Artificial intelligence is pushing global electricity demand to levels that existing power grids were never designed to handle, and a small but growing number of researchers and policymakers are asking whether the answer lies not on Earth but in orbit. Data centers worldwide consumed roughly 415 terawatt-hours of electricity in 2024, and projections suggest that figure could more than double by the end of the decade. Against that backdrop, space-based solar power has re-entered serious policy discussion as a possible way to deliver clean, constant electricity to energy-hungry AI data centers, with NASA publishing a formal assessment of whether orbiting solar arrays could eventually beam power to the surface.
AI’s Appetite Is Outpacing the Grid
The scale of the problem is no longer speculative. Global data-centre electricity consumption hit approximately 415 TWh in 2024, accounting for about 1.5% of total worldwide electricity use, according to the International Energy Agency. Under the IEA’s base-case scenario, that demand is projected to climb to roughly 945 TWh by 2030, driven largely by the computational intensity of training and running AI models. To put that in perspective, 945 TWh would rival the total annual electricity consumption of Japan and would arrive on top of electrification pushes in transport and heavy industry that are already stretching grids in North America, Europe, and parts of Asia.
The strain is especially acute in the United States. Data centers consumed approximately 4.4% of U.S. electricity in 2023, and the U.S. Department of Energy projects that share could reach 6.7% to 12% by 2028. That range translates to between 325 and 580 TWh annually, a jump so steep it has drawn cabinet-level attention. The DOE report frames the surge as both an infrastructure challenge and a climate risk, because meeting that demand with fossil fuels would undercut federal decarbonization targets. For ordinary ratepayers, the consequences are tangible: utilities in data-center-heavy regions are already requesting rate increases and delaying coal plant retirements to keep the lights on, while local permitting battles over new transmission lines and substations are becoming more contentious.
Why Terrestrial Renewables Cannot Close the Gap Alone
Ground-based solar and wind are expanding rapidly, but they face hard physical limits that make them an incomplete answer to AI’s always-on power needs. Solar panels produce nothing at night, and their output falls sharply on cloudy days or during winter at higher latitudes. Wind farms depend on weather patterns that shift seasonally and can go calm for days at a time. Battery storage can bridge short gaps, but storing hundreds of terawatt-hours worth of energy for weeks or months remains prohibitively expensive with current lithium-ion technology. Data centers, by contrast, run around the clock and cannot tolerate brownouts without risking service failures for millions of users. The mismatch between intermittent generation and constant demand is the central tension that has revived interest in alternatives once considered science fiction.
Nuclear power is one candidate, and several tech companies have signed agreements with reactor operators, including exploratory deals around advanced small modular reactors. Yet new nuclear plants take a decade or more to permit and build in the United States, a timeline that does not match the pace at which AI workloads are growing. Geothermal and hydropower are geographically constrained and often face local opposition or environmental limits. Each of these options will play a role, but none individually can absorb the projected doubling of data-center load within five to six years. That gap is precisely what makes space-based solar power worth examining seriously, even if the technology is far from ready and would need to complement, rather than replace, aggressive deployment of terrestrial renewables and efficiency measures.
NASA’s Case for Orbital Solar Arrays
NASA’s Office of Technology, Policy, and Strategy published a dedicated space-based solar power assessment that evaluates whether orbiting collectors could become cost-competitive and climate-relevant by mid-century. The concept is straightforward in principle: satellites in geostationary orbit would capture sunlight 24 hours a day, unobstructed by clouds or nightfall, and beam the energy to ground receivers using microwave or laser transmission. The report finds the idea technically viable but identifies steep capability gaps, including autonomous in-space assembly of structures far larger than the International Space Station, efficient long-distance power beaming, and launch costs that would need to fall by at least an order of magnitude. NASA has been using its broader science communications platforms to explain how such ambitious infrastructure might fit alongside more traditional space missions.
What makes the NASA analysis relevant to AI specifically is the reliability argument. A single orbital array in geostationary orbit could deliver power continuously, eliminating the intermittency problem that plagues terrestrial renewables. For a hyperscale data center running thousands of GPUs on AI training jobs, even brief power interruptions can waste millions of dollars in lost compute time and delay product launches. An orbital source that operates independent of weather, season, and time zone would, in theory, match the always-on profile of AI workloads better than any ground-based alternative. The report stops short of recommending a crash development program but frames the technology as worth sustained investment, particularly as Earth observation missions refine climate models and launch economics continue to evolve under commercial pressure.
The Hard Realities Standing in the Way
Despite the appeal, space-based solar power faces obstacles that no amount of enthusiasm can wish away. The NASA report’s lifecycle cost and emissions comparisons show that current launch vehicles would make orbital solar far more expensive per kilowatt-hour than terrestrial wind or solar paired with storage. SpaceX’s Starship and other next-generation rockets could change that math, but they have not yet demonstrated the flight cadence or payload capacity needed to loft power-station-scale hardware. Autonomous robotic assembly in orbit, another prerequisite, has been demonstrated only in small-scale experiments. Scaling those techniques to structures spanning kilometers is an engineering challenge with no guaranteed timeline, and failures in orbit would be costly and difficult to repair.
There is also a regulatory and safety dimension. Beaming gigawatts of microwave energy from orbit to a ground rectenna raises questions about spectrum allocation, aviation safety, and land use that no international framework currently addresses. National security agencies would likely scrutinize any system capable of directing concentrated energy beams, even if they are designed within strict safety margins. The ARPA-E program office has funded exploratory energy research across many domains, but specific large-scale funding for space-based solar hardware tied to AI demand has not been publicly announced. Meanwhile, the Department of Energy’s own national laboratory network and its grid modeling efforts continue to focus primarily on terrestrial infrastructure and efficiency, reflecting where federal research dollars are concentrated today.
Bridging the Gap Between Vision and Policy
For now, the most immediate response to AI-driven electricity demand is likely to be a mix of conventional grid upgrades, efficiency improvements, and incremental innovation rather than a leap to orbital power stations. The Department of Energy and utilities are using tools such as the GENESIS platform to simulate how new data centers, renewables, and transmission lines interact under different scenarios, helping planners avoid bottlenecks and blackouts. At the same time, researchers are mining the federal research archive for past studies on wireless power transmission, high-efficiency photovoltaics, and autonomous robotics that could inform a future space-based solar program. These efforts are incremental, but they lay the groundwork for more ambitious projects if political and economic conditions shift.
On the financing side, policymakers are experimenting with new ways to align private capital with long-term infrastructure needs. The Department of Energy’s infrastructure exchange is one example of how federal support can de-risk large, capital-intensive projects in clean energy and grid modernization. While these mechanisms currently prioritize terrestrial projects such as transmission corridors, storage, and industrial decarbonization, they could, in principle, be adapted to support early demonstrations of space-based solar power if the technology matures. In that sense, the debate over orbital solar is less about replacing existing climate strategies and more about expanding the option set for a world where AI, electrification, and climate resilience are converging on the same stressed grids.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
