Data Centers in Space: A New Frontier for Artificial Intelligence Infrastructure
The U.S. Government Accountability Office’s April 2026 Science & Tech Spotlight, Data Centers in Space, provides a concise but important assessment of an emerging idea at the intersection of artificial intelligence, space systems, energy policy, and digital infrastructure. Giving credit to GAO and its Science, Technology Assessment, and Analytics team, including Karen L. Howard, PhD, Katrina Pekar-Carpenter, Nathan Hamm, Marie Armbruster, Jehan Chase, Mark Kuykendall, and Anika McMillon, the report examines whether data processing and storage systems could be placed in orbit to support AI, cloud computing, and space-generated data processing.
The report begins from a practical premise: terrestrial data centers require substantial land, electricity, cooling, and water. As artificial intelligence expands, these demands are expected to increase. GAO notes that the Department of Energy projects data centers could account for up to 12 percent of U.S. electrical demand by 2028, driven in large part by AI development. Space-based data centers are proposed as one potential response. Instead of locating servers, storage systems, and network equipment in facilities on Earth, these systems would be housed in satellites, most likely in low Earth orbit.
The concept is appealing because certain low Earth orbits could provide near-continuous access to solar energy, while space-based processing could reduce some pressure on terrestrial electrical grids, land use, and water consumption. The report’s figure comparing terrestrial and hypothetical in-space data centers illustrates this contrast. A large terrestrial data center may require 100 megawatts of power and more than 200,000 gallons of water per day, while a proposed satellite design would rely on large solar arrays and radiator surface area to support computing and cooling needs.
However, GAO is careful not to overstate the maturity of the technology. The underlying components—satellite power, communications, computing hardware, and thermal management—are familiar, but using them at the scale necessary for large data centers remains unproven. Smaller orbital systems designed to process data collected in space may be closer to practical deployment than large platforms intended to train AI models. The distinction matters because the power, cooling, and communications requirements for AI-scale computing are substantial.
The most significant engineering challenge may be heat. Although space is often thought of as cold, it does not cool computer hardware efficiently because heat does not disperse easily in the vacuum of space. Large data centers generate significant waste heat, and that heat must be radiated away to avoid damaging equipment. GAO notes that cooling solutions at this scale remain unproven, and the solar arrays required for major systems could be larger than any assembled in space as of April 2026.
The report also identifies broader policy and operational risks. Large constellations of data-center satellites could make low Earth orbit more crowded, increasing collision risks, interfering with astronomical research, and complicating management of radio frequencies. Computing in space also faces radiation risks that can corrupt data and degrade hardware. If these systems require frequent replacement or decommissioning, they could contribute to orbital debris or reentry risks.
For policymakers and contractors, the report’s importance lies in its framing. Space-based data centers are not science fiction, but neither are they near-term substitutes for terrestrial infrastructure. They are an emerging technology with potential benefits, especially for processing space-generated data and reducing some terrestrial resource burdens. Yet their deployment will require advances in launch economics, thermal control, satellite servicing, communications, orbital governance, and international legal frameworks. GAO’s central contribution is to place this technology in the proper policy context: promising, technically plausible, but not yet proven at scale.
Disclaimer:
This article is for general informational purposes only and does not constitute legal, technical, engineering, procurement, or policy advice. Readers should consult qualified professionals regarding specific technology, contracting, regulatory, or investment decisions.