SpaceX's recent IPO filing reveals a significant shift in Elon Musk's energy strategy. While Musk built Tesla on the promise of electrifying the economy and moving away from fossil fuels, his AI venture xAI is now relying heavily on natural gas to power its data centers. The company has deployed dozens of unregulated natural gas turbines and plans to purchase an additional $2.8 billion worth of the equipment. This marks a notable departure from Musk's original vision articulated in Tesla's first Master Plan, which stated the company's purpose was "to help expedite the move from a mine-and-burn hydrocarbon economy towards a solar electric economy." Despite Musk's clean energy legacy, xAI has shown limited interest in Tesla's solar products. The IPO filing shows xAI spent $697 million in the past two years on Tesla Megapacks—grid-scale battery storage systems—while purchasing no significant number of solar panels. Meanwhile, SpaceX itself spent $131 million on 1,279 Cybertrucks from Tesla. The cross-company purchasing demonstrates Musk's willingness to direct business between his ventures, yet solar panels remain conspicuously absent from xAI's procurement. SpaceX's filing does address solar power, but only in the context of space-based applications. The company argues that orbital solar arrays can generate "more than five-times the energy" of terrestrial installations due to 24/7 sunlight exposure. As AI data centers face increasing opposition on Earth due to power demands and environmental concerns, SpaceX is positioning space-based solar as the solution for future data center power needs. This aligns with broader Silicon Valley interest in placing large server facilities in orbit. However, the economics of space-based data centers face substantial hurdles. Power costs for existing Starlink satellites already run multiples higher than typical terrestrial data center expenses. Protecting sensitive computing hardware from radiation and other space hazards would add significant expense. Additionally, it's unclear whether AI training workloads can be distributed across multiple satellites, which may keep much of the computationally intensive work bound to Earth regardless of where solar panels are located.