Meta reserves up to 1GW of space-solar capacity for AI data centers

Meta is reserving future power from a space-solar startup.

Overview Energy and Meta announced a capacity reservation agreement that would give Meta early access to up to 1GW from Overview’s planned space solar energy system. Overview says it plans an initial orbital demonstration in 2028 and expects commercial power delivery in 2030.

The proposed system would collect solar energy in space, convert it to near-infrared light, and beam it to existing solar facilities on the ground so those facilities can generate power beyond normal daylight hours.

Why it matters

This is an AI infrastructure story.

Meta’s model roadmap depends on data centers, and data centers depend on power. The company has already made large nuclear and renewable energy moves. This agreement shows how far hyperscalers are willing to look for capacity that could support future AI training and inference.

The interesting part is not only space solar. It is that Meta is reserving future energy capacity years before commercial delivery. AI infrastructure has become a planning problem across chips, land, interconnects, cooling, grid access, and power-purchase agreements.

Tool impact

There is no immediate Meta AI product change.

The downstream impact is capacity. More power can mean more training runs, more inference budget, fewer rollout bottlenecks, and less pressure to ration expensive model features.

For AI tool buyers, this kind of deal is an indirect signal. If hyperscalers are fighting for future power, hosted AI pricing, rate limits, model availability, and regional capacity will remain strategic constraints. The best products may still be shaped by where compute and energy are available.

What to watch

The timeline is long and the technical risk is high.

The key milestones are Overview’s 2028 orbital demo, regulatory approval for beamed energy, commercial-scale satellite deployment, and whether Meta turns a reservation into delivered power.

Also watch whether existing solar farms can practically receive and convert the beamed energy, whether ground operators and regulators accept the safety case, and whether launch, maintenance, and transmission costs beat other long-duration power options.

Aipedia take

This is not near-term power for today’s Meta AI products. It is a marker of how aggressive AI infrastructure planning has become. The AI race is no longer only about models and GPUs; it is also about who can secure enough reliable energy to run them.