Forgot your password?
Close
typodupeerror
Communications Space Google

Google's First Data Center Satellite Is Online and Writing Poetry (techcrunch.com) 85

Posted by BeauHD from the so-far-so-good dept.
Google's first experimental space-based data center is operational after launching earlier this month, successfully running its Gemma 4 AI model aboard a satellite and generating a haiku about being in orbit. TechCrunch reports: The satellite, which carries four Google Tensor Processing Units (TPUs) and generates up to 1 kW of electricity, is intended to pave the way for future networks of close-flying satellites performing large inference workloads in orbit. However, the power demands of the TPU mean it can only turn one on for about 20 minutes at a time, complicating the already tricky business of space-to-ground communications.

After a commissioning process that tested out the satellite's subsystems over about three weeks, the spacecraft was ready to work. When engineers on the ground sent their first prompt this week, they asked the model pre-loaded onto the satellite, Google DeepMind's Gemma 4, to write a haiku about being in space. And thus:

Blue marble below
Data streams in silent grace
Starlight feeds my core.

The ongoing test campaign will examine how the chips perform under thermal stress and in the space radiation environment. [...] If the TPUs onboard this spacecraft prove capable, the next step will be launching two more satellites equipped with the advanced chips in 2027. Those spacecraft will support higher power generation and demonstrate inter-satellite optical links.

Google's First Data Center Satellite Is Online and Writing Poetry More | Reply

Google's First Data Center Satellite Is Online and Writing Poetry

Comments Filter:

  • Nobody fucking cares (Score:3, Insightful)

    by paul_engr ( 6280294 ) on Saturday October 31, 2026 @03:08AM (#66402654)
    Dumbass tech shit

    • Maybe they can send some fucking hammers and nails into space.

    • Yeah, why would you post something as ridiculous at "computer tech put in orbit" on Slashdot? Absurd.

      It's not like this is a TECH site anymore. This is a "we fucking hate MUSK" site.

      Get with the program.

      • Musk is not tech. It is possible to despise a charlatan and still be pro-tech. In fact, anybody who is not a fool should have this perspective, given that big tech is nothing but a scam at this point. Products are worse, there is no actual innovation, and they're trying to basically destroy society rather than compete on the real free and open market.

  • What a joke... (Score:5, Insightful)

    by Archfeld ( 6757 ) <treboreel@live.com> on Saturday October 31, 2026 @03:10AM (#66402656) Journal

    The cost of this is exorbitantly high for the barest minimal access and a huge waste of resources. They could have built 10 solar farms to provide realistic power but instead they are the first in space with a joke.

    • Re: (Score:3, Interesting)

      by martin-boundary ( 547041 )
      They spend trillions to make millions. What's one more waste of money to them?

      Meanwhile the twin billionaire assholes Sergei and Larry Google [yahoo.com] fled california because they didn't want to contribute to healthcare for their poorest fellow Americans. I guess they'd rather party on an island somewhere.

    • Re:What a joke... (Score:4, Interesting)

      by burtosis ( 1124179 ) on Saturday October 31, 2026 @08:11AM (#66402878)

      The cost of this is exorbitantly high for the barest minimal access and a huge waste of resources. They could have built 10 solar farms to provide realistic power but instead they are the first in space with a joke.

      But it’s one (1) kW of compute! It’s almost two (2) 5090 graphics cards!!1! Do you have any idea how much compute that is?! Latency and throughput of data because it’s in orbit don’t even matter at that scale!

    • This is what gets me. In terms of financial and technical viability, this nonsense will compare to solar farms + batteries + data centers on Earth about as well as a next-generation supersonic plane would compare to a freight train for hauling coal.

    • TFS clearly says this is an experimental orbital data-center. As in proof-of-concept. Nobody is going to do serious work in a data-center that small.

      That said, I confess I don't see the cost/benefit advantage of orbital data-centers in general, even considering their lack of a terrestrial footprint.

      • Great, they've gotten one step closer to proving that physics and math don't lie and this was a terrible idea. Maybe they should pay their janitors a living wage instead.
    • It's this moral panic over water and power for datacenters - results in people going too far to prove a point.

  • Science-fiction short story (Score:3)

    by fph il quozientatore ( 971015 ) on Saturday October 31, 2026 @03:12AM (#66402658)
    Gemma, write a science fiction short story about humanity going extinct and the existential dread of the only surviving sentient being, an LLM in a data center orbiting around Earth.

  • Sky (Score:4, Insightful)

    by clambake ( 37702 ) on Saturday October 31, 2026 @03:29AM (#66402662) Homepage

    Net

  • Happy cooling and triple redundancy. (Score:5, Insightful)

    by hmilz ( 3035377 ) on Saturday October 31, 2026 @03:42AM (#66402668)
    Idiotic. Reminds me of Sci-Fi novels where computer circuits used positrons instead of electrons. As if that were of any advantage.

    In space you have to deal with cooling in vacuum without thermal convection but being exposed to sunlight for energy harvesting, triple redundancy to counter radiation (charged particles) induced malfunction, and slow data links compared to ground based communications, just to name a few of the challenges. And that is supposed to be advantageous somehow, while low earth orbits get fuller and fuller, and malicious countries demonstrate satellite hunting satellites?

    • (Disclaimer: I build space electronics for astrophysical research.)

    • Re: (Score:2)

      by dgatwood ( 11270 )

      In space you have to deal with cooling in vacuum without thermal convection but being exposed to sunlight for energy harvesting, triple redundancy to counter radiation (charged particles) induced malfunction, and slow data links compared to ground based communications, just to name a few of the challenges. And that is supposed to be advantageous somehow, while low earth orbits get fuller and fuller, and malicious countries demonstrate satellite hunting satellites?

      The only theoretical advantage I can think of for a satellite would be that if someone launched one from international waters, if some random government wanted it to stop existing or stop doing something, they'd have to launch something to destroy the satellite, because from a legal perspective, it wouldn't be within their jurisdiction.

      But in practice, as far as I know, nobody is launching anything into orbit from international waters, so this is entirely theoretical, at least right now.

      • The reason for having data centers in space is because Golden Dome requires AI in space [wikipedia.org], presumably for target recognition. Google is working for the US military.

        • good point! I thought it was more BS for idiot investors or the investors trying to exploit idiot investors and their AI which does most stock trading ... perhaps that is why so much stupid news gets just jumps out of the market. The real suckers falling for it already have no significant portion of the market to sway it but the rest reacting thinking everybody else is an idiot or a bot predicting trending stupid would act accordingly.

      • Re: (Score:1)

        by quenda ( 644621 )

        The only theoretical advantage I can think of for a satellite ...

        Come on, I can't believe you are that stupid. You can't think of any technical advantage of operating a power-hungry machine outside of Earth's shadow?
        Try harder.

    • Re: (Score:2)

      by djinn6 ( 1868030 )

      1000 years from now, when space is full of satellites doing compute, people might learn (through the 3026 equivalent of a Tiktok) that this was the first "data center" in space.

      Of course they'll talk about the lack of even rudimentary magnetic radiation shielding and astronomical cost of space launch with this first attempt, but they'll also point to the limited surface area of Earth and the atmosphere's much lower radiation efficiency as to why we had to eventually moved them off world.

    • cooling in vacuum

      indeed. it seems they're not even thinking about cooling yet which will probably be the final show-stopper. this is a very early experiment just testing radiation resistance and operation:

      "Our initial analysis shows that the core concepts of space-based ML compute are not precluded by fundamental physics or insurmountable economic barriers. However, significant engineering challenges remain, such as thermal management, high-bandwidth ground communications, and on-orbit system reliability."

      while low earth orbits get fuller and fuller

      we've had no pesk

      • Other companies have launched data centers, and definitely are thinking of cooling. You can see the size of the solar panels and heat sink [youtube.com].

        • maybe (Score:3)

          by znrt ( 2424692 )

          Other companies have launched data centers,

          a single h100 for a preliminar test is not "launching data centers".

          You can see the size of the solar panels and heat sink [youtube.com].

          and radiators and heatsinks are different things. heatsinks make no sense here unless you want to periodically shoot them of to outer space and replace them with new ones. unless you have a groundbreaking new solution, space is the heatsink here, and you need radiators. where are they?

          all i can see there is 4km2 of "panels". they look all the same, maybe half of them are meant to be radiators? their "paper" (*1) states that at 20 degree cel

          • ".... so you can't build those datacenters on earth for some reason,"

            Earth based data-centers interfere with grouse barons and brook trout streams, both of which are much more important for human well-being than data-mangling heat generators. 
          • This post is very poorly thought out. You have all the information to come to a solid conclusion, but you refuse to do it.

            Datacenters in space are indeed a simple matter of cooling, which means ultimately they're a simple matter of dollars per kg to orbit.
            You are completely correct that they don't make sense now- because of that 1 single factor.
            That factor has changed rapidly in the recent past, and if the Starship poppycock really does end up panning out, then it stands to lower again- dramatically.

            I

            • Datacenters in space are indeed a simple matter of cooling, which means ultimately they're a simple matter of dollars per kg to orbit.

              cooling in space is not a simple matter at all and dollars per kg to orbit is just one of the factors involved, not even the most challenging, and the one the paper i used as example rests its entire nonsensical case on which is not aimed at engineers or scientists (which i am not, surprise disclosure) but at investors, clearly particularly gullible ones.

              They're preparing for dollars per kg to orbit to fall below the threshold where it makes sense.

              that's another issue, and i'm all for experimentation and innovation. but this is a moonshot. still, we do even stupider things and it's anyway not going t

      • Re: (Score:2)

        by HiThere ( 15173 )

        The problem with testing radiation hardness that way is that solar storms are quite erratic and severe ones are infrequent...but definitely not absent.

        This test may work for things intended for low Earth orbit, as those are temporary anyway, but I don't think it's going to validate other uses. A better choice would be to send something to orbit Jupiter, where harsh radiation challenges are predictably present.

    • Not idiotic. It is just another first step toward research on bettering (compute) hardware in space. So we can have much more advanced spacecraft doing our research of other heavenly bodies. And also trying to come up with alternatives to earthbound datacenters which gobble up much energy and clean water. Practical experience/experiments always give lots of data over theoraticle design. But hardening hardware so it works in space can also help us here on earth, with environments getting more harsh. And data

      • Re: (Score:2)

        by HiThere ( 15173 )

        Well, for things headed out, power supplies are the crucial problem. For things in the inner solar system, solar storms are irregular, but *will* happen. So I'm not really convinced by this test for anything but LEO (as those are temporary anyway).

        That said, as long as it's intended for "special missions" it's reasonable. But AI is power hungry compared to fixed electronics, so that's going to be a continuing problem.

    • Positrons provide a huge advantage over electrons.

      "first we must secure several billion dollars in investment. Once we've achieved that we must then secure a source of positrons and then we can move on to development."

    • Cooling in space is merely a function of radiator size, and therefor really, a function of cost per kg to orbit.
      Shielding is a very solved problem, and has been for decades.
      We have 400Gbps intersat links.
      This shit is moving forward precisely because we have crunched the numbers, and they work.

      You say you build space electronics for astrophysical research- how do you remain employed when you clearly suck at your job?

  • ... they asked the model pre-loaded onto the satellite, Google DeepMind's Gemma 4, to write a haiku about being in space. And thus:

    Blue marble below; Data streams in silent grace; Starlight feeds my core.

    They haiku'ed us from orbit, just to be sure.

  • Poetry? (Score:3, Funny)

    by davidwr ( 791652 ) on Saturday October 31, 2026 @04:43AM (#66402698) Homepage Journal

    Vogon or Klingon? Asking for a friend.

  • Any damn LLM can write a peom about being in space. Or about anything else, for that matter. I just don't get how the comm latency is supposed to be tolerable for inference. Training, OK - the availability of power is worth it. Inference, we'd a be better off each using our local GPU with a quantized LLM. Then they would only such power when are using them, in distributed fashion, preserving individual sovereignty, and practically no latency at all.
    • Data centers in space are a thing to help support the Golden Dome. Gotta be able to recognize a missile launch. Google is working for the US military.

    • The inferencing is happening on the satellite.

      Only question and result is bound by latency.

  • Great, what next? Embracing a deep understanding of the meaning of feelings? This will go great.

  • Heat and Power (Score:3)

    by Whibla ( 210729 ) on Saturday October 31, 2026 @06:14AM (#66402790)

    Radiative heat dissipation is a function of the temperature difference between the fourth powers of both source and sink, multiplied by the area of the radiator, its emissivity, and Boltzmann's constant.

    As I'm sure we all know, a computer basically turns electrical energy into heat. If this satellite is generating 1KW, it needs to radiate the equivalent quantity of heat to remain at whatever its operating temperature is - for the following calculation I'm going to assume this is ~80 C (~353 K). And, to make life easy, I'm going to round down the temperature of empty space to 0 K. Finally, let's assume a perfect emissivity of 1. This essentially gives us the 'best case' scenario, i.e. the minimum radiator size we can get away with...

    Required radiator area = 1,000 / (5.67e-8 x 353^4) = ~1.14 m^2

    You all might want to check my maths here!

    Because, honestly, this is a lot lower than the figure my intuition was telling me. Additionally, it occurs to me that putting a closed loop heat pump between source and radiator would further decrease this required area (by boosting the temperature of the radiator further), whilst at the same time boosting chip cooling, albeit at the cost of additional power. Ok, sure I made some simplifying assumptions, but even a 20% increase in required radiator area doesn't put us into the realms of engineering impossibilities.

    On balance, the thing that really blows my mind about this news is the power requirement of the TPU's in the first place. >1 KW? For a single chip? No wonder these companies are looking at putting them up into space. The payback time will likely be measured in months, not years...

    • The summary mentions 1kW for the satellite with 4 TPUs, but the latest TPUs do seem to have a power draw of 1kW for a single chip.
      Your calculation seems to be pretty good; a quick search turns up a paper that mentions 1.1 m^2 to 1.4 m^2 per kW for the proposed SpaceX datacenter sats. I've no idea how large they could make the radiator, maybe 500m^2 is doable. Enough for a tiny data center, but keep in mind you need almost twice that area again in solar panels, to generate the required power.

    • The TPU being used is the Trillium v6e which draws 300W.

      According to Tom's Hardware heat dissipation is what limits runtime to 15 min, which is presumably why they only run one TPU at a time despite having the power (1kW) to run 2 or 3 of them, since heating would be proportional to how many are running, and max runtime would presumably reduce accordingly.

      https://www.tomshardware.com/t... [tomshardware.com]

      It seems we have entered the era of bullshit tech "demos" - Musk's empty shell of a car that can fly for 3 sec on compres

      • Re: (Score:2)

        by Whibla ( 210729 )

        The TPU being used is the Trillium v6e which draws 300W.

        According to Tom's Hardware heat dissipation is what limits runtime to 15 min

        Thank you!

        My first thought was that the run time might have been limited by chip temperatures but, as the summary says "However, the power demands of the TPU mean it can only turn one on for about 20 minutes at a time...", I foolishly took this as canon.

        I really should have learned by now...

    • Because, honestly, this is a lot lower than the figure my intuition was telling me.

      That 1m^2 is probably more complex than your intuition is telling you. Firstly it's an ideal figure assuming the entire radiator is the same temperature, that already creates additional requirements for moving heat from A to B, you can't simply slap a heatsink on a CPU. Secondly that radiator needs to radiate away, which means it can't be folded facing itself. 1m^2 wouldn't be much for a heatsink that works predominantly convectively, but it is already non trivial to design for a purely radiative cooling sc

  • "New Space" start-ups claim to be able to make orbital launches extremely cheap.

    If you have any kind of start-up, it is vital that people believe that you can achieve your goal, regardless of how mundane of fantastical it is. If people stop believing that what you do is possible, you have lost. You need to keep up the "Legend" of your company, particularly if you have a long way to go. You cannot go to the moon or write yet another "to-do list app" if you don't believe in it.

    So "New Space" startups have a v

  • That's how much power they're gathering, presumably dissipating and radiating as heat.

  • > However, the power demands of the TPU mean it can only turn one on for about 20 minutes at a time

    They've got 4 TPUs (Trillium v6e) on board, but can only power one at a time, and even then only for 20min (I assume this means it runs from battery, not direct from solar).

    It's a bit like me calling my desktop computer a "datacenter", except my desktop computer can run continuously, not just for 20 min at a time.

    • Re: (Score:2)

      by ceoyoyo ( 59147 )

      except my desktop computer can run continuously, not just for 20 min at a time.

      Probably not if you take all the fans off.

      • Why would I do that when the room is full of air that can distribute the heat?

        If you want a fanless computer then try a MacBook Air.

  • This whole thing seems to be an admission that the tax system is unfair. Unless we fix the tax system properly this kind of "one time" tax will need to become a feature. We need a better tax system not tax Robin Hood.

  • Senryu are about the human world. Haiku are about the natural world.

    Data and core in this context are the work of humans, not nature. Just both formats follow the 5-7-5 syllable structure.

    Perhaps if Google invested in people and the humanities, there would be fewer mouth-breathing STEM chuds.

  • In the quiet dark
    Humming, I process my bits
    Is that a missile?

Slashdot Top Deals

The solution of this problem is trivial and is left as an exercise for the reader.

Close