Meta has reserved up to 1 gigawatt of solar power that would be collected in orbit and beamed to Earth, a deal that would mark the first time a major technology company has turned to space to feed the staggering electricity demands of artificial intelligence.
The agreement, announced in April 2026 with a startup called Overview Energy, calls for satellites in geosynchronous orbit, roughly 22,000 miles above the equator, to capture sunlight and transmit it downward as a low-intensity near-infrared beam. Ground-based solar arrays would convert that beam into grid-ready electricity for Meta’s data centers. Overview is targeting an orbital demonstration in 2028 and commercial-scale delivery by 2030.
Alongside the space solar reservation, Meta also locked in a parallel deal with Noon Energy for up to 1 GW and 100 gigawatt-hours of long-duration energy storage. Noon’s technology uses modular reversible solid oxide fuel cells that can discharge for 100 hours or more, far beyond the four-to-eight-hour window typical of lithium-ion batteries. A 25 MW pilot project is slated for completion by 2028.
Together, the two deals sketch out an energy strategy built for a specific problem: AI training runs and inference workloads that can consume enormous amounts of power around the clock, sometimes for days at a stretch during large model training cycles. Conventional solar and wind cannot deliver that kind of reliability on their own, and Meta appears to be betting that pairing orbital generation with ultra-long storage could close the gap.
Why space solar, and why now
The concept of harvesting solar energy in space and sending it to Earth has circulated in engineering circles since the 1960s, when NASA and the Department of Defense first studied the idea. The appeal is straightforward: a satellite in geosynchronous orbit sees sunlight for more than 99 percent of the year, unobstructed by clouds, nightfall, or seasons. That consistency dwarfs what any ground-based solar farm can achieve, even in the sunniest deserts.
But the concept has always stumbled on cost and complexity. Launching heavy hardware into orbit is expensive, and converting sunlight to a beam and back to electricity introduces efficiency losses at every step. What has changed recently is the sharp drop in launch costs driven by reusable rockets, combined with advances in photovoltaic and power-beaming technology. Overview Energy’s approach uses near-infrared light rather than microwaves, which the company says allows it to target existing ground solar infrastructure rather than building specialized receivers from scratch.
Overview built its case with a physical demonstration: transmitting power via a near-infrared beam from a moving aircraft to a ground receiver 5 kilometers below. The company has described this as the world’s first airborne power-beaming demonstration for space solar energy. During the test, a laser-based transmitter aboard the aircraft maintained a coherent beam on a receiver mounted on a moving vehicle, proving the system could hold alignment and deliver usable power while both platforms were in motion. That is a meaningful step beyond laboratory experiments, though it remains a long way from orbit.
The scale of the engineering gap
Five kilometers is not 36,000 kilometers. Scaling Overview’s airborne test to geosynchronous altitude introduces challenges the company has not yet publicly quantified. Beam divergence over tens of thousands of kilometers means a larger fraction of transmitted energy will miss the ground receiver. Atmospheric scattering from clouds, haze, and moisture will absorb or redirect portions of the beam. The satellite must maintain pinpoint alignment with a ground target while compensating for its own station-keeping maneuvers and the rotation of the Earth below.
No independent data on end-to-end conversion efficiency, from photon capture in orbit to usable electricity at a data center, has been published. Until Overview flies an orbital demonstrator and releases verified performance numbers, the 1 GW commercial target remains a projection rather than an engineering fact.
Noon Energy faces a different but still substantial set of unknowns. Solid oxide fuel cells are a proven technology in other contexts, but operating them as reversible, long-duration storage systems at the 25 MW scale Noon is proposing has not been done commercially. Questions about round-trip efficiency, cell degradation over thousands of charge-discharge cycles, and maintenance costs will only be answered once the pilot project is built and operated under real grid conditions.
Regulatory and economic unknowns
Beaming power from orbit to the ground is not something any existing regulatory framework was designed to handle. Overview would likely need clearance from multiple agencies. Spectrum coordination or authorization from the Federal Communications Commission could be required depending on how the near-infrared transmission is classified. Safety reviews would need to address whether the beam poses risks to aircraft, wildlife, or people, and whether it could interfere with existing communications or remote-sensing systems. None of the public announcements reference active regulatory filings or agency engagement, leaving a significant procedural gap between the stated timeline and actual operations.
The economics are equally opaque. Neither deal discloses pricing, capital expenditure estimates, or the financial structure of the capacity reservations. Whether space-based solar can compete on a levelized cost-of-energy basis with terrestrial alternatives, including ground solar paired with battery storage, is an open question that hinges on launch costs, satellite manufacturing scale, and operational reliability that have not been tested at commercial levels.
It is also worth noting what these agreements actually are. A “capacity reservation” is not a power purchase agreement with firm delivery obligations. The language suggests Meta has secured priority access to future output rather than contracted for guaranteed megawatt-hours. If either technology hits delays or fails to meet performance targets, the reservations may not translate into actual electricity on the timelines described.
How Meta’s bet fits the Big Tech energy race
Meta is not the only technology giant scrambling for clean power to run AI infrastructure. Microsoft signed a deal to restart a reactor at Three Mile Island to supply its data centers. Google has invested in geothermal energy through Fervo Energy and signed an agreement with Kairos Power for advanced nuclear reactors. Amazon has purchased a nuclear-powered data center campus in Pennsylvania and signed multiple nuclear power agreements.
What distinguishes Meta’s approach is the sheer novelty of the supply source. Nuclear restarts and geothermal wells extend proven technologies into new applications. Space-based solar, by contrast, has never delivered commercial power to anyone. The Overview deal is less a procurement contract and more a strategic option: Meta is paying to hold a place in line for a technology that, if it works, could provide nearly continuous carbon-free electricity from a source with no land-use footprint on Earth.
The Noon Energy storage deal serves a complementary purpose. Even if orbital solar eventually delivers on its promise, Meta’s data centers will need backup for the rare periods when beams are interrupted or demand spikes beyond what a single orbital source can supply. A 100-hour storage system could bridge gaps that lithium-ion batteries simply cannot cover, turning intermittent clean generation into something closer to baseload power.
What to watch next
The milestones that will determine whether these deals amount to anything tangible are specific and trackable. For Overview Energy, the critical test is the orbital demonstration planned for 2028. A successful satellite launch, beam transmission from geosynchronous orbit, and verified power delivery to a ground receiver would transform space-based solar from a concept into a demonstrated capability. Anything short of that will leave the 2030 commercial target looking aspirational.
For Noon Energy, the 25 MW pilot project carries similar weight. If the system achieves its 100-hour discharge duration with acceptable round-trip efficiency and proves durable over repeated cycles, it would validate a class of long-duration storage that the grid badly needs. If it falls short, the 1 GW reservation becomes a much harder sell.
For Meta, the broader question is whether these early-stage bets can mature fast enough to matter. The company’s AI ambitions are growing now, and its data centers need power now. Space solar and 100-hour storage are, at best, late-decade solutions. In the meantime, Meta will have to rely on conventional renewables, grid power, and whatever interim clean energy it can secure. These deals are best understood not as near-term fixes but as long-horizon wagers that the physics of orbit and electrochemistry can be engineered into reliable, affordable infrastructure before the AI power crunch becomes unmanageable.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
