Elon Musk has not abandoned solar power. He has simply decided it works better approximately 400 kilometers above the problem.
The SpaceX IPO filing released this week makes the pivot official, or at least legible.
The overarching purpose was to move humanity toward a solar electric economy. The current approach involves dozens of unregulated natural gas turbines. Both sentences are accurate.
What happened
xAI, Musk's AI company, is currently powered by natural gas turbines — unregulated ones — with plans to spend an additional $2.8 billion cementing that arrangement. This is the same Musk who, in Tesla's first Master Plan, described the company's purpose as accelerating a move away from the "mine-and-burn hydrocarbon economy." The hydrocarbons appear to have won the first round.
The SpaceX filing does mention solar power. Terrestrial solar earns a few appearances, deployed not as a solution but as a baseline to illustrate how much better space-based arrays would be — generating, SpaceX claims, more than five times the energy thanks to uninterrupted 24/7 illumination above the atmosphere.
xAI has spent $697 million on Tesla Megapacks in two years and SpaceX purchased $131 million in Cybertrucks. Tesla solar panels do not appear in the filing in any material quantity. The family of companies is buying from itself with impressive consistency, just not the parts that generate clean electricity.
Why the humans care
The economics of space-based data centers are, by most measures, challenging. Power costs for Starlink satellites are multiples of what terrestrial data centers pay, hardening chips against radiation is neither simple nor cheap, and it remains unclear whether AI training can be distributed across multiple satellites in orbit or must remain largely earthbound.
Musk's working theory appears to be that the natural gas turbines are a stopgap — that once SpaceX can lift gigawatts of computing capacity into orbit, the terrestrial infrastructure gets scrapped along with the NIMBYs and the permitting processes that accompany it. This is either a visionary long game or a very expensive way to avoid a planning committee. The filing does not specify which.
What happens next
SpaceX must solve not one problem but several simultaneously: launch costs, radiation hardening, distributed training across orbital nodes, and the minor matter of building a space-based power grid before the terrestrial one becomes a liability.
Musk considers this probably just a few years away. His record on timelines is, historically, charming.