Repairability and depreciation are the main problems. A earth data center can be repaired, depreciated, and recycled at EoL recovering some of the costs. SpaceX datacenters are a total write off from the moment they are launched.
The biggest issue with space is not repairability but heat - when you’re in a vacuum the only way to disperse heat is through black body radiation and that’s horribly slow compared with normal mechanisms. It means you need giant physical structures whose sole job is to accept heat from the processing core and radiate it away and have so much more material that you can radiate it at the speed you generate. It’s a huge unsolved physics problem which is why everyone is skeptical.
The big win of being in space is just a worse alternative to using an intermediary heat transfer medium.
The problem with data centers in space is one of materials science and engineering: how to make radiators large enough and effective enough to cool it while also being economically feasible, both in terms of construction and getting them up there in the first place.
We can make a space data center right now. It would just be terrible and expensive.
COTs GPUs throttle at about 90-100C but that's because they have plastic parts and solder that melts. Those are relatively easy to eliminate.
We haven't tried much to scale up operating temp.
(Electronics that works at 200C do exist, but they're not gonna meet a modern smartphone's specs, let alone a datacenter)
The story I heard was intel failed to deliver on die shrinks that were supposed to increase thermal efficiency. The fan also had poor routing and basically did nothing.
Apple was clearly not at the top of their game for that design, nor would you expect them to be as the decision was likely already made to move to apple silicon.
Musk’s proposal is to put 1 rack per satellite drawing 150kw using 110 m^2 of radiators (1/4th that of the ISS). The only way to do that is by running the satellite at 71-100C which is a problem for the chips to actually run.
So the open questions are: * can they actually dump the heat and can they have the racks running so hot for so long * is it economical to in one year replace all the capacity you launched 5 years ago (you save up to 0.5B in electricity costs over that time frame but that doesn’t seem like a lot for having to replace that much capacity buildup)? And given they’re running racks way hotter than has been validated on earth, will those chips end up lasting 5 years (including space radiation).
Pretending like it’s a solved problem is neat but I’m not saying sure it’ll be so easy.
In other words, run fully lights-out datacentres that no humans will access during normal operation - this allows you create more extreme temperature gradients and to replace the atmosphere with CO2 (or even He2), both of which will make your cooling solution much more efficient as well as increasing rack density and perhaps even taking the place of a fire suppression system.
The reason that no-one's actually doing that at scale is that it's currently significantly more expensive than traditional datacentre designs. And yet, that's still much less expensive than boosting racks into space and then letting them burn up in the atmosphere five years later.
The problem is not solved for something of this scale, I think its solvable, but no has done anything at the size these would have to be. And if the physics and engineering can be made to work, they are unlikely to be solved AND ALSO be economical.
It's classic type problem where the choices are:
- economical
- in-space
- big
Now pick two.
There is also the unsolved problem that the radiators would be absolutely massive in area. Making damage from flying objects, meteorites, space jump, close to 100% likely over any reasonable life span. Repairs would be expensive AF. Again economics is the driver of why this is dumb.
I've done builds that ran for 5+ years with virtually no physical attention, just continual degradation as hardware is taken out of service. There's also not much money to recover from 5+ year-old hardware.
I used to run AI inference GPU servers in road vehicles, which is probably an even harsher environment than a single rocket launch, and the vibration problems are real but solvable.
Also space has more radiation
I think this could be done at an interesting scale even on Falcon 9 alone. If Starship does even 20% of its early design goals, it'll beat Falcon 9 and we could see orbital servers being demised and replaced every 3 years, maybe even 2, for ones with abnormally high failure rates.
Now, whether or not this will all make money in the end has a lot to do with what's going on down here on terra firma and how long it takes to get useful capacity into orbit.
(It's taken 7 years to get Starlink capacity enough for serving 10M customers. Verizon FiOS did 10M in 5 years. AT&T Fiber took 4-5 years to deploy to 10M. So, space isn't a lot slower than terrestrial.)
But it depreciates faster. That fiber run is lasting for 50 years, not 5. You need 10x the installation capacity just to keep up.
If the orbit isn't full, all they are is additional compute.
Definitely not definitive but it's plausible current hardware could survive with minimal modification
my question was more whether the hardware would need extra redundancy or shielding in order to not have unacceptably high error rates
https://arstechnica.com/space/2026/07/how-hard-is-it-to-buil...