Elon Musk says it. Jeff Bezos believes it. And now Sundar Pichai is betting big on it. By 2030, it could be cheaper to run massive data centers in space than on Earth.
Sound like science fiction? It's not. Google just announced Project Suncatcher—a plan to launch solar-powered satellites equipped with AI chips into orbit by 2027. SpaceX says they'll be doing it too. And startups are already lining up to make this happen.
So what's going on? Why would anyone want to put data centers in space? Let's break it down.
When we say "space," we're not talking about Mars or the Moon. We're talking about Low Earth Orbit (LEO)—roughly 200 to 2,000 kilometers above Earth's surface.
This is where most satellites already operate. The International Space Station orbits at about 400 km. SpaceX's Starlink satellites are around 550 km. This is the sweet spot: close enough to Earth for fast communication, far enough to avoid atmospheric drag.
Google's Project Suncatcher plans to use sun-synchronous orbits—a special type of orbit where satellites stay in constant sunlight. No day. No night. Just 24/7 sunshine.
Here's where it gets interesting. You can't just launch a giant building into orbit. Instead, the plan is to create constellations of small satellites that work together.
Think of it like this: Instead of one massive data center, you have dozens or hundreds of smaller satellites flying in tight formation—separated by just hundreds of meters—all connected through ultra-fast optical links.
Each satellite would contain:
Google's research shows they can achieve 800 Gbps to multiple terabits per second of data transfer between satellites. That's comparable to the speeds inside terrestrial data centers.
The satellites would fly in formations so tight—just kilometers or even hundreds of meters apart—that they can maintain these high-bandwidth connections using free-space optical links (basically, lasers).
This is the trillion-dollar question. Here's the math:
Problem #1 on Earth: Power is expensive and dirty
Data centers consume enormous amounts of electricity. Currently, global data centers use about 415 terawatt hours annually—that's 1.5% of all electricity consumed worldwide.
With AI models getting bigger and more power-hungry, this is only going up. Google, Microsoft, Amazon—they're all scrambling to secure power. Some are even restarting old nuclear plants.
In space, power is free and unlimited.
Solar panels in orbit are 8 times more productive than on Earth. Why?
Starcloud, a startup working on orbital data centers, projects that energy costs in space could be 10 times cheaper than on Earth, even including launch costs.
Problem #2 on Earth: Cooling is a nightmare
Data centers generate massive heat. Cooling them requires enormous amounts of water and electricity. Large data centers can use 5 million gallons of water per day.
In space, cooling is free.
The vacuum of space is the ultimate heat sink. You don't need water. You don't need air conditioning. You just use radiative cooling—heat radiates directly into the cold void of space.
According to research from Nanyang Technological University, space-based data centers could save 10 times the carbon emissions compared to terrestrial operations.
Problem #3 on Earth: Real estate and infrastructure
Building a data center on Earth means:
In space, you skip all of this.
No land purchase. No grid connection. No water supply. No local permits. You just launch and operate.
The launch cost math
Jeff Bezos said at a tech conference: "We will be able to beat the cost of terrestrial data centers in space in the next couple of decades."
Here's why he's probably right: Launch costs have been falling dramatically. Google's research suggests that with sustained improvement, launch prices could drop to $200 per kilogram by the mid-2030s.
At that price, the cost of launching and operating a space-based data center becomes comparable to the energy costs of running an equivalent facility on Earth.
Beyond just cost, there are some compelling technical reasons:
1. Lower latency for satellite communications
Right now, if you're using satellite internet (like Starlink), your data travels: Device → Satellite → Ground station → Data center → Ground station → Satellite → Device
That's a lot of back-and-forth. With orbital data centers, you cut out the middle steps. Data goes: Device → Satellite data center → Device
This cuts latency in half. For applications like autonomous vehicles, drones, ships, and remote areas, this matters.
2. Better for space-based applications
Satellites already collect massive amounts of data—Earth observation, weather monitoring, GPS, imaging. Right now, all that raw data gets beamed down to Earth for processing.
With orbital data centers, you could process data in orbit and only send back the results. This is far more efficient.
For example: Instead of sending terabytes of raw satellite images to Earth, you process them in orbit and just send back "Found 50 ships in this area" or "Detected forest fire at these coordinates."
3. Carbon neutral operations
After the initial carbon cost of launch (which gets offset in about 5 years), these facilities run on 100% renewable solar energy with zero ongoing emissions.
Compare that to terrestrial data centers, which are responsible for about 2% of global carbon emissions—roughly the same as the aviation industry.
4. No geographical constraints
On Earth, data centers are limited by where power, water, and internet connectivity exist. In orbit, you can scale infinitely (well, until you run out of orbital slots, but that's a problem for later).
5. Security and redundancy
Space-based data centers are physically isolated. No risk of physical break-ins, natural disasters like floods or earthquakes, or geopolitical interference. They're also naturally distributed, providing built-in redundancy.
This isn't easy. Here's what needs to be solved:
1. Radiation
Space is full of radiation—cosmic rays, solar particles, etc. This can cause computer chips to malfunction or degrade.
The good news: Google tested their Trillium TPU chips in a 67MeV proton beam and found they're surprisingly radiation-tolerant. They only showed irregularities after three times the expected 5-year mission dose.
Still, radiation-hardened chips may be needed for long-term operations.
2. Heat dissipation
While space is cold, getting rid of heat is actually harder than on Earth. There's no air for convection cooling. Heat can only escape through radiation, which requires large radiator surfaces.
This means satellites need huge radiators, adding weight and complexity.
3. Maintenance
On Earth, if a server fails, a technician replaces it. In space, you can't do that easily.
You need autonomous robotic systems for repairs, or you accept that satellites have a limited lifespan and will need replacement every few years.
4. Space debris
More satellites mean more collision risk. Orbital data centers would need sophisticated tracking, maneuvering, and shielding to avoid debris. The industry needs responsible "end of life" protocols to prevent creating more space junk.
5. Communication with Earth
Getting data up to and down from orbit requires high-bandwidth ground stations. While satellite-to-satellite communication is solved with optical links, the Earth-to-space link is still a bottleneck.
6. Economics still need to prove out
Launch costs are falling, but they're not there yet. The entire business case depends on launch costs continuing to decline as projected.
Google (Project Suncatcher): Planning to launch two prototype satellites by early 2027 in partnership with Planet. Testing TPUs and optical inter-satellite links.
SpaceX: Elon Musk tweeted that SpaceX will be doing this, likely by scaling up Starlink satellites to include computing capabilities.
Amazon (Blue Origin): Jeff Bezos said massive gigawatt-scale data centers powered by space-based solar energy will be operating in orbit within 10-20 years.
Starcloud: An Nvidia-backed startup that launched a satellite with an Nvidia H100 GPU. They predict that "in 10 years, nearly all new data centers will be built in outer space."
Orbit AI and PowerBank: Announced plans to launch the first orbital cloud satellite in 2026, with solar-powered infrastructure in low Earth orbit.
Thales Alenia Space: The French aerospace giant published a feasibility study on space data centers last year.
Timeline predictions:
Elon Musk recently said space AI data centers could be realized within five years.
Jeff Bezos predicts 10-20 years for gigawatt-scale operations.
Google's research suggests economic feasibility by the mid-2030s.
In the near future, when you:
...there's a good chance your request will be processed not by a server farm in Virginia or Oregon, but by a satellite orbiting Earth at 17,000 miles per hour.
You won't notice any difference. Except maybe your query gets answered a bit faster. And the carbon footprint is zero.
This isn't just about data centers. It's about humanity's expansion beyond Earth.
Building computing infrastructure in space forces us to solve hard problems:
These technologies will be essential for:
Space-based data centers are a stepping stone to becoming a multi-planetary civilization.
Plus, there's a philosophical elegance to it: The Sun produces more energy than 100 trillion times humanity's total electricity production. We're floating in a sea of unlimited clean energy. Why not use it?
The idea of data centers in space sounds crazy. But the physics makes sense. The economics are getting there. And the world's richest, smartest tech leaders are betting billions on it.
Within a decade, answering your AI queries from orbit might not just be possible—it might be cheaper, faster, and cleaner than doing it on Earth.
The cloud is moving to space. Literally.
Sources: Google Research Blog, Forbes, SingularityHub, WSJ
Note: This is an educational article exploring emerging technology. The timeline and feasibility of space-based data centers remain subject to technological and economic developments.
