Meta space-based solar power deal

The most consequential technical detail in the Meta-Overview deal is what isn't being built. Classical space-based solar power, proposed by Peter Glaser at Arthur D. Little in 1968 and stress-tested by NASA and the DOE through the late 1970s and early 1980s, depends on dedicated ground infrastructure: large rectennas tuned to receive microwave transmissions from geosynchronous orbit. That ground footprint — and the political, regulatory, and capital requirements that come with it — has historically been one of the heaviest anchors on the concept's economics.

Overview Energy is trying to lift that anchor. Its satellites, which will sit in geosynchronous orbit roughly 22,000 miles above the equator where sunlight is constant, convert solar energy into low-intensity, near-infrared light and beam it directly at conventional, utility-scale ground-based solar farms. Existing photovoltaic panels — already permitted, sited, and grid-connected — become the receivers. Marc Berte's pitch of "reliable siting, and speed to power" hinges on this idea: if a hyperscaler doesn't need to build a new ground station for every gigawatt, the orbital fleet plugs into infrastructure that already exists. Whether ground arrays designed for diffuse sunlight can efficiently absorb a focused near-infrared beam at scale is the open engineering question the 2028 demonstration is meant to answer.

Meta's data centers consumed more than 18,000 GWh of electricity in 2024 — roughly the equivalent of 1.7 million U.S. homes for a year — and that load is climbing as AI training and inference workloads scale. The company was already the largest corporate clean-energy offtaker globally in 2025 at 10.24 GW, and over 2024 and 2025 it stacked roughly 6.6 GW of nuclear PPAs with Constellation, Vistra, Oklo, and TerraPower. The Overview agreement, plus the parallel 1 GW / 100 GWh ultra-long-duration storage reservation with Noon Energy, slots into that portfolio rather than replacing any piece of it.

The strategic logic is diversification across both technology and timeline. Nuclear deals address baseload but face long permitting cycles; terrestrial renewables plus storage handle near-term load growth; space solar and 100-hour storage are bets on the back half of the decade and beyond. Nat Sahlstrom's framing of space solar as "leveraging existing terrestrial infrastructure to deliver new, uninterrupted energy from orbit" is the cleanest articulation of why a hyperscaler would pay to reserve a technology that doesn't yet exist commercially — it's an option on 24/7 carbon-free generation that doesn't require Meta to abandon its existing solar siting strategy.

The cleanest counterweight to Meta's announcement is a January 2024 NASA Office of Technology, Policy, and Strategy report that concluded space-based solar would still be more expensive than terrestrial alternatives by 2050 unless major capability gaps close. Latitude Media reinforces the point with the blunt observation that "shooting solar panels into space and maintaining them for several decades is very, very expensive," and notes that the IPCC's net-zero pathways do not include the technology at all. None of those baselines have shifted because of a corporate capacity reservation.

The critique gets sharper when it turns to motive. Caroline Golin, formerly Google's global head of energy, frames space solar less as an engineering breakthrough than as a political maneuver: "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." In that reading, Meta's deal does work that goes beyond electrons — it lets a hyperscaler tell a clean-power story that doesn't require contentious local fights over transmission lines, land use, or interconnection queues. The ESG framing comes prepackaged with the orbital narrative.

Overview is targeting a fleet of roughly 1,000 spacecraft in geosynchronous orbit with more than 10-year lifespans, eventually covering about one-third of the planet. Whatever the merits of the near-infrared architecture, the financial spine of the project is launch cadence and cost. 24/7 Wall St. captures this directly: "Meta's pact to explore space-based solar power for its data center fleet is fundamentally a launch story." Without sustained access to cheap, frequent heavy-lift capacity to GEO, a thousand-satellite constellation moves from ambitious to uneconomic regardless of how efficiently the beam reaches the ground.

That dependency makes Overview a downstream beneficiary of every reusable-rocket and small-launcher cost curve in the industry, and it ties Meta's ESG narrative to a launch market that neither company controls. The November 2025 demonstration — power-beaming from a Cessna aircraft at 16,500 feet down to a ground solar installation — proved the optics work in atmosphere. It says nothing about the unit economics of putting the transmitter 22,000 miles up, where every kilogram of satellite mass is priced by the launch market on the day it ships.

Berte's marketing line is "speed to power," but the announced schedule reads more cautiously. The first orbital power-transmission demo is planned for low Earth orbit in January 2028. Satellite launches for Meta begin in 2030. The Noon Energy storage pilot is also a 2028 milestone, at 25 MW / 2.5 GWh — a fraction of the headline 1 GW / 100 GWh figure. None of the gigawatt numbers reserved this April are scheduled to meet a real load this decade.

That gap matters because Meta's AI compute build-out is happening now. The 6.6 GW of nuclear PPAs and the 10.24 GW of clean-energy contracts signed through 2025 are doing the heavy lifting for near-term demand. The Overview agreement is best read as an option on a different post-2030 generation mix — one that, if it works, gives Meta access to a generation source that doesn't compete with anyone else for terrestrial siting. If it doesn't, the company has lost a capacity reservation, not a power plant. The asymmetry of that bet is probably the single best explanation for why a hyperscaler would sign first.