Space data centres: Can orbiting AI infrastructure solve Earth’s computing crisis?

The artificial intelligence boom is creating a problem that few predicted at such scale: where will all the computing power come from? As AI models become larger and more computationally intensive, data centres are consuming unprecedented amounts of electricity, land, and water. The International Energy Agency (IEA) has warned that AI-driven computing is set to dramatically increase electricity demand from data centers over the coming decade. At the same time, many communities are increasingly resisting the construction of large-scale facilities due to concerns over energy consumption, water use, noise, and environmental impact.

This challenge has prompted an idea that sounds like science fiction but is increasingly being taken seriously by industry: building data centres in space. The concept is attracting attention because it appears to address some of the most pressing constraints facing terrestrial computing infrastructure. In orbit, solar energy is abundant, land is effectively unlimited, and operators would not need to compete with local communities for resources.

Why look ‘up’ to space?

Modern society depends on data centres. These facilities support cloud computing, banking, streaming services, scientific research, and increasingly AI platforms. The rapid growth in AI applications has dramatically increased demand for processing power. According to analysis from JPMorgan, AI workloads are accelerating demand for advanced computing infrastructure at a pace far beyond traditional digital services.

On Earth, this expansion comes with costs. Data centres require vast quantities of electricity and often consume substantial volumes of water for cooling. New facilities can also place pressure on local power grids, while residents frequently object to their environmental footprint. An orbital alternative promises a different approach. Solar panels in space receive uninterrupted sunlight for much of their orbit and are not affected by weather or seasonal fluctuations. Unlike terrestrial facilities, orbital data centres would not require valuable real estate, zoning approvals, or extensive local utility connections.

One of the greatest challenges facing terrestrial data centres is heat management. Almost every watt consumed by a server eventually becomes heat. If the heat cannot be removed efficiently, equipment performance declines and system failures rise.

At first glance, space appears to offer an elegant solution. The cosmic background temperature is close to absolute zero, creating the impression of an endless cooling reservoir. However, reality is considerably more complicated.

On Earth, cooling systems benefit from moving air and evaporative processes. In space, there is no atmosphere. Heat cannot simply be blown away using fans. Instead, it must be radiated away as infrared energy. This requires enormous radiator structures. Engineering studies suggest that removing around 10 megawatts of heat, roughly equivalent to a modest modern data centre, could require radiator surfaces approaching the size of two football fields.

As a result, space-based facilities may eliminate cooling towers and water consumption, but they replace them with large and complex thermal management systems. It also needs to be noted that orbital data centres would require frequent rocket launches, which create emissions and other environmental impacts. Large satellite constellations also raise concerns about orbital debris, atmospheric pollution from re-entry, ozone effects, and impacts on astronomy.

Surviving a hostile environment

Unlike a terrestrial server farm, an orbital data centre must operate in one of the most hostile environments imaginable. Radiation presents a continual threat to electronic systems. Cosmic rays and solar particles can damage chips, corrupt data, and reduce equipment lifespans.

Temperature fluctuations are also severe. Depending on orbital position, components may repeatedly cycle between intense solar heating and extreme cold several times each day.

Micrometeorites and orbital debris represent another significant risk. A collision with even a relatively small object can disable equipment or generate additional debris that threatens other spacecraft.

The orbital environment is already becoming crowded. Thousands of additional satellites dedicated to computing could intensify concerns surrounding space traffic management and the long-term sustainability of near-Earth orbit.

Perhaps the greatest obstacle is not building orbital data centres, but maintaining them.

Today’s data center industry relies on regular hardware upgrades. Servers are commonly replaced every three to five years as processors improve and workloads evolve. In a terrestrial facility, technicians can simply remove outdated equipment and install modern replacements.

Any repair mission would be expensive, complex, and dependent upon robotic servicing capabilities or crewed operations. Equipment failures that would be routine on Earth could become major logistical exercises in orbit.

This raises a critical economic question. Computing technology evolves rapidly. If orbital infrastructure cannot be upgraded easily, operators may find that their systems become technologically obsolete long before the physical platform reaches the end of its useful life.

A communications challenge

Even if these facilities could operate effectively, they would still need to exchange vast quantities of information with Earth. Satellite constellations such as Starlink have demonstrated that high-speed communication from orbit is possible. However, transmitting the enormous data volumes required for mainstream cloud computing would require significant advances in both radio-frequency and laser communication technologies.

Many modern computing applications, including interactive AI services, financial trading systems, and cloud-based enterprise applications, depend upon minimal delays. Physical distance inevitably introduces transmission latency, making certain workloads less suitable for orbital deployment.

The most realistic near-term opportunity may not be replacing Earth-based hyperscale facilities but serving space-based customers. Potential applications include processing Earth observation imagery directly in orbit, supporting military and intelligence systems, powering scientific missions, and providing computational resources for satellite networks. By analysing data closer to where it is generated, operators could reduce transmission requirements and improve operational efficiency.

This approach aligns with recent industry developments. SpaceX has unveiled plans for its AI1 Compute Satellite, an early attempt to bring computing infrastructure into orbit. While its capabilities remain significantly smaller than those of leading terrestrial data centres, it represents a notable step toward testing the concept in practice.