• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Data center in Space

erek said:

“Nvidia is sending GPUs to the moon​

“What we at Lunar Outpost are currently working on is how do we go from exploration to permanence, how do we actually build that outpost on the moon?” Cyrus said. “We do think combining our deterministic models plus the physical AI layer — that’s what allows us to make a human presence in space sustainable, actually having that robotic workforce.”

In the next few years, Lunar Outpost has several smaller autonomous rovers scheduled for launch to the moon, and one larger one, Pegasus, that is intended to carry astronauts. Pegasus is waiting on a rocket built by Jeff Bezos’ space company, Blue Origin, to carry it to the moon, but that launch vehicle suffered an anomaly this summer and it’s not clear when it will return to flight.

“It does seem like they’re making pretty darn good progress on the pad,” Cyrus said. “I can’t really answer Blue Origin’s timelines or NASA’s timelines on the pad reconstruction, but everything we’ve been told, we’re on the same timeline [for 2028].”

That leaves futuristic visions of space data centers and fleets of robots at work on the moon both waiting on the same thing: a bigger rocket.“

https://techcrunch.com/2026/07/23/nvidia-is-sending-gpus-to-the-moon/
Click to expand...
At least on the moon, they can probably leverage geothermal cooling in a closed loop setup. I won't knock them for dreaming, and despite the distance involved, I think that a moon data center makes more sense than one that orbits the earth to eventually become a giant man made asteroid.
 
SticKx911 said:
At least on the moon, they can probably leverage geothermal cooling in a closed loop setup. I won't knock them for dreaming, and despite the distance involved, I think that a moon data center makes more sense than one that orbits the earth to eventually become a giant man made asteroid.
Click to expand...

It's only talking about rovers/crafts powered by Nvidia Jetson going to the Moon - no Moon based DC planned yet I'm aware of (outside future plans for eventual US Moon base which will have some form of compute to start) - all space based DCs still focused on LEO first, with MEO next

Edit - there is a DC but more just a data-locker/backup on the Moon: https://interestingengineering.com/space/lonestar-first-solar-powered-moon-data-center
 
SticKx911 said:
At least on the moon, they can probably leverage geothermal cooling in a closed loop setup. I won't knock them for dreaming, and despite the distance involved, I think that a moon data center makes more sense than one that orbits the earth to eventually become a giant man made asteroid.
Click to expand...
The moon is ~ 239k miles. c is a vacuum is 186,282.4 mi/s. How much latency is acceptable here?
 
kram182 said:
It's only talking about rovers/crafts powered by Nvidia Jetson going to the Moon - no Moon based DC planned yet I'm aware of (outside future plans for eventual US Moon base which will have some form of compute to start) - all space based DCs still focused on LEO first, with MEO next

Edit - there is a DC but more just a data-locker/backup on the Moon: https://interestingengineering.com/space/lonestar-first-solar-powered-moon-data-center
Click to expand...

I think the real value chain is LEO edge compute for FSD. An all electric fleet of tractor trailers with RT inner loop on board (controls are sub 5ms). For logistics planning, you have space edge compute that can see the full picture of traffic, routes, recharge stations, across both trucking and cars. Near real-time planning and path navigation forms a space based outer loop can run at 5 Hz (this is effectively the global optimizer outer loop).
 
1.1.2.3.5... said:
The moon is ~ 239k miles. c is a vacuum is 186,282.4 mi/s. How much latency is acceptable here?
Click to expand...
Offsite backup doesn't have to be fast. It's not a gaming server. Even for remote processing, cooling equipment in orbit in near vacuum at best has to be harder than a geothermal loop through the lunar surface that's probably super cold.
 
SticKx911 said:
Offsite backup doesn't have to be fast. It's not a gaming server. Even for remote processing, cooling equipment in orbit in near vacuum at best has to be harder than a geothermal loop through the lunar surface that's probably super cold.
Click to expand...
brb starting a vacuum gapped data protection startup.
 
4saken said:
brb starting a vacuum gapped data protection startup.
Click to expand...
Everything around us today was science fiction 50 years ago. 🤷

I'd also rather them build them on the moon than in every city in the country. I'll happily encourage them to dream as their alternative is to continue to build everywhere they can here.
 
SticKx911 said:
Offsite backup doesn't have to be fast. It's not a gaming server. Even for remote processing, cooling equipment in orbit in near vacuum at best has to be harder than a geothermal loop through the lunar surface that's probably super cold.
Click to expand...
Lunar surface ranges from 120 C when facing the sun to -253 C in the permanently shadowed polar craters. Also, one lunar day is about 30 earth days since the moon is tidally locked to us. Cooling and power is not going to be simple on the moon either.
 
Tsumi said:
Lunar surface ranges from 120 C when facing the sun to -253 C in the permanently shadowed polar craters. Also, one lunar day is about 30 earth days since the moon is tidally locked to us. Cooling and power is not going to be simple on the moon either.
Click to expand...
Dream crusher. 😂😂
Good to know though. They'll just have to dig very very deep for those ground loops.
 
You know what? I guess I actually support this. I always support more work and research into space. I didn't think it would be due to a bunch of idiots trying to build datacenters on up there to escape regulations, but hey if it gets us a moon base... yeah, I guess I don't give a shit.
 
1.1.2.3.5... said:
The moon is ~ 239k miles. c is a vacuum is 186,282.4 mi/s. How much latency is acceptable here?
Click to expand...
For many things minutes could be acceptable, from training models to run big batch/tasks.

People wait 6-8-12 hours for a Claude task or specialized to do medical research model response now, they probably want some feedback on how its is going some call back home is preferable, but latency in seconds instead of milliseconds is not a big deal once those agentic models work by their own.

AlphaFold type project have been running for 20,000 gpu-years or something during very verbose light and hand off mostly affair for that type of things even a week long latency could work.
 
StoleMyOwnCar said:
You know what? I guess I actually support this. I always support more work and research into space. I didn't think it would be due to a bunch of idiots trying to build datacenters on up there to escape regulations, but hey if it gets us a moon base... yeah, I guess I don't give a shit.
Click to expand...
True. Theoretically, they could set up in the permanent polar shadows and erect solar panels on the poles, that way they would just need to rotate the panels to keep them in sunlight. The lunar surface is also supposed to be an excellent source of tritium, currently the best fuel for nuclear fusion.
 
sc5mu93 said:
Paywall. Eventuality (or sooner) Earth is going to become a prison for humanity due to all the crap in orbit because of this garbage.
Click to expand...

It isn't inevitable. They've always been making sure this doesn't become an issue.

All of the SpaceX satellites deorbit at EOL. All of their orbits are made pubic information.


Even if there were a million data centers in space, in the same orbit, the chances of someone hitting one by launching their own in the same orbit just randomly without paying attention to anything else already there is less than 1 in a million.
 
sc5mu93 said:
Paywall. Eventuality (or sooner) Earth is going to become a prison for humanity due to all the crap in orbit because of this garbage.
Click to expand...
sounds Terminator Judgment Day scenario

Skynet prison and internment camps
 
https://www.spacex.com/spacexai/starmind

seem to be going for around 160 meter square of radioator for, using all tech they already have pretty much, a bit simpler than the starlink.

Rubin being made to be cooled with mild-hottish water making thing easier and made to run hot if needed make thing easier here, that pretty much 2 whole NVL72 per satellite, will have to see how much you save by no electricity bills-land-water-municipal-state giant gift you have to give to get accepted to have anything on earth make up for the extra cost.
 
Last edited:
There is no way to cool a data center in space really. Where is the heat going to go? Space is cold but not cold air cold. There is no medium to wick away the heat. Unless someone invents heat sink tech that is reusable from Elite Dangerous the how do we cool stuff? And thats make believe. Maybe on the moon because then we could bury coolant lines and let the dirt and rock on the moon absorb heat.

We could use lasers to transmit back and forth but there is a 3 second latency so we could use moon DCs for say.... heavy compute workloads and then package final outputs to storage and retrieval facilities on earth for low latency. I mean you could send terabytes per second to the moon using lasers but its 3 second per direction latency at speed of light.
 
Last edited:
tangoseal said:
There is no way to cool a data center in space really. Where is the heat going to go? Space is cold but not cold air cold. There is no medium to wick away the heat. Unless someone invents heat sink tech that is reusable from Elite Dangerous the how do we cool stuff? And thats make believe. Maybe on the moon because then we could bury coolant lines and let the dirt and rock on the moon absorb heat
Click to expand...
Just need giant, low speed fans with refrigerant piped through the blades. ;)
 
tangoseal said:
There is no way to cool a data center in space really. Where is the heat going to go? Space is cold but not cold air cold. There is no medium to wick away the heat. Unless someone invents heat sink tech that is reusable from Elite Dangerous the how do we cool stuff? And thats make believe. Maybe on the moon because then we could bury coolant lines and let the dirt and rock on the moon absorb heat
Click to expand...

Radiation and then keeps radiating away into the cold dark of space until it hits something that can absorb it, space itself is the heatsink, the radiators on craft etc are just more like thermal transfer/paste of a sort
 
tangoseal said:
There is no way to cool a data center in space really.
Click to expand...
radiation, the giant 160 meters of surface for the amount of watt used work fine here, with the good angle with the sun and material used you can optimise the surface needed. They already do this right now (they do cool compute in space this moment, like this, just different scale)
 
Last edited:
LukeTbk said:
radiation, the giant 160 meters of surface for the amount of watt used work fine here, with the good angle with the sun and material used you can optimise the surface needed. They already do this right now (they do cool compute in space this moment, like this, just different scale)
Click to expand...
Were not talking about 500 watts for a 5090 rig in orbit, which would probably cost 5 Million bucks to send up already. Were talking MW scale usage here. Anything that we can put in space now is useless for data center compute tasks. The space station radiates 70 kw of thermal energy using liquid ammonia cooling loop. A data center in space would require at least megawatt scale reactor tech and miles of radiation surface grids to dissipate infrared thermal energy.which a totally could be done by burning off the heat through changing the heat into some big badass lightbulb that shoots off energy in a beam into the black. (OK just using imagination unrealistically here haha)

My degree is in biochemistry, not thermodynamics outside of biological reaction sized stuff. But im.pretty sure im relatively close to reality.
 
Last edited:
tangoseal said:
Were not talking about 35 watts of cooling power flr.some cube sat. Were talking MW scale usage here. Anything that we can put in space now is useless for data center compute tasks. The space station radiates 70 kw of thermal energy using liquid ammonia cooling loop. A data center in space would require at least megawatt scale reactor tech and miles of radiation surface grids to dissipate infrared thermal energy.
Click to expand...

The datacenters on Earth and hardware within =/= the datacenters of space and hardware within - even 'one large continuous building' is thinking about it wrong as the plan for most companies doing this is distributed compute/datacenterS in orbit beaming/sharing data back and forth and down to Earth

Nvidia, AMD, Intel, all make 'space variants' of their chips for reduced power/cooling constraints in space

SpaceX is starting Starmind (putting up in orbit) in 2027

Axiom already launched nodes to the ISS and has free standing in orbit nodes (that all communicate with each other)

Lonestar has/had test servers on the ISS too

Multiple other companies have built compute/compute units into satellites already in orbit specifically for the testing/start of 'datacenters'/distributed orbital workloads (aside from 'normal' compute satellites used for other reasons already have that are also up there)
 
kram182 said:
The datacenters on Earth and hardware within =/= the datacenters of space and hardware within - even 'one large continuous building' is thinking about it wrong as the plan for most companies doing this is distributed compute/datacenterS in orbit beaming/sharing data back and forth and down to Earth

Nvidia, AMD, Intel, all make 'space variants' of their chips for reduced power/colling constraints in space

SpaceX is starting Starmind (putting up in orbit) in 2027

Axiom already launched nodes to the ISS and has free standing in orbit nodes (that communicate with each other)

Lonestar has/had test servers on the ISS too

Multiple other companies have built compute/compute units into satellites already in orbit specifically for the testing/start of 'datacenters'/distributed orbital workloads (aside from 'normal' compute satellites used for other reasons already have that are also up there)
Click to expand...
Well I fear space will be so cluttered well never be able leave earth. We need to stop.putting junk in orbit and focus on the moon. Think of the vast sums of silicon, gold, silver,Water ice etc... buried in our moon we've not been able to survey. Weve only surveyed a proverbial paper thin layer barely using radar.
 
Last edited:
tangoseal said:
Well I fear space will be so cluttered well never be able leave earth. We need to stop.putting junk in orbit and focus on the moon.
Click to expand...

All the immediate/upcoming ones for the foreseeable future and most of all of them are planned to be in LEO - they'll just burn up in the atmosphere when retired

Only Bezos' one has a MEO component just for networking - the compute part of it will still be in LEO
 
tangoseal said:
Well I fear space will be so cluttered well never be able leave earth. We need to stop.putting junk in orbit and focus on the moon. Think of the vast sums of silicon, gold, silver,Water ice etc... buried in our moon we've not been able to survey. Weve only surveyed a proverbial paper thin layer barely using radar.
Click to expand...
It is easy to underestimate the scale of the size of the earth and even more the much bigger space, if they send 120 millions satellites of A1 size, it would be around 1% of 1% of a single low earth orbit amount of space and that would be 4 time the current world generation of electricity. you have millions of available orbits for satellite of that small thickness.

There is a reason we do not even see the space station from space picture of the earth, even with a good sun angle hitting on it, the station is about 30 times smaller than a single pixel of a 4k picture and does not emit enough light reflecting the sun being that small.

tangoseal said:
Were not talking about 500 watts for a 5090 rig in orbit, which would probably cost 5 Million bucks to send up already. Were talking MW scale usage here.
Click to expand...
In aggregate sure, but each satellite has only 150 kw (not megawatt) peak compute payload (120kw average) too cool, more akin to each having one NVL72 node running around their best efficacy window. The math is relatively easy to make for this and it is "old" tech, at a certain size you have X radiation possible with the best angle and so on, what they have work here. You just need a lot of A1 to reach gigawatt and terawatt later on.

Price is high to send, but the cost of the compute and the cost of the infrastructure on earth to build is also high and the cost of the electricity-battery system that it would use on earth during its lifetime much higher on earth than space.

Now that a single NVL72 are about to cost around 5-7 millions, an high cost to install them become more natural commercially to be viable the more $ per gram compute get the better you amortize at space launch for it. On earth it already starting to cost like about 1 millions per rack apparently for the infrastructure around them (cheaper than space launch maybe, but not that much cheaper) and maybe 0.8 to 2 million in electricity during say 6 years of use depending of the rate you got if you run them in earth vs being independent in space with the cost all upfront.

1- 3 millions per satellite purely for the launch in comparison if they use a working Starship that incredibly much cheaper than anything before, is not that far from the electricity price of running an NVL72 for 6 years that you do not need to pay. It does not work before Starship ~10x price reduction in cost, but does not sound out of place with it.
 
Last edited:
LukeTbk said:
It is easy to underestimate the scale of the size of the earth and even more the much bigger space, if they send 120 millions satellites of A1 size, it would be around 1% of 1% of a single low earth orbit amount of space and that would be 4 time the current world generation of electricity. you have millions of available orbits for satellite of that small thickness.


In aggregate sure, but each satellite has only 150 kw (not megawatt) peak compute payload (120kw average) too cool, more akin to each having one NVL72 node running around their best efficacy window. The math is relatively easy to make for this and it is "old" tech, at a certain size you have X radiation possible with the best angle and so on, what they have work here. You just need a lot of A1 to reach gigawatt and terawatt later on.

Price is high to send, but the cost of the compute and the cost of the infrastructure on earth to build is also high and the cost of the electricity-battery system that it would use on earth during its lifetime higher.

Now that a single NVL72 are about to cost around 5-7 millions, an high cost to install them become more natural on earth it cost about 1 millions per rack apparently for the infrastructure around them (cheaper than space, but not cheap) and maybe 0.8 to 2 million in electricity during say 6 years of use depending of the rate you got if you run them in earth vs being independent in space with the cost all upfront.

1- 3 millions per satellite purely for the launch in comparison if they use a working Starship that incredibly much cheaper than anything before, is not that far from the electricity price of running an NVL72 for 6 years. It does not work before Starship ~10x price reduction in cost, but does not sound out of place with it.
Click to expand...

It doesn't have to be cheaper than it is on earth for it to be worth doing in space. There is a bottleneck to making them on earth so it just has to cost less than the money they can make from it.

It's a political struggle to build a datacenter on earth. People are saying no.
There is also a bottleneck to building out power for them on earth. They can't build turbines even close to fast enough.

So even if it costs 10X as much to build a space data center if it is cheap enough to still make them money they're going to do it.
 
tangoseal said:
Well I fear space will be so cluttered well never be able leave earth. We need to stop.putting junk in orbit and focus on the moon. Think of the vast sums of silicon, gold, silver,Water ice etc... buried in our moon we've not been able to survey. Weve only surveyed a proverbial paper thin layer barely using radar.
Click to expand...

Why we need Space Debris Collectors (Definitely worth a watch, good Anime)

1786120607164.png
 
Being buried/jettisoned into space will be huge in the future. I bet low orbit degrade then burn up through the atmosphere will be half off price.
 
Balkroth said:
Why we need Space Debris Collectors (Definitely worth a watch, good Anime)

View attachment 819019
Click to expand...

This is actually a thing already, debris removal as a service or something. They'll go up into LEO and push things into descent to burn up or a further out orbit.

Kessler Syndrome really only matters for things in LEO, after/above/beyond that there's so much 'space' in-between things and 'space' to navigate around things it's no longer an issue.
 
Last edited:
kram182 said:
This is actually a thing already, debris removal as a service or something. They'll go up into LEO and push things into decent to burn up or a further out orbit.
Click to expand...
I believe there's been some trials and people have offered it for some in the future (along with a "refuelling" possibility), but I don't think there's been a real service launch so far for it.
 
Balkroth said:
I believe there's been some trials and people have offered it for some in the future (along with a "refuelling" possibility), but I don't think there's been a real service launch so far for it.
Click to expand...

Gotta get to step 0, or step 0.5, before you get to step 1
 
Well what do you know, apt timing for an article on CNBC.

https://www.cnbc.com/2026/08/08/space-debris-junk-spacex-rocket-moon-crash.html


How cleaning up space debris could grow to become a big business​

Published Sat, Aug 8 20268:04 AM EDT
Elsa Ohlen
ShareShare Article via FacebookShare Article via TwitterShare Article via LinkedInShare Article via Email
Key Points
  • The space around Earth is becoming increasingly congested, as the number of satellites and pieces of space debris grows, raising the risk of collisions.
  • Companies including Astroscale and ClearSpace are betting that debris removal will develop into a broader business servicing, repairing and extending the lives of satellites.
  • The market is still nascent and largely government-backed, with regulation likely to play a key role in creating commercial demand.
Thousands of satellites are circling the Earth, and when they die, they don’t just disappear. Neither do the rockets that put them there.

This week, a spent SpaceX Falcon 9 upper stage crashed into the moon, more than a year after it was left on an uncontrolled trajectory following a lunar mission.


Closer to Earth, defunct satellites, spent rocket parts and other debris can remain in orbit for years, adding to an increasingly congested environment and an elevated risk of collisions.

Now, a growing group of companies is betting that cleaning up this orbital junk, and servicing satellites before they become it, could become a business in its own right.
Click to expand...
 
You must log in or register to reply here.
Back
Top