There’s a tension at the heart of Elon Musk’s empire, and it’s getting harder to ignore.
For years, Musk has positioned himself as the clean energy visionary. Tesla’s Master Plans, stretching back over a decade, all centered on the same core mission: accelerating humanity’s shift from “a mine-and-burn hydrocarbon economy towards a solar electric economy,” as Musk framed it in the original plan. It’s a compelling narrative. It’s also what built his fortune.
But then xAI showed up. According to reporting based on SpaceX’s recent IPO filing, the AI company is running dozens of unregulated natural gas turbines to power its data centers, with plans to buy another $2.8 billion worth of fossil fuel infrastructure. That’s not some minor operational detail. That’s a direct contradiction to the foundational principle that made Tesla relevant in the first place.
The contradiction gets sharper when you look at how Musk’s companies actually do business with each other. SpaceX dropped $131 million on Cybertrucks. xAI spent $697 million over two years on Tesla’s Megapacks, those grid-scale battery systems designed to manage power loads. But Tesla solar panels? Barely a blip on xAI’s shopping list.
If Musk truly believed in his own technology roadmap, you’d expect his companies to be his first customers.
The Space Escape Route
Here’s where it gets interesting. The SpaceX filing doesn’t ignore solar power entirely. It just relocates it. Rather than powering xAI’s terrestrial data centers with panels on the ground, SpaceX is pitching space-based solar as the real future. The company claims orbital arrays can generate “more than five-times the energy” of earthbound ones, thanks to 24/7 sunlight without atmospheric interference.
It’s a neat vision, and the logic is tempting. As AI compute demands have exploded, data center opposition has grown fiercer. NIMBYs fighting local projects. Towns worried about water use and heat. Grid strain. The solution? Move the problem into orbit where nobody lives and nobody can complain.
Musk is clearly thinking in terms of first principles here. SpaceX’s filing repeatedly references “terawatt-scale annual AI compute growth.” For context, all the world’s data centers today use around 40 gigawatts. A terawatt is 25 times that. It’s a stunning number, the kind that justifies radical rethinking.
But there’s a hitch. Actually, several hitches.
The Economics Don’t Work. Yet.
Bringing a data center into space sounds like something from a sci-fi screenplay, not a near-term business strategy. The power costs for Starlink satellites run several multiples higher than what terrestrial data centers typically pay. Space-ready solar panels would need to be manufactured at unprecedented scale. Protecting semiconductor chips from the radiation and thermal rigors of orbit won’t be cheap or easy. And there’s still no clear answer to whether AI training can be effectively distributed across multiple satellites, which means much of the work would stay earthbound anyway.
That’s not one problem to solve. That’s an interconnected web of challenges, each dependent on the others.
Musk likely views xAI’s current natural gas-powered data centers as temporary placeholders. The thinking probably goes: suffer through the near term with fossil fuels, crack space-based solar power in a few years, then obsolete the whole terrestrial operation and move on. But that calculation assumes success on a timeline that’s extremely aggressive, even for Musk.
The bigger risk: he’s wrong about the timeline. Or wrong about the feasibility. Or wrong about both.
The Distraction Problem
Here’s what strikes me most about this strategy. We’ve barely scratched the surface of what terrestrial solar can do. Solar efficiency keeps improving. Installation costs keep falling. Grid integration gets better every year. Battery storage, the piece xAI is already buying from Tesla, continues its exponential cost decline.
Musk is chasing what could be a perfect solution while ignoring that there are plenty of good ones available right now.
The energy math is worth noting too. Humanity uses roughly 4 terawatts continuously, or about 35,000 terawatt-hours annually. Energy demand is rising, and AI’s contribution is probably in exponential growth right now. But growth curves flatten. Efficiency improves. Systems adapt. The “terawatt-scale annual compute growth” SpaceX warns about reads less like prophecy and more like extrapolation without guardrails.
The Master Plan Disconnect
Three years ago, Tesla released Master Plan Part 3. It outlined a roadmap to eliminate fossil fuels. Reading that document next to xAI’s natural gas procurement strategy creates a kind of cognitive dissonance. One sounds like someone serious about climate impact. The other sounds like someone optimizing for short-term efficiency while betting the farm on a speculative long-term moonshot.
Business pragmatism and ideological consistency don’t always align. But when a founder has built his entire public brand on a particular vision, the gap between what he says and what his companies do becomes a story worth telling.
Musk is betting that space-based solar will eventually solve the data center power problem. Maybe he’s right. But in the meantime, his companies are doing the exact opposite of what his Master Plans promised. That’s not a minor inconsistency. It’s a fundamental question about whether the vision was ever the point, or just the marketing.
