Space-based solar, which comes with the promise of 24/7 sunlight, has occupied a niche corner of the energy and tech worlds for several decades. But given significant cost and efficiency barriers, the concept has remained stuck in the demonstration stage.
The data center boom and the resulting capacity crunch in the United States have changed the equation for many next-gen sources of electricity by introducing an apparently insatiable appetite for power from some of the most deep-pocketed players in the market. Hyperscalers are turning to everything from massive off-grid gas plants and nuclear restarts to fuel cells and enhanced geothermal in a bid to get as many electrons as they can, as fast as possible.
As a result, space-based solar may be poised for new momentum — not in the “traditional” sense of beaming energy back down to the ground, but rather co-locating solar power with space-based compute. Today, that momentum is largely concentrated around Elon Musk and SpaceX, though Google and Meta have both dipped a toe in as well.
According to two recent analyses by Bank of America and Boston Consulting Group, this approach of sending not energy but data back to earth via optical links is likely a better fit for the AI era, avoiding the cost and conversion losses associated with building gigawatt-scale arrays in space and then beaming microwave energy back to the ground. In the right orbit, satellites could access near-continuous sunlight, and potentially circumvent land or queue constraints on the ground.
Such data centers wouldn’t be a good fit for latency-sensitive applications, or for training large language models, because those tasks require fast response times and high densities of both chips and power that orbital systems would struggle to supply. But there are several more plausible use cases, including batch generation of documents, images, and videos; enterprise work; tagging and classification; and perhaps most straightforward, the processing of data already generated in space, like satellite observation imagery, climate monitoring, and disaster-response information.
However, even when co-located with space-based data centers, the cost challenge remains. Bank of America estimates that, at around $170 billion, a one-gigawatt orbital data center is still three times more expensive than a similarly sized facility on Earth.
That said, the market potential of getting space-based solar for data centers right, the analysts found, is immense. These projects could scale global solar cell sales from $60 billion in 2025 to $700 billion by 2035, capture as much as 15% of the global AI data center market by 2040, and reach $320 billion in annual revenue. That hypothetical, however, is heavily dependent on steep reductions in launch costs, major gains in satellite performance, and new approaches to thermal management.
Betting on SpaceX
Any movement would likely depend heavily on Elon Musk and SpaceX. That company has a rare combination of capabilities and access that could make it well-positioned to scale space-based solar and data centers: launch capability, satellite-manufacturing, an existing optical communications network via Starlink, and close ties to both AI chip and model development.
Earlier this year, SpaceX filed with the FCC for permission to deploy up to one million satellites, part of an orbital data center “constellation.” Dubbed “Starmind,” the group of satellites would be designed to carry up to 250 kilowatts of peak computing power.
Musk, for his part, estimates his company could eventually deploy up to 500 GW per year of those solar-powered AI satellites — roughly equal to around 60% of average U.S. electricity demand, according to BofA’s calculations.
Getting costs down will hinge on making it cheaper to launch the infrastructure itself into space. In BCG’s report, the consultancy estimates that at current launch costs, putting one GW of compute capacity in orbit would cost roughly $30 billion in launch expenses alone. BCG argues those costs need to drop from $1,500 per kilogram to around $100 per kilogram to make orbital computing economically plausible.
That said, if capacity were available in space, even at exorbitant prices, there may well be a market for it. After all, OpenAI and SoftBank’s planned 10-GW data center campus in Ohio, set to be powered by a whopping 9.2 GW of new gas-fired power, could cost as much as $500 billion.
Infrastructure challenges
In addition to getting infrastructure to space, there are also some significant technical changes that will need to be made.
Standard satellites rely on highly efficient but very expensive gallium-arsenide solar cells. Eventually, space-based co-located data centers will need to leverage perovskite-silicon tandem cells, which have more than six times the power-to-mass ratio of the gallium-arsenide cells, and which offer up to a 15-fold reduction in cost-per-watt. And cost is ultimately all about the mass: Every kilogram saved on the solar panels themselves would yield millions of dollars in launch savings, the report explains.
But perovskite-silicon cells are untested for large-scale use in space, and there’s a production hitch: BofA estimates China currently controls over 90% of global cell manufacturing, including 60% of perovskite capacity and 75% of perovskite patents. The country is roughly a year or two ahead of the U.S. in scaling perovskite production, the report added.
The U.S., meanwhile, has the edge when it comes to getting more materials into space more cheaply. SpaceX’s launch costs around $1,500 per kilogram, compared to the approximately $7,000 per kilogram spent on even the cheapest Chinese launches. And while SpaceX continues to reuse its boosters, Chinese companies are still testing recovery and reuse systems.
Scaling space-based solar and data centers is ultimately a broad industrial contest, the analysts conclude. It spans not just solar cells and satellite production, but laser communications, thermal management, and radiation-proof components. China is relying on its ability to produce lightweight solar hardware at volumes that no U.S. supplier can even come close to matching, while U.S. progress relies heavily on SpaceX and Musk’s ecosystem of companies being able to execute on his promises for massive-scale solar production, cheap and reusable Starship launches, and Starlink’s optical communications network.
