Meta has signed an agreement with Overview Energy, a space-based solar power startup, to receive up to one gigawatt of orbital solar power for its data centers, the companies announced April 27. The deal positions Meta as an early anchor customer for a technology that has spent decades on engineering wish lists and is now being pulled forward by the energy desperation of the AI buildout.

Overview’s approach differs sharply from competitors chasing the same problem. Rather than putting compute in orbit, the company plans to beam solar energy collected in space down to existing terrestrial solar farms via infrared lasers, allowing those facilities to keep generating electricity after sunset. The company targets commercial service to begin as soon as 2030, following an in-space demonstration in 2028.

The energy math behind the deal

One gigawatt is a meaningful number. It is roughly the output of a large nuclear reactor, and it is the kind of baseload Meta now needs to feed AI training clusters that consume electricity at industrial-utility scale. The Overview agreement sits inside a broader Meta procurement push that, on the same day, included a separate 100 gigawatt-hour storage deal with battery startup Noon Energy.

Meta has also been signing for geothermal and nuclear capacity over the past year. The pattern is clear. Hyperscalers are no longer treating clean power as a sustainability line item. They are treating it as a supply-chain bottleneck.

According to Meta’s energy sustainability team, space solar technology could leverage existing ground infrastructure to receive continuous orbital energy.

Why Overview’s architecture is different

The competitive context matters. SpaceX, Blue Origin, and Starcloud have all filed applications for large-scale orbital satellite constellations that would function as data centers. Starcloud raised $170 million in late March as part of its orbital compute ambitions.

These competitors are pursuing the same logic: orbital sunlight is nearly continuous, and putting compute in space sidesteps terrestrial grid constraints. Overview’s pitch inverts the architecture. Keep the compute on the ground, where it can be cooled, repaired, and upgraded. Send only the energy generation to orbit.

Overview CEO Marc Berte has explained the rationale: put the thing that doesn’t need maintenance and can last for a long time in space—the energy generation part—rather than the processing infrastructure that requires constant updates and repair.

Industry observers have largely backed that framing. The Overview model makes sense because it beams power to existing solar facilities, letting them produce electricity past sundown without requiring entirely new ground infrastructure.

What Washington and the FCC are actually being asked to approve

The regulatory subtext here is significant. The FCC is evaluating filings for large-scale orbital satellite constellations. Even if a fraction of those proposed constellations are approved and deployed, the spectrum, debris, and astronomy implications are substantial. Overview’s approach asks the FCC for something narrower: laser power transmission licensing, not a megaconstellation.

That distinction will likely matter as regulators weigh the cumulative orbital impact of AI’s energy hunger. A power-beaming demo satellite is a smaller policy ask than tens of thousands of compute nodes in low Earth orbit. For a Meta legal team thinking about regulatory risk over a decade-long procurement horizon, that is not a trivial consideration.

The Los Angeles Times reported that the deal reflects Meta’s broader bet on diversifying clean energy sources before grid bottlenecks force AI capacity decisions.

The political backdrop: data centers as a public fight

The energy procurement scramble is colliding with rising public resistance. Local opposition to data center expansion has surfaced across multiple states over the past 18 months, driven by concerns about water use, electricity rates, and grid reliability for residential customers. Some utilities have proposed cost structures that would shift infrastructure expenses toward AI tenants. Others have proposed the opposite.

Meta’s energy team is reading the politics. Geothermal contracts, small modular reactor agreements, battery storage deals, and now space-based solar — the procurement portfolio is being built less around cost optimization than around defensibility. Each contract is also a story Meta can tell regulators, governors, and ratepayer advocates about why its data centers will not strain local grids.

That story has limits. Overview’s first commercial service date is 2030 at the earliest. The gigawatt is a forward commitment, not delivered power. In the meantime, Meta’s actual electrons will keep coming from gas, nuclear, and conventional renewables, much like the rest of the AI sector. Hyperscaler clean-energy commitments frequently coexist with near-term fossil fuel purchases.

The technical bet that has to pay off

Space-based solar power has been studied since the 1960s. The physics works. The economics never have. What changed is the cost of orbital launch and the willingness of a customer like Meta to pre-commit to gigawatt-scale offtake before the technology has flown.

Overview’s specific approach — infrared laser transmission to existing photovoltaic arrays — is engineering-clever because it reuses ground infrastructure that is already permitted, interconnected, and depreciating on someone’s balance sheet. A solar farm that runs at night because of an orbital laser is, from the grid operator’s perspective, just a solar farm with an unusually high capacity factor. That is easier to integrate than an entirely novel generation asset.

The risks are non-trivial. Atmospheric attenuation, beam pointing accuracy, eye-safety regulation, aviation deconfliction, and the question of what happens when a laser-transmitting satellite passes over a jurisdiction that has not authorized the transmission — these are all unresolved. The 2028 demo will start producing real data on most of them.

What the deal signals

For the space industry, the contract is validation that orbital energy delivery has crossed from speculative to procurement-ready in the eyes of at least one hyperscaler. For the AI industry, it is another data point that compute companies are now functionally also energy companies, with multi-decade infrastructure obligations they did not have five years ago.

The orbital data center crowd will argue Overview is solving the wrong half of the problem. Overview will argue that putting servers in space is engineering theater. Meta, characteristically, has hedged. The company has not signed with SpaceX or Blue Origin’s orbital data center pitches. It has signed with the company that promised to leave the servers on the ground.

That is a meaningful institutional choice. It suggests Meta’s energy team believes the constraint worth solving in 2030 is power availability, not compute siting. Whether Overview can actually deliver a gigawatt from orbit by then is a separate question, and the answer will not be known until the 2028 demonstration. The contract is a bet, not a fait accompli.

Photo by Jake Heinemann on Pexels