Meta Signs Deal to Power AI Data Centers With Solar Energy Beamed From Space

Meta has struck a deal that sounds like science fiction but is now a signed commercial agreement: the company will purchase up to 1 gigawatt of space-based solar power from Overview Energy, a startup building a constellation of satellites designed to collect sunlight in orbit and beam it down to Earth as near-infrared light, which terrestrial solar farms then convert to electricity — including in the dark.

The deal, announced April 27, marks a significant escalation in how the largest AI companies are approaching the energy problem that underpins their growth ambitions. Meta has committed to spending between $115 billion and $135 billion on AI infrastructure in 2026 alone, and the bottleneck to that buildout is increasingly not capital or chip supply — it’s the availability of reliable, low-carbon power on demand.

The Technology: Solar at Night

The concept of space-based solar power has existed in theoretical form since the 1970s, when NASA first commissioned studies into whether placing solar collectors in orbit could provide continuous power generation unconstrained by Earth’s day-night cycle or weather patterns. What changed is that the economics of satellite launch have collapsed — SpaceX’s Falcon 9 and Starship have reduced the cost of putting mass into orbit by more than 95% over the past decade — making what was once a purely theoretical energy architecture increasingly plausible to build commercially.

Overview Energy’s approach works in three phases. First, a constellation of satellites in low-to-medium Earth orbit collects sunlight using large, deployable photovoltaic arrays. The satellites operate in orbits and inclinations chosen to maximize the fraction of time they are illuminated by the sun — well above 90% in the right orbital configuration, compared to terrestrial solar panels that generate power only during daylight and at reduced efficiency under cloud cover. Second, the collected energy is converted and transmitted to Earth as a directed beam of near-infrared electromagnetic radiation. Near-infrared was chosen for its high efficiency in passing through the atmosphere with minimal absorption losses. Third, specialized terrestrial receiver farms — essentially large-area solar panels tuned to the incoming wavelength — capture the beam and convert it to grid electricity.

The “at night” capability is the critical differentiator. A satellite constellation in a suitably designed orbit can beam power to a specific ground location regardless of local time, because the satellites themselves remain in sunlight while the receiving site is in darkness. This solves one of the fundamental limitations of terrestrial solar: the need for large-scale, long-duration energy storage to cover generation gaps.

Why Meta Is Betting on This Now

Meta’s interest in space solar is a function of simple arithmetic that is becoming increasingly uncomfortable for all major AI operators.

Training frontier AI models and running large-scale inference at the scale Meta requires is extraordinarily power-intensive. Meta’s data centers collectively consume roughly 10 gigawatts of power today, and the buildout required to support the company’s AI ambitions — training successive generations of Llama models, running Meta AI across WhatsApp, Instagram, Facebook, and Messenger at global scale, and supporting Reality Labs compute — is projected to require significantly more.

The problem is that grid power is not uniformly available. In most of the United States and Europe, adding large new loads to the grid requires years of interconnection queue processing, new transmission infrastructure, and increasingly, public permitting battles as communities push back against data center construction near residential areas. The result is that even companies with the capital to build new data centers often cannot get power to them on the timeline their AI roadmaps require.

Meta has responded with an unusually aggressive portfolio of energy strategies. The company has now contracted more than 30 gigawatts of clean and renewable energy across its global portfolio, and is one of the largest single corporate purchasers of nuclear energy in American history — holding 7.7 gigawatts in nuclear agreements with Vistra, TerraPower, Oklo, and Constellation Energy. It has also signed a 20-year nuclear power agreement.

Space solar is being positioned as a complement to these existing commitments — specifically for situations where terrestrial clean energy supply cannot be sited or permitted quickly enough to meet demand at specific data center locations.

The Noon Energy Storage Partnership

Alongside the Overview Energy space solar deal, Meta simultaneously announced a partnership with Noon Energy to deploy up to 1 GW / 100 GWh of ultra-long-duration energy storage.

Noon Energy’s technology is materially different from the lithium-ion battery systems that currently dominate grid-scale storage. The company uses modular, reversible solid oxide fuel cells paired with carbon-based storage media, enabling more than 100 hours of energy storage — roughly 4 to 5 times the practical limit of lithium-ion systems, which top out around 4 to 8 hours at grid scale. This is the difference between smoothing out overnight generation gaps and weathering multi-day periods of low renewable generation — what the energy industry calls “dark doldrum” events, when solar and wind output simultaneously drops due to weather patterns.

Together, the Overview Energy and Noon Energy deals represent a coherent strategy: space solar provides continuous generation independent of local weather and time, while Noon Energy’s long-duration storage handles the gaps when the satellite constellation is not in optimal position over a given data center location.

The Timeline and Commercial Reality

The space solar deal is real, but the timeline is measured in years, not months.

Overview Energy’s orbital demonstration is planned for 2028 — the first commercial test of beaming power from orbit to a terrestrial receiver. That is not a science project; it is the system working end-to-end at small scale. Commercial delivery to the U.S. grid could begin as early as 2030, which aligns with the multi-year buildout timelines that characterize large data center campuses anyway.

For perspective, a data center facility that breaks ground today and requires a new 500-megawatt power connection may not receive that connection until 2028 or 2029 regardless of the energy source. From Meta’s planning perspective, a space solar supply that begins commercial delivery in 2030 is not far outside the planning horizon of infrastructure being committed to today.

What This Signals for the Industry

Meta is not alone in the fundamental energy problem it is trying to solve. Microsoft, Google, and Amazon are each spending $175–200 billion on AI infrastructure in 2026, and each faces versions of the same grid access and clean energy supply constraints.

The difference is that most are pursuing solutions that work within the existing energy system — nuclear power purchase agreements, long-term renewable contracts, advanced geothermal, grid-scale batteries. Meta, by committing to a space solar deal, is making a bet that terrestrial energy solutions will be insufficient to meet the pace of AI infrastructure demand, and that the timeline for new energy architecture to come online is short enough to be commercially relevant.

That bet may or may not prove correct. Space-based solar has a long history of being perpetually 15 years away from viability. What is different in 2026 is that launch costs have been transformed, Meta has the capital to backstop a commercial partner through development risk, and the demand signal — the extraordinary power requirements of frontier AI — has finally made the business case legible.

Overview Energy secured the Meta agreement as a customer anchor. The rest of the industry is watching.