Can data centers leave earth?

Artificial intelligence is driving an unprecedented expansion in digital infrastructure. The United States alone hosts 5,427 data centers, more than 10 times as many as any other country, according to the 2025 Stanford AI Index, while the U.S. Department of Energy projects that electricity demand from U.S. data centers could nearly triple by 2028. The challenge extends beyond the United States. Governments from Europe to the Gulf and East Asia are racing to expand computing capacity without overwhelming electricity grids, land and water resources. As AI models grow larger, attention is shifting from algorithms to the infrastructure needed to power them.

That search has brought new attention to an idea that once seemed far removed from commercial reality. Researchers, space agencies, and private companies are now exploring whether some future data centers could operate in Earth orbit, powered by abundant solar energy and connected to satellites and ground stations through high-speed communications systems. The concept remains in its infancy, and formidable technical and economic barriers remain. Yet growing investment from both the public and private sectors suggests that computing in space is evolving from a theoretical concept into an active area of research, raising new questions about where the next generation of digital infrastructure might eventually be built.

Why now?

The prospect of placing data centers in orbit is not new, but high launch costs and limited commercial demand kept it largely impractical. Today, several technological and economic trends are changing that equation.

The rapid commercialization of space has lowered the cost of reaching orbit through reusable launch vehicles, while satellite constellations have expanded dramatically in both size and capability. At the same time, artificial intelligence has become one of the world's fastest-growing sources of demand for computing power, prompting governments and technology companies to explore new ways of expanding digital infrastructure beyond traditional terrestrial facilities.

Researchers also point to several characteristics that make space an attractive environment for certain computing applications. Satellites in orbit have access to abundant solar energy for much of their operational life, while the cold environment of space could help dissipate the large amounts of heat generated by high-performance computing systems.

Processing data where it is created

The strongest argument for computing in orbit is not that every cloud server should be launched into space. Rather, researchers see the greatest potential in processing information closer to where it is generated.

Today, Earth observation satellites generate enormous volumes of imagery and sensor data that are typically transmitted to ground stations before analysis, consuming bandwidth and introducing delays. As satellite constellations grow, that approach is becoming more difficult to sustain.

The European Space Agency has been among the organizations examining an alternative. Through its Cognitive Cloud Computing initiative and broader Space Cloud research, ESA is exploring whether artificial intelligence could process data directly aboard satellites or within orbital computing networks before transmitting only the most relevant information to Earth.

Rather than sending every image captured by an Earth observation satellite to a terrestrial data center, onboard AI could identify wildfires, floods, methane leaks or other significant events in orbit, dramatically reducing the amount of data that needs to be transmitted. ESA has also examined applications including autonomous lunar exploration, space traffic management, in-orbit robotics, satellite health monitoring and communications between future lunar missions.

According to ESA, processing information in orbit could improve response times while making more efficient use of limited satellite communications capacity by allowing satellites to share data and computing resources before transmitting only the most relevant information to Earth.

From concept to commercial development

Research is now being matched by growing private investment.

Houston-based commercial space company Axiom Space announced in January that it had launched two orbital data center nodes into low Earth orbit following earlier demonstrations aboard the International Space Station. The company says the system is designed to test technologies for future cloud computing, AI and cybersecurity applications in orbit.

Another company attracting attention is Starcloud, a California startup developing computing infrastructure specifically for AI workloads in orbit. Earlier this year, the company raised $170 million at a $1.1 billion valuation, underscoring growing investor interest in orbital computing.

Starcloud's first demonstration satellite carried Nvidia's H100 graphics processing unit, one of the processors widely used to train and operate today's most advanced artificial intelligence models. Testing the chip in orbit allows engineers to evaluate how commercial AI hardware performs under the harsh conditions of space, including exposure to radiation, temperature extremes and prolonged autonomous operation.

The company's next planned mission is expected to include Amazon Web Services' AWS Outposts, extending the experiments beyond individual processors toward cloud computing infrastructure capable of managing data and applications in orbit. Together, the demonstrations represent an effort to determine whether technologies that underpin modern terrestrial data centers can be adapted for use beyond Earth.

The engineering challenges remain formidable

Yet the same environment that makes orbit attractive creates new engineering constraints. While abundant solar energy could power future facilities, researchers note that supporting AI workloads would require solar arrays far larger than those used today. Cooling also remains difficult because spacecraft cannot dissipate heat like terrestrial data centers, while communications between Earth and orbit remain far more limited than terrestrial fiber networks.

According to the U.S. Government Accountability Office and the 2025 Nature Electronics study, overcoming these challenges will require significant advances in power generation, thermal management and communications before large-scale orbital data centers become commercially viable.

The practical realities of operating hardware in space present another layer of complexity. Servers must withstand radiation, micrometeoroids and years of continuous operation without the routine maintenance that keeps terrestrial data centers running. Replacing failed equipment or upgrading computing hardware would require servicing missions that remain technically demanding and costly.

Those obstacles mean orbital data centers remain an emerging technology rather than a commercial reality. Yet as artificial intelligence accelerates demand for computing infrastructure, governments, researchers and private companies are beginning to explore a question that would have seemed far-fetched only a few years ago. Could part of the world's digital infrastructure one day operate beyond Earth?