For decades, the cloud has been a metaphor for boundless digital capacity — a weightless abstraction where data lives in endless abundance. But that cloud is getting heavy. Artificial intelligence (AI), high-frequency trading, streaming and autonomous systems are devouring bandwidth, power and land. The global data center industry already consumes nearly 3% of the world’s electricity. By 2030, that figure could double, surpassing the power use of Japan, according to the International Energy Agency (IEA).
Now, a new idea is taking shape in boardrooms and laboratories from Redmond, Washington, to Turin, Italy: moving the cloud into actual space. Amazon founder Jeff Bezos, SpaceX’s Elon Musk, Alphabet CEO Sundar Pichai and a swarm of startups are betting that the next generation of computing power will orbit Earth — or even sit on the moon. These companies see space-based data centers not as science fiction but as the logical next step in humanity’s digital evolution: limitless solar energy, zero land costs and near-constant uptime.
“It will soon make much more sense, with the abundant energy and cold temperatures, to build data centers in space than it does to build them on Earth,” Philip Johnston, CEO of Redmond, Washington-based startup Starcloud, told the October 2025 Future Investment Initiative (FII) in Riyadh, Saudi Arabia. “For the first time, the technology really is there to make data centers in space our reality.”
From Ground to Orbit
The concept sounds over the top: Take the vast server farms that hum quietly in places like Ashburn, Virginia, or Dublin, Ireland, and move them into orbit. The idea isn’t unprecedented. In 2021, Hewlett Packard Enterprise’s Spaceborne Computer-2 began operating on the International Space Station, processing workloads ranging from DNA sequencing to image analysis. The project proved that commercial-grade computing hardware could function in microgravity without catastrophic failure.
Since then, the vision has grown exponentially. In November 2025, Starcloud launched an Nvidia H100 graphics processor — the same kind used in terrestrial AI training — aboard a satellite roughly the size of a minifridge. The company says the processor is 100 times more powerful than any previous space-based computer. Sometime in 2026, Starcloud plans to launch its first commercial “micro data center” constellation.
Others are not far behind. Florida-based Lonestar Data Holdings successfully tested a shoebox-size storage device on the moon in early 2025. Its next step: six satellites to serve as secure off-world data vaults. Axiom Space is building “orbital nodes” for AI processing, while Abu Dhabi’s Madari Space is preparing to launch toaster-sized computing payloads for on-orbit data analytics in 2026.
Europe, too, is moving quickly. The ASCEND project, led by aerospace giant Thales Alenia Space and funded by the European Union, envisions a 200-megawatt constellation of orbiting data platforms — roughly equivalent to a medium-sized terrestrial data center — assembled on orbit by 2035. ASCEND stands for Advanced Space Cloud for European Net Zero emission and Data sovereignty. The consortium believes such infrastructure could transform the European digital landscape while cutting emissions from Earth-based computing.
Why space? The answer lies partly in Earth’s limits. Data centers are energy gluttons, requiring vast tracts of land and millions of gallons of water for cooling. In Europe and parts of the United States, local resistance to new facilities is growing. Power grids are straining under the load, and environmental regulators are pushing back.
An October 2025 Microsoft financial report noted rising “community opposition, local moratoriums and hyper-local dissent” to many such projects. Alongside peers including Meta Platforms Inc. and Alphabet Inc.’s Google, Microsoft has embarked on a massive build-out of server farm infrastructure to support its cloud computing and AI businesses. Yet, like others, the world’s largest software company is facing pushback. In October, it nixed plans for a Wisconsin data center after residents and elected officials opposed its construction, Bloomberg reported.
Elsewhere, plans for centers are triggering protests, environmental impact reviews and grid capacity debates. In space, there are no neighbors and no zoning laws. Solar power is constant — 24 hours a day, 365 days a year, with no clouds, no night and no atmospheric loss. Energy from the sun is 30% more intense than at ground level, offering free and limitless power once infrastructure is in place.
That constant energy could make orbital data centers ideal for AI training and inference workloads that demand enormous, uninterrupted power. “There’s a physics-based elegance to it,” Jason Aspiotis, global director of in-space data and security at Axiom Space, told the Massachusetts Institute of Technology’s (MIT) Technology Review. “In space, you don’t compete with cities for electricity or water. You can radiate heat directly into the vacuum, and solar energy is infinite.”
Space also offers potential security advantages. With data stored above the atmosphere, cyberattacks, natural disasters or physical sabotage become exponentially harder. Subsea fiber-optic cables, on the other hand, which carry 95% of global data traffic, are vulnerable to both physical breaks and geopolitical manipulation. Laser-linked constellations on orbit could create a secure, self-contained network beyond the reach of terrestrial interference.
“Data stored in space is accessible regardless of on-the-ground challenges and provides a secure connection regardless of borders, natural disasters, cut cables, power blackouts or wars,” Lonestar chairman and founder Chris Stott told Apogee. “It’s the ultimate way to secure data — out of reach, but accessible when needed.”
The Big Bet: AI on Orbit
AI is the accelerant behind this new frontier. Training large language models like ChatGPT or Gemini requires petawatt-hours of energy and warehouses of graphics processing units (GPUs). A petawatt is 1 trillion kilowatt hours. The IEA estimates that AI-related electricity use could grow by 160% by 2030.
Space-based computing could help relieve that burden — and, paradoxically, speed it up. By processing satellite imagery, sensor data or AI workloads directly on orbit, companies could cut latency and reduce the costly backhaul of raw data to Earth. “Processing data in orbit reduces the need for extensive bandwidth and alleviates bottlenecks,” Kepler Communications’ Ryan White, director of software and digital design, explained on the company’s website. “On-orbit computing minimizes latency, which enables faster insights and actions.”
There’s another reason tech giants are looking skyward: AI’s insatiable hunger for electricity. Training and running large language models, image generators and autonomous-system algorithms consume staggering amounts of power — sometimes rivaling that of small nations. Goldman Sachs projects that global data center electricity demand will climb 165% by 2030, driven largely by AI. Each new generation of chips — like Nvidia’s H100 and H200 processors — delivers greater performance but also draws far more energy and generates more heat.
That surge in demand is colliding with a strained power grid. Utilities are struggling to meet the needs of hyperscale facilities, delaying projects and raising prices. In Virginia, a joint legislative report in December 2024 predicted a doubling of the state’s power demand over the next 10 years, driven primarily by data centers. Increases in the projected energy demand have “raised concerns about whether enough infrastructure can be built to keep pace,” the report said.
Space, by contrast, offers unlimited solar energy. A single square kilometer of solar array on orbit can generate several times more power than the same surface on Earth, free from clouds and nightfall. For companies betting their future on AI, that promise of clean, continuous energy is irresistible. “In space, we don’t need permitted land … we don’t need battery storage because we have 24/7 solar power,” Starcloud’s Johnston told potential investors in Riyadh.
Musk’s SpaceX, whose Starlink satellites already blanket space with internet coverage, is quietly planning to integrate graphics processing unit clusters into future satellites. Musk said on the social media platform X in October 2025 that upgraded Starlink V3 satellites — each roughly the size of a Boeing 737 — could double as orbital data nodes to build data centers. Powered by Starship — SpaceX’s next-generation superheavy launch system — these satellites eventually could form a solar-powered AI network spanning Earth’s orbit and, later, the moon.
Alphabet’s Project Suncatcher, announced in November 2025, takes a similar approach. Google plans to launch prototype satellites carrying its custom Tensor Processing Units (TPUs) in 2027, testing how AI models function in microgravity. “Our new research moonshot, Project Suncatcher, envisions compact constellations of solar-powered satellites, carrying Google TPUs and connected by free-space optical links. This approach would have tremendous potential for scale and also minimizes impact on terrestrial resources,” Google senior director Travis Beals wrote that month in a Google Research blog post.
Lonestar is similarly poised for on-orbit testing, teaming up with another company to launch and evaluate in-space data storage and processing in low Earth orbit (LEO). Potentially, Lonestar is eyeing data centers on the moon and in cislunar space, generally the region between Earth and the moon — or what Lonestar’s Stott calls “the ultimate secure Fort Knox solution … due to the physical distance and delayed tolerant network.”
“This will be the most secure place to store data for resiliency purposes,” he said.
Bezos has joined the chorus. Addressing an audience during Italian Tech Week on October 3, 2025, the Amazon and Blue Origin founder predicted “gigawatt-scale” data centers in space within 20 years. “We have solar power there 24/7,” Bezos said. “No clouds, no rain, no weather. We’ll be able to beat the cost of terrestrial data centers.”
If that sounds optimistic, that’s because it is. The economics remain daunting. Launching a kilogram into orbit can cost tens of thousands of dollars, and space-based data centers must haul heavy shielding, cooling systems and networking gear.
Yet those costs are dropping fast. SpaceX’s Falcon 9 has slashed launch prices by 90% compared with two decades ago. The company’s Starship, once operational, would reduce them further — to about $100 per kilogram for payloads of 100 metric tons, possibly by the mid-2030s, according to statements by Musk and analysts’ projections. At that price, operating costs of orbital data centers could approach those of terrestrial facilities on a per-kilowatt-year basis.
Companies like Starcloud believe they can achieve 10 times the energy cost savings over a data center’s lifetime. “Instead of paying $140 million for electricity, you can pay $10 million for a launch and solar,” Johnston told TechCrunch in December 2024. Still, the upfront costs are staggering. Under an agreement finalized in November 2025 with Florida-based Sidus Space Inc., Lonestar’s six-satellite LEO network is expected to cost $120 million.
Then there’s the issue of heat. Without air or water, cooling electronics in vacuum conditions requires enormous radiators to dissipate thermal energy. Starcloud’s engineers are designing radiators the size of football fields for what eventually would be a massive, multigigawatt orbital data center. “Cooling is by far the biggest challenge we are solving,” Johnston told the “Thinking On Paper” podcast in March 2025. With no convection or conduction, “everything needs to be dissipated with infrared, like black body radiation,” which drives “very large, low-cost, low-mass deployable radiators.”
The vacuum of space may be rich with sunlight, but it’s also flooded with radiation. High-energy particles from the sun and cosmic rays can fry conventional computer chips within months. To survive, systems must rely on radiation-hardened hardware, which is vastly more expensive and less powerful than Earth-based processors.
Axiom Space is experimenting with commercial off-the-shelf chips adapted for radiation tolerance based on lessons from human spaceflight. Other companies, such as Lonestar, have touted the potential benefits of housing data centers in lava tubes beneath the moon’s surface to shield them naturally.
Maintenance presents yet another hurdle. “Fixing problems in orbit is far from straightforward,” Dr. Domenico Vicinanza, a data science expert at Anglia Ruskin University in the United Kingdom, warned in an April 2025 BBC article. “Even with robotics and automation, a big hardware failure might require a human mission, which could mean downtime for weeks or months.”
Space debris is another hazard that could sideline transmissions. More than 40,000 tracked objects already orbit Earth. A collision with even a tiny fragment traveling at over 27,000 kph could destroy a satellite and trigger an expanding cascade of debris — known as the Kessler Syndrome — that renders entire orbits unusable.
Security, too, is double-edged. While orbital data centers are difficult to hack physically, they could become prime targets for counterspace technologies — satellite jammers, lasers or even anti-satellite missiles developed by adversaries such as the Chinese Communist Party (CCP) and Russia. In an era of geopolitical tension, storing sensitive data on orbit might provoke more risk than it solves.
Environmental Paradox
Advocates argue that space-based data centers could dramatically reduce Earth’s carbon footprint. Continuous solar power, no water use and zero land consumption sound like an environmental win. But the reality is more complicated. Rocket launches remain carbon-intensive, releasing soot and aluminum oxides into the upper atmosphere, where they linger for years. Europe’s ASCEND study concluded that space data centers will only be eco-friendly if rocket emissions per launch fall tenfold from current levels — a goal still years away.
“We need to make sure that a suitable launch solution can be developed and that the overall structure is as light as possible to minimize the carbon footprint of launch operations,” Damien Dumestier, ASCEND project architect at Thales Alenia Space, said on the company’s website. “Optimizing payload weight and volume will also be a key challenge.”
A Quiet Space Race
Behind the technical debate lies a strategic one. Data sovereignty — who controls and processes data — has become a geopolitical flashpoint. The European Union’s ASCEND program explicitly frames orbital data centers as a path toward “digital sovereignty,” freeing Europe from reliance on U.S. hyperscalers like Amazon Web Services, Google Cloud and Microsoft Azure. For its part, the U.S. has likewise remained guarded about sovereignty. “U.S. federal law prohibits U.S. data being stored overseas. It’s all about jurisdiction and control,” Lonestar’s Stott told Apogee. “There is no ‘perfect’ place to store data on the planet for resiliency purposes.”
The CCP, meanwhile, already has launched a dozen satellites as part of a proposed 2,800-satellite “space computing” constellation. Analysts view it as both a commercial and military play: The ability to process Earth-observation data on orbit could offer tactical advantages in surveillance and defense.
The economics of space data centers mirror those of early commercial aviation: high costs, enormous risks and transformative potential. Venture capital is pouring in. Andreessen Horowitz, Sequoia Capital and even In-Q-Tel — the CIA’s venture arm — have backed Starcloud.
Former Google CEO Eric Schmidt has invested in launch startup Relativity Space to lower costs for orbital infrastructure. And Bezos’ Blue Origin is developing next-generation reusable rockets that could haul multigigawatt data centers into orbit by the 2040s.
Axiom Space’s Aspiotis echoes those views. “There is a point in the not-too-distant future where data centers in space are as economical as they are on the ground,” he told the MIT Technology Review. “In which case, do we want them on the ground, where they are consuming power, water and other kinds of utilities, including real estate?”
Despite the excitement, experts urge caution. Quentin Parker, director of the Laboratory for Space Research at the University of Hong Kong, says that off-world data center supporters may be “overselling the idea.” He notes that space brings its own risks, including radiation, space junk and the consequences of sending more human-made objects into orbit without a solution to the debris problem. Other issues include solar flares that could disrupt data transmissions and the challenge — and expense — of making repairs in space. In addition, some nations are developing jamming technologies that can target satellites and kinetic weapons to take out spacecraft altogether, and in the process cause more debris.
“The terrestrial solutions are still there, and they’re still probably a lot cheaper than trying to put anything into space,” Parker told CNN in October 2025. “Putting them into space has all sorts of problems associated with it.”
Others question the timeline. The ASCEND team projects commercial viability before 2037, assuming breakthroughs in launch efficiency, orbital assembly and cooling. Even Bezos and Musk see a 10- to 20-year horizon before space-based data centers beat their earthly counterparts.
For now, orbital computing remains an experiment — one that blends equal parts engineering ingenuity, financial speculation and planetary necessity. Still, history suggests today’s moonshots often become tomorrow’s infrastructure. In the 1950s, satellites were fantasy. By the 1980s, they were indispensable. The same could happen with data. As the appetite for AI grows, the case for orbital data centers will only strengthen. Unlimited solar power, untapped real estate and a near-perfect vacuum make space the ultimate clean room for computation.
But it’s also a gamble — on physics, economics and humanity’s ability to manage the heavens responsibly. “Space is hard, and yes, there are technical challenges, but they are all solvable at a cost that is either equal to or lower than the cost on Earth,” Johnston told investors at the FII in Riyadh, adding that breakthroughs around laser communication, orbital debris mitigation and robotic maintenance are turning space-based data centers from a concept into a tangible opportunity. “For the first time, the technology really is there to make data centers in space our reality.”