These are the tech minds of Silicon Valley who control trillion dollars in capital flows who are listening to this podcast. There's value in that.
Hello everyone and welcome to another Semi Doped podcast. I'm Austin Lyons with Chipstrat, and with me is Vik Sekar from Vik's newsletter. All right, hot off the press, Vik. I saw a tweet this morning, 8 a.m. Central time, and it's a picture of Lip Boutan shaking Elon Musk's hand, and Intel said, 'Intel is proud to join the Terrafab project with SpaceX, XAI, and Tesla to help refactor silicon fab technology. Our ability to design, fabricate, and package ultra-high-performance chips at scale will help accelerate Terrafab's aim to produce one terawatt per year of compute to power future advances in AI and robotics. It was fun hosting Elon Musk at Intel this past weekend.' Did you see this? This is hot off the press. Did not see this coming myself.
No, I did not actually. This is the first time you're reading out the news to me.
Okay, perfect. Well, let me keep going. So then, Lip Boutan retweeted it, quote-tweeted it, and said, 'Elon has a proven track record of reimagining entire industries. This is exactly what is needed in semiconductor manufacturing today. Terrafab represents a step change in how silicon, logic, memory, and packaging will get built in the future. Intel is proud to be a partner and work closely with Elon on this highly strategic project.' Rocket emoji. Rocket emoji.
Yeah. Yeah. And which of course then I thought, I wonder if Lip Boutan actually wrote this or if he has someone write it for him. But okay. So let's unpack this quick because it's hot off the press and it's sort of going to be top of mind for people. So, as a reminder of the Terrafab, I think the first question, if people haven't been tracking this closely, because I think the Terrafab announcement was only like two weeks ago or something, but Elon had a little live video presentation that he made from a power plant in downtown Austin, Texas, and he basically laid out a problem that there's not enough chip capacity for his ambitions going to space, going to Mars, all these things. He laid it out nicely actually, working from first principles of like what his vision is and then working backward through all the bottlenecks, and it boiled down to where he wants to get long-term is going to require like 1 million wafer starts per month, but the industry is not there, and therefore they wanted to build their own fab to try to increase the amount of chips, the wafers produced per year essentially. And this Terrafab concept that he put forth is sort of touted as like a fully vertically integrated fab. So design, lithography, fabrication, memory, advanced packaging, everything under one roof. And I think the idea is to, by putting it all centrally located, which is probably a conversation for another time if how true this is, but that you could iterate really quickly and just like cut out a ton of inefficiency if you can build and design and build everything soup to nuts under one roof. And so Elon's vision was initial target, let's do something. We'll start at the Tesla Gigafactory in Austin as like a prototype and we'll try to ramp up to a production facility, but we'll try to hit 100K wafer starts per month. And that was about it. So it's a lot of just vision casting. So Terrafab alone has a lot of questions of like, is this possible? How are they going to do this? Does it actually solve the problem? So on and so forth. We won't get into that. But what I found interesting is then this announcement just a couple weeks later, and there's very little details in those tweets, but Intel Foundry really saying we're going to come along and help accelerate the Terrafab vision.
Okay, nice. I mean, I guess this is a second American fab if it comes to fruition. But I heard this whole thing about Terrafab and I was like, aren't we going backwards? Like there were companies who used to make their own chips and then sell their own products, aka Intel, and then you know, now we're like, okay, no, we'll just make the chips separate and we'll make the product division separate. And then AMD, you know, the what, who's an earlier CEO of AMD? He's like, he's like, 'Oh,' he said something like, 'Oh, we won't have a fab over my dead body.' Or something like that. He's like, he was a real fab guy.
Jerry Sanders. He said...
Yes. Yes. He said, 'Real men have fabs.' Real men have fabs. And yeah. So, so we're going back now. So, not only will we have fabs, we'll have memory, we'll have packaging, we'll have everything. We'll even put them on rockets, ship them to space, all in one roof. This is like going back but with steroids on. Like this is like, I don't know.
Yeah. No, let's unpack that. So this is actually really interesting. So back in the day when the semiconductor manufacturing industry first started, it was actually beneficial if you could design your own chips and manufacture your own chips in one house because you could do extreme co-design from design through manufacturing. So the question is, well, why don't we still do that? And obviously the economics of producing chips has gotten so expensive that you couldn't support the cost of a new fab with just your own chips alone. So people ended up needing to decouple and say, okay, well, we're going to manufacture for hundreds of chip designers, you know, and to support the huge fixed costs. So then the interesting question is like, okay, well, actually if it's best to do everything under one roof, the question is how can it economically be possible? Like what's going to be different? Is it going to actually have that volume of chips from one customer or is there some sort of like manufacturing unlock that makes this possible? Now, the next natural question is like, what about lithography? Because that's where the cost is. Like, you're going to have to buy a ton of ASML EUV machines. And unless you can really get rid of that huge expense, like it just doesn't seem to add up as we're talking about it out loud. But that's where my head goes is like, well, maybe there's something economically where they think it actually could make sense to bring this all back and be an integrated device manufacturer. The other thing I don't understand is like bring it all in, but at what level? Like, are you going to do the chip manufacturing and packaging? Okay, I get it. Then what are you going to do? Are you going to do the assembly, like component assembly? Like let's say you're making an optical transceiver now. Are you going to start assembling components? Okay, then what about, you know, are you going to source things like resistors and capacitors to build your PCBs? Oh, by the way, are you going to make your own PCBs?
Yeah. Well, okay. So, let's... Okay. I need to go deeper on this topic. I'm just speculating, but I have a feeling that Elon would say, 'Yeah, manufacture stuff. No problem. We do that. That's what Optimus is for in the long run. So, we will try to set it up and automate it in the long run.' I'm pretty sure SpaceX actually has some sort of like PCB manufacturing plant. Have you heard this before?
No. Oh, so they already have that. Okay. Yeah, they have a Texas facility for Starlink. So they actually, so they have some experience here. But so in theory maybe they know enough and have a vision for trying to automate a lot of it that they think that it's possible. But then it raises the question, what's Intel have to do with it? What does Intel, if you're going to do it all, what does Intel Foundry have to do with it? So, I have also a hunch that there's a long-term vision for doing it all, but a realization that in the near term, somehow they need to partner with an existing foundry that's American-made that has fabrication, that has advanced packaging, for example, maybe in Arizona and New Mexico like Intel does, Intel Foundry does. It will be fascinating to understand what's Intel's play here because best case scenario for Intel, this is actually Foundry's potential first public customer proclamation of a whale customer, not just like a little government project or a little customer that's kicking the tires, but this could literally be the first like big customer if it ends up as like these are chips for Optimus or these are chips for, you know, even for Elon's like space ideas, like that could be, you know, hundreds of thousands of chips. And then I think what I wrote, I tweeted about this this morning was just like, 'Wow, this was not on my bingo card that Intel Foundry's first flagship customer publicly is Elon Musk.' You know, you thought eventually that it would be like Google or Amazon or someone, Nvidia, right? Right. Apple, who knows, right? But not Elon. But then I thought about it and I was like, 'Wait a minute. All of those companies don't really want to ever say anything publicly that they're working with Intel Foundry because like they don't want to lose their place in line with TSMC and they don't want to maybe like give it away to their competitors.' And so like they have like zero reasons to come out and say they're working with Intel Foundry, which is a very tough thing for if you're Intel Foundry and you're trying to like get people to believe that you're legitimate. But then Elon like does his own thing, says his own thing, he is his own thing, it's xAI. So you know, in all this conversation about OpenAI and all the crazy stuff they do, which we'll get to. Then there's Anthropic killing it on the enterprise side. Sometimes we forget that there's this thing called Grok and there's this thing called xAI which is like occupying so many data centers and he needs a lot of chips too to handle that thing.
Totally. Totally. Which that we should have a conversation on that sometime because the question is what is xAI's long-term vision and are they going to continue to fill data centers?
Yeah. Yeah. It's, I don't know. I feel like it's not been much in the news these days, right?
Agreed. Agreed. I know that I'm pretty sure a lot of people left after the merger, so there's a lot of questions about what they're going to do. And then I know that they're saying like, 'Hey, we need to make sure that the next Grok is good at coding,' just like OpenAI is saying because everyone's realizing that the stable outsized revenue today is enterprise, as Anthropic is showing with Claude.
But okay, let's move on. I guess very last thought from the tweet with Lip Boutan and Elon shaking hands. I thought, 'Wow, Lip Boutan and Elon see eye to eye quite literally because Elon's 6'2", I think, and Lip Boutan must be like 6'1". I know he played basketball in college and...
Oh, literally. I thought you were saying something like, 'Wait, did they have a beef, like an argument in the past that they're looking eye to?' No, you mean literally eye to eye.
Okay. Okay. What's next? So, what should we talk about next?
Uh, yeah. So, as a, you know, people who run podcasts, I'm sure you've seen that, uh, OpenAI bought TBPN, which is a podcast for an estimated hundred million, probably 300 million.
Yes. Low single hundreds of millions or something like that. Low hundreds of millions. Yeah.
Yeah. So, I don't know. 100 to 300. That's still a lot of zeros, you know.
Has Sam called you to buy our podcast or anything? Just check in because...
Not yet. Not yet. Sam, if you're listening, uh, you know, I'm sure you can figure out how to find me. I use ChatGPT all the time. Just have it message me.
You have all his info. So, speaking of listeners, TBPN stands for, you know, Technology Business Programming Network, which, uh, this podcast does a three-hour live stream every day covering tech, AI, and, you know, the business of all of this stuff. They do a really good job. John Kugan and Jordi Hayes are like amazing hosts. You know, they're dressed like, you know, it reminds me of like Lex Fridman. They're dressed in like suits and stuff and it's a great show. I mean, they've had some really good guests on. Name it, you know, Sam Altman, Satya Nadella, all these people have been on there. Great podcast. It's an 11-person team, so you can kind of divide it up and see like...
million better than the Drop deal, I think, per person. Think about it.
Is it? I don't know. Anyway, but yeah, so it looks, so this is very interesting because the question that comes up is why would OpenAI buy a podcast? Like why? I mean, it's like, is it literally required for what they do? Aren't they planning to go IPO? Don't you want to be a little responsible about what kind of stuff you buy? Didn't they just kill Sora too, by the way?
Right. Right. Right. Yes. So, yes, they're trying to kill side quests. They're trying to double down on Codex. They're trying to increase revenue ahead of an IPO. And so then to your point, it's a very interesting question to say how does acquiring TBPN help those goals, increasing revenue, preparing for an IPO.
And so on. And it looks like the show will continue to go on as is, which is great. And you know, the hosts now, John and Jordi will report to their chief global officer, I guess, to assist with their like OpenAI's communications or something. It seems like a rather expensive acquire to get a comms team. Now there are plenty of people who know how to do comms.
Well, okay, let's... No, that's actually really interesting. There is this new media as you and I know through writing Substacks, having a podcast. Obviously, these guys are probably like the foremost experts in sort of new media meets Silicon Valley meets AI meets technology. So, it's not a bad idea from an acqui-hire perspective to say we should get closer to these guys and basically try to get them to help us rethink all of our communications. I hadn't thought about that. I thought about it more of just like what is the benefit of OpenAI to owning TBPN, but maybe it's more like what is the benefit of like John and Jordi essentially being OpenAI employees?
Yeah, I think they have their autonomy as to what they're going to do and say on the podcast, so that doesn't change. But yeah, this creator-led tech media is what we are also a part of, I guess, now.
Yeah. Seems to be gaining strategic value in the sense that people are more prone to listen to a few voices who sort out the complexity in this scene because of their past experience or history or the ability to do it. Dylan and SemiAnalysis has done really well as well. So this is like the rise of a bunch of specialists or experts in this particular field which is really hot in the world right now. But the general mainstream media finds it hard to navigate because of the technical complexity involved and the many orders of implications involved with every piece of news or, you know, collaboration or partnership that comes out. Like look at how long we spoke about one tweet that Elon and Lip Boutan are like shaking hands and we analyze it for like 10 minutes. So I think, you know, that is the value that this kind of media brings to the table. I don't know if it's 300 million but it's some value.
No, totally. No, you're spot on. Like when you compare and contrast like John and Jordi versus traditional media, like traditional media for better, for worse, I subscribe to Wall Street Journal. I get the print edition. I read it every morning. Me and my son sit down, eat breakfast, read it. It's journalists first writing about technology, writing about AI. They're trained as journalists generally.
And what you're saying is like for John and Jordi, they were entrepreneurs. They were operators. They were builders first. And now they're becoming essentially like media creators, influencers, you know, second. And so they have like the best insights because they've lived and worked in that world. And by the way, why can they have a three-hour podcast? Not only can they talk about things ad nauseam like we can, but also they have a ton of connections so they can interview really interesting people. Traditional media doesn't have those connections. And then also the incentives aren't necessarily aligned with traditional media, but obviously John and Jordi just like they love entrepreneurship, AI, startups, technology, like they love it. And last thing I'll say is circling back to the point you made is like, well, we are trying to take the same approach which is like we have been engineers, I've been in a product management role, I've been in an entrepreneurship role, and so like we are first and foremost like the technical experts and then can also write and communicate these things to everyone else. So I do think that there is a premium now being put on personable technical experts or startup experts or operators who can in real time like see and navigate what's happening and talk about it as opposed to going to traditional media and hoping that you can convince someone that lives on the East Coast that, you know, went to an Ivy League school, studied journalism, and then convince them to understand your story and have enough passion about it that it will come through in their traditional route. Yes, it's an easier sell for companies doing AI hardware or software or all these LLMs to talk to people like Jordi and John who have the background in Silicon Valley in tech. So that the communication medium is like much simpler. They speak the same language.
Exactly. Exactly. Yes. So then the question is back to OpenAI. Why does OpenAI want to own that? Like I get if you're OpenAI you're like, great, these are people who get us talking about AI, like this is good for us for like awareness, adoption, that kind of thing. Of course OpenAI already has like a ton of awareness and especially within the audience that John and Jordi have, those people already know OpenAI really well. The first place, the like less gratuitous place that my mind went to first was like, oh, I don't think OpenAI would say like, let's buy these so that we can control them and have them talk about us. That's very obvious that that would be not genuine and people would just stop listening. But I did think, oh, maybe they're trying to say indirectly, don't talk about Anthropic. That's where my head went at first was like, okay, keep talking about AI. Don't talk about Anthropic.
Speaking of, I'm very happy for TBPN. This is a great outcome to be honest. Love it. Congrats to John and Jordi and everybody on the team. The problem that I have is what OpenAI is spending money on because it is not... if Anthropic did the same thing I would have less of a problem with what they do because they are reporting a lot of revenue right now. Exponential growth of revenue. They have nailed the enterprise aspect. Everybody's top-of-mind LLM model for enterprise is Claude. The coding agents is like Claude Code being widely used in enterprises. You know, OpenAI's strategy is like a bit all over the place. They tried this like advertising thing. I'm not sure where that's going. They're bleeding money. They're making $2 billion a month in revenue, but they're signing compute deals out the wazoo because they have no conservatism in like compute spending. Whereas Dario, you know, from Anthropic is like very, very conscious about how much compute he's building out to make sure that he's just building the right amount of compute for the demand he's seeing. So like there's fiscal responsibility on the Anthropic side.
Totally. Yes. So, Anthropic has product-market fit with the enterprise customer. OpenAI always was going after consumers first and that seemed very exciting and got them a ton of users, but ultimately obviously it's hard to pull revenue out of seven to eight billion people around the world. Like, you're going to mostly make $20 a month except for some people you might make $200 a month or $1,000 a month. Now, obviously, Anthropic said, 'Oh, well, for enterprise use cases where it's tied closer to creating value or reducing cost, like we can charge people $200 a month, but we could also charge them like thousands of dollars a month in tokens,' right? So, Anthropic has figured out they're a very good clientele and they've got the right business model and they just have product-market fit and they're taking off. So if you said to me Anthropic is buying TBPN, I'd go, 'Oh, interesting. Okay, they are trying to speed up adoption and awareness amongst enterprises.' And maybe they want someone like John and Jordi to help take that like very forward-thinking mindset, but maybe they're going to say, 'We're buying you and we're going to put you in front of enterprises. You focus on Silicon Valley people, but we're going to like help you talk to the rest of the United States, for example. Talk to people in Europe.' Like talk to John Deere or whoever, that would actually make a ton of sense. Now, if you lift that strategy and then you plop it down to OpenAI, you could say, 'Okay, maybe OpenAI realizes consumer was wrong and they should go after enterprise.' So maybe the like galaxy-brain thinking here is we're going to try to get in front of enterprises and we don't know how to talk to them and we're hoping John and Jordi can help us because they're so personable that they could help us talk to more enterprisey type people, maybe 30s and 40s year olds that are still kind of like forward-thinking in enterprises. But even that's a bit... that's like the most gratuitous thing I can come up with for the strategy. Even that feels a little bit of a stretch because that's not who they talk to today necessarily. So, it's still asking them to do something new to find a new audience and make sure they can talk to them.
Yeah. All right. Anyway, we'll see. This is an exciting development to be honest. Because I looked up what Joe Rogan was paid for the, you know, the Joe Rogan Experience on Spotify in 2020 before the AI era, you know, when we lived in the stone ages, it was 200 million paid out over multiple years. And I think in 2024 there was a refresher again. But yeah, so this is about as big as it gets. And their viewership isn't in the millions or hundreds of millions that Joe Rogan has. You know, their viewership is in the tens of thousands. I think 70,000 viewers. But it shows you what the value of that audience is because these are the tech minds of Silicon Valley who control, you know, now trillion, trillion dollars maybe in capital flows who are listening to this podcast.
