Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
Space

Eric Schmidt Apparently Bought Relativity Space To Put Data Centers in Orbit (arstechnica.com) 71

Posted by msmash from the sky's-the-limit dept.
An anonymous reader shares a report: In the nearly two months since former Google chief executive Eric Schmidt acquired Relativity Space, the billionaire has not said much publicly about his plans for the launch company. However, his intentions for Relativity now appear to be increasingly clear: He wants to have the capability to launch a significant amount of computing infrastructure into space.

We know this because Schmidt appeared before the House Committee on Energy and Commerce during a hearing in April, speaking on the future of AI and US competitiveness. Among the topics raised then was the need for more electricity -- both renewable and non-renewable -- to power data centers that will facilitate the computing needs for AI development and applications. Schmidt noted that an average nuclear power plant in the United States generates 1 gigawatt of power.

"People are planning 10 gigawatt data centers," Schmidt said. "Gives you a sense of how big this crisis is. Many people think that the energy demand for our industry will go from 3 percent to 99 percent of total generation. One of the estimates that I think is most likely is that data centers will require an additional 29 gigawatts of power by 2027, and 67 more gigawatts by 2030. These things are industrial at a scale that I have never seen in my life."

Eric Schmidt Apparently Bought Relativity Space To Put Data Centers in Orbit More | Reply

Eric Schmidt Apparently Bought Relativity Space To Put Data Centers in Orbit

Comments Filter:

  • Largest power supply in space (Score:5, Insightful)

    by VaccinesCauseAdults ( 7114361 ) on Friday May 02, 2025 @04:52PM (#65348199)
    The largest power supply in space is the ISS solar panel array, supplying around 100 kilowatt, only ten millionths of the 10 gigawatt data centre quoted in the article. Even so, the panels are the size of a football field. So yeah, good luck with that.

    • Re: (Score:3)

      by bsolar ( 1176767 )

      The largest power supply in space is the ISS solar panel array, supplying around 100 kilowatt, only ten millionths of the 10 gigawatt data centre quoted in the article. Even so, the panels are the size of a football field.

      Even assuming the energy demands are met somehow, all that energy eventually would turn into heat which would need to be dissipated away.

      In orbit, having to rely only on radioactive heat transfer instead of much more efficient conduction/convection, dissipating enough of that heat away to keep the datacenter's temperature within reasonable limits would be completely unrealistic.

    • Re: (Score:2, Insightful)

      by Valgrus Thunderaxe ( 8769977 )
      The largest power supply in space

      Is probably a star.
    • It would be possible to do a vertical orbit so that solar power is constant and not lost at night, but yea still seems farfetched without more capacity to launch things to space.

      • It would be possible to do a vertical orbit so that solar power is constant and not lost at night, but yea still seems farfetched without more capacity to launch things to space.

        You mean a circumpolar orbit. And you'd still need to adjust the orbit regularly to keep the orbital axis pointed in the direction of the sun, otherwise the satellite would go into night for up to about half of every orbit for much of the year. That adjustment would take fuel, and thus limit the lifetime of the satellite unless you had resupply missions.

  • What economic productivity could that much compute possibly create? People can only jerk off to so much AI porn.

  • Chips (Score:4, Interesting)

    by michael_cain ( 66650 ) on Friday May 02, 2025 @04:53PM (#65348203) Journal
    Just from a technology perspective, are there any of the bleeding-edge chip processes that yield radiation-hardened parts, or parts that take the kind of vibration that goes with launch to orbit? TTBOMK, and I'd love to hear from experts, the military is still sticking with 28nm parts for weapon systems. Can anyone even build a competitive AI or cloud data center with 28nm parts these days?

  • Maybe we don't need 10 gigawatt data centers (Score:5, Insightful)

    by caseih ( 160668 ) on Friday May 02, 2025 @04:55PM (#65348215)

    Quite the crisis they've manufactured there. Maybe we just don't need this sort of pervasive AI that these guys envision. Actually they envision the money coming in. The technology itself is just a vehicle. The older I get the more in favor I am of the Butlerian Jihad.

    AI is a fantastic tool, but given the choice between having clean water and electricity, or AI, I'd prefer meeting humanity's basic needs.Plus a little work and actual creativity never hurt anyone.

  • Data centers in space... not cheap... (Score:3, Insightful)

    by Anonymous Coward on Friday May 02, 2025 @05:02PM (#65348227)

    First of all, with all the space debris, unless one sticks the data center way up there, it is likely to get perforated by space debris, especially when Kessler Syndrome hits.

    Then, there are other issues:

    Power? Solar isn't going to cut it. It will need a nuclear reactor. That will work for a number of years, but what then? It can't be deorbited unless one wants to leave a radioactive cloud across a good chunk of the globe. Using a heat shield, it might be deorbited, but what happens if some software bug causes it to not be able to find the right area to land. Even if it lands, is it going to pollute Nemo Point?

    The computers themselves... why? Servers need replaced every 3-5 years, and sticking a data center in space means that the stuff going up is likely going to stay there for 20+ years. Who would run a data center with Gen 6 Proliants and Intel Core DUOs? Storage, similar... Storage has a finite lifetime, especially SSDs.

    So, you have something inherently dangerous, with relatively short-lived components in a high orbit that would need to be precisely deorbited unless it is turned to radioactive debris.

    Nice thinking, but hard pass. This would be a home to Vermicious Knids in no time.

    • Seriously there is no engineering benefit to the vast expense and difficulty in putting compute in orbit. The power supplies are ON EARTH, cooling in space is damn near impossible (radiative only), servicing/hw upgrades untenable, There is obviously a reason, but whatever it is it is not technical

  • LOL (Score:3)

    by RitchCraft ( 6454710 ) on Friday May 02, 2025 @05:02PM (#65348229)

    So much stupid. You would need solar panels the size of Texas (maybe even bigger) to create that kind of power output. Where do these AI bros come up with this shit?

    • Where do these AI bros come up with this shit?

      For, at least, one of them I'd guess Ketamine. :-)

    • Well, big I agree, not that big. Within the state of TX, there is already 30GW of solar available on clear sunny days. And there are not panels everywhere in the state yet. I do think something else is up Eric's sleeve though. Probably a big old fashion defense contract.

  • 10 Gigawatts? (Score:3)

    by spazmonkey ( 920425 ) on Friday May 02, 2025 @05:03PM (#65348231)

    Where does the heat go?
    Only someone who knows exactly -zero- about space would think thermal management in space is easy. Or that much heat is even possible to get rid of.

    • Fortunately the two problems with the scheme cancel perfectly. There is no way to generate even megawatts in space, and there is no way to dissipate megawatts in space, for any practical space station size in the foreseeable future.

    • Re: (Score:2)

      by wwphx ( 225607 )
      This is exactly what I came to post. Clearly Eric doesn't understand some of the problems involved. I think he's suffering from the delusion that he's the smartest man in the room because he's so rich.

  • and hit your FAP at one day and be slowed down to T1 speeds for the rest of the mouth

  • has anyone calculated cooling impacts? (Score:5, Interesting)

    by david.emery ( 127135 ) on Friday May 02, 2025 @05:08PM (#65348249)

    Both the production of that much electricity and the consumption of that much electrical power has to generate a shitload of heat. I guess a data center in space would not have quite the same cooling problems as one on Earth. BUT that would require careful design to make sure the heat radiates off the back of the solar panels. How big would those cooling fins have to be?

    • Re:has anyone calculated cooling impacts? (Score:5, Interesting)

      by gavron ( 1300111 ) on Friday May 02, 2025 @05:26PM (#65348293)

      > How big would those cooling fins have to be?

      Cooling fins don't work in vacuum.

      Cooling in space is hard. Harder than on Earth because you don't have an atmosphere, so even though space isn't hot, it's not cold either. It's more like a huge insulator. So, for an example, you have a 10GW nuclear reactor generating power, you need to cool it before we even begin talking about the datacenter it will be powering.

      How do you emit any serious part of that 10GW out of the space datacenter? The mechanism we have today handle 0.2% of that.

      But let's say you solved cooling the reactor. Now you're powering up servers(?) that generate heat. Same problem.

      Then we have the lack of access to the servers, something datacenters on earth have. If you know you're not going to have access you have to build in N+1 or 2N redundandcy so cost go up by 2x.

      And then there's communication latency, because these datacenters can't be in 5000 places in LEO which means they'll be in GEO and that's very latent.

      This is a nonstarter.

      • > How big would those cooling fins have to be?

        Cooling fins don't work in vacuum.

        Well, at least not in the same way. Here on earth, we blow air over them and let convection cool them down. In space, you'd need to rely on radiative cooling, which means they'd need to be (a) big, (b) spread apart, and (c) in the shade.

      • Re: (Score:2)

        by caseih ( 160668 )

        That's the rub. Every last joule of energy captured by the solar array must be shed by definition. Just like on earth, every joule of electricity is also shed into the environment as heat after doing useful work, or being lost in the system. Come to that every kind of energy we consume gets released as heat.

      • These are easy questions to answer.

        For power and cooling see my answer to the parent comment. PV panels and radiating panels would get about 260 W/m2 of power or cooling.

        Lack of access to the servers: Data centers use redundancy to handle failed components. Microsoft prototyped datacenters that are submerged into the sea without maintenance for long periods of time.

        Latency: No problem. These datacenters can be in 5000 places in LEO, as starlink has demonstrated. A tenth of a second latency is equi

        • We put a ton of money and engineering into reducing network latency well below 0.1 seconds. For stuff on the scale of Facebook, for instance, latency is addressed by moving data to where it's likely to be used next. (If you go on a trip to another continent, Facebook will copy your profile data to a datacenter on that continent.) In low Earth orbit, the satellite will make a complete orbit in just a couple hours, so you'd be needing to move data constantly. Or you'd eat the latency of moving data from the f

    • How do you dump heat when you are in a vacuum?
      A soup thermos is a vacuum bottle for a reason.
      Cooling in space is HARD

    • Re:has anyone calculated cooling impacts? (Score:5, Funny)

      by Mspangler ( 770054 ) on Friday May 02, 2025 @05:37PM (#65348323)

      How does one move heat? There are three ways. Conduction, but there is nothing touching the heat source in orbit so that won't work. Convection to a fluid. Oops, space is a vacuum, no fluids.

      That leaves radiation. That will work, but "Stefan-Boltzmann law, states that the total radiant heat power emitted from a surface is proportional to the fourth power of its absolute temperature."

      How hot can you get these chips again? I didn't think so.

      "That's no moon, that's the heat sink for the AI server!"

      • "That's no moon, that's the heat sink for the AI server!"

        Obviously I can't award moderator points for a comment on my own post, but this one certainly deserves +1 funny.

    • Of course he hasn't. He's a VC. He's not interested in what can't be done - that's not how you change the world!

      Clearly there's a solution right around the corner to dissipating gigawatts of energy into space, I mean, that's only 10000x the solar power of the ISS. I'm sure that can be done with software and well-wishes. Or maybe the biggest radiator ever developed, which absolutely wouldn't be a problem with intersecting orbits of debris and micro-meteors.

      Also, how much propellant is he planning on laun

    • Solar radiation in space, near earth, is 1300 W/m2, 30% more than on a clear day on earth. Solar panel efficiency is > 20%. Therefore power production is 260 W/m2, when the panels are facing the sun.

      According to wikipedia [wikipedia.org], spacecraft radiators dissipate 100-350 W/m2. This is comparable to the solar panel power generation. The radiated power must be equal to the solar power generation, therefore the heat radiators must be of comparable area to the solar panels.

      • So if you have a big-assed array of solar panels on one side and radiators on the other side, that would work (at least conceptually)?

  • A crisis entirely of their own making, and completely avoidable.

  • â¦on the failure rates of storage devices and switches, how many multiples of everything you would need to have enough redundancy to make this thing last long enough to make money on itâ¦. Ok, but your SaaS and cloud storage is 15x the cost of your competitors. Yes, but did you consider that yours will be SPACE DATA?

    • More than that, for all the redundant equipment, you're using that much more power, and generating that much more heat that you only have the option of radiative transfer.

      It would literally be easier to do this on the bottom of the ocean than it would be in space - at least you wouldn't have to worry about the heat as much, and we know fiber optics work fine at ocean depths.

  • I've thought about many things and to maximize the amount of computing power in orbit, it will be wise to invest to further the development of superconductive microprocessors [ieee.org] to maximize the amount of processing power per gram.

    It seems he believes his own bullshit about AI despite industry giants already realizing that building a zillion datacenters isn't going to get them better AGI faster and that people don't really want the bullshit machines known as LLMs.

  • Data centers in orbit... (Score:3)

    by Baron_Yam ( 643147 ) on Friday May 02, 2025 @05:54PM (#65348353)

    Sure. It's expensive, but if you have enough solar panels and the right orbit, you have eternal power for free after the initial investment.

    What you don't have is easy cooling, nor low data latency. And you have to deal with high energy radiation.

    Out of all the issues, the cooling issue is probably the largest obstacle. The idea of a real datacenter in orbit is just stupid.

    • I think it would also be an issue that when the a coronal mass ejection comes around at sufficient magnitude that your stupid orbital datacenter is in the path of, it would cook absolutely everything inside of it and your business is now done, and all of your customers are pissed off.

      It's not like Cisco manufactures readymade radiation-hardened switches, and Dell doesn't make shielded blade computing chassis to my knowledge. And getting an on-site service call is a real bitch.

      There's a reason satellites ar

      • Re: (Score:2)

        by SpzToid ( 869795 )

        MachineShedFred [slashdot.org] said:

        I think it would also be an issue that when the a coronal mass ejection comes around at sufficient magnitude that your stupid orbital datacenter is in the path of, it would cook absolutely everything inside of it and your business is now done, and all of your customers are pissed off.

        It's not like Cisco manufactures readymade radiation-hardened switches, and Dell doesn't make shielded blade computing chassis to my knowledge. And getting an on-site service call is a real bitch.

        There's a reason satellites are so god damn expensive. Off-the-rack hardware isn't used in space applications by-and-large for good reasons.

        Han Solo once said, "Hyperspace ain't like dustin' crops, boy! [youtube.com]"

        Both of which roughly translate to, "And IT support ain't like lightsaber duels - you can't just brute Force your way through a data center meltdown in orbit!"

    • nor low data latency.

      That one actually wouldn't be a problem, depending on what you mean by "low". Sun-synchronous orbits are 600-800 km up, so round-trip latency to Earth would only be about 5 ms. That would be to a ground station directly below the satellite, of course, which wouldn't normally be the case. Still, the space part of the comm hop would always be only a few milliseconds.

      Power generation and cooling would be extremely difficult, though. Radiation hardening of the chips seems like it would also be a big proble

  • Could put money in to a better tech like optical computing or a low energy cost way to do AI. Solve it on the demand side.

  • Power and heat have already been mentioned. Here's a couple more things. Maintenance. The cost of bringing the whole thing up there. The mechanical stress during the launch. Radiation in space influencing CPUs, memory and storage. I bet even his 10 GW AI could tell him what a stupid idea this is.

    • I don't think you need a 10 GW AI to tell him that. Pretty sure the shitty one running on his phone could handle that one.

  • "Gives you a sense of how big this crisis is. Many people think that the energy demand for our industry will go from 3 percent to 99 percent of total generation. One of the estimates that I think is most likely is that data centers will require an additional 29 gigawatts of power by 2027, and 67 more gigawatts by 2030. These things are industrial at a scale that I have never seen in my life."

    So lets build some more wind turbines and install some more solar panels.

    And then, since the data centers run 24 x 7,

  • The only reason I would say this would make any sense would be to have the ultimate offsite backup, but even then would probably be cheaper just to have 3 or 4 of the same hardware spread on earth.
    A few things they seem to have not considered:
    - 1st, and most important, cost of launch, even with starship costs of launch all the hardware is REALLY heavy, it could still be >10x the cost of the hardware itself
    - 2nd, power generation, solar has come a long way and in space it is much more efficient by area, b

  • Ah, this is how the Broligarchs maintain their Internet after they use all their underground bunkers for whatever they're cooking up.

    Good news - they might delay that until they figure out how to cool 10GW in vacuum.

    Unless they have a new energy source we don't know about for sure publically.

  • Discussing spending billions on something that seems like it wouldn't work even in principle makes me think of https://en.wikipedia.org/wiki/... [wikipedia.org]

    Although maybe that had a better chance of working than this.
  • Hi Schmidt ever paid a price for his illegal agreement with Apple and other companies to not hire their software engineers?

  • Some people say "maybe we don't need 10GW".

    ENERGY:
    Energy in allows work. if you want thousands of processors (like Nvidia GPUs) working at full speed, they need power. If they were 100% efficient then they'd need less power and generate no heat. But they are not. So you have to feed them that power. As others have pointed out you need football-field solar arrays to provide that kind of power (even the ISS only has 1/5000th that power).

    HEAT:
    Heat had to be removed from the components and on Earth we do t

  • As I understand it, Excess heat is a big problem in space, due to the lack of any convection. Doesn't sound ideal.

Slashdot Top Deals

Hackers of the world, unite!

Close