Here’s a wild fact that should terrify every utility company in America: a single AI training cluster can suck down 1 to 2 gigawatts of electricity. That’s enough juice to power a mid-sized city. And Meta, Microsoft, Google, and Amazon? They’re each trying to build dozens of these things simultaneously. The grid wasn’t built for this. Interconnection queues now stretch 5 to 10 years in many regions. You literally cannot run a 2-gigawatt data center if the local grid only has 500 megawatts available. Physics doesn’t negotiate.
This is the real constraint on AI—not chips, but watts.
So what does Meta do? It goes to space.
This week, Meta announced a commercial deal with Overview Energy to beam solar power collected in orbit back to Earth to power its AI data centers. Not a PR stunt. Not a feel-good ESG announcement. A binding commercial contract with a delivery schedule. Meta is spending $60-65 billion on AI infrastructure in 2025 alone. The company doesn’t sign orbital energy contracts for fun. It signed this one because its engineers ran the numbers and concluded that Earth—the entire planet—can’t supply enough electricity for their ambitions.
Let that sink in.
The Orbital Compute Thesis: Why Space Wins
Here’s the elegant insight: terrestrial infrastructure is resource-bound. You hit ceilings that don’t scale with innovation. You can’t Moore’s Law your way past a decade-long utility queue. But orbital infrastructure is technology-bound. Costs fall with engineering progress, just like semiconductors did.
Space is actually a remarkable place to generate power. Solar panels in orbit face the sun 24/7 with no clouds, no night, no weather—producing 8 to 10 times more electricity per panel than ground-based systems. Cooling is easier too: heat just radiates into the cold of space. And for AI jobs that don’t need real-time processing (training models, crunching datasets), it doesn’t matter where the computer sits. Why not run those jobs in orbit where power is essentially free and unlimited?
The Stocks Positioned to Win
The pure-play space names: NASA ETF (SpaceX exposure), Rocket Lab (RKLB), Planet Labs (PL), AST SpaceMobile (ASTS), and Redwire Space (RDW) are your direct orbital infrastructure plays.
But here’s the thing—orbital compute doesn’t replace the terrestrial AI infrastructure trade. It extends it. While Overview Energy preps for its 2028 demo, hyperscalers are still building like crazy on Earth. The near-term dollars flow into terrestrial infrastructure: Nvidia (NVDA), Broadcom (AVGO), Marvell (MRVL), Eaton (ETN), Coherent (COHR), Lumentum (LITE), and Astera Labs (ALAB).
The Bottom Line
Meta’s deal signals something profound: the hyperscalers have looked at their options and decided going to space is the rational choice. This isn’t speculation. This is capital allocation at the highest level. The companies solving Earth’s power problem from orbit will be among the most important infrastructure investments of the next two decades.
The question is whether you’re positioned before the rest of the market notices.
