Meta has signed a deal that sounds like science fiction: put roughly 1,000 satellites in geostationary orbit, convert sunlight into infrared beams, and send that power down to solar farms on Earth.
This is not a slide-deck concept. It is a signed cooperation agreement.
1 GW of solar power from orbit, with delivery targeted for 2030
The partner is Overview Energy, a Silicon Valley startup. Meta has reserved up to 1 GW of orbital solar capacity, roughly comparable to the full output of a mid-sized nuclear plant.
The technical plan is straightforward in outline:
- Satellites would sit in geostationary orbit about 22,000 miles above Earth, with about 1,000 spacecraft forming one constellation.
- Each satellite is designed to operate for more than 10 years.
- The system would collect solar energy in space, convert it into low-intensity near-infrared light, and beam it to ground receivers tied into existing solar farms.
- One constellation could cover about a third of the planet, allowing ground stations from the U.S. West Coast to Western Europe to receive power.
The timetable is also explicit: the first satellite demonstration is planned for 2028, with commercial power delivery to the U.S. grid targeted for 2030.
Overview Energy CEO Marc Berte framed the business logic this way:
"There's a big difference between being in any one energy market, and being in all of the energy markets."
In plainer terms: ground solar sells power when the sun is up locally; space solar can chase every time zone. That is why the model is commercially interesting.
The second leg: 100-hour energy storage
On the same day, Meta also announced an energy-storage agreement with Noon Energy:
| Item | Figure |
|---|---|
| Total capacity | 1 GW / 100 GWh |
| Pilot project | 25 MW / 2.5 GWh |
| Pilot launch | 2028 |
| Storage duration | More than 100 hours |
| Technology | Modular reversible solid-oxide fuel cells plus carbon-based storage |
A four-hour lithium battery is already considered useful. One hundred hours is a different order of magnitude. Combined with power beamed from orbit, the goal is clear: keep AI data centers running around the clock.
Why Meta, and why now
Start with the scale. Meta used more than 18,000 GWh of electricity in 2024, about as much as 1.7 million U.S. households use in a year. That was before LLaMA 4 was fully trained.
Next-generation model training, inference growth, real-time voice, and video generation all push electricity demand sharply higher.
Meta has already signed a daunting stack of clean-power agreements:
- More than 30 GW of clean-energy contracts in total.
- 7.7 GW of nuclear power, including deals involving Vistra, TerraPower, Oklo, and Constellation Energy.
- Next-generation geothermal work with Sage Geosystems and XGS Energy.
- Now, orbital solar power and 100-hour storage.
Put nuclear, geothermal, space-based power, and very long-duration storage together, and the message is obvious: Meta does not want the grid to become the bottleneck for AI.
Can it actually work?
Large-scale power beaming from space by 2030 is a hard target. Laser conversion efficiency, safety rules for near-infrared beams, orbital slots, and permits for ground receiving stations could all slow the plan.
The interesting question is not only whether the technology works, but who is willing to pay for it before it is built.
AI companies are increasingly using early purchase agreements as a hedge: nuclear, geothermal, and now space solar follow the same pattern. Energy projects take a decade; AI compute demand can move month by month. When those two curves do not line up, early access matters.
Meta is not just betting that the technology lands. It is betting that by 2030, ground-based energy supply may not be enough.
If that turns out to be true, today’s contract will look valuable.
Sources: Meta inks deal for solar power at night, beamed from space (TechCrunch); CocoLoop; Powering AI, Strengthening the Grid (Meta Newsroom)