A new preprint (arXiv:2603.28829v1, not yet peer-reviewed) models the astronomical, technical, and cultural consequences of placing large data centers in sun-synchronous low-Earth orbit to meet exploding AI power demand. The authors use orbital mechanics, photometric calculations, and basic collision-risk modeling rather than observational data or large-scale samples. They conclude that a single 5 GW facility would require a 4 km × 4 km solar array at roughly 500 km altitude. This structure would span 0.4 degrees of sky—comparable to the Moon—and shine at magnitude g = -5 to -7, roughly 100 times brighter than the brightest stars. Dozens of such platforms would form visible north-south chains for about three hours daily around twilight, blocking stars and increasing collision probabilities that could trigger debris cascades.

Mainstream technology coverage has largely celebrated the concept: unlimited solar power without terrestrial cooling demands or grid constraints. What it has missed is the full suite of environmental, thermal, and orbital externalities. Similar to how SpaceX’s Starlink constellation triggered widespread complaints from astronomers (documented in a 2020 Astronomy & Astrophysics study on satellite trails contaminating observations), these far larger structures would create persistent, moving glare that disrupts both professional telescopes and cultural experience of the night sky. The preprint’s brightness estimates rely on assumed albedo and geometry; real-world specular reflections or attitude adjustments could make them even more intrusive.

Thermal impacts receive less attention in the source but are critical. Earth-based data centers consume enormous energy and water for cooling. In vacuum, waste heat must be radiated via large panels, increasing the overall cross-section and making the platforms even more visible and collision-prone. This heat radiation could also produce subtle upper-atmospheric effects not modeled here. Orbitally, the proposal accelerates the industrialization of low-Earth orbit already underway with thousands of communication satellites. Patterns seen in recent mega-constellations—rapid deployment followed by belated mitigation—suggest that astronomical and sustainability concerns are treated as secondary. A 2021 Nature Astronomy perspective on space environmentalism warned that without deliberate governance, low-Earth orbit risks becoming a polluted commons, exactly the trajectory these data-center platforms would reinforce.

The preprint’s limitations are clear: it is a theoretical desk study without prototype testing, real reflectivity measurements, or detailed debris-propagation simulations. Yet its core warning is robust. Projections of U.S. data-center demand exceeding 100 GW by 2035 and global demand nearing 1 TW make the temptation to move compute to space understandable. However, this merely displaces environmental burden from Earth’s surface to the shared orbital and sky environment. Sustainable computing cannot be achieved by exporting problems to a domain with even weaker governance. The night sky is cultural heritage and a scientific instrument; treating it as industrial real estate risks irreversible damage to both astronomy and the long-term viability of space operations. Genuine solutions likely require rethinking AI energy efficiency, terrestrial renewable integration, and stricter orbital traffic rules rather than assuming space can absorb unlimited growth.