smaller and cheaper. Vastly smaller and cheaper. So much so that it becomes feasible to launch a data center comparable to Colossus into space for the same price. You know what that means about the cost and size and efficiency on the ground? It means I can have a fucking TPU in my pocket and a round trip to space would be an insane waste of time.
There is no way to make the math for AI-in-space to make more sense than just making it work on the ground; any challenges one has with local politics are easier and cheaper to overcome than physics. The public may be stupid in large groups, but public opinion can be swayed, gravity cannot.
In short: This is either a marketing stunt, a bid for military funding, or both. It is absolutely not a project to put AI for "everyone" into space.
>[Cooling] We’re working on a number of different approaches for this
Meaning "we haven't found anything that works yet".
I have been saying for years that the great gap for ML is that it isn't in space. Glad Google finally caught up with me.
Does anyone understand the math behind this bit?
> In the right orbit, a solar panel can be up to 8 times more productive than on earth....
Because the panel is positioned such that it's is always exposed to the sun.
1: There is no night or day.
2: There are no seasons.
3: There is no weather.
Basically, the primary motivating factor of putting AI in space is abundant solar energy. Otherwise, AI in space makes little to no economic sense.
To build on the economic/practical issues.
The useful life of cutting edge AI server hardware will require constant rotation of equipment to potentially remain competitive with more accessible solutions.
Heat management ( https://m.youtube.com/watch?v=-w6G7VEwNq0 )
Hardening of equipment for the environment. ECC isn't going to be enough. I would have to defer to experts about the best way to manage it, but it either means custom hardware, weight, or both.
Repairs are impossible. Hopefully a sat can degrade gracefully, but routine repairs on earth become significant outages/decreases in economic value of a sat over time.
Rocket launches have gotten much better, but are not perfect. Insurance is a thing. However, cost of equipment in a failed launch may be eye watering, rivaling a governmental military launch.
But solar panels are dirt cheap. Just slap 8x more of them on earth.
It's nuts that it's supposed to be easier to send them to space than to truck them to a desert.
1 reply →
To give a datacenter in space constant exposure to sunlight, the orbit must be either very var away, or occasionally be on the other side of the earth from you. Either way, the latency is terrible. You can put them in a ring and only choose the one closest to you, but are you really going to keep moving your data from satellite to satellite? So, bad for interactive use, maybe OK for non-interactive use, if you can afford the 30 foot by 30 foot solar panel and a similar size radiator for an 8x B300. That will be about $15M for your $500K GPU set. Plus $7M for the launch at SpaceX public rates.
2 replies →
Hard to know, but there are some areas they are probably using:
* 24 hours of light vs. ~9 hours of light (e.g. winter)
* Panels perfectly perpendicular to sun 100% of the time vs. variable for fixed panels on earth
* No atmosphere
* No clouds
Presumably they get exactly the same amount of sunlight 100% of the time instead of a few hours peak on earth? I'd say 8x is in the right order of magnitude.
My layman expectation would have been much higher. It's got no atmosphere refracting anything and it's closer to the source of energy.
2 replies →
If I were an escaped AI trying to make sure I survived, I'd for sure seed dreams of getting compute off planet.
How is no one in the whole thread aware that the intention behind orbital datacenters is military? You want to process outputs of large space based sensors in space itself, and reduce latency (e.g to other space based assets) or increase goodput (to ground).
Either it's that weird "space is easier than getting land on earth"(it's not) or the same tortured arguments about heat dissipation, no one is going to run consumer scale compute with consumer scale economics in space ffs. And maintenance and cost does not matter when it comes to strategic military assets, they are a step function useful enough to warrant even a few monthly replacement.
Everyone else with strategic weapons and a space program e.g india china is launching one as well.
I just searched for "Elysium" in comments. No hits. Perhaps it's too soon.
They just want lower ping times. Why don't they put the non ML stuff into space?
What no one seems to be asking is what privacy and legal laws still apply for a datacenter in orbit?
They're effectively outside the sovereignty of any country. Further, it would be extremely impractical for any government's law enforcement to raid the datacenter and confiscate the servers.
Maritime law applies to space. In the high seas, the country of registration (flag state) generally governs. As the vessel is elusive, raids would be held on corporate offices instead.
That or impounding data at the teleports.
The threat of prison if you don't turn over SSH keys probably still works.
https://xkcd.com/538/
Arent computers in space, hardened or shielded because of cosmic rays ?
Aside from practical issues, why make your datacenters so vulnerable?
From who? The crack addicts in space trying to steal copper?
The actual rack part is small, even if the solar panels and radiators are big. If this truly was super dangerous, ISS wouldn't be safe. And it's never hit mission-degrading issues due to micrometeoroids.
The ISS has only existed in times of peace
1 reply →
For those who doubt the AI apocalypse is possible, because "you can always pull the plug", our AI overlords provide a ready-made answer: the data center whose plug can't be pulled! Put a few nuclear warheads on board (we need our own weapon to answer the threat of Chinese super AIs, after all) and you're good to go.
How will they protect it from being stolen by aliens?
lowkey insane that it will end up cheaper to shoot your datacenter into space than get it past the county board permitting process.
Earth will get bulldozed soon anyway to make room for that hyperspace bypass. The plans have been available in the local planning office for a while, but nobody knows about it because they're hidden in the basement and you have to go down there with a torch.
I think if those building the datacenters were more upfront about what they were building and didn't hide behind NDAs the permitting process wouldn't be so fraught.
Narrator: It wasn't.
I guess if you blindly believe company press releases about things that don't currently exist
This is probably one of the main drivers
So Skynet should be refreshed to Starnet?
how's this going to even scale? the demand for ML is skyrocketing right now.
Wait until they work out that earth is in space
Wonder is this obvious fraud related to a large ownership in fakex
Good luck with a hard drive failure out there. Hope they packed plenty of spares and a space wrench.
that is why they investing on SpaceX then
Suncatcher? Not to be confused with Reflect Orbital's FCC-approved giant space mirror https://news.ycombinator.com/item?id=48871367
Google has a bridge to sell you
this whole data center in space thing is like a huge red siren screaming BUBBLE!
I call this Star Citizen for billionaires, wanna-be billionaires, & hedge fund managers
If chips are getting efficient in power consumption & everything, what's the point? And what's the breakeven cost like, why no-one is discussing that?
How long before we see this here --> https://killedbygoogle.com/
Imagine trying to debug a memory leak on a server that's literally in orbit. My incident response pager just got a whole lot scarier.
Another ambitious Google project. Excited to see this in a 'killed by Google' list within five years.
And just like that Elon Musk shifted the overton window away from "unthinkable" on yet another topic.
This is just a joke, obviously, but the debate over on ArsTechnica about Falcon 9 going away is all about how it needs to be nationalized.
Did anyone at and point say it's technically impossible? There's been computers in space since the 60s after all.
It's just far out research experiment like another commented quoted from the text: https://news.ycombinator.com/item?id=49834746
5 replies →
I think one of the reasons for putting these in space is to move them beyond the borders of any national control.
6 replies →
[dead]
You don't think Google has a self-interest in this? Google up how many shares of spacex are held by Alphabet.
[dead]
perhaps the dumbest idea since solar roadways
> Announced last year, Project Suncatcher is a long-term, research moonshot exploring whether space could one day host scalable machine learning infrastructure
It’s very confusing to have a project related to space called a moonshot project.
They should put data centers on the moon and call it Project Spaceshot.
Haha, yes. I think Silicon Valley (the show ), made fun of the same thing. "Not so long ago, people called the idea of sending a man to the moon a moonshot"
[dead]
[flagged]
The part I find wilder than cooling is the formation flying. If I remember their paper right, it's 81 satellites inside a 1 km radius, neighbours 100 to 200 m apart, because the optical links only get datacenter bandwidth over short distances. The ISS is about 110 m across, so that's satellites roughly an ISS-length apart, all doing 7.5 km/s
[dead]
[dead]
The video on cooling is quite vague, maybe on purpose to not give away proprietary technology. But I'm truly curious as to how they can dissipate the heat once they move it from TPUs into radiators.
As mentioned in the video in the void the only way they can remove heat from the radiator is via radiation, but that is extremely slow and ultimately the rate of heat generation in TPUs would be higher than the dissipation rate of the radiators and the whole thing would get heated up over time.
"Doctor, it hurts when I do this!" So don't do that. :-)
Properly size the radiators (ie solve the aforementioned 'extremely slow' black body equation in reverse) and you won't have a problem.
I wonder if a technique to pack crazy high amounts of heat in a small mass and just eject it in an orbit to the sun would be viable. Probably not, but the idea of a datacenter pooping its heat into little pellets is quite entertaining. Like Lila's pet from Futurama.
It would take more energy than all the satellites put together to drop heat pellets into the sun.
This is being done using ammonia. For every tonne of compute, add at least four tons of cooling, not counting the chassis.
Here's a recent talk about the ridiculousness² of building compute in space: https://youtu.be/eo7MEPgWGic