The Next Data Center Might Be in Space
Because apparently Earth is running out of room for our AI addiction
Imagine this
It is 2035
You wake up, grab your glasses, and ask your AI assistant
Plan my entire trip to Japan?
A few seconds later, it has booked your flight, found a hotel, checked the weather, created an itinerary, and somehow remembered that you hate early morning flights
You smile
You probably don’t think about where all that computing happened
Maybe you should
Because somewhere hundreds of kilometers above your head, a computer might be doing part of the work
Not in an office
Not inside a giant building surrounded by cooling towers
In space
Yes, we may actually start putting data centers in orbit
And no, this isn’t a plot twist from a science fiction movie
Companies are already working on it
Why would anyone put a data center in space
Because AI has developed a small problem
It wants a ridiculous amount of computing power
Every few months, we seem to get bigger AI models, more powerful GPUs, more servers, and more electricity consumption
So we build bigger data centers
Then bigger ones
Then even bigger ones
At some point, someone naturally asks
Where are we going to put all this
And suddenly Earth looks surprisingly small
Modern data centers need land, electricity, cooling, networking infrastructure, and a lot of equipment
Space, meanwhile, has one thing in absurd abundance
Sunlight
Google’s Project Suncatcher is exploring whether solar powered satellites carrying AI processors could perform large scale computing in orbit, with prototype satellite testing planned with Planet as early as 2027
So the basic idea sounds almost ridiculously simple
Take computers → put them on satellites → power them with sunlight → make them communicate with each other
Sounds easy
It isn’t
Space has no electricity bill
Imagine opening your electricity bill and seeing
₹0
Beautiful
Unfortunately, your landlord has been replaced by orbital mechanics
One of the biggest attractions of space based computing is access to solar energy
Satellites can spend significant periods above Earth’s atmosphere with access to sunlight, making solar power especially attractive for orbital infrastructure
But there is a catch
Computers don’t just consume electricity
They produce heat
And computers are apparently very committed to this tradition
So how do you cool a computer in space
On Earth, data centers can use air conditioning, water cooling, and massive mechanical systems
Space doesn’t have air
You can’t simply install a giant fan and point it toward the universe
Instead, spacecraft use radiators to release heat into space
This means an orbital data center has to be designed very differently from a normal one
The more computing power you add, the more heat you have to deal with
Google’s research includes studying thermal management and radiation effects on computing hardware in orbit
So yes, the future might contain a data center floating peacefully above Earth while its engineers are arguing about radiator size
Some things never change
There is another problem
Space hates electronics
Earth is comparatively comfortable
Space is not
Satellites have to deal with radiation, launch vibrations, extreme temperatures, and hardware failures that are much harder to fix
On Earth, if a server dies, someone replaces it
In orbit
“Can you restart it”
“Sure”
“Physically”
“…Let’s not talk about that”
That is why orbital computing needs highly reliable hardware and increasingly autonomous systems
And this is where things become interesting
Your satellite could become a computer
Today, a satellite might take a photograph of Earth and send that data back to a ground station
Then the information travels through terrestrial infrastructure
Then some computer analyzes it
But imagine the satellite doing the analysis itself
It sees a huge forest fire
Instead of sending thousands of raw images to Earth, it could process them onboard and send
Large fire detected, location here
Much less data
Much faster response
Axiom Space is already developing orbital data center infrastructure designed to process data in space, including AI workloads and optical communication between orbital nodes
This is the really important idea
Space data centers don’t necessarily have to replace Earth data centers
They can process information where it is created
Now imagine hundreds of them
One computer in space is interesting
One hundred is a network
Get Sanchit’s stories in your inbox
Join Medium for free to get updates from this writer.
Now imagine thousands
Satellite A collects data
Satellite B processes it
Satellite C stores it
Satellite D sends the result toward Earth
All of them communicate with each other
Suddenly, you don’t have a satellite anymore
You have something that looks a lot like a cloud computing network in orbit
And that’s exactly why some companies are taking this seriously
Google isn’t the only one
Google is exploring Project Suncatcher
Axiom Space is developing orbital data center nodes
And startup Starcloud is building its own orbital computing infrastructure
In November 2025, Starcloud launched a satellite carrying an NVIDIA H100 GPU
Yes
An actual AI GPU
In space
And in August 2026, the company announced another $250 million in funding to expand its orbital data center ambitions
At this point, space is starting to look suspiciously like the IT department
But why not just build more data centers on Earth
Good question
Maybe that’s exactly what happens
Space data centers have massive challenges
Launching hardware is expensive
Spacecraft are difficult to repair
Radiation is a problem
Thermal management is difficult
Communication is complicated
And then there is the small issue of space debris
You probably don’t want your million dollar AI server becoming another piece of junk flying around Earth at several kilometers per second
So this isn’t
Earth data centers are finished
It’s more like
Could some workloads be better handled in orbit
That is a much more reasonable question
And then AI makes the idea even more interesting
AI is changing the economics of computing
Ten years ago, nobody needed a giant amount of computing power just to ask a chatbot to write an email
Today, AI is being used for code, video, scientific research, robotics, drug discovery, simulations, and increasingly autonomous systems
The demand for compute keeps growing
That makes people look for new sources of power and new places to put computers
Space happens to offer both a completely different environment and a very large supply of solar energy
That doesn’t guarantee success
But it makes the experiment worth trying
One day, the cloud may literally be above you
We’ve become so used to saying
It’s in the cloud
But the cloud is not actually a cloud
It is a warehouse full of machines
Maybe one day, part of that cloud really will be above us
You ask an AI something
The request goes to a network
The network decides where the workload should run
Maybe it is in a data center in Mumbai
Maybe it is in Singapore
Maybe it is on a satellite
You probably won’t know
And you probably won’t care
Just like you don’t think about the thousands of kilometers of fiber optic cable carrying your message today
Technology becomes truly successful when it becomes boring
That’s when we stop noticing it
So will data centers really move to space
Maybe
And that’s the most honest answer
The technology is still developing
The economics aren’t proven at massive scale
There are serious technical and environmental challenges
But something interesting has already happened
People stopped asking
Can we put computers in space
Because we’ve already done that
Now they’re asking
Can we build a computing industry in space
That’s a much bigger question
And if the answer eventually becomes yes, the implications could go far beyond AI
Earth observation
Scientific computing
Communications
Space exploration
Autonomous spacecraft
Lunar infrastructure
Maybe eventually even computing networks extending beyond Earth
The internet connected computers
The cloud connected data centers
AI connected intelligence to software
Perhaps the next step is connecting computing itself to space
And honestly, after spending decades putting everything in the cloud
we may have finally decided the cloud wasn’t high enough
