AI’s Electron Problem — Solved by Energy from Space

The energy impact of data centers is now front-page news. It’s the clearest signal that our energy investments aren’t moving fast enough. Electricity demand is rising faster than we’ve been able to build grid capacity, and the mismatch is no longer abstract.

Data centers are the first sector to visibly hit the energy wall, but they won’t be the last.

Electricity isn’t like data. You can’t reroute it instantly. It’s bound to a more physical infrastructure: where power is generated, where transmission exists, and how quickly new capacity can be built. Our approach to space solar energy changes that dynamic by making electricity directable across grids, countries, and continents.

AI data centers are a clear case study for why that shift matters.

AI’s energy wall at scale

Energy availability is constraining AI development. Just the top five hyperscalers alone are planning roughly $450B in AI infrastructure spending in 2026 alone, yet the limiting factor is no longer chips or capital. It’s electricity.

Interconnection queues tell the story. In Texas, proposed projects now total more than 430 GW, up 43% from two years ago. Across the U.S., queues stretch years into the future, delaying projects that are otherwise ready to proceed.

Executives are blunt about the bottleneck. “Power is my problem today,” Microsoft CEO Satya Nadella said on the BG2 podcast. “It’s not a supply issue of chips; it’s the fact that I don’t have warm shells to plug into.”

The pressure is so acute that companies are exploring proposals to move data centers into orbit to access space’s continuous solar energy. That line of thinking reflects the real constraint: power on Earth is increasingly hard to deliver at the pace AI requires.

Relocating compute, however, introduces a new set of technical, economic, and operational tradeoffs, from launch and maintenance cycles to thermal management and latency. It makes more sense to keep compute here while still taking advantage of the benefits of space by changing how energy is delivered.

Space solar energy addresses the energy constraint at the grid level, and not just for AI.

Making electricity behave differently

Overview’s space solar energy works with the grid that already exists. The receivers are existing utility-scale solar projects on the ground, day or night. Solar—already the fastest-deploying source of new generation—can reach 70–80% capacity factors* with our beam, versus 25-30% with sunlight alone.

The mechanics matter because they change how energy is used. Overview’s satellites will operate in geosynchronous orbit (GEO), collecting sunlight and transmitting it as low-intensity, invisible infrared light down to the ground (see our launch blog for details). Instead of pushing electricity through wires across long distances, the energy is delivered as light to solar projects that are already powering data centers, and that light is converted into electrons exactly where they’re needed.

The critical property is directability: each satellite can continuously see a third of the Earth and can route energy to where electricity is most valuable at a given moment. When solar is curtailed in California on a summer afternoon, that energy can be redirected to support a training workload in Virginia after sunset, or to displace fossil generation on a nighttime grid in Spain when the wind isn’t blowing.

Directability decouples electricity supply from location and timing. Instead of overbuilding generation or waiting years for new infrastructure, electricity can be shifted across the system to meet demand as it emerges. We’re choosing where and when energy is created on the grid.

What changes when energy is directable

Once electricity can move, the tradeoffs that define today’s grid begin to change for AI and beyond.

Lower, rather than raise, energy prices.
At the proposed scale, data center growth would strain the grid and risk higher electricity costs—a concern already influencing public and regulatory scrutiny. Directable power flips that dynamic. When the grid is congested, data centers can self-power with space solar energy without adding demand. When renewable generation would otherwise be curtailed, satellites can redirect that energy elsewhere, reducing system congestion rather than compounding it. The result is higher utilization across the grid and lower system costs, instead of expensive overbuild that drives prices up for households.

Maximize existing infrastructure.
Amazon, Microsoft, Meta, and Google have contracted roughly 30 GW of solar capacity, generating about 80 TWh annually. Powered from space, those same assets could generate up to 500 TWh, which means 3-6 times as much energy from infrastructure that’s already being built.** Space solar energy offsets tens of billions in additional investment and reduces land requirements for new solar developments.

Collapse time to power.
Time to power determines time to revenue. For example, Meta has started deploying compute in tents to reduce construction time at sites with power. Space solar energy can make that power available in days at sites with existing solar projects. If a data center needs more power during a specific time period (e.g., to support a training load), space solar energy can deliver incremental capacity on demand. And when new long-term capacity is needed, greenfield solar remains the fastest generation asset to build, and every new project doubles as a future receiver for space solar energy.

Use space and Earth for what each does best.
We should play to each environment’s strengths: build energy infrastructure in space and keep data centers on Earth. Orbital data center concepts would rely on low Earth orbit (LEO), where sunlight isn’t continuous and large constellations create congestion and light pollution. Space solar energy in geosynchronous orbit (GEO) offers constant, 24/7 sunlight without adding to LEO crowding or dark-sky impacts. Earth remains the better place to run data centers: lower latency, simpler cooling, and easy chip replacement.

Close the electron gap.
The United States leads the world in AI development and data center deployment. Maintaining that leadership relies on solving what OpenAI calls the “electron gap.” Space-to-grid energy scales with demand, works with existing infrastructure, and builds on American strengths in launch and space systems. It turns energy delivery into a strategic advantage rather than a bottleneck.
 

From AI to the rest of the grid

AI is the first sector to collide with the limits of fixed electrical infrastructure.

By the end of this decade, the mismatch between how fast demand is growing and how slowly capacity can be delivered will become acute across the economy. Data centers will represent about 10–15% of total grid demand. Manufacturing, logistics, cold storage, critical facilities, and transportation will face the same constraints and the same risks to growth.

Delivering energy from space to the grid addresses that system-wide challenge. By making electricity directable, it strengthens the grid underneath every sector it serves.

That’s Overview’s value proposition: changing how energy moves across the grid. Fast enough to meet demand, flexible enough to avoid overbuild, and scalable across industries.

###

Notes

(*) Our near-infrared beam is blocked by clouds, which results in a 70-80% capacity factor depending on geography. The remaining gap can be closed with a modest amount of storage, on-site back up generation, and/or other assets on the grid

(**) Baseline assumes 30% capacity factor. Space solar energy enables 3-6x increase in energy density per unit area (i.e., 3-6x less land for the same power) as a function of 2-3x higher capacity factor and 2x higher nameplate capacity for the same PV panels. This is conservative in practice, as space solar energy optimized solar projects are higher density (e.g., lower row spacing or Earth mount panels) to reduce beam waste