SpaceX and Google Race to Build Data Centers in Orbit

As local governments impose restrictions on ground-based data centers from New York to Florida, major technology companies are pursuing an unconventional solution: building computing infrastructure in space.

SpaceX, Google, and several startups are actively developing orbital data centers, with Google planning prototype satellite launches in 2027 and SpaceX filing applications for up to one million satellites. Startup Starcloud has already deployed an Nvidia H100 GPU to orbit, where it is processing AI workloads, according to Quartz.

The energy equation

The primary attraction of space-based data centers centers on power generation. Solar panels operating in orbit produce five to eight times more electricity than equivalent ground installations, benefiting from constant sunlight without atmospheric interference or nighttime interruption. For companies facing five-to-ten-year permitting timelines for new terrestrial power plants, this represents a significant advantage.

Yet the vacuum of space introduces substantial engineering challenges. Without air or water for cooling, orbital data centers must dissipate heat exclusively through large radiating panels. Cosmic radiation poses another risk, as high-energy particles can corrupt data by flipping bits in semiconductor chips.

The cost barrier

Economics present the most formidable obstacle. Andrew McCalip, head of research and development at space startup Varda, developed a model showing space infrastructure currently costs approximately four times more per watt than conventional data centers. The gap depends almost entirely on launch expenses.

SpaceX's Falcon 9 currently charges around $2,700 per kilogram to reach orbit. Google researchers estimate break-even requires costs near $200 per kilogram, while broader industry analysis suggests $500 per kilogram would suffice. Citigroup analysts project SpaceX's Starship could reach $100 per kilogram by 2040, with an optimistic scenario approaching $33 per kilogram.

However, terrestrial data center costs continue declining as well, driven by decades of engineering refinement and economies of scale.

Operational realities

Maintenance presents another complexity. Replacing a failed server on the ground takes minutes, while orbital repairs require launching an entirely new satellite. Matching the capacity of a single large terrestrial facility demands thousands of satellites, raising concerns about debris accumulation in already-congested low-Earth orbit.

Jeff Bezos, developing orbital computing capabilities through Blue Origin, recently characterized two-to-three-year deployment timelines from competitors as "a little ambitious."

Why it matters

Orbital data centers represent more than technical ambition—they reflect genuine constraints facing AI infrastructure expansion. Terrestrial facilities encounter mounting opposition over energy consumption and environmental impact, while AI companies require exponentially growing compute capacity. Whether space-based solutions prove viable depends on launch costs falling faster than ground-based alternatives improve. SpaceX's pending IPO, potentially the largest in history, positions orbital data centers as central to its investor pitch, adding financial urgency to technical development.

These details were first reported by Quartz in its AI & Tech newsletter.