Elon Musk gave up solar energy (on Earth)

Has Elon Musk abandoned Tesla’s master plans, the electrified economy, and solar energy as we know it? SpaceX’s IPO filing released this week certainly looks like it.

Recap for those not caught up in the Musk-verse: Tesla has released four major plans over the years, and while the details have varied, the dividing line has been the electrification of the economy. Musk put it best in his first version: “The overarching purpose of Tesla’s engines…is to help accelerate the transition from an economy of extracting and burning hydrocarbons to an economy of solar electricity.”

But more recently, one of Musk’s companies, xAI, has embraced the economy of extracting and burning hydrocarbons, using dozens of deregulated natural gas turbines to power its data centers with plans to buy an additional $2.8 billion, effectively cementing the role of fossil fuels in the company’s AI operations.

It’s a strange turn for an entrepreneur who built his empire on clean energy — and who has no qualms about steering his companies to buy from each other. SpaceX has spent $131 million on 1,279 Cybertrucks, and xAI has spent $697 million in the past two years on Tesla Megapacks, grid-scale battery storage systems that the company will use to manage peak loads. But so far, xAI has not purchased a significant number of solar panels from Tesla.

Solar energy is not missing in SpaceX’s portfolio, it is focused solely on space, which the company is touting as the future of data center energy. Ground-based solar power is getting some mention — not as a power source for xAI data centers, but rather to show how much better space-based solar power SpaceX thinks it will be.

It’s no secret that Musk and other Silicon Valley executives have become obsessed with space solar energy. SpaceX says solar arrays in space can generate “more than five times the power” of ground-based arrays thanks to 24/7 lighting. As AI data centers face opposition here on Earth, CEOs like Musk are starting to think about creating large server racks in space that run on sunlight 24/7. Hammer, meet nail.

Even if SpaceX is able to lower the cost of boosting an on-orbit data center, the economics are challenging at best. Power prices for Starlink satellites are multiples higher than what a terrestrial data center would typically spend, and protecting the chips from the rigors of space won’t be easy or cheap. It’s also not clear whether AI training can be distributed across multiple satellites, leaving much of the AI ​​work on Earth. It’s not just one problem SpaceX needs to solve, there are many.

Musk will likely view existing The danger, of course, is that he is wrong.

However, Musk’s concern isn’t just about NIMBYs. He is clearly concerned that the computing demands of AI will quickly exceed what we can provide here on Earth. Scattered throughout the SEC filings are references to “annual AI computation growth on a terawatt scale,” which would require power to match. This is an amazing number when we consider that all data centers in the world use about 40 gigawatts today.

These are the “first principles” that Musk thinks about while working. At some point, he assumed that the world would need the equivalent of an additional terawatt of computing each year, and worked back from there. “We believe that third-party estimates of data center demand are constrained by practical supply constraints that exist in a terrestrial context, and energy shortages may be much greater than research estimates indicate,” the company says.

maybe? Sure, I think. But let us take into account that humanity today uses about 35 thousand terawatt-hours of energy annually, or about 4 terawatts on an ongoing basis. Energy demand has been rising recently, and for AI, it is likely in a phase of exponential growth, which may continue or stabilize. We have no way of knowing at this point, but if there’s one thing Musk is good at, it’s spotting a trend at its inflection point and extrapolating wildly.

And here Musk’s problems come back to earth. I’m not a rocket scientist, but I suspect that shipping solar panels on a flatbed truck uses less energy than sending them into orbit. In addition, space-ready solar panels must be manufactured on an unprecedented scale. They are not insurmountable problems, but they may also be a distraction. For example, we’ve barely scratched the potential of solar energy here on Earth.

Perfect doesn’t have to be the enemy of good. There is so much room for improvement here on Earth even as we chase our dreams in the stars.

Just three years ago, Musk and his Tesla colleagues released “Master Plan Part III,” which thoughtfully outlined a “plan to divest from fossil fuels.” xAI data centers may be a good starting point.