Computing Beyond the Atmosphere

The space economy is entering a data-centric era. In March 2026, the first cluster of 'Orbital Data Hubs' has reached full operational capacity in Low Earth Orbit (LEO). These aren't just communication satellites; they are high-performance computing nodes equipped with specialized AI chips designed for the vacuum of space.

Why Space?

By processing data in orbit, companies can drastically reduce the latency of global AI services. High-resolution satellite imagery, autonomous maritime routing, and planetary defense monitoring now happen 'at the edge'—thousand of miles above the surface. Furthermore, the natural cooling provided by the space environment offers a unique solution to the massive heat generated by earth-bound data centers.

The Lunar Gateway Expansion

As part of the broader Artemis mission infrastructure, these data hubs are also acting as critical relays for the growing number of lunar surface operations. As we look toward the end of the decade, the expansion of these networks is expected to form the backbone of a true 'Interplanetary Internet'.

The Latency Paradox and Who Benefits

The instinctive assumption about space-based computing is that physical distance from Earth increases communication latency, making it unsuitable for time-sensitive applications. The reality is more nuanced. For certain workloads — specifically, processing data that originates in space, such as Earth observation imagery, satellite telemetry, or inter-satellite communications — processing at the edge in orbit eliminates the latency and bandwidth cost of transmitting raw data to ground stations. A single high-resolution Earth observation satellite can generate terabytes of imagery per day; processing that data in orbit and transmitting only the analytically relevant results to ground stations reduces downlink bandwidth requirements by orders of magnitude and accelerates the delivery of actionable intelligence from days to hours.

Power and Cooling Challenges

The fundamental engineering constraint for orbital data centres is power and thermal management in the space environment. Solar panels provide intermittent power dependent on orbital position relative to the sun, creating duty cycles that must be managed carefully for compute-intensive workloads. Heat dissipation in vacuum requires radiative cooling rather than the convective cooling used in terrestrial data centres, making thermal management engineering significantly more complex and expensive per unit of compute. Current orbital data hub designs achieve roughly 20–50 kilowatts of compute capacity per orbital node, compared with megawatts in a terrestrial data centre, meaning that orbital computing is currently suited for preprocessing and edge inference rather than large-scale training workloads.

Why Anyone Is Seriously Pursuing This

Orbital data centres have two genuine advantages over terrestrial installations. The first is cooling: in space, excess heat can be radiated to the surrounding environment (which is effectively a heat sink at near absolute zero) without the water-intensive cooling towers or air conditioning systems that ground-based data centres require. A 2024 study from the European Space Agency estimated that a well-designed orbital data centre could operate with essentially zero water consumption and significantly lower cooling energy overhead.

The second advantage is solar power density. In low Earth orbit, solar panels receive approximately 1,360 watts per square metre of unobstructed sunlight — roughly 40% more than on the Earth's surface (where atmosphere, weather, and day-night cycles reduce effective irradiance). An orbital facility operating 24 hours a day on solar power would have a fundamentally different energy economics than any ground-based facility.

The Launch Cost Problem

Neither of these advantages currently justifies the launch costs. Getting one kilogram of payload to low Earth orbit costs approximately $2,000–$3,000 with SpaceX's Falcon 9 (a dramatic reduction from the $50,000+ per kilogram costs of the early 2000s, but still prohibitive for large-scale compute hardware). GPU clusters dense enough to be cost-competitive with ground-based facilities would require thousands of tons in orbit — economically impractical for at least another decade even with continued reductions in launch costs.

Starship, SpaceX's fully reusable next-generation rocket, has a target cost of $10–100 per kilogram in high-volume operation. If those targets are achieved, the economics of orbital data centres change significantly — which is why the serious players in this space (Axiom Space, Jeff Bezos's Blue Origin via AWS partnership discussions) are building models based on Starship-era launch economics rather than current costs.