• A Great friend to the HardForum with a great kid that he is trying to get a scholorship to continue his schooling. Please give hime a vote! Only 24 hours left! Thanks.
    If you have an VOTE FOR KEENAN!

Data center in Space

AceGoober said:
Then there is the major relevant question:

Which country owns the debris and has the money to pay for the cleanup?
Click to expand...
The one thing that is certain about this is that it will be the taxpayers that has to fund the clean up.
 
erek said:
OFF-PREM

“SpaceX claims it will put a Vera Rubin NVL72 rack-scale system into orbit next year​

And fly ‘significant scale’ in 2028 – a claim that isn’t entirely bonkers, but needs a solid grain of salt


SpaceX will send at least one datacenter rack into space next year, according to Elon Musk, with more to follow in 2028.”

https://www.theregister.com/off-pre...ack-scale-system-into-orbit-next-year/5292067
Click to expand...
Like he was going to send huam to Mars righ?
Elon likes to talk....
 
How many thousands of people in their infinite wisdom have told Elon "you can't do that, that will not work, unachievable, it's gonna fail, it's never been done before, impossible, you don't understand"?!!!!!!!!!!!!!!!!1
 
mullet said:
How many thousands of people in their infinite wisdom have told Elon "you can't do that, that will not work, unachievable, it's gonna fail, it's never been done before, impossible, you don't understand"?!!!!!!!!!!!!!!!!1
Click to expand...

And again, Elon's not the only one doing this so you don't even have to care or pay attention to what Elon says for this to happen.
 
kram182 said:
And again, Elon's not the only one doing this so you don't even have to care or pay attention to what Elon says for this to happen.
Click to expand...
He is a genius, I do listen to him and everyone else will have to use his rockets anyway.
 
Tsumi said:
A genius inventor, or a genius at manipulating people and the system to his advantage?
Click to expand...

Yep. He's a smart investor but by no means a brilliant engineer or scientist. He's said many things that turned out to be true, yes, but also said many things that turned out to be absolute bullshit. He's made many predictions that have yet to come true even though he said we'd have them years ago. If you're a fan of his you remember all his guesses that turned out true and forget the timeline between when he said they'd happen and when they actually happened. I could have said, when the first iPhone came out, that we'd have modern (today) smartphones in two years time. I would have been absolutely wrong but if people liked me they'd just remember I predicted it, not how poorly I estimated the timeframe.

Will we have data centres in space? Probably. Will we have self-driving cars? Sure. Will we have them in the absurd timelines that Elon predicts? No. Will he make bags of money off both? Yep. His predictions are largely to attract investors and since most investors are not technically advanced folks they fall for the bravado. Once they're invested they don't want to admit that maybe they jumped the gun so they lock in and stick with it waiting for that ultimate payoff. If it never shows up they quietly slink away and go find something else to invest in. Not all the investments in data centres will pay off but some of them will. The investors who get lucky will be hailed as geniuses and the ones who lose their shirt a bunch of idiots all while neither being true. They both rolled the dice and one got lucky and one didn't.
 
His clearly a genious entrepeneur (with good and impressive skill in science/engerining/physic for that position, just not near top of the field as an mono expert) and for this, working really hard is a major skill, inspiring and attracking talent has well. There is an certain amount of company you elevated or started where it stop to be a debate, when having started OpenAI is just one bullet point on your resume...

Ididar said:
Will we have self-driving cars? Sure.
Click to expand...
That already here in 2026, China has Level 4 autonomous going in Bejing, Shanghai, Shenzhen and many other major cities. If taken a fully driveless taxi in Shenzhen, from one spot to the other with zero human intervention does not count has having self-driving car, we are just pushing what a self-driving car mean more and more.

Not sure a single NVL in 2027 and expension for 2028 datacenter in space is that absurd of a timeline, the robots timeline feel more aggressive (it involve more third party affair that can slow it down and yet to be fully solve challenge around the hands).
 
LukeTbk said:
His clearly a genious entrepeneur (with good and impressive skill in science/engerining/physic for that position, just not near top of the field as an mono expert) and for this, working really hard is a major skill, inspiring and attracking talent has well. There is an certain amount of company you elevated or started where it stop to be a debate, when having started OpenAI is just one bullet point on your resume...


That already here in 2026, China has Level 4 autonomous going in Bejing, Shanghai, Shenzhen and many other major cities. If taken a fully driveless taxi in Shenzhen, from one spot to the other with zero human intervention does not count has having self-driving car, we are just pushing what a self-driving car mean more and more.

Not sure a single NVL in 2027 and expension for 2028 datacenter in space is that absurd of a timeline, the robots timeline feel more aggressive (it involve more third party affair that can slow it down and yet to be fully solve challenge around the hands).
Click to expand...

Again, the point is not that he hasn't made predictions but that the timeline on all his predictions is horribly wrong to the point that his ability to predict the future is horrible. What he's great at, as I said, is attracting investors with those wild-assed promises and getting their money locked in before they realize he's never going to deliver on time. Deliver? Yes. Deliver return on their investment when he said they would? Definitely not.
 
Ididar said:
that the timeline on all his predictions is horribly wrong to the point that his ability to predict the future is horrible.
Click to expand...
That could be a bit over blowing it, do you start openAI if your ability to predict the future horrible, would spaceX made any sense, considering the business he started, the ability to predict the futur did matter a good deal.

Ididar said:
Deliver return on their investment when he said they would? Definitely not.
Click to expand...
I feel paypal delivered faster than expected with the fast ebay buyout and after that, it was a lot of who knows, get in for the long ride people.

Ididar said:
What he's great at, as I said, is attracting investors
Click to expand...
I would say, attracting high end talent and working extremelly hard are 2 big one (and yes despite time horizon issue directionnaly correct about where the futur goes), he does not need investor much with is willingless to risk all is own fortune many times, say OpenAI extreme success story, very little to do with Musk ability to attract investor (he left before they rose a lot of investor capitals, it was all vision of the future and capacity to attract best in the world to a project).

Sleeping on the floor of factories, is own way to have the best in the world working hard at something.
 
LukeTbk said:
That could be a bit over blowing it, do you start openAI if your ability to predict the future horrible, would spaceX made any sense, considering the business he started, the ability to predict the futur did matter a good deal.
Click to expand...

If I tell you I'll deliver your goods in 6 months and it takes 3 years, are you going to call me a success? Are you going to hail me as a visionary or a genius? Likely you're going to be livid and look for your money back.
 
Ididar said:
If I tell you I'll deliver your goods in 6 months and it takes 3 years, are you going to call me a success? Are you going to hail me as a visionary or a genius? Likely you're going to be livid and look for your money back.
Click to expand...

Except his late predictions, even when late, still make people money. That's why people are willing to buy in time and time again no matter how late he is.
 
Google is joining the game now

https://x.com/ns123abc/status/2103143161727992118

1790275959399.png


 
You know back in the 1990's I believe this very same machine they have built so much trust in (AI) use to be called the answering machine hooked right up to your tele - phone and now they are all standing around waiting for this thing to give them answers to everything it has been listen too.
 

“Google Tests AI Data Center In Space​

Geoffrey.landis 8 hours ago		57
Longtime Slashdot reader Geoffrey.landis shares a report from NPR: Google just put a refrigerator-sized satellite into space, part of a research project the company hopes can pave the way for orbiting AI data centers powered by the sun. The prototype satellite has four specialty chips in it called Tensor Processing Units, or TPUs, that Google designed for machine learning and has already deployed in data centers on Earth. The chips will run a version of the company's Gemma AI model for 15 minutes at a time due to heat management constraints, using the open weight model to answer simple queries.

In a blog post, Google said the mission will gauge how the TPUs handle "the physical stress of spaceflight and the radiation and thermal extremes of space." The satellite launched October 1 is a prototype that Travis Beals, senior director and lead of Project Suncatcher, called "a very minimal test" to make sure the chips can run in space. A SpaceX rocket carried it into orbit, and Planet will get it up and running. Beal said Google will then fire up and test the TPUs. The goal is for the mission to be operational for a year.”
 
this shit is never happening.. if anyone at all understands physics... they know this is not happening. for so many reasons. the sheer amount of rocket fuel that it would take... we will have to abandon propulsion the way we've been doing it.. to get that much equipment up there..(talking large scale datacenters rivaling the ones we have now on earth)

OH and EDIT: if anyone knows what a bit-flip is due to radiation.. just imagine how bad that would be in space. even with protection.. the sheer amount of testing.. and personnel on site to deal with issues... and no robots are not ready for this.
 
Last edited:
revenant said:
this shit is never happening.. if anyone at all understands physics... they know this is not happening. for so many reasons. the sheer amount of rocket fuel that it would take... we will have to abandon propulsion the way we've been doing it.. to get that much equipment up there..(talking large scale datacenters rivaling the ones we have now on earth)

OH and EDIT: if anyone knows what a bit-flip is due to radiation.. just imagine how bad that would be in space. even with protection.. the sheer amount of testing.. and personnel on site to deal with issues... and no robots are not ready for this.
Click to expand...
That "protection" is going to add weigth too...
 
revenant said:
this shit is never happening.. if anyone at all understands physics... they know this is not happening. for so many reasons. the sheer amount of rocket fuel that it would take... we will have to abandon propulsion the way we've been doing it.. to get that much equipment up there..(talking large scale datacenters rivaling the ones we have now on earth)

OH and EDIT: if anyone knows what a bit-flip is due to radiation.. just imagine how bad that would be in space. even with protection.. the sheer amount of testing.. and personnel on site to deal with issues... and no robots are not ready for this.
Click to expand...

They'll put a very small "data centre" up there and declare victory to keep the investors happy and to appease the fans. They'll quietly deorbit it after a while and never launch any more of the things but people will have stopped paying attention by that point.
 
I think that under estimate how much concentrated compute is getting, how cheap sending things in space is getting and how expensive building datacenter on earth could get.

The people that know the most about sending stuff in space and building datacenter like Amazon-Google-SpaceX seem to all think it will work (about everyone with serious actual experience at both), the people saying it wont does not tend to have a lot of real life experience of sending large compute in space like those peoples have done for years.

And the math seem to work, it could use only 100,000 KG of rocket fuel per rack.

The specs are 75kw of Rubin a ton and sending a kilo in space at the current pace a kilo to low orbit should cost under $200 (raw cost) by 2028, so we are talking about ~2.7 billion per gigawatt of space compute in raw flight cost....

The earth version for 6 years of usage, buying lands, right to the grids, water rights, local incentive to get accepted, can easily cost 6 billion per gigawatt of compute plus 2-3 billion in electricity, in space you do not pay the giant municipal taxes.

Napkin math, seem to make space cheaper and the more dense-expensive the compute unit get, the cheaper space become.

One other way to look at it, there is 4,500-5,000 tons of Starlink satellites currently active, that would be already a good sized datacenter and that was made by much costlier Falcon 9 flights (they could have send 2+ gigawatt at the upcoming starship rate with that payload cost).
 
Ididar said:
They'll put a very small "data centre" up there and declare victory to keep the investors happy and to appease the fans. They'll quietly deorbit it after a while and never launch any more of the things but people will have stopped paying attention by that point.
Click to expand...
Probably. Google is sending up test equipment now or did recently. And they’ll not be anything like our traditional dcs are now. The cooling alone in zero atmosphere will have to be all different.. maybe something peltier based? Idk. So many hurdles to get through.

LukeTbk - yeah I was thinking about the footprint of starlink in space now.. it’s a valid point.. unless they have plans to really shrink down the dc equipment … lighter, completely hands off, modular… in the end the weight might be the least of the issues here.. other hurdles like thermal management and radiation are pretty steep.
 
Last edited:
LukeTbk said:
I think that under estimate how much concentrated compute is getting, how cheap sending things in space is getting and how expensive building datacenter on earth could get.

The people that know the most about sending stuff in space and building datacenter like Amazon-Google-SpaceX seem to all think it will work (about everyone with serious actual experience at both), the people saying it wont does not tend to have a lot of real life experience of sending large compute in space like those peoples have done for years.

And the math seem to work, it could use only 100,000 KG of rocket fuel per rack.

The specs are 75kw of Rubin a ton and sending a kilo in space at the current pace a kilo to low orbit should cost under $200 (raw cost) by 2028, so we are talking about ~2.7 billion per gigawatt of space compute in raw flight cost....

The earth version for 6 years of usage, buying lands, right to the grids, water rights, local incentive to get accepted, can easily cost 6 billion per gigawatt of compute plus 2-3 billion in electricity, in space you do not pay the giant municipal taxes.

Napkin math, seem to make space cheaper and the more dense-expensive the compute unit get, the cheaper space become.

One other way to look at it, there is 4,500-5,000 tons of Starlink satellites currently active, that would be already a good sized datacenter and that was made by much costlier Falcon 9 flights (they could have send 2+ gigawatt at the upcoming starship rate with that payload cost).
Click to expand...

How does that stack up beyond 6 years though? The building doesn't change if the hardware changes. The launch costs have to happen all the time as satellites de-orbit, hardware degrades, etc. Yes, it is getting cheaper to send it up but the costs keep coming. Once you've gotten your data centre and power figured out on the ground that is locked in for decades, not 6 years. The short term makes ground look bad, of course, because the costs of buildings are rarely, if ever, amortized across that short a time frame.
 
Ididar said:
Once you've gotten your data centre and power figured out on the ground that is locked in for decades, not 6 years. The short term makes ground look bad, of course, because the costs of buildings are rarely, if ever, amortized across that short a time frame.
Click to expand...
But electricity and giant taxes on earth continue, everytime you update, while the price of sending stuff in space continue to go down and the compute concentration would have continued to go up, I can see the race between the 2 to be competitive enough to get better earth deals at least.

It is probable that both will co-exist and try to win the battle and keep the other honest.

If there is some YIMBY movement turn around, adoption of a reasonable rest of the world VATS type taxation (the current american tax system make no sense for this and people get angry when you must have them tax rebate are used) and faster access to grids power that could it of course, but that does not seem to be the direction it is going.

It just need to be competitve enough to give better earth deal to be worth it, would be my guess
 
Ididar said:
How does that stack up beyond 6 years though? The building doesn't change if the hardware changes. The launch costs have to happen all the time as satellites de-orbit, hardware degrades, etc. Yes, it is getting cheaper to send it up but the costs keep coming. Once you've gotten your data centre and power figured out on the ground that is locked in for decades, not 6 years. The short term makes ground look bad, of course, because the costs of buildings are rarely, if ever, amortized across that short a time frame.
Click to expand...

It isn't just a cost comparison.

There is a bottleneck in building out new power plants on earth. If they want more than that (they do), they need to build datacenters in space where solar can give them 24/7 coverage.
 
revenant said:
Probably. Google is sending up test equipment now or did recently. And they’ll not be anything like our traditional dcs are now. The cooling alone in zero atmosphere will have to be all different.. maybe something peltier based? Idk. So many hurdles to get through.
Click to expand...
The cooling is well know and used right now for cooling compute in space, it is just a bigger version.

The 72 gpu rack satellite design as 160 square metere of liquid radiator, this involve zero new tech from them (well hot water cooling was a key important variable, but that Nvidia). it is about a volleyball court size affair.

320meter when you double side of simple deep space passive radiation can cool 150kw around 35-38C, Rubin is made to work with 45C water no problem...

That sound big, but the starlink v3 are already quite big and doing 20-30kw of constant power with giant solar panels already, apparently those satelitte will be simpler in many ways.

Has for radiation, AI inference is less of challenge than Starlink compute and signal would be my guess (which already work right now), there is a reason it still work at 2-3-4 low precisions bits that can run on cheap gpus instead of fancy cpu, it is not an exact value critical form of compute, probabilistic affair vs determinisc exact algo.

you have hundreds of billions of weight in attention, some mistake on some of them and end up with 97.2% instead of 97.15% likelyhood score is not a big deal.
 
what about some kind of vapor chamber cooler filled with ammonia in a thermal loop for cooling discrete chips and such.. or just using ammonia as the coolant in the "bigger version" ..

re: radiation - Yeah maybe some robust ECC type error correction at the silicon / memory layer ... or silicon insulator.. or both.. Seeing how these will cope with geomagnetic storms/cme will be very interesting.. throttle everything down maybe ahead of those events.
 
revenant said:
what about some kind of vapor chamber cooler filled with ammonia in a thermal loop for cooling discrete chips and such.. or just using ammonia as the coolant in the "bigger version" ..

re: radiation - Yeah maybe some robust ECC type error correction at the silicon / memory layer ... or silicon insulator.. or both.. Seeing how these will cope with geomagnetic storms/cme will be very interesting.. throttle everything down maybe ahead of those events.
Click to expand...
Probably some sort of dead man mechanism which deorbits the satellite if it can no longer function as intended. Maybe a hardened transceiver is activated and can be used to control thrusters in that sort of situation.
 
I worked in multiple datacenters for 14 years in the bay area, like it was my job to build out everything from unboxing to racking and cabling everything up... I saw the dc go through all kinds of struggles with thermals for dense rows of hadoop compute etc... finally they had to employ cold aisle containment in the fronts of the cabinets making the whole dc a hot aisle basically.. that was when I said "no more" to that role. at any rate contemplating all that in space really had me thinking it would be too hard in orbit.. even a highly modified variant of it.. but these will be more like satellites with highly evolved blade type servers plugged into them, or just modular.. highly specialized for the environment.. including weight. So yeah I guess I can see this working but that is one harsh environment... Google did just launch a test unit... I imagine there will be a fair amount of testing...

I'd rather create a WALLE situation vs having these things planted all over rural America putting stress on our power grids and more.

So yeah going to back pedal on my previous post. I'll be interesting to track the progress of this.
 
revenant said:
what about some kind of vapor chamber cooler filled with ammonia in a thermal loop for cooling discrete chips and such.. or just using ammonia as the coolant in the "bigger version" ..
Click to expand...
The cooling problem people are describing is not really related to the cooling technology used. It's related to the fact there is no conductive or convective cooling in space. Only radiative. It means you need large surface area of radiative panels (so extra weight to launch).

If you find images of the ISS for example, the panels closer to the middle of the ISS are all heat rejection panels, radiative.
 
serpretetsky said:
The cooling problem people are describing is not really related to the cooling technology used. It's related to the fact there is no conductive or convective cooling in space. Only radiative. It means you need large surface area of radiative panels (so extra weight to launch).

If you find images of the ISS for example, the panels closer to the middle of the ISS are all heat rejection panels, radiative.
Click to expand...
ah so the best option is a big ol heatsink.. crazy. I was thinking about mercury in a vapor chamber also.. then a rad exposed to the space temps to shed the energy.. but it's not great at grabbing the energy apparently.. but it has a high boiling temp.. probably overly complicated and underperforming.. idk so heatsink it is.

So maybe like Dell m1000 blade chassis with stacks of compute nodes installed.. with a very different cooling solution, solar panels, thrusters, and good shielding... just spit-balling how I envision it very roughly.
 
revenant said:
ah so the best option is a big ol heatsink.. crazy. I was thinking about mercury in a vapor chamber also.. then a rad exposed to the space temps to shed the energy.. but it's not great at grabbing the energy apparently.. but it has a high boiling temp.. probably overly complicated and underperforming.. idk so heatsink it is.

So maybe like Dell m1000 blade chassis with stacks of compute nodes installed.. with a very different cooling solution, solar panels, thrusters, and good shielding... just spit-balling how I envision it very roughly.
Click to expand...
Im not saying your original idea is bad, i honestly dont know what they use for the coolant loop (water? Ammonia? Freon?). Just wanted the highlight the main issue.
 
serpretetsky said:
The cooling problem people are describing is not really related to the cooling technology used. It's related to the fact there is no conductive or convective cooling in space. Only radiative. It means you need large surface area of radiative panels (so extra weight to launch).

If you find images of the ISS for example, the panels closer to the middle of the ISS are all heat rejection panels, radiative.
Click to expand...
Yeah, It may have been discussed in this thread before, but the vacuum of space is cold, but temperature can't actually conduct in a vacuum, so a heatsink would work better in the Sahara desert than in orbit, as the Sahara actually has a conductive medium (air) and despite it being deadly to us, the daytime air temperature is actually MUCH cooler than a fully-loaded CPU.
 
KazeoHin said:
Yeah, It may have been discussed in this thread before, but the vacuum of space is cold, but temperature can't actually conduct in a vacuum, so a heatsink would work better in the Sahara desert than in orbit, as the Sahara actually has a conductive medium (air) and despite it being deadly to us, the daytime air temperature is actually MUCH cooler than a fully-loaded CPU.
Click to expand...
Keep in mind, a heatsink and radiator are different, serve different purpose, and there are different kinds of radiators which operate better under different conditions.
 
KazeoHin said:
Yeah, It may have been discussed in this thread before, but the vacuum of space is cold, but temperature can't actually conduct in a vacuum, so a heatsink would work better in the Sahara desert than in orbit, as the Sahara actually has a conductive medium (air) and despite it being deadly to us, the daytime air temperature is actually MUCH cooler than a fully-loaded CPU.
Click to expand...

Space itself is the heatsink, the radiators are akin to thermal paste in this regard/environment/function (through radiation), just getting heat 'into' the heatsink essentially
 
KazeoHin said:
Yeah, It may have been discussed in this thread before, but the vacuum of space is cold, but temperature can't actually conduct in a vacuum, so a heatsink would work better in the Sahara desert than in orbit, as the Sahara actually has a conductive medium (air) and despite it being deadly to us, the daytime air temperature is actually MUCH cooler than a fully-loaded CPU.
Click to expand...
Why does capsule in space with astronomers inside need to run heater constantly to keep temperature up? If there is no conductive medium in space, where is heat going?
 
Bezos is building 'long term' infra out in MEO for networking/communication, not compute (which is being treated as disposable when it falls out of orbit)
 
You must log in or register to reply here.
Back
Top