ELON MUSK anuncia TERAFAB: JOINT VENTURE da TESLA com SPACEX para PRODUÇÃO de CHIPS em LARGA ESCALA

Kind: captions Language: en Elon Musk announced the Terafab project today . Terafab is Tesla and XAI's project for a chip factory, right? So, you know, Tesla, SpaceX, and XAI have a lot of artificial intelligence projects and stuff like that. And even now they're using chips from traditional manufacturers, Nvidia, and so on. Elon Musk is saying that this doesn't work for his projects because he needs a scale that these companies are currently unable to achieve. Furthermore, he expects that 90% of the factory's production will be for space chips and 10% for terrestrial chips, okay? It seems like another extremely risky gamble by Elon Musk. Building a chip factory is not a simple thing. If that were the case, China would have managed to do it a long time ago, right? TSMC, the world's largest chip manufacturer based in Taiwan, recently attempted to establish a factory in Arizona, USA, but is still struggling to achieve the same level of production quality and price as in Taiwan. The challenges in chip production are enormous. Will Elon Musk be able to master this? He, who until now has at least no chip production capacity, will dominate this entire market to make chips in such a way that he will produce only for himself, only for SpaceX, for Tesla, and for XAI. Will he be able to do it? Let's understand what this project of Elon Musk is all about, which is once again a risky bet, but Elon Musk has made risky bets before that have paid off very well, right? I am Peter Turgunev, this is the Ancapsul channel, and this news was not suggested by anyone. I found this myself, everyone's talking about it here in the United States, on X too, on the internet, and stuff like that . It's not every day that someone announces a brand new $20 billion factory, is it? Let's understand what's going on here. Anyway, I thank everyone who suggests news on our website, on Visão, and thank you also to you who are watching our video. If you enjoy our content, please leave a like and subscribe to the channel! Well, Elon Musk already has experience, some experience, at least with chip design, right? He's already made several of his own chips for Tesla, right? For Tesla's artificial intelligence, basically for the dojo, which is the training computer for Tesla's chips, right? You all know that artificial intelligence has two phases, right? One phase that we do n't really see is the phase where the model is trained, the phase you use, when you go to the GPT chat and ask a question , and it answers. This is the model application phase, right? The application phase is simpler. The training is much more intensive in computing and artificial intelligence training. So that's what a dojo does. And Tesla has been producing the chips for some time now; they design the chips. TSMC in Taiwan is the company that actually makes the chips , right? And now Elon Musk, he's got this idea for data centers in space. We've already talked about this in another video, right? He integrated XI with SpaceX, precisely to produce cloud computing, literally in the cloud. This has some interesting aspects, but it's also very challenging, right? A challenge that many people think is technically impossible, for example, due to the issue of heat dissipation, right? On planet Earth, it's relatively easy to dissipate the heat from the chip. You use water, you use air, you have a lot of ways to dissipate the heat from the chip. In space. There are ways , of course, but they're not as efficient, because in space you're in a vacuum, so you can't pass water through, for example, OK? Go ahead, the water will cool the chip. How do you cool the water down afterwards, right? Well, the only way you can dissipate heat in space is through radiation. And for that, you would need a huge heat dissipation area to avoid, you know, to be able to dissipate all that energy from the data centers, right? But data centers in space have another advantage, right? We've seen this here. Here in the United States, it's a very big discussion. This hasn't really caught on in Brazil yet, but here in the United States there's a huge debate about it because nobody wants to have data centers nearby, right? Here in the United States, energy operates on a free market, so if a data center were to appear near your home, that data center would consume a lot of electricity, and guess what, your electricity bill would go up too, right? Demand increases for the same supply, so the electricity bill goes up. Furthermore, some people are concerned about pollution and such, although of all types of human endeavors, a data center is one of the least polluting. In fact, it has air conditioning to keep the chips cooler, but apart from that, there's nothing that causes pollution per se, right? It would not be an industrial process that causes pollution. Despite that, you know, they're ugly too, they're huge, closed warehouses where practically nobody works, there's no advantage in terms of work, in the end, right? Brazil has entered the race to host some of these data centers, but so far I don't know if it has been very successful, because the truth is that electricity in Brazil is very expensive, so it does n't end up having that much of an advantage, right? So, and it's a complicated thing, nobody wants to have a data center nearby, but data centers are necessary, right, for artificial intelligence, which will certainly play a very important role in the future. E. Musk's idea is to have SpaceX have data centers in space. Without a doubt , it solves the problem of the local eh, right? In space, nobody owns the space, nobody has to pay land tax, nobody has to pay state tax, nobody has to pay a bunch of taxes in space. Yes, in that respect, Elomask's data center idea is truly excellent . Will he be able to do it? I don't know, but that's where his second bet comes from. He wants to produce his own chips, right? So, here it is. El Musk reveals $20 billion plan for Terafab factory. A project to manufacture chips, memory, and processors, all in a single building, to achieve 1 terabyte of annual computing power. It's a joint venture between Tesla and SpaceX. And remember, XAI, which owns Twitter, is also part of SpaceX. Well, both would make these chips, both terrestrial and terrestrial, right, to create their own chips, right? Tesla and SpaceX CEO Elon Musk announced Saturday night that Terafab's semiconductor manufacturing project will be built on the Traves campus in Austin, Texas, where SpaceX and other Elon Musk ventures are already located. It will be a joint venture between the two companies. He did a live stream yesterday, right? He talked about it, okay? That's his live stream right here. It's still here. Ready. It says here, his live stream starts, and he's talking about Trafábica and stuff like that. I'm not going to play the entire live stream because it's 48 minutes long , but basically he describes this whole process that he hopes to complete with this large factory, right? We either need to build Terafab or we won't have the chips we need to carry out our projects, right? The project is projected to cost $20 billion . This factory in Austin would have the equipment, manufacture logic, memory, packages, testing, lithography, mask production, all in a single building. Mask said that this capability doesn't exist in any other facility in the world and that he's waiting to do all of this under one roof to allow for quick interaction. In other words, make the chip, test it, see if there are any problems, modify it, improve it, correct it, and make it again, right? The owner of Nvidia, James Wang, who would be a competitor of theirs, warns that this is an extremely difficult challenge, right? Remember? Nvidia manufactures the chips, but it does n't actually produce the chip itself. Nvidia designs the chip, right? The company that produces Nvidia's chips is TSMC, which is a large Taiwanese company, a huge global bottleneck, right? So, uh, Elon Musk is thinking about outsourcing some of Tesla's chip production to Intel's foundry and even establishing another company of his own to operate it, he said at that shareholder meeting, right when he launched this, right? We're going to create Terafab, which will be precisely Tesla's grand vision , right? So, it's an interesting thing, it shows an audacious vision from Elon Musk, but the big question is whether he'll be able to do it because it's not an easy thing. What the Nvidia guy said makes perfect sense , right? Here, Terafab, is the next step in becoming a galactic civilization together with SpaceX and X. We're building the largest chip production facility in history, 1 TW per year, right? Combining logic, memory, and packaging under a single roof, right? To achieve as much power as quickly as we do to get as much power from the Sun, we need to send 100 million tons of equipment into space per year. This requires a massive scale, the ability to send millions of tons of mass into orbit and also to make satellites and, uh, with solar energy, to create artificial intelligence and also thousands of Tesla Optimus robots to do this, right? All of these needs, uh, between 100 and 200 GW of chips for Timus alone, plus this satellite artificial intelligence capacity, all of this requires a lot, and we are building Terafab to close the gap between the current capacity demand in the world and our future demands. So, what he's saying is that the companies that exist today can't produce as much as he wants, right? Yeah, it's a complicated thing, really, isn't it? People here point out four important things about that, right? Well, first thing, Elon Musk's first challenge, right? He wants to do the whole process in a single plant. It's not very clear yet up to what point in the process this goes. For example, TSMC is the foundry, right? She's the one who makes the chips themselves, but she uses them for masking. They use the ASML machine, right? which is a Dutch manufacturer, the only one in the world that produces this type of mask, right, for printing the chip, right? So, we don't know if Elomas, who is also going to enter this area , will also produce his own mask or if he will buy it from SML, right? So, in other words, fitting everything into a single floor plan is a huge challenge, right? Well, to the best of our knowledge, this doesn't exist anywhere in the world. It's such a targeted technology that there are several specialists in each part of the process. Nobody does everything alone. Elon Musk is proposing to do the whole thing himself, right? There will be two main types of chips , right? Well, according to Musk. One will be designed primarily for Optimus and Tesla vehicles, which are built and trained to be fully autonomous. Musk said the chip will be especially for Optimus Prime because he expects the volume of units to be 10 to 100 times greater than the volume of cars. Wow, Elon Musk is hoping to produce more optimal robots than cars, right? The other chip, called D3, will be specialized for space environments and will be made for data centers that will be placed in low Earth orbit, requiring them to be powered by solar power, right? One way to mitigate the problem of heat dissipation in space is precisely by making chips with very low power requirements. If your chip uses very little power, it has very little power to dissipate. So that's one way you can do that kind of thing, right? Perhaps that's the direction Elon Musk is taking in this story, right? He's talking here about artificial intelligence satellites, which is his vision of putting satellites into orbit and so on. So, those are the main points here, right? Now, remember, this is a very big challenge, right? Yeah, like the guy from NVIDIA pointed out, right? NVIDIA specializes in artificial intelligence chips . Google is currently developing the design of its chips, but even Google contracts TSMC to manufacture them. Elon Musk wants to do everything himself, right? So, remember, TSMC has been producing, trying to set up a factory in Arizona for some time now. Here's a timeline of their factory. They started in 2020 and still have n't begun large- scale production, okay? They're already producing something, but it's still almost a test production of chips there in Arizona, in Phoenix, Arizona. And they also hope to begin large-scale production. Here, look. Well, from October 2025 until CMC Arizona, volume production of Nvidia chips and Blackwell GPUs for N4P processors will begin. So, uh, production started in October 2025, it took 5 years to produce this and a total investment of 16 million dollars. Not here, no, no. Look, the investment expanded from the original 12 billion to 165 billion dollars to build this factory in Arizona. Elon Musk is talking about 20 billion dollars. He might be being too optimistic, right? I do n't know. He might have some kind of secret there. And remember that TSMC already has the expertise to produce chips; it already produces these chips in Taiwan. She just has to bring it to the United States, and that alone is proving incredibly expensive . The sheer size of the challenge that El Musk is facing is truly daunting, isn't it? He 'll manage it, I don't know, he's already achieved things that impressed me, okay? Starlink, I didn't imagine it would be as successful as it's turning out, did you? When he started talking about, "Oh, I'm going to build antennas to put in rural areas," I said, "Damn, there won't be that many people there, there aren't that many rural customers ." But when you think about rural areas all over the world, right? The entire rural Brazil, it's true, has many more customers in São Paulo than in the rural region, but now you have a single provider that serves all these rural people in Brazil, it's become a mega-service already, if I'm not mistaken, it's already one of the largest internet providers in Brazil, it's Starlink, right? And the same is true for many countries around the world. So, it was something I didn't imagine would work out so well, and it worked out incredibly well, this Starlink thing. Tesla cars are a success too, right? You may like Elon Musk, you may not like him, but he is undoubtedly a smart guy who gets things done. Will Tera Fab work or will it fail? I don't know. I do n't know. Let's see. Thank you for watching our video until the end. There are some links on the screen and in the video description to help you.