A focused beam of near-infrared light, shot from a satellite floating 35,000 kilometres above the equator, is about to become the most expensive extension cord in history. Meta has officially signed a deal with the startup Overview Energy to source up to one gigawatt of electricity from a space-based solar system, a move that signals just how desperate the world’s tech giants are for power that doesn’t blink when the sun goes down.
The deal isn't just a flight of fancy for Mark Zuckerberg. It is a tactical response to a brutal reality: the artificial intelligence boom is cannibalising the world’s power grids. To keep its Llama models dreaming and its Reels algorithms churning, Meta needs a constant, unwavering stream of electrons that the traditional terrestrial grid is increasingly unable to guarantee. By tapping into space-based solar power (SBSP), Meta is attempting to bypass the atmosphere entirely, harvesting sunlight in a place where the concept of 'night-time' doesn't exist.
For the uninitiated, one gigawatt is a staggering amount of energy—roughly the output of a large nuclear reactor or about 3.1 million solar panels. Meta isn't just looking for a green PR win; they are looking for industrial-scale baseload power that can run 24 hours a day, 365 days a year, without the intermittency problems that plague wind and ground-based solar.
The sun that never sets
The physics of the project are as ambitious as the price tag. Overview Energy plans to place massive satellite arrays in geosynchronous orbit. At this altitude, the satellites stay fixed over a single point on Earth and remain in direct sunlight for 99% of the year. Down here, even the best solar farms are at the mercy of the Earth’s rotation, cloud cover, and seasonal shifts. In orbit, the sun is always shining, and the light is roughly 30% more intense because it hasn't been filtered through our thick, messy atmosphere.
The clever part of the Overview Energy concept is what happens to that light. Rather than trying to beam high-energy microwaves—a concept that has historically spooked regulators and the public alike—the system converts sunlight into low-energy near-infrared light. This beam is then aimed at existing solar farms on the ground. These terrestrial sites, which usually sit idle and useless the moment the sun dips below the horizon, act as receivers. They catch the infrared beam and convert it into electricity using the same photovoltaic technology they use during the day.
This 'dual-use' approach for existing ground infrastructure is the secret sauce. It means Meta doesn't have to fight for new land permits or build massive new receiver stations (rectennas) from scratch. They can effectively 'wake up' a sleeping solar farm at 2 AM by shining a giant, invisible torch on it from space.
Why the chips are staying on the ground
The announcement sets up a fascinating philosophical clash between the world’s two most prominent tech billionaires. While Zuckerberg is looking to space for energy, Elon Musk has been quietly floating the idea of putting the data centres themselves into orbit. The logic behind the Musk-adjacent SpaceX pitch is simple: if the power is in space, why bother beaming it down? Just put the H100 GPUs next to the solar panels.
However, Meta’s strategy suggests they think that plan is a logistical nightmare. Shortly before the Meta announcement, SpaceX itself warned investors in a private document that orbital AI computing might not be commercially viable anytime soon. The reasons are stubbornly physical. Data centres generate an incredible amount of heat, and in the vacuum of space, getting rid of that heat is notoriously difficult. On Earth, you can use fans, water cooling, or even just the ambient air. In space, you’re stuck with radiators that have to be massive to be effective.
Then there is the issue of latency and maintenance. If a server rack fails in a facility in Virginia, a technician can swap it out in twenty minutes. If it fails in orbit, you’re looking at a multi-million-dollar repair mission or a very expensive piece of space junk. By keeping the 'brains' on Earth and only outsourcing the 'battery' to space, Meta is betting that the cost of beaming energy down is lower than the cost of keeping hardware up.
A billion-dollar safety net
It’s important to note that Meta isn't just cutting a cheque and hoping for the best. The agreement is structured around 'preferred access,' which is corporate-speak for a sophisticated waitlist. Meta has committed to taking the power once Overview Energy hits specific technological milestones. It’s a way of providing the startup with the 'bankability' it needs to secure further investment without Meta having to bear the entire risk of the satellites blowing up on the launchpad.
This is a pattern we’re seeing across the sector. Microsoft recently signed a deal to resurrect the Three Mile Island nuclear plant, and Google is backing small modular reactors (SMRs). The common thread is a total abandonment of the 'wait and see' approach to energy. Big Tech has realised that if they want to dominate the AI era, they have to become energy companies that happen to write code.
Meta’s portfolio now includes over 30 gigawatts of clean energy projects, ranging from traditional wind and solar to more exotic geothermal and nuclear bets. The addition of space-based solar is the wildest card in the deck, but it’s one they feel forced to play. If the ground-based grid can’t keep up with the demand for AI training, the only place left to look is up.
The 100-hour battery problem
Even with a giant space flashlight, you still need a way to buffer that energy. Alongside the space deal, Meta also partnered with Noon Energy to build a gargantuan long-duration storage system. We’re talking about 100 gigawatt-hours of capacity—enough to keep a small city running for days. Unlike the lithium-ion batteries in your phone, which are great for a few hours of discharge, Noon Energy uses modular, reversible solid oxide fuel cells with carbon-based storage.
This technology is designed to store energy for more than 100 hours, bridging the gap if the orbital beam is eclipsed or if terrestrial weather turns truly foul. A pilot project for this storage is slated for 2028, with the full gigawatt-scale rollout expected to follow shortly after. It represents one of the largest commitments to ultra-long-duration storage ever made by a private company.
The combination of these two deals—energy from the stars and storage in carbon cells—paints a picture of a company trying to build a 'closed-loop' energy ecosystem. Meta is effectively trying to insulate itself from the volatility of the global energy market and the fragility of the ageing US power grid.
Can the economics ever actually work?
The elephant in the room is, as always, the cost of getting things into orbit. While companies like SpaceX have drastically lowered the cost per kilogram to reach Low Earth Orbit (LEO), Overview Energy’s satellites need to sit much higher, in Geosynchronous Orbit (GEO). Reaching GEO is significantly more expensive and requires more fuel.
There is also the regulatory hurdle. Shouting 'don't worry, it's just low-energy infrared' might not be enough to satisfy government agencies concerned about what happens if a beam wanders off-target. While infrared won't 'fry' a bird or a plane in the way a high-frequency microwave beam might, the optics of 'lasers from space' are a tough sell for any PR department.
Comments
No comments yet. Be the first!