Meta is looking to orbit because AI data centers are outgrowing easy power
A geosynchronous solar-power platform beaming a controlled near-infrared energy column toward a ground solar farm beside a large data-center campus at dusk.📷 AI-generated image / TECH&SPACE
- ★Meta gains early access to up to 1 GW of Overview Energy’s planned space-based solar capacity.
- ★The concept uses geosynchronous satellites that would beam near-infrared light to solar facilities on Earth.
- ★A demonstration is targeted for 2028 and commercial delivery for 2030, with major technical and regulatory questions still open.
Meta’s latest energy move is not just another clean-power contract with a futuristic label attached. According to PV Magazine, the company has secured early access to up to 1 GW of planned space-based solar power from Overview Energy, aimed at feeding the kind of dense AI workloads that now strain ordinary grid planning.
The basic idea is elegant and unforgiving: collect sunlight in geosynchronous orbit, convert it into a controllable beam of near-infrared light, and send that energy down to existing solar projects on Earth. Overview says the approach could boost output without requiring new land or fuel, which is exactly the kind of sentence that sounds simple until physics, regulation, orbital deployment, and utility integration enter the room.
The timeline is the sharpest part of the claim. The system is expected to demonstrate in 2028 and deliver commercial power in 2030, giving Meta a long-range option as data-center power demand collides with slow transmission buildouts and crowded interconnection queues.
The Overview Energy deal puts space-based solar power on a 2028 demo clock and a 2030 commercial target.
Ground-level receiving solar arrays glowing under a narrow orbital energy beam, with a distant data-center load zone and grid equipment showing the infrastructure consequence.📷 AI-generated image / TECH&SPACE
The operational context matters as much as the orbital hardware. Hyperscalers are increasingly looking for energy that sits closer to their load, or at least outside the slowest parts of the public grid process, and Meta’s agreement with Overview Energy fits that shift neatly.
There are still large blanks in the picture. The specific receiving sites have not been identified, and the feasibility of large-scale near-infrared beaming from space remains something to prove, not something to treat as settled infrastructure. A 1 GW reservation is meaningful, but it is not the same thing as a working orbital power plant.
That is why the 2028 demonstration carries more weight than the headline number. If it works, it would suggest space-based solar can become part of the AI-era energy stack, not as spectacle, but as a practical attempt to route sunlight around terrestrial bottlenecks. If it fails, the lesson will still be useful: orbit is patient, but data centers are not.
