Meta plans to power part of its massive artificial intelligence ambitions using solar panels in outer space.
The tech giant announced today that it has placed a pre-order for up to a gigawatt of space-based solar to power future data centers here on Earth. The “capacity reservation agreement” puts Meta first in line when — and, crucially, if — Overview Energy’s satellite-mounted panels are fully operational. The company anticipates that will be sometime in 2030.
Overview, a Virginia-based startup that emerged from stealth last year, takes a unique approach to space-based solar. It plans to beam energy down in the form of near-infrared light to existing solar farms on the ground, maximizing the output of existing infrastructure.
Details on the startup’s agreement with Meta are sparse so far. The hyperscaler declined to comment on investment terms, including whether the tech giant has made any upfront financial investments or what it anticipates the eventual cost-per-megawatt of the power to be beamed down will be. As far as whether Meta is including this gigawatt of space-based solar in its data center power generation plans, the company said it is continuing to evaluate how that energy will be deployed.
This is a departure from the predominant approach to space solar. Both universities and government research institutions are pursuing the idea of a solar satellite, in which panels in orbit fully convert sunlight into microwaves or lasers, then transmit that energy to ground-based receivers. In Overview’s case, the solar farms themselves become the receivers.
It’s a long shot. The promise of space-enabled 24/7 solar has long faced skepticism, in large part because shooting solar panels into space and maintaining them for several decades is very, very expensive. Overview faces most of the same core hurdle as its space solar peers: namely, that given the high costs of operating in space, the process of converting energy into infrared light and back again would need to be very efficient to have a hope of competing with increasingly cheap grid-scale batteries on the ground.
The company says it plans to launch a satellite into low orbit by 2028 — far lower than the 36,000 kilometers above the planet that it eventually plans to operate from.
The space-based compute comparison
Despite the hurdles, the deal is emblematic of the industry’s growing enthusiasm for space-based solar. That’s in part because it’s getting cheaper to get infrastructure into space — but it’s also thanks to the AI boom. Much of the attention in recent months has focused on putting compute power itself into space alongside solar.
Elon Musk’s SpaceX is all-in on this approach, after acquiring xAI last year. The company has filed with the FCC to deploy satellites for that very purpose, which it plans to launch later this year. Google’s Project Suncatcher, focused on scaling machine learning in space, is planning to launch prototype satellites by early next year.
The immense cost of all this may have been harder to justify before the AI era. The original concept was a decarbonization gambit, but it is not included in any of the net zero pathways considered by the Intergovernmental Panel on Climate Change, ostensibly due to the tech and cost barriers.
Now, though, tech companies are spending hundreds of billions of dollars every year on speed to power for the proliferation of data centers they need. Caroline Golin, host of the Open Circuit podcast and former global head of energy at Google, said on a recent episode that the sheer quantity of available capital is changing the landscape of what’s possible. But the push, she added, probably isn’t really about energy availability here on Earth: “I think the real motivation is less about harnessing the sun’s power and avoiding the power crunch on Earth and more about getting out of the politics of completely transitioning this economy.”
And motivations aside, significant barriers remain, Golin added: “I would argue that the one thing we aren’t really talking about is the materials that would be needed to do that and how you’re going to extract and refine and process those materials. That, to me, is the sleeping Achilles heel.”
