4 Minutes
AI does not sleep. Neither do the data centers training it, serving it, cooling it and quietly pulling more electricity from the grid with every new model release. That uncomfortable reality is pushing Meta into territory that still sounds half like science fiction: solar power collected in orbit, paired with batteries designed to keep clean energy flowing for days.
The company has signed new partnerships with Overview Energy and Noon Energy, two startups working on very different pieces of the same puzzle. One wants to harvest sunlight in space. The other wants to store renewable power for more than 100 hours. Together, they point to the same conclusion: the AI boom is becoming an energy problem as much as a computing one.
Meta’s move comes as hyperscale infrastructure gets bigger, hungrier and harder to hide from public debate. Proposed projects such as the 9 GW Stratos campus in Utah show just how massive the power requirements around AI and cloud computing are becoming. For companies racing to build the next generation of artificial intelligence, access to chips is only part of the battle. Access to reliable electricity may prove just as decisive.
Sunlight, without the sunset problem
Conventional solar power has a beautifully simple weakness. The sun sets. Clouds arrive. Output drops. That intermittency is manageable at small scale, but it becomes a serious headache when the customer is a data center expected to run around the clock.
Overview Energy is trying to move the solar farm above the weather entirely. Its plan involves satellites in geosynchronous orbit, positioned to collect near-continuous sunlight and transmit energy back to Earth. Ground stations would then convert that beamed signal into usable electricity, allowing solar infrastructure to generate power even when it is night on the ground.
It is a bold idea, and not a small one. Meta has reportedly reserved up to 1 GW of capacity from Overview Energy’s future system. The first orbital demonstration is targeted for 2028, while early commercial deployment could follow around 2030 if the technology performs as promised.
Noon Energy is tackling the other side of the equation: what happens when clean power is available, but not exactly when the grid needs it. The company is developing long-duration storage designed to hold energy for more than 100 hours, far beyond the typical window of lithium-ion battery systems. Its approach uses solid oxide fuel cells and carbon-based storage to stretch renewable electricity across multiple days.
Under its agreement with Meta, Noon Energy has reserved up to 1 GW and 100 GWh of storage capacity for the tech giant, with a pilot project expected in 2028. If it works at scale, this kind of storage could help data centers rely more heavily on clean power without needing fossil-fuel backup every time renewable generation dips.
None of this is guaranteed. Space-based solar power has been discussed for decades, and the engineering challenges are enormous. Long-duration storage is also a crowded field where promising lab results do not always translate into cheap, durable infrastructure. Launch costs, transmission efficiency, permitting, grid integration and plain old economics will all matter.
Still, Meta’s willingness to reserve capacity this early says plenty about where the industry is heading. AI infrastructure is no longer just a matter of servers, GPUs and networking gear. It is becoming a full-stack energy challenge, stretching from orbit to the electrical grid.
If these projects succeed, Meta may not simply buy cleaner power. It could help reshape how clean energy is generated, stored and delivered for the AI era.
Leave a Comment