Hello, it's Transistor Radio. It always will be. It always will be. And I guess I'm trying to talk because Jordan's going to upload it. So welcome back to another Chip Chat with your four favorite horsemen of the semiconductor apocalypse. This week, we don't want to talk about DeepSeek. I'm sure we'll talk about some AI labs. We'll talk about all kinds of things. It's middle of earning season. But the one thing we don't want to talk about is Deep Seek. So we're going to preface it with that. How's that sound? Let's

kick it off with Dylan's star turn. Give us a little bit of color on the Lex experience. Well, okay, not just Lex, dude. He did BG2.

Like, what's the – because you've been on a streak, right? It's BG2. A generational run. A generational run. Let's get it.

Can we talk about how Semi Analysis did not call the Luka Doncic trade? Fucking insane. How did we miss this? I'm sorry. I want to say sorry to everyone here that we did not call the Luka Doncic trade.

This is very bad intelligence. Let's talk earnings. Okay, we can talk earnings. Google came out of nowhere, and they actually did a Chad thing for the first time in a long time, because they haven't been Chad's this entire time. They've been super conservative. And instead of doing $55 billion of CapEx, they said, actually, we're doing $75 billion. That's probably the biggest piece of news in earnings so far. Yeah. Meta doing $65 billion. Who's even counting? Is it really

Chad if Microsoft is still five higher? Yeah.

Well, I mean, that's a great question. That's a question between you and I guess our next week with Dylan on this historic streak. We'll get Satya on the pod. We can ask him himself. I

thought it was Joe Rogan next. It's pretty Chad of Google, though, yeah. I think Sundar woke up and decided he wanted to be CEO instead of being a watcher. He's one of those guys who likes to sit in the chair and watch, and this time he decided he's going to be the CEO. You

guys think Waymo had anything to do with it? So Waymo CapEx, I don't think they... So one, just like within the last quarter, they've got it to $5 billion of Waymo CapEx. So the big differential cannot be just data centers and AI. I do think... $5

billion of Waymo CapEx? What does that mean in terms of car count? I

mean, you know, I think it's like $300K a car or something like that. That's only 25k cars. That's pretty... Oh, that's nothing. Oh. Oh, yeah. Yeah, I think they're doing it pretty slow. But I will say something on the call. Sundar seems to take it pretty personally. This is the first time I feel like... Because, like, okay, at least in many places, semi-analysis has been like, come on, where you have big dog, you have the capability to totally change the game, given that they have TPUs, right? And I think on the call, it felt very personal almost. Like he was very much like, no, I'm here. I'm not going to let this one pass me by. It's time for $75 billion. Don't forget about Google. What do they spend that on? TPUs, baby. And GPs what? For their crappy AI Google search? You know, actually, it's... Whoa, whoa, whoa. Whoa, whoa, whoa. Whoa, it's actually not that bad. It's terrible, man. It got a lot better, John. So let me put some asterisks. Google just sucks at, like, making the product into... Like, they're good at technology and pretty dog at making it into a product right like that's the issue like the notebook lm team left right but notebook lm is like a pretty good example of like a pretty realized version of the future where it's like a relatively fixed blueprint of what you want to happen there's a really good one they have too of like uh plus sound plus like vibe, and it gives you like essentially like a dynamic thing back to you. But like, they're really good at like creating that kind of recipe and deep research, they did it first, right? Like there was some search that they were doing with essentially like a perplexity clone at OAI, but OpenAI really like copied deep research. Now deep research 03 is pretty cracked. It's pretty freaking good. It's very good. But Google did it first, and I've been doing Google Deep Research for a hot second, and it's a pretty good product. They just are not good at being all in.