What If We Moved Data Centers Into Space?
AI is creating an extraordinary appetite for energy. Could part of the solution eventually be above us?
Published by Norm Levy: Septemer 2026
Every time we ask an AI a question, something happens somewhere in the physical world.
Computers process it.
Servers generate heat.
Cooling systems remove that heat.
Power plants, solar farms, wind turbines, nuclear reactors, batteries, transmission lines, and electrical grids provide the energy that keeps the whole system running.
The cloud, it turns out, isn’t really a cloud.
It’s infrastructure.
And we are about to need an extraordinary amount of it.
AI Has an Energy Problem
Artificial intelligence may feel weightless when we use it.
Type a question.
Generate an image.
Translate a document.
Ask a model to analyze a million lines of data.
Seconds later, an answer appears.
Behind that seemingly effortless interaction are increasingly enormous data centers filled with specialized processors operating around the clock.
The International Energy Agency estimates that data centers consumed roughly 485 terawatt-hours of electricity worldwide in 2025. By 2030, that figure could roughly double to around 950 terawatt-hours.
AI-focused data centers are growing even faster.
In the United States, the Department of Energy cites estimates suggesting data centers could consume roughly 12 percent of the nation’s electricity by the end of the decade.
That doesn’t mean AI is going to consume all of our electricity.
But it does mean something important is changing.
Computing is becoming an infrastructure problem.
And increasingly, an energy problem.
The New Factories
For much of the industrial age, economic power was measured in physical production.
Steel mills.
Oil refineries.
Automobile plants.
Ports.
Railroads.
The infrastructure of the twenty-first century looks different.
Rows of servers.
GPU clusters.
Fiber-optic networks.
Transformers.
Substations.
Cooling systems.
The factories of the AI age manufacture something less visible.
Compute.
And compute requires electricity.
Lots of it.
The International Energy Agency reported in 2026 that electricity demand from data centers increased 17 percent in a single year. At the same time, the enormous buildout is beginning to encounter physical bottlenecks, including transformers, turbines, chips, grid connections, permitting, and power generation.
Suddenly the AI race isn’t only about who builds the best model.
It’s also about who can power it.
The Cloud Comes Down to Earth
There is an irony hidden in the word “cloud.”
Our digital world feels increasingly detached from physical reality.
But the infrastructure supporting it is extraordinarily physical.
Data centers require land.
They require transmission infrastructure.
They generate enormous amounts of heat.
Many cooling systems require water, although technologies and designs vary significantly.
And because enormous data centers concentrate electricity demand in particular locations, their impact can be much greater on a local grid than their share of global electricity consumption might suggest.
Communities across America are beginning to confront questions that would have sounded strange a generation ago.
How much electricity should a data center receive?
Who pays for new transmission infrastructure?
Where does the additional generation come from?
How much water should computing consume?
What happens when the infrastructure required for the digital economy begins competing with the infrastructure required for everyday life?
There may be many answers.
Build more renewable energy.
Build nuclear power.
Improve chip efficiency.
Develop better cooling systems.
Strengthen electrical grids.
Build data centers where abundant power already exists.
But there is another possibility that sounds considerably more like science fiction.
Move some of the computing somewhere else.
Why Space?
Space offers something data centers desperately need.
Energy.
Above Earth’s atmosphere, sunlight can be extraordinarily abundant. Solar arrays do not have to contend with clouds or weather, and depending on the orbit and architecture, systems could achieve extremely high solar availability.
That has led engineers and technology companies to begin asking a remarkable question:
Could we eventually build computing infrastructure in orbit?
In 2025, Google announced Project Suncatcher, a research initiative exploring constellations of solar-powered satellites carrying its Tensor Processing Units.
The idea is to place computing hardware where enormous amounts of solar energy can potentially be harvested and connect those satellites through high-speed optical links.
Google plans to launch two prototype satellites in partnership with Planet in early 2027 to begin testing the concept.
And Google isn’t alone.
In November 2025, Starcloud launched Starcloud-1 carrying an NVIDIA H100 GPU into orbit. The company later reported running an AI model and training a small language model aboard the spacecraft.
These are experiments.
We are nowhere near relocating the world’s data centers into orbit.
But something important has happened.
The question has moved from:
That’s impossible.
to:
Could this work?
Space Doesn’t Make the Problems Disappear
Of course, putting computers in space introduces an entirely new collection of problems.
Hardware has to survive radiation.
Getting equipment into orbit remains expensive.
Processors eventually fail and must somehow be repaired, replaced, or deorbited.
High-performance computing generates heat, and getting rid of heat in the vacuum of space is not as simple as opening a window. Without air or water flowing past equipment, heat must ultimately be radiated away.
Information also has to travel between Earth and the orbital infrastructure quickly and reliably.
Then there is the environmental question.
Moving infrastructure off Earth doesn’t automatically make it sustainable.
Someone still has to manufacture the satellites.
Someone has to launch them.
Someone has to replace them.
And eventually, someone has to decide what happens when thousands of pieces of computing infrastructure reach the end of their useful lives.
Space isn’t a magic solution.
It is another engineering environment.
But What If It Works?
This is where the idea becomes fascinating.
Imagine a future in which some of humanity’s most energy-intensive computing no longer needs to sit beside communities on Earth.
Enormous solar arrays collect energy in space.
Orbital computing platforms process AI workloads.
Manufacturing facilities eventually produce components beyond Earth.
Asteroid resources provide some of the raw materials.
Earth remains the place where people live.
But an increasing share of the industrial infrastructure supporting civilization begins migrating outward.
Not because Earth has been abandoned.
Because Earth is worth preserving.
That idea has fascinated me for years.
The Idea Behind H.O.M.E.
When I began building the world of Band on the Run, I kept returning to a simple contradiction.
Humanity wants more.
More computing.
More energy.
More devices.
More transportation.
More resources.
More everything.
Yet we live on a planet with finite land, finite ecosystems, and communities that bear the consequences of the infrastructure required to satisfy that demand.
That contradiction helped inspire Miles Wentworth’s vision for H.O.M.E., the High Orbit Manufacturing Enterprise.
Wentworth doesn’t believe humanity should stop growing.
He believes we should move some of the consequences somewhere else.
In his world, orbital solar infrastructure produces enormous amounts of energy. Asteroids supply industrial materials. Manufacturing increasingly moves beyond Earth.
And eventually, so does computing.
His philosophy can be summed up in one line from the novel:
“Not to save the Earth, but to spare it.”
That distinction matters.
Wentworth isn’t an environmentalist in the traditional sense. He doesn’t believe humanity will willingly consume less.
He assumes we’ll continue wanting more.
So his answer is to expand the resource base of civilization itself.
Science Fiction Has a Funny Way of Moving
When I first began imagining this world, orbital data centers felt comfortably speculative.
Then the AI boom arrived.
Data-center construction accelerated.
Electricity demand surged.
Technology companies began searching for new energy sources.
And now engineers are actually putting advanced AI processors into orbit and studying whether networks of solar-powered computing satellites could work.
That doesn’t mean Band on the Run predicted the future.
Science fiction doesn’t have to predict anything to be useful.
Sometimes its job is simply to take something happening today and ask:
What happens if this keeps going?
AI will require more computing.
Computing will require more energy.
Energy will require infrastructure.
And infrastructure has to exist somewhere.
For the moment, that somewhere is Earth.
But perhaps it won’t always be.
Maybe the cloud really will end up in the sky.
Which, come to think of it, is where the name always made it sound like it belonged.