Meta bets on orbital solar to power AI data centers as hyperscalers chase energy sovereignty
The 1 GW space-based solar deal marks a strategic pivot from grid dependency to unconventional power infrastructure as AI compute demands approach 3% of global electricity by 2030.
Meta signed a capacity reservation for up to 1 gigawatt of space-based solar energy from Overview Energy, becoming the first hyperscaler to publicly commit to orbital photovoltaics as AI compute demand forces tech giants to secure power sources beyond traditional grids.
The deal, according to Meta’s announcement, targets an orbital demonstration in 2028 with commercial delivery expected by 2030. Overview Energy will beam solar power from satellites directly to Meta’s receiving stations, bypassing grid interconnection queues that stretch 5-10 years and increasingly strain under datacenter loads that consumed 18,000 gigawatt-hours across Meta’s facilities in 2024 alone.
The move follows similar infrastructure plays by Amazon and Google, signaling how Hyperscalers are no longer competing purely on algorithms or chip access. The real constraint is energy. IEA projections show global data center electricity consumption reaching 945 terawatt-hours by 2030 — just under 3% of total global electricity — with AI-optimized servers accounting for 432 TWh of that demand, up from 93 TWh in 2025.
Why orbital photovoltaics now
Space-based solar addresses the intermittency problem that has limited terrestrial renewables. Solar panels in orbit can produce up to 8 times more electricity than identical panels on Earth, according to Google’s Project Suncatcher research. They operate 24/7 without weather interference, eliminating the storage burden that makes grid-scale renewables expensive at high penetration.
Meta’s deal includes a parallel agreement with Noon Energy for up to 1 GW and 100 gigawatt-hours of ultra-long-duration energy storage — batteries designed to discharge for over 100 hours. The pairing suggests Meta views orbital solar not as a replacement for nuclear baseload but as a complement to intermittent renewables, with storage bridging gaps when satellite coverage shifts or maintenance occurs.
“Space solar technology represents a transformative step forward by leveraging existing terrestrial infrastructure to deliver new, uninterrupted energy from orbit.”
— Nat Sahlstrom, VP of Energy and Sustainability, Meta
The technology remains unproven at commercial scale. Overview Energy has yet to launch a demonstration satellite, and cost estimates remain speculative. Analysis from IEEE Spectrum in March 2026 estimated a 1-gigawatt orbital data center at approximately $51 billion for five-year operation — roughly three times the $16 billion cost of an equivalent terrestrial facility. However, those figures assume data processing in orbit, not just power generation. Overview’s model transmits power wirelessly to ground stations, potentially reducing costs significantly.
The hyperscaler energy arms race
Meta’s space solar bet is one move in a broader portfolio strategy. The company has contracted over 30 gigawatts of clean and renewable energy across multiple sources, including 7.7 GW of nuclear capacity through agreements with Vistra, TerraPower, Oklo, and Constellation Energy. Those nuclear deals target small modular reactors starting in 2030, though no commercial SMRs are yet operational in the United States.
Google announced Project Suncatcher in late 2025, targeting launch of two prototype satellites by early 2027. The project pairs orbital solar arrays with tensor processing units in space, testing whether AI training can occur in orbit to reduce latency and cooling costs. Amazon’s Starcloud-1 satellite launched in late 2025 with an Nvidia H100 GPU and successfully trained AI models in orbit, demonstrating proof of concept for space-based compute infrastructure.
Microsoft took a different path, signing a 20-year deal with Constellation Energy in September 2024 to restart the shuttered Three Mile Island Unit 1 reactor, delivering over 800 megawatts by 2028. That approach prioritises speed — the reactor already exists and requires refurbishment rather than new construction — but ties Microsoft to a single geographic site and exposes it to nuclear regulatory risk.
The grid can’t keep up
U.S. Data Centers consumed 4.4% of total electricity in 2023. The Department of Energy projects that share will reach 6.7% to 12% by 2028, per Lawrence Berkeley National Laboratory analysis. In absolute terms, U.S. data center electricity demand is expected to increase by 130% by 2030, with the country accounting for 45% of global data center consumption.
That growth is concentrated in AI-optimized facilities. Gartner estimates AI-optimized servers will consume 432 terawatt-hours by 2030, representing 44% of total data center power. Traditional Grid Infrastructure was never designed for this density of load or this pace of growth. Interconnection queues stretch years, with renewable projects waiting an average of five years from application to energisation.
Grid interconnection is the process of connecting a new power source or load to the transmission network. In the U.S., applications are managed by regional transmission organisations (RTOs) and require engineering studies to ensure grid stability. The queue has grown from 800 gigawatts in 2020 to over 2,600 GW in 2024 as solar, wind, and data center projects compete for access.
Hyperscalers are bypassing that bottleneck by securing dedicated power. Lori Bird, director of the U.S. Energy Program at the World Resources Institute, described the dynamic: “Companies are scrambling to try to get as much power as they can as quickly as possible. It’s a mad rush and a lot of competition for resources.”
Economic and regulatory hurdles
Space-based solar faces significant cost and regulatory barriers. Launch costs must fall below $200 per kilogram to make orbital power economically viable at scale, according to industry estimates. SpaceX’s Starship aims to reach that threshold by the mid-2030s, but reusability at that price point remains unproven.
Regulatory frameworks are equally uncertain. The Federal Communications Commission regulates orbital spectrum allocation, the Federal Aviation Administration oversees launch licensing, and the Department of Energy may claim jurisdiction over wireless power transmission if it impacts grid stability. No comprehensive regulatory regime exists for commercial space-based solar, creating execution risk for early movers like Meta.
- Meta’s 1 GW space solar reservation is the first public hyperscaler commitment to orbital photovoltaics, targeting 2030 commercial delivery after a 2028 demonstration.
- AI compute is driving data center electricity demand toward 945 TWh globally by 2030 — nearly 3% of total consumption — with U.S. data centers projected to consume 6.7%-12% of domestic electricity by 2028.
- Hyperscalers have collectively secured over 50 GW of nuclear, geothermal, and orbital capacity to bypass grid interconnection queues and secure energy sovereignty.
- Space-based solar remains unproven at commercial scale, with cost estimates 3x terrestrial equivalents and unclear regulatory pathways.
What to watch
Overview Energy’s 2028 orbital demonstration will be the first real test of whether space-based solar can deliver reliable power at a price competitive with terrestrial alternatives. If the demo succeeds, expect accelerated timelines from Google and Amazon, both of which have signaled interest in orbital infrastructure.
Nuclear remains the more immediate bet. Microsoft’s Three Mile Island restart is scheduled for 2028, and Meta’s Oklo contracts target 2030. If either delivers on time, it will validate nuclear as the fastest path to baseload AI power and shift capital away from riskier orbital plays.
Regulatory clarity is the wildcard. The FCC, FAA, and DOE have yet to establish coordinated oversight for space-based power transmission. Any delays in spectrum allocation or safety certification could push Meta’s 2030 timeline into the mid-2030s, eroding first-mover advantage and forcing reliance on storage and grid power in the interim.
The hyperscaler that secures reliable, dedicated power first gains a structural advantage in AI development — more compute, faster iteration, and independence from grid constraints that will only tighten as electrification accelerates. Meta’s space solar bet is a hedge on that future.