SpaceX's Space Data Centers: The Multi-Trillion Dollar Gamble on Orbital AI

Uh, Sam Altman, um, used the word homeboy, uh, when tweeting at Elon Musk, and, uh, them two went back and forth a few times where Elon Musk said that, uh, SpaceX is gonna be worth more than the entire planet, um, if they hit all their goals, which seems unlikely. But, um, uh, the, the main crux of their discussion/argument was about space data centers, which is one of the big selling points that SpaceX are trying to, um, sell you with their multi-trillion dollar IPO, uh, deal that went off, uh, a couple of... well, a month or so ago. Um, and in all of that, uh, you know, there's, there's an expectation that, uh, SpaceX's valuation hinges on the ability to, for them to be able to commoditize hardware in space. And not just commodity hardware, but commodity hardware in space, because they want to launch a fleet of, uh, in-orbit data centers that do AI inference tasks, um, whilst, you know, using solar energy rather than, um, uh, you know, energy sustainable or not on planet Earth to be able to, to be able to do the work.

Um, so there's a few things to unpack in all of that because obviously from Elon Musk's perspective, um, they want to be able to provide on-orbit compute, and, you know, clearly you can do it. Um, but the question is, is how do you make it so that it ends up being cost-effective and sustainable? Now, you have to be able to launch rockets, and you have to be able to launch them in a volume that's large enough to put enough compute in space to make it a viable solution. Problem is, with his, uh, latest Starship rockets, uh, they have not yet on repeat often enough. Um, but then the other part of that is really you have to end up making the rocket sort of fully, uh, reusable.

And even during the IPO event, they pointed out that that may also not be the case because there's still gonna be elements of, like, the second stage where it's gonna have to get, for now at least, thrown in the bin, um, and not a reusable rocket, uh, component. And so that sort of makes it quite hard to do the math and come up with an economically viable way of sending, um, silicon chips into space on a repeatable basis until they figure out a chunk of what that's gonna look like. And it's not to say that it's not gonna happen. I, you know, I know enough about SpaceX to [laughs] to say that that, you know, at some point I'm sure it will. Um, but the other aspect of it is, like, how soon compared to the rising energy prices and the chip shortage that we see in consumer goods because people are, like, losing out to AI data centers, both from an electricity cost perspective, but also from a access to actual c- you know, hardware, silicon.

Um, yeah. And so that becomes, that becomes, uh, a slightly different conversation, which I'm sure we'll have in the coming days, weeks, which is of course, you know, trying to offset the huge loads that are required from a compute perspective, both from an energy and compute point of view, to what we can actually supply as a planet, because they're not necessarily aligned. Anyway, that's enough from me. Happy Monday. I'll be back, uh, tomorrow with some more AI news and gossip.

Um, and thank you very much for listening to the AI Briefing. My name is Tom. I will speak to you then. Bye for now. [upbeat music] Why hire when you can partner?

Concept Cloud's leading engineers build your startup's prototype without the overhead. Launch faster. Conceptcloud. com.