Space Data Centers: Four Engineering Hurdles to Orbital Computing

SpaceX filed an application with the US Federal Communications Commission in January to place up to one million data centers in Earth's orbit — a move that has forced engineers and policymakers to confront just how far the technology needs to travel before space-based computing becomes operational.

The filing, made by Elon Musk's SpaceX, follows years of quietly building interest among hyperscalers and defence contractors in moving computing infrastructure off-planet. Proponents argue that orbital data centers could reduce latency for global users, sidestep land and water constraints that limit terrestrial construction, and leverage the near-unlimited solar energy available in space. But according to a technical analysis published by MIT Technology Review, four foundational problems must be solved first — and none of them are straightforward.

Power Without a Grid

The first challenge is energy. Terrestrial data centers draw from established electrical grids; in orbit, every watt must come from onboard solar panels. Space does offer one genuine advantage here: solar irradiance in low Earth orbit is roughly 30% stronger than on Earth's surface, and there is no weather to interrupt generation. But converting that energy efficiently, storing it during orbital night periods, and transmitting it within the facility without enormous losses remains a significant engineering problem. Current solar panel technology and battery storage systems are not designed for the scale that even a modest data center demands.

Space offers abundant solar energy, but converting, storing, and distributing it at data-center scale is an unsolved engineering problem.

The second challenge compounds the first: heat. On Earth, data centers rely on air cooling, liquid cooling, or proximity to cold water sources to manage the enormous thermal load generated by servers running at capacity. In space, there is no air and no convective cooling whatsoever. Heat can only leave a spacecraft through radiation — a far slower and less efficient process. Any orbital facility would require purpose-built radiator arrays, potentially vast ones, dramatically increasing the size, weight, and cost of each unit launched.

Getting the Data Down Fast Enough

The third problem is connectivity. A data center is only useful if data can move to and from it quickly. Terrestrial facilities connect via fiber optic cables capable of moving data at extraordinary speeds with minimal latency. An orbital data center would depend on laser-based optical links — technology that SpaceX has already deployed across its Starlink constellation, but not yet at the bandwidth densities that hyperscale computing demands. Atmospheric interference, orbital geometry, and the sheer number of simultaneous connections required would all stress current laser communication systems well beyond their tested limits.

The fourth and perhaps least-discussed challenge is radiation. Earth's magnetic field shields ground-level electronics from the worst of the solar and cosmic radiation environment. In orbit, that protection is substantially reduced. Consumer and commercial-grade processors — the kind that fill today's data centers — degrade and fail far faster when exposed to high-energy particles. Radiation-hardened chips exist, but they are expensive, less powerful, and produced in limited quantities compared to the mainstream silicon that makes large-scale computing economically feasible.

The Economics Behind the Ambition

Understanding why this idea is attracting serious attention requires stepping back from the engineering. Terrestrial data center construction is running into hard physical limits. Water scarcity is already forcing some local governments to restrict new builds; suitable land near dense population centres is increasingly scarce; and the energy demands of AI workloads have pushed grid operators in several US states to warn of capacity shortfalls. Against that backdrop, orbit starts to look less exotic and more like a legitimate, if distant, alternative.

SpaceX's FCC application does not commit the company to an immediate build — regulatory filings of this kind often function as placeholders or signals of intent rather than firm deployment plans. The company has not disclosed financial projections or a construction timeline associated with the proposal, according to available public records. However, the application does indicate that SpaceX is thinking about orbital infrastructure at a scale — one million units — that would dwarf anything currently in orbit by several orders of magnitude.

Other players are circling the space. Microsoft has previously explored the concept of underwater data centers through its Project Natick programme, demonstrating appetite for unconventional deployment environments. Several defence-oriented startups have received funding to explore space-based edge computing for military applications, though none have announced commercial-scale orbital data center plans publicly.

Radiation, Reliability, and the Road to Commercialisation

For orbital data centers to move from concept to commercial product, the industry would need simultaneous advances across at least four distinct engineering disciplines — power systems, thermal management, optical communications, and radiation-hardened semiconductors. Historically, waiting for multiple independent technology tracks to mature simultaneously is one of the slowest paths in engineering.

That does not make it impossible. The cost of launching mass to orbit has dropped dramatically over the past decade, largely driven by SpaceX's own reusable rocket programme. Falcon 9 and Starship have reduced per-kilogram launch costs to a fraction of what they were in 2010, and further reductions are expected as Starship reaches full operational capacity. Cheaper launches do not solve the thermal or radiation problems, but they do change the economic equation enough to make serious engineering investment worthwhile.

The timeline most frequently cited by researchers working in adjacent fields places viable small-scale orbital computing demonstrations somewhere in the early-to-mid 2030s, with commercial-scale operations — if the engineering problems are solved — potentially following later in that decade.

What This Means

For businesses planning data infrastructure over a ten-to-fifteen year horizon, orbital computing is no longer purely speculative — but the four unsolved engineering challenges identified by MIT Technology Review mean that terrestrial alternatives will remain the only practical option for the foreseeable future.