The Growing Energy Demands of AI
As artificial intelligence (AI) models grow more complex and power-hungry, the question is shifting from whether they can be trained to where they should be trained. This has prompted researchers at Google to propose an ambitious idea: building data centers in space. Known as “Project Suncatcher,” the initiative explores whether space-based infrastructure could help mitigate the soaring energy demands of AI systems.
Outlining their vision in a study published on the arXiv preprint server on Nov. 22, Google researchers suggest deploying constellations of satellites equipped with AI accelerators and powered by solar energy. These satellites would operate in low Earth or sun-synchronous orbits, where solar panels could function almost continuously, avoiding many of the limitations terrestrial data centers face, such as night-day cycles, atmospheric interference, and grid constraints. Additionally, space-based systems could reject heat via radiative cooling, eliminating the need for water-intensive cooling methods on Earth.
Why the Interest in Space-Based Solutions?
Data centers already consume a significant portion of global electricity. In 2024, worldwide data center energy usage is projected to reach around 415 terawatt-hours—roughly 1.5% of total global consumption. As AI use surges, these figures are expected to more than double by 2030. In the United States alone, data centers could account for up to 12% of regional electricity demand by 2028, leading some utility executives to warn that current infrastructure cannot keep up without major expansions.
Given this backdrop, space-based data centers are attracting attention not just as a novel idea, but as a possible necessity. With Earth-bound energy and cooling capacity nearing their limits, moving computation off-planet is beginning to look less like science fiction and more like pragmatic innovation.
Technical and Practical Challenges
Despite the promising theory, many experts are skeptical. Joe Morgan, COO of Patmos, a data center infrastructure company, is blunt in his assessment. “What won’t happen in 2026 is the whole ‘data centers in space’ thing,” he said. “One of the tech billionaires might actually get close to doing it, but aside from bragging rights, why?”
Morgan points out that the industry has previously tested extreme cooling solutions—from mineral oil immersion to underwater data centers—only to abandon them due to operational challenges. He emphasizes that hardware in data centers must be frequently upgraded or replaced, a task far more complicated in orbit where every repair requires a costly and logistically complex space mission.
Latency is another major issue. Most AI systems rely on high-speed communication between servers. While Google’s plan includes laser-based inter-satellite links to replicate these connections, latency in even low Earth orbit remains an unavoidable constraint. “Putting the servers in orbit is a stupid idea, unless your customers are also in orbit,” Morgan added.
Arguments in Favor of the Concept
Not everyone is ready to dismiss the idea. Paul Kostek, a senior IEEE member and systems engineer at Air Direct Solutions, believes the proposal is worth exploring. “The interest in placing data centers in space has grown as the cost of building centers on Earth keeps increasing,” he said. Advantages include continuous solar power and the ability to expel excess heat directly into space, avoiding the environmental impact of terrestrial cooling systems.
Heat rejection remains one of the most significant barriers to computational scaling, and Earth-based facilities are increasingly strained by water scarcity and environmental opposition. For example, residents near the xAI data center in Memphis have reported health issues and poor air quality, raising concerns about the local impact of massive data installations. Similar opposition is growing across the U.S., with critics citing risks to air and water quality as reasons to halt new data center projects.
Engineering Hurdles and Future Vision
Kostek acknowledges that many technical questions remain unanswered. “Can the processors used in Earth-based data centers survive in space? Will they withstand solar storms or heightened radiation?” he asked. Google has begun testing its Tensor Processing Units (TPUs) for space viability and is modeling satellite network configurations to support distributed computing. Still, these efforts are in the early stages, and operational deployment is years away.
For some experts, the most compelling use case isn’t solving Earth’s energy crisis, but building infrastructure for a space-faring civilization. Christophe Bosquillon of the Moon Village Association argues that space-based data centers could serve as the backbone for lunar industries and the cis-lunar economy. “With humanity on track to establish a permanent lunar presence, we need infrastructure capable of supporting data-intensive operations in space,” he said.
From this vantage point, Project Suncatcher is not merely a solution to current terrestrial limitations but a step toward enabling future space exploration and development. Such infrastructure could handle non-latency-sensitive workloads, freeing up Earth-based resources and even serving as secure storage for critical data.
Conclusion: A Long-Term Bet
While space-based data centers remain a speculative concept, the idea reflects deeper concerns about the sustainability of AI’s rapid expansion. As global demand for computational power reaches planetary scales, researchers must consider all environments—however harsh or remote—where energy is abundant. Whether or not Project Suncatcher ever launches, it forces an important conversation about the future of computation in an energy-constrained world.
This article is inspired by content from Original Source. It has been rephrased for originality. Images are credited to the original source.