Tesla has just signed a $16.5 billion contract with Samsung to make their next generation of AI chips. And I think that there's actually a bigger strategy here that is going to apply to a lot of different companies. I think the reason why they're making these chips is also going to be fascinating. So I want to dive into not just the deal, but why the deal is being made and what this means for the overall industry.
Obviously $16 billion is a lot of money, but we know that a lot of money gets thrown around in the chip space, especially when you're making things for AI. So we're going to get into all of that on the podcast today. But before we did, I wanted to mention if you ever wanted to try all of the different AI models I talk about on the podcast, I'd love for you to go check out my platform, which is called AIbox.ai. Basically we allow you to compare all the top models side by side with each other, all on one platform.
So you get one subscription, you get 40 different models, you get a bunch of image models. I recently, if you're seeing my screen share, asked it for a LinkedIn influencer traveling to Japan with fans, but I got it to generate the image with Flex 1.1 Flow, with Ideogram, with ChatGPT Image 1, and with Flex 1 Chanel. Anyways, all different models come up with different like images and different image qualities and concepts for a prompt. And so I love the ability to be able to try one prompt and get it from a whole bunch of different AI models.
And you can also do this with text, you can also do this with audio. So we have a ton of different models, you can compare them all side by side, figure out what model basically does what you need to best, and you can chat with them all AIbox.ai, but let's get into what's going on with Tesla. And the first thing I'll do is share the tweet that Elon Musk recently made. He said, Samsung's giant new Texas fab will be dedicated to making Tesla's next generation A16 chip.
The strategic importance of this is hard to overstate. Samsung currently makes AI4. So basically, this is a chip that goes inside the Tesla help itself driving. Samsung was making their AI4 chip, but for their AI5 chip, they actually went to Taiwan Semiconductor Company.
And then for their AI6 chip, they went back to Samsung. Now, what's going on here is a good strategy. I guess one other thing that I'll bring up here is, I think Elon's trying to do like the, you know, tariffs and other reasons, trying to get a lot of stuff made inside of America. So that could be an element to this, but basically he said, TSMC will make the AI5, which just finished design, initiated in Taiwan and then in Arizona.
So TSMC is building a ton of fabs in Arizona. I used to live kind of in that area. And honestly, it's pretty insane what they're doing down there. And so basically what it feels like though, is that Elon is kind of bidding out his suppliers against each other, right?
So he's not just picking TSMC to do everything. He's not just picking Samsung to do everything. For like different models of chips, which will go in different cars and different generations of cars, he's going back and forth between different companies. So I think some people would like basically say, this is a bad idea because you're not going to get, you know, super, you're not going to get super sustainable or like very similar results.
There's going to be issues from one company to another. Like I think a great example of someone that doesn't do this is Apple, right? I'm sure some people know more about Apple supply chain than me will complain and say like, yes, they do. They use a whole bunch of companies.
But like basically for iPhones, Apple is deeply embedded with Foxconn in China and they're not bidding this out to, okay, now this company building something in America, we're going to just like ship all of our iPhones to get built in America this year. And then in the next season, we're going to go over to Vietnam where there's new, right? Like Apple doesn't want to do that. Their supply chain is so massive.
This project is so massive. They don't just flop around. And of course the demand for an AI four chip, AI five chip, AI six chip isn't going to be exactly what the demand for an iPhone, which is way higher. But still, I think just conceptually, you see some companies do this and some companies do not take the risk.
What is the benefit of it? Obviously the risk is like quality or, you know, some sort of like bugs or issues that might not happen from one to the next, right? Some sort of operational inefficiencies. What is the benefit?
The benefit is the cost of this. It goes down a lot when you're getting these companies to bid against themselves. And also if you have a look at the backstory, basically Samsung was building out this whole new fab in Texas and allegedly, I'm not, I don't know, allegedly they were struggling to get people to basically secure any big suppliers or to buy what they were building there. And so, um, and this is according to an article on TechCrunch, I don't know how legit it is, but in any case, that's the allegation.
And it would make sense that if that was the case, Tesla could go in and negotiate a very good rate, right? They're struggling to get any big people. They're like, all right, you make a whole new chip. They've worked with them in the past.
It's not a brand new person. They're capable of doing it. Um, but, uh, but like how good are they? Are they as efficient as TMC?
All right, I'm going to get to that in a second because Elon followed up on the street by saying Samsung agreed to allow Tesla to assist in maximizing manufacturing efficiency. This is a critical point is I will walk the line personally to accelerate the pace of progress and the fab is conveniently located not far from my house. Okay, it's in Texas. So yes, that sounds like it would be more, uh, that'd be easier for Elon to be able to get over there and, uh, walk the lines like he does the Tesla line.
But I think more importantly, the bigger story here is the fact that they agree to let Tesla maximize their manufacturing efficiency. So basically Tesla probably has some concept of what's going on at Taiwan Semiconductor and Taiwan Semiconductor Manufacturing. And also they have their own manufacturing for their cars. So like they're manufacturing, they're, you know, they're all about manufacturing efficiencies.
This is the big thing they do. And they're literally going there to help. Now, is this just because Elon's like, well, we just want the chips better and faster and we want to make sure the quality control? Probably an element of it, but Samsung should be happy about this as well because if there's another company that has mastered, you know, some sort of manufacturing efficiency thing and they're willing to come bring it to your fab, why would you say no, right?
If they leave you tomorrow and they gave you that efficiency and you can go sell off, you know, chips or, uh, you know, manufacturing capacity for another company, like that makes perfect, that's amazing, that's brilliant. So it seems like a real big win-win for both companies there and Tesla's still probably saving a ton of money by bouncing back and forth between their suppliers. Some people in the comments are saying things like, um, they said, uh, you know, Elon controls the silicone, he controls the future and they're talking about like, but like it's, I mean, controls it, but like Samsung's building it. It's not like he's literally building the fab like some other companies are.
Um, some people said this Tesla-Samsung move feels like a power play to localize critical AI infrastructure in the US. AI6 won't just be a chip, it's a geopolitical asset. I mean, other than the fact that sounds like it's probably written by ChatGPT, though, it won't just be a chip, it's a geopolitical asset. Yeah, that's probably written by ChatGPT, but the concept is good.
Um, so I think this is pretty interesting and, uh, definitely going to be a big deal. $16 billion. Now, what's interesting is Elon also said that this isn't going to be, uh, by the end, it might actually get, uh, get bigger than the $16 billion. So this could be interesting.
He said, quote, actual output is likely to be several times higher. Um, so this should be quite interesting. Now, what's interesting to note is that Tesla actually switched from NVIDIA's drive platform to their own custom chip back in 2019. So it's not like a brand new thing they're doing.
They've been working on, you know, their own chips for like six years almost. And that new chip that they built, kind of their own in-house thing, is called FSD computer, um, or hardware three. And it's, in 2019, it was in all of their EVs. So that was actually made by Samsung.
But what's interesting about it was it was actually two duplicate systems right next to each other on the same board. Basically, it's this redundancy where one of your chips, like fries or brakes or short circuits, has some sort of issue. There's two right next, two systems right next to each other. And so there's built-in redundancy.
If you're self-driving, you don't want some sort of issue to make the self-driving completely stop working. So there is a backup built into it, which is pretty interesting. Um, since then, they're basically all their focus on these new custom chips, um, has grown a lot. AI chips are basically the center of not just the cars, but also their push from, um, to get into AI and also into robotics with their Optimus robot.
So these chips theoretically will be used in Optimus robots and Tesla cars. It's kind of interesting as they design newer, newer, newer versions, what they have to build there, uh, for all of this to work. So thank you so much for tuning into the podcast. I hope that you learned something new about basically what's going on in the industry and with Tesla.
If you learned anything new or if you appreciate the podcast, I would love a rating and review, a rating over on Apple. I read all the reviews and, uh, on Spotify. If you hit the about cap, you can leave some stars. Thanks so much for tuning in.
Make sure to go check out AIbox.ai, uh, as it is new with a beta. And I hope that you can find a ton of amazing new, uh, AI models on there. The link is in the description for that. Thanks so much for tuning in and I will catch you in the next episode.