Has Elon Musk given up on Tesla’s Grasp Plans, on the electrified financial system, on solar energy as we all know it? From the SpaceX IPO filing launched yesterday, it certain looks as if it.
A recap for these not enmeshed within the Musk-verse: Tesla has launched four Master Plans over time, and whereas particulars have assorted, the by line has been electrification of the financial system. Musk put it greatest in his first version: “the overarching function of Tesla motors…is to assist expedite the transfer from a mine-and-burn hydrocarbon financial system in direction of a photo voltaic electrical financial system.”
However lately, considered one of Musk’s corporations, xAI, has embraced the mine-and-burn hydrocarbon financial system, utilizing dozens of unregulated natural gas turbines to energy its information facilities with plans to buy $2.8 billion more, successfully cementing the fossil gasoline’s position within the firm’s AI operations.
It’s a curious flip for a businessman who constructed his empire on clear power — and who has no qualms directing his corporations to purchase from each other. SpaceX spent $131 million on 1,279 Cybertrucks, and xAI has spent $697 million within the final two years on Tesla Megapacks, it’s grid-scale battery storage techniques that the corporate will use to handle peak masses. However to this point, xAI hasn’t purchased a materially important variety of photo voltaic panels from Tesla.
Solar energy isn’t lacking within the SpaceX submitting, it’s simply all focused on house, which the corporate touts as the way forward for information heart energy. Terrestrial photo voltaic garners just a few mentions — not as an influence supply for xAI information facilities however as an alternative to indicate how a lot better SpaceX thinks space-based photo voltaic will probably be.
It’s no secret that Musk and different Silicon Valley executives have change into obsessive about space-based solar energy. SpaceX says that space-based photo voltaic arrays can generate “greater than five-times the power” of terrestrial ones because of 24/7 illumination. As AI information facilities have run into opposition right here on Earth, CEOs like Musk have began mulling huge server racks in house powered by that 24/7 sunshine. Hammer, meet nail.
Even when SpaceX is ready to convey down the price of boosting a knowledge heart into orbit, the economics are challenging at best. Energy costs for Starlink satellites are multiples larger than what a terrestrial information heart usually spends, and defending chips from the trials of house gained’t be straightforward or low cost. It’s additionally not clear whether or not AI coaching may be distributed throughout a number of satellites, leaving a major chunk of AI work earthbound. It’s not only one drawback that SpaceX wants to unravel, however many.
It’s doubtless that Musk considers xAI’s present information facilities as stopgaps, that after SpaceX is ready to loft gigawatts price of servers into orbit — most likely only a few years away, in his thoughts — he’ll scrap what’s right here on the bottom, pure fuel generators included and never have to consider NIMBYs anymore. The danger, in fact, is that he’s improper.
It’s not simply NIMBYs that Musk is apprehensive about, although. He’s clearly involved that computing calls for from AI will shortly outstrip what we will present right here on Earth. Sprinkled all through the SEC submitting are references to “terawatt-scale annual AI compute development,” which would require energy to match. That’s a surprising determine when you think about that every one the world’s information facilities use round 40 gigawatts today.
That is Musk’s “first ideas” pondering in motion. Sooner or later, he assumed the world will want an extra terawatt price of compute yearly, and he labored again from there. “We consider that third-party estimates on information heart demand are constrained by the sensible provide limitations that exist in a terrestrial context and the facility scarcity could also be far higher than what analysis estimates counsel,” the corporate argues.
Attainable? Certain, I suppose. However contemplate that humanity right this moment makes use of about 35,000 terawatt-hours of power yearly, or about 4 terawatts on a steady foundation. Vitality demand has risen currently, and for AI, it most likely is in an part of exponential development, which might both proceed or stage off. We have now no approach of figuring out at this level, but when there’s one factor Musk is nice at, it’s recognizing a development at its inflection level and extrapolating wildly.
Right here’s the place Musk’s issues settle again right down to Earth. I’m no rocket scientist, however I think that transport photo voltaic panels on a flatbed truck makes use of much less power than sending them into orbit. Plus, space-ready photo voltaic panels will should be manufactured at unprecedented scale. Not insurmountable issues, but in addition possibly a distraction. We’ve barely scratched photo voltaic’s potential right here on Earth, for instance.
The proper doesn’t need to be the enemy of the nice. There’s loads of room to enhance issues right here on Earth even whereas we chase after our desires within the stars.
Simply three years in the past, Musk and his colleagues at Tesla launched the “Grasp Plan Half 3,” which thoughtfully outlined a “plan to remove fossil fuels.” A great place to begin is perhaps xAI’s information facilities.
If you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
