Meta Signs Deal for Space-Based Solar Power to Feed AI Data Centers

Meta has quietly inked a supply deal with Bay Area startup Overview Energy that would let the company tap solar power even at night. According to TechCrunch, the agreement calls for a future constellation of roughly one-thousand small satellites to harvest sunlight in geosynchronous orbit and beam it to existing solar farms using infrared lasers. The farms would then feed Meta’s AI data centers with round-the-clock renewable electricity. Overview has already raised $20 million and demonstrated the core laser-beaming concept in airborne tests last year.

Meta’s appetite for electrons is staggering. The company disclosed that its global data centers burned through 18,000 gigawatt-hours in 2024, equal to the annual demand of 1.7 million U.S. homes. With AI training clusters still expanding, Meta has committed to building or contracting 30 GW of renewable capacity by 2030. The Overview contract is the first that explicitly promises power after sunset without batteries or fossil-fuel backups.

If proven, the space-to-ground model could slash the carbon footprint of open-source model training. Meta’s Llama family is already downloaded millions of times; cheaper, cleaner baseload power would let independent researchers spin up larger clusters without blowing sustainability budgets. Open-source communities have long complained that energy costs favor well-funded labs; orbital solar might level the field.

Overview’s plan is elegantly retro. Each satellite unfolds a lightweight photovoltaic array nearly the size of a football field. Electricity is converted to coherent infrared light, aimed at ground receivers no larger than conventional solar trackers. A pilot receiver in Nevada has shown 85 % end-to-end efficiency during clear nights. The constellation will orbit at 22,000 miles, matching Earth’s rotation so every receiver sees a fixed “power satellite” hanging overhead.

Amazon, Google and Microsoft have all pledged carbon neutrality but still struggle with nighttime power. Bloomberg notes that Meta’s move pressures rivals to adopt similar exotic sources or risk higher operating costs as grid prices spike. Cloud providers that can advertise “space-solar-backed compute” may command a premium from climate-conscious customers.

The first three demonstration satellites launch in Q1 2027 on a SpaceX rideshare. If the pilot beams 5 MW reliably, Meta will exercise an option for the full thousand-satellite fleet. Regulatory filings show the constellation could deliver up to 2 GW by 2032, enough for roughly one-tenth of Meta’s projected data-center load. Until then, expect more conventional PPAs: the company just added another 176 MW Texas solar plant and is eyeing 1 GW of long-duration storage from Noon Energy as bridge capacity.

Space-based solar still faces wild uncertainties. Launch costs must fall another 50 % to hit price parity with terrestrial solar plus batteries. Atmospheric interference could drop efficiency below viable thresholds. And the geopolitical optics of orbital power stations are untested; countries may balk at foreign lasers crossing their airspace. Meta’s contract is structured as a power-purchase agreement, not an equity investment, limiting its downside if the tech stalls.

Even as Meta bets on satellites, it’s hedging across every clean tech available. In the past twelve months the company has signed deals for 791 MW of new wind and solar through Invenergy, 176 MW from Zelestra’s Texas plant, and a memorandum for up to 100 GWh of iron-air storage from Noon Energy. The strategy mirrors that of a hedge fund rather than a single-project developer, spreading risk across orbit, rooftops, plains and salt caverns.

Utility analysts told Latitude Media that large-scale space solar could suppress wholesale power prices during off-peak hours, hitting gas-peaker revenues hardest. If Meta and its peers soak up midnight electrons from orbit, traditional generators lose the lucrative arbitrage window they rely on. Regulators in ERCOT and PJM are already modeling scenarios where 10 GW of orbital capacity enters the stack by 2035.