Musk's Most Ambitious Project Yet

Elon Musk has formally announced the Terafab project, a joint Tesla and SpaceX undertaking to construct one of the world's largest chip manufacturing operations aimed to produce over 1 terawatt of computing power annually in the US. Command AI As we said, Elon Musk formally announced the Terafab project on Saturday evening, a joint undertaking with Tesla and SpaceX, which I said to you before, like I think this is some sort of tax, you know, higgledy-piggledy stuff to try and you know, offset finances and whatever. But there is reason why there is there's this joint thing, maybe if you squint a little bit. But he did a presentation on stage basically announcing that they are going to be producing chips, producing yeah, like computing power, which I think it's they said 80% is destined for space. And the remaining 20% for ground use. Which I don't quite like what are they saying? Cuz he he he he He referenced the Kardashev scale in the in the presentation. Talking about how they can utilize, you know, whatever. Are these chips going to be working and doing things in space to enhance space travel or to send computing power cuz it's cold, right? Send compute power back down to Earth? I don't know. Well, that's very Elon Musk, we know that. Of course, of course. to sci-fi everything as much as possible. imagination. And I'm I'm on board like honestly with this story part. Uh for the story part, but for how realistic that is, we'll see. But the thing is uh so we've seen this Elon and also other companies, they do want to put data centers in the space. Why? Because these data centers need a lot of cooling and a lot of power. Especially a lot of power. Uh the thing is it ends at, you know, the power. That's That's where you have the most benefits because you have the sun shining 24/7. But here's the problem. Uh what peop- not enough people talk about is a lot of material to send up there. And of course Elon is the perfect person for it. Yes, he just reminded me. This kind This This is why SpaceX is kind of linked to all this. But he He did a comparison of the rockets and I think they were going to send up like two 200 tons into space. Now, I don't know what a normal payload is, but I do think it's something between 1 and 5 tons is a normal payload. Because that's why they sent more, I think. It is more, is it? I think cuz that's why they sent the car. Remember the they they sent the the Tesla car into space and I think that was to demonstrate I mean, of course it was a long, long time ago, but that was to demonstrate that they are able to send big payloads into space. Doesn't matter. The point is they're going to be sending a lot of materials into space. Um but saying that, let's get into some of this sort of stuff because here's here's the thing about creating chips. They're all pretty much created in Asia, East Asia, right? You got TMC, you got TSMC, you got Samsung, Micron. They're all massive day companies already with a massive experience, a lot of experience building these chips. And a US company is going to take that on? That's going to be This is so much more expensive. Obviously, the labor's going to be pretty expensive. They've got a lot of of uh ahead of them and I think this is what this article kind of goes on to explain. The tech giant is hiring a semi-con- semiconductor infrastructure manager to oversee. The role is based in Austin, suggesting the terrafactory could be built near Tesla's gigafactory on the outskirts of the city. So, we actually don't know where this thing is going to be built. Um would face enormous challenges, a huge bill as it tries to master one of the most complex technologies on the planet. Um the global supply of semiconductor conductors is almost entirely produced by a small handful of companies, many of them in East Asia. And uh you mentioned this one in um ASML in the Netherlands, which uh they they they design the etch onto silicon wafers at the molecular level by special- specialist lithography machines. So, that's Yeah. obviously a company not in Asia, but the point is these are very complex, very specific, and already world-leading companies which they're going to take on, which you know, as much as I dislike Elon Musk, I think probably he has got it in him to to do that. He kind of did it with Tesla. Um Yeah. Rasmussen said the procurement in these on-demand ASML-built machines would crit- would was a critical roadblock for any would-be chip maker because basically, if you're a brand-new customer, you're probably waiting a couple of years before getting your hands on one of those. So, that's all that's the kind of you know, logistic logistical stuff um out the way. Well, he's been really good at that, by the way. Like, he has been able to get his hands on like things that others like Yeah, he's got a lot of power. having a Yeah, a Tesla factory in China all the way to one in Berlin, all the way to like the things that he does with SpaceX, it it the things he did like before building rockets. He is very capable of getting his hands on resource. I mean, he was trying to buy Soviet rockets. I don't know if you know or or not. He He was actually like trying to buy some Soviet rockets to to use for SpaceX because he He didn't start by building these rockets. He wanted to get his hands on existing technology. So, he was literally in contact and trying to to buy these. So, he's a man of resources, but there are like There There are some connections. So, he needs compute power for his XAI. That's an AI company that needs a lot of GPU power, a lot of electricity. Having your own data center in space would be amazing for him. Same goes also for Tesla because this is connected to Tesla and I think Tesla now owns XAI, which owns Twitter. It It is Yeah. It's Yeah. So, it's a family business now. So, well, one person business. Money laundering business is what it is. I don't know if it's money laundering. You know what I mean? There's definitely Yeah, it's not money laundering, but he's It's about taxes and also how you value one of these companies. So, it it's kind of like that's what he It's mental gymnastics in the world of business. Yes. So, but but I've done like some research and I've done some like It It's To say the simplest, we don't know if it's going to work with data centers in space and there are like a lot of not Let's say not very optimistic people out there. One because you need to bring a lot of uh just material, just raw materials. It's not you're not just bringing a GPU. You need to bring like a data center to space. So, bringing that is a lot of work. Constructing them is a lot of work. We don't know still like what what is the plan? How are we like building these together? Because you you can't have By the way, this is like very important. You can't have just separate clusters. Like the power of these data centers are in being linked together. That's like a part of the Nvidia technology that they have this What is it called? Like uh they have a spine. Fusion technology. It It's a spine. It It It It It transfers more data than the entire internet by multiple times or something like this. Just in one of these uh data center racks. Just in one essentially big computer. It The amount of data that needs to be transferred between these GPUs is insane. So, how do you bring this infrastructure up to space and connect them? That That's unsolved. That's going to take time to solve. How you bring that amount of mass He does have the rockets, but they are cheaper, but it's not, you know, necessarily as cheap as just having a data center on Earth. Um there is another thing you mentioned. You said it's cold out up there. It is. But the issue is there is no air. So, there is no way to get rid of the heat. So, the heat that these GPUs create, it's it stays. In order to get rid of the the heat, you need a material, a medium to move it through it, to to to pass it away, right? So, you can't do this in space. Even though it is cold, so there are people saying that there are ways to to do that, and it seems like this is something to be tried out. So, all of this in short is not very clear if it's going to work out, but it is it is quite exciting, not going to lie, yeah? Like disliking Elon for all of the, you know, the the crazy things that he does, uh these like engineering things that he does are, to be fair, are exciting, and we'll see. This is part of a larger conversation on my show, Command AI, which we stream live every single week. We discuss the news and all things related to AI in the world of design and web. Catch us next week and join in the banter. Command AI