Odd Lots - Apple Is at the Cutting Edge of a Revolution in Chips

Episode Date: December 17, 2020

On a recent episode of Odd Lots, we talked about Intel, and how the former dominant American semiconductor company was stumbling. But big things are happening in the chip industry beyond the manufacturing woes of one company. As it turns out, we're seeing a dramatic rethink of chip architecture, and what they can do, with more emphasis on specialized semiconductors that are really good at performing a specific task. One company that's blazing new ground is Apple, whose M1 chip is earning rave reviews online. We speak with Doug O'Laughlin, a former buy-sider, who now writes the newsletter Mule's Musings, on the industry and other things in tech.Correction: A previous version of this description misspelled Doug O'Laughlin's name.See omnystudio.com/listener for privacy information.

Thanks for listening to Odd Lots. Follow the show on Amazon Music for more future episodes or just ask. Alexa, play the Odd Lots podcast on Amazon Music. Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Wisenthall. And I'm Tracy Allaway. Tracy, I had a thought, and it might be a little controversial, but what if... You with a controversial thought, never. No, it's like, I know, I'm just going to throw this out there and tell me what you think. What if we made it so that odd lots wasn't about market and finance anymore, but just about the semiconductor industry?

Could we maybe start with a semiconductor series before we completely rebrand the podcast? All right. We'll revisit the question of a total rebrand. But I don't know about you, but I found our episode a couple of weeks ago with Stacey Razgan from Bernstein about the decline of Intel to be fascinating. And after that, I was like, I just wanted to talk more at ship. Like, I just, that really sort of like, it really wet my appetite. It was definitely a lot more interesting than I had been expecting. I think I mentioned when we recorded it that I hadn't been following all the ins and outs of the semiconductor industry,

except, of course, as it pertained to the U.S.-China trade tension. So I wasn't aware of some of the competition between the big makers, like just how bad things had gotten for Intel. And I wasn't aware of some of the strategic decisions that were going into the making of their chips and the way that some other chip makers have proceeded in a very different direction. Yeah, right. It's that different direction that's interesting because, I mean, we talked with Stacy a lot about the sort of manufacturing issues that Intel has been facing as it tries to understand. the veil, it's a 7 nanometer chips. But there's more to the story than just faster and smaller chips.

There's also a fundamental shift happening in what chips are and how they're used and how they're designed that goes beyond just sort of manufacturing problem. Yeah, and I guess the big news on that front is Apple and its new series of Mac processors. That news came out this month. And everyone, as far as I can tell, all the technologists that have YouTube channels are very, very excited about what exactly this means. I don't completely understand the technology, but I am interested in learning why everyone else is so excited and why a lot of people are saying this is a game changer. I heard one, there was one guy on YouTube who was talking about how this is going to make Apple computers basically on a different level to other types of computers. So in the way that we don't necessarily compare Apple phones with

Android's anymore, we will no longer be comparing Macs with your average PC. It's just like two different types of equipment. Yeah, there was like a medium article about the Apple chip and went viral on like articles about chips. Don't typically go viral. But that I think sort of gives you an indication of the enthusiasm, the excitement that people have for all this stuff.

definitely a lot to talk about, but I have to admit, I still don't, like, really get it. Like, I'm trying to read about this stuff and understand it, but I don't totally get it. Yeah, I'm in that same place. So, today we are going to be talking more, as anyone could have guessed, about CHIP. And I'm very excited because in addition to being a sequel of sorts to that other CHIP episode, our guest who writes on this stuff, and he's prominent on Twitter, et cetera, we're going to do something we've never done before. He's going to dox himself on this episode.

So he's been writing under and tweeting under a pseudonym, but he's agreed to come on and with this episode actually tell us who he is. So this is a totally new thing for us. I don't think we've done this before. No, we definitely haven't. And we should just stress this is a voluntary self-doxing. We haven't forced him into it or anything like that. That's exactly right.

So we're going to be speaking with the author of the newsletter called Mules Musings. He tweets under the handle at full all the time. Great stuff on the matter of chips. But we are going to find out now who he is. So Mule, thank you for joining the Outlods podcast. Who are you and why are we listening to? Hey, Joe. My name is Doug O'Waflin and I am just a huge nerd with an investing background.

I guess I previously worked at a byside firm in Dallas and kind of just taking a break between things, started writing a newsletter. And clearly there's been a lot of interest in semiconductors, not just from you guys, but from the investing community at large. And it kind of just hit a note. So I'm really obsessed with the whole, you know, the whole revolution, all the YouTubers are going crazy about. I'm obsessed.

Like it's not just, it's not just all the technology stuff, but there's a lot of business and investing things that could happen. And I think it's, I think it's truly a revolution. Like it's, it's about as exciting as it's ever been in semiconductors, which is frankly not that exciting, but it's still very exciting to me. So, yeah, that's what I've been doing recently. I have to say that while you were introducing yourself, I just had a look at your Twitter profile page. And I saw the binary code on your profile. And I looked it up.

And you must be the first all-thoughts guess that we've ever had who's been. basically done like a practical joke in binary code. So kudos for that. And I guess that gives some indication of your interests. And everyone else can go and check it out and translate it into English to see what it actually says. But before we begin, I'm curious, why did you decide to start the substack and why were you tweeting under the pseudonym to begin with? well um especially in the past past it's it's not you know i was employed and it's um it's not you're not supposed to talk publicly about public markets if you're in the business so that was that was the big reason for a pseudonym and frankly finfoot is an amazing community there's a lot

of um there's more information there than anywhere else so it's it's an awesome place to learn but obviously you have to be extremely private if you choose to do so so That was the reason why under a pseudonym. And then the substack, I guess I'm just kind of like in between things next year. If everything goes right, I'll take a little bit of a gap. But, you know, I'm like born to do or keep researching and doing stuff like this. Like this was pretty much what I was doing in my free time. And I was like, heck, a lot of other people could really learn a lot from this as well.

And so then I started writing for a broader audience. And writing in general has been really fruitful. Like you, it's a crazy skill to have. And I'm frankly not, I'm getting a lot better at it, but it's still, it's been a learning journey. So that's why I started the substack and I've just been kind of riding the wave. I have to say, I feel like a deep emotional resonance connection with your story because I got my start. I worked for a small portfolio management company. I also started, I started blogging in 2005 under a blog called The Stallway.

which is now defunct, but that's why I got my Twitter handle. And it was kind of the same thing. I just wanted to practice writing for the skill of writing and keep up to date. And it was between jobs, so kind of like you. So I'm a big fan of people with your arc, and I see big things. But let's move on and let's talk chips. I mentioned there's this a viral article about the Mac chip that came out. And articles about chips don't particularly go viral,

but everyone's raving about this new chip from Apple. So why is a chip going viral? Okay, so then, yeah, now I can talk ages about this. So the M1 chip is really exciting. So the reason why is it's kind of, it's kind of like a hybrid. And we've seen it before, frankly, like your iPhone is powered by something that's similar to an M1 chip, but we've never seen it for a Mac or a desktop or any other form factor other than an iPhone. And part of it is because up until this point, all the laptops in the world have been mostly running X-86.

So X-86 is like a, just think of it as like a language of processors. And for a long time, that was kind of like, you know, the de facto standard. But the iPhone getting a lot better has pretty much made a second language become a lot more popular, which is Arm. And now that it's become a lot more mature, it's actually kind of in the past, it never was fast enough to really compute in a actual computer. But now it seems to be that the M1 chip is like super mature. The arm infrastructure is ready and the laptop looks really compelling. And a big reason for what that is is it's not really a CPU. Like it's actually a lot of different dyes on one giant die. And it looks like there's a lot of benefits of arm over X86 that we didn't even know were possible up until

the M1 came out. So. So I mean, I think from what I understand, this is kind of the heart of it. So normally when we talk about chips, you know, certainly when we're talking about chips made by Intel or AMD or something like that, we're talking about CPUs, these central processing units. But the Apple chip, the M1, is not a CPU. It's something else. Or only one component of it is a CPU.

Can you explain that in a bit more detail? Yeah. So it's called a system on chip or sock. And what that means is that it has a traditional dye like any other piece of silicon, but there are little pieces on the silicon that each have different jobs. And so instead of having one giant piece of silicon that does everything kind of well, pretty much this silicon each has little parts that do their job really well. And then together, it does a lot better as a system because it's like, hey, instead of having a ton of random, instead of having a general computer that can do everything kind of well, they have like, you know, 10 specials in the house. And when you look at a laptop, really, you're only doing 10 types of workloads or however many number of workloads.

And so they said, instead of trying to make just a faster CPU, why don't we specialize a little bit? And we'll break down all the things that the laptop has to do and we're going to make them do it really well with these specialized chips. So that's pretty much what the difference is. And so this is kind of, in the picture of the broader things, kind of been a revolution anyways. Like getting off the CPU has been happening for a long time. And part of the reason why now it's become such a big shift is because, you know, in the last episode with Stacey, you guys talked about Moore's Law slowing down or maybe it's even, you know, they love to say it's dead. It's not quite dead, but it's slowing down quite a bit. And so now the road forward definitely seems to be specialization.

And so the M1 did this really well. It specialized in a lot of different ways, put it all onto one piece of silicon, and also couple the memory really close together, which is a whole, you know, kind of in the same vein, but a little different. And now the M1 just rocks, frankly. It's not, yeah, it's a little bit shaky as a first product, right? Like the Apple Watch 1 or the iPhone 1 wasn't the best product ever, but it still is a miracle, quite frankly. San Francisco.

On April 4th, 23, around two in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. So, you know, you talk about how this sort of emerged out of Apple's chips for the iPhone and something I think, you know, like I buy a new iPhone every few years like anyone else. I don't buy every single one that comes out, but, you know, when mine feels like it's getting

old or whatever, I buy a new one. But honestly, the only thing I do on my iPhone besides tweeting is taking photo. And so even though, like, I have this like incredible computer in my phone. pocket basically is just a camera with Twitter attached or Twitter with a camera attached. And so I'm curious, like, how much has this sort of like shift towards these sort of focus tasks instead of generalist chips come out of this fact that, yes, you have, I have this whole computer, but there's one thing that it's really important for the iPhone to do for a lot of people, and that is take really clear cameras, really, really clear photos.

and this sort of recognition that there are just a few simple tasks that on a phone you have to get done extremely well for most people. Yeah, that's actually the perfect, that's like the perfect thing to bring up because on the Apple phone, right, the A14 or A15 or whatever iteration of the A chip, the biggest dye, the biggest specialized dye is something called an IPU or an image processing. Sorry, can you dye? What you said? A dye is just like a piece of silicon. Sorry. No, no, that's just wanted to me. Yeah, just think of it as like a unit of silicon, if that makes sense.

And the biggest piece of like that's been taking share from everything else is an IPU. And the CPU has become a smaller and smaller part. And similar in that same vein, right? Like computers, one of the things that we've been using the most is GPUs to play video games. So just like the CPU is kind of losing relevance or the, you know, the CPU is losing relevance in your phone to an IPU or for your images. In a desktop computer, for example, the CPU seems to be losing. relevance to a GPU for a lot of people who are using it for games. So kind of that same, that same energy is, you know, disrupting one and the other. And I think you ask a question about,

like, why Apple, Apple in the iPhone specifically kind of like shifted over to the Mac. What's interesting about that is in the iPhone itself, they actually had a lot of requirements that were not, that you didn't really need to use in a, in a computer, right? Like, it has a lot smaller form factor, it has a battery life constraint, it has to also not have a fan to actively cool it. And so all these things in the beginning were just like, you know, a pain for them to to manufacture around. But as they got better and better and better, they pretty much got really close to a laptop in an iPhone. And that's while having a smaller form factor, while having no active fan. And then they're like, wait, why don't we just do this for a laptop? And so that's kind of what they did. And the M1 really is like a super souped up

Apple chip. And the announcement that you guys just broke the story on is that they're going to actually be making even bigger, beefier versions of this because the M1 is actually made to be put in a laptop with no active cooling. Like that's also a big part of it. There's like it's supposed to not heat up. It's supposed to not have a fan. But what if we, what if we just like let it run loose and, you know, actively cooled it, put a fan, see how high, how hot we can get this thing going. And I think what's going to happen is we're going to be really shocked by the outcome. And so all the manufacturing and the constraints that they had to do to make the chip fit into a small phone, pretty much when you actually scaled it up into a laptop, it made it a lot better. And that's

kind of part of the M1, you know, the Apple Silicon Magic. And Apple isn't the only company pursuing this. Like, I think, you know, almost every company is kind of looking at Intel and saying, like, okay, we need to transition off this vendor. And whether it's a data center or whether it's a phone or whether if it's anything else, like everyone is kind of going. this way. And that's what's exciting is this isn't just an Apple story. Qualcomm mentioned they might do heterogeneous chips, which is like one way to call this whole thing is heterogeneous as in many different types of chips put together into a larger package. And so, you know, Android might have a similar M1 like chip very soon. That's probably, you know, a few years away, but the writing's on the wall

that everyone has to shift into this kind of M1 style of chip, if that makes sense. So, So it does make sense. This was actually what I wanted to ask you next. So I think I understand why Apple sort of struck upon this design, given its experience with the iPhone, why it thought that specialized chips for specialized tasks would be more efficient. But why was it able to actually produce the chip ahead of time? Or maybe another way of saying this question is, why was Apple able to do this? as you just mentioned, the entire industry seems to, you know, maybe at different speeds, be moving in this direction. So Apple Silicon has always been a leader. They've been crushing it for a long time. But in particular, this kind of goes back to the last episode you guys had. TSM really enabled them to do this. And Apple has been buying and building silicon expertise in house for years. This is not an overnight success story. The original chip, like the A7 or whatever, was when they first started to really make this more heterogeneous. And every year, it adds more specialization. And it's gone to this point where now the CPU is like one of the, not the least important,

but it's just one aspect of the chip in the phone. And so Apple's been doing this for a long time. TSM has been a huge part in enabling this. And then other companies kind of look to Apple and are seeing like, okay, we can. can do this too. So I hope that kind of answers the question. When it comes to manufacturing, and of course that was the big thing we talked about on our last chip episode and this idea that it's like getting more and more difficult to sort of shrink these chips, have the smaller nanometer form factor. Is that an issue here or just sort of rethinking the fundamental architecture

of the chip such that it is a heterogeneous chip? Does that sort of sidestep some of these issues? Does that question make sense? Yeah, yeah. Actually, I think that's the huge driver of why this whole thing exists, right? Moore's Law slowing down and the struggle to shrink smaller and smaller pretty much made them think, hey, we have to improve this somehow. We can't do it through the ways we used to do it.

Why don't we specialize a little bit? So it's kind of like a mix of both that's happening at the same time. I definitely think that the desire for better, faster chips was what helped the heterogeneous revolution began. But TSM being there to essentially make it so that anyone who isn't, because in the past, Intel also owned the method to get smaller, right? So it's like, you know, you can't compete against a general chip that's getting smaller, getting faster, faster than you could specialize your chips faster. But now that TSMC is clearly in the lead and anyone can reach.

out to them and work with them, now everyone's like, well, why don't we just make our own special little chips? And so that's really been kind of all the things that's made this happen at this right moment right now. I mean, on that note, how easy is it going to be for competitors to make these specialized chips then? Are they going to be able to do the same thing that Apple's doing? And in particular, is Intel going to be able to do it given that, as we discussed in the last episode, their business seems to be almost entirely based on this idea of selling, you know, general purpose CPU, which then can go into a large variety of desktop PCs or whatever. Yeah, so Intel is already pursuing this. I mean, the people at Intel are smart and they know what

they're doing, but they're obviously tied to a very large platform. It's something called Fovarose, which is like kind of their method of stacking the chips together. They have this new platform and their Intel Architecture Day, they're like, okay, we're going to go all in on shiplets. Like, we're going to do the same thing that everyone else is doing. But that kind of brings it back to the point. Like, Intel is going to be doing the same thing that everyone else is doing. And if they don't have the manufacturing advantage like they used to, pretty much everyone's on the same level playing field. And it's, it's for real right now. AMD, for example, does have like a big, a big reason why they won was kind of like a process heterogeneous thing where it's like, okay, we don't

have to be the fastest, you know, we don't have to have the smallest chip, but what we're going to do is we're just going to make sure this really important part is super small. And that way, you know, it's like getting it across the finish line by having the best of the little pieces that really matter and having a lot of like not as good pieces that didn't really matter as much. And so everyone is kind of going to this like hybrid playbook. And every company to a certain extent is pursuing this. And yeah, I mean, I can talk to individual. companies in particular, but I know it's a lot to wrap your mind around sometimes. So when we think of Intel, we think of this sort of like basic business model, which I guess,

you know, still exists. It still probably makes tons of money, whether you have the one company makes the chips and then another company puts together boxes or laptops or servers where they assemble various components, taking Intel chips or chips from elsewhere, and then packaging them and distributing them, et cetera. Apple has always kind of astrued that model being much more vertically or horizontally integrated. I can never remember which is which and having a sort of custom set.

Is that fundamental model going to remain or is there going to be pressure on the very premise of having a separate computer that's distinct from a chip? So, yeah, that's a great question. I think, I mean, I can't, I think something that's happening in the near term is definitely verticalization. Like you're seeing it with Apple and they're probably furthest down the road. I actually wrote about this quite a bit.

But like Amazon in some ways is making sure that their whole data center is pretty much owned by them using arm chips that they license manufactured on TSM. And in some ways, their whole data center will be vertical. And they'll start to, like, think of the data center as a giant computer. They'll own the entire thing. And it won't be like how it used to be where it could build your own computer or in this case, a data center. Instead, you're going to be forced to be vertical just like Apple is kind of for your phone, if that makes any sense.

That's a huge, that's a huge new trend that is like right now happening. The AWS announcements last, like the last week or so, pretty much they announced like three pieces of custom silicon, which is, which is crazy. So they're doing this, but think of it in a much bigger picture versus Apple at a much smaller picture. And same with Microsoft, same with, same with Google especially. And Facebook has even recently hired someone to probably start this path for them as well, too. So I have a dumb question kind of going in the opposite direction. But if Apple is making, you know, this new type of chip that everyone is very excited about,

could they ever sell it on the market in the way that Intel sells its chips or does it just not make any sense for them to do it? I think it makes a lot of sense. The YouTube community is freaking out right now about if M1 can run Windows because if it can, that would be awesome. And so Apple could actually, they could sell it. I don't know if that would be kind of like in their ethos, right? Like they make very, very defined and like a very close. closed ecosystem. But yes, they could. And even Microsoft would buy it for their, for their high-end surface books. A lot of the laptop OEMs would definitely be an insta buyer.

So yeah, if they wanted to, they could instantly, they could open it up and sell it. And heck, I think it would be really good for their business. Like, it'd be gross margin, accretive. It's actually probably more profitable than selling the phones outright, which is crazy to think about. I have two questions. is very short and one is a little bit following on to what you just said. A, you've said the term arm or you talked about arm a few times. What does arm mean or what does it stand for? What does it like represent? And two, just there you said the question is can it run windows? What do we really talk about when like here's this powerful chip? Can it run a piece of software? What would be the

tension or the difficulty in running windows? And how much does that sort of add to all this? This is need to sync the chip with the, with the operating system. Okay, so Arm is, is kind of one of those languages. I, I don't remember what it stands for off the top of my head, but it's, it's actually, it's, it actually got sold to

Navidia. Well, it's in the process of a deal to Navidia, which is like, you know, super high stakes. Is Arm a company? Yes, Arms, arms a company, yeah. I mean, I know that Arm is a company, but it's also a language. Like, this is where I always, Yeah, Arm is a company, and they pretty much, they pretty much licensed their software, their language of silicon out to anyone who wants to use it. So whenever Arm license it,

they essentially say like, hey, you can, you can use this. We'll even help you tinker with it a little bit, make sure it works out. And you're going to have to like 2% of revenues or something like that. It's usually a percentage of, and it's pretty cheap for a lot of companies. Like, right, If you don't have any silicon expertise in house, like building anything from the ground up is, you know, probably a billion dollars plus enterprise. And so Arm just does that all for free. You can essentially just buy off the shelf Arm, you know, cores or IP. And then you can kind of build your own little piece itself. And, oh, and so the software side of it. So X86 is pretty much what everything for the PC has been built on. Arm obviously is everything mobile has already been built on.

And there's kind of been this push and pull, right? Because Arm right now definitely has the ecosystem momentum. But people are really nervous. Like even the laptops, the M1, the biggest problem with the new Mac laptops with the M1 in it, is that some of the programs won't work with it. And you pretty much have to redo it to make it into an arm compatible style. However, I mean, a lot of companies kind of can see what the writing on the wall is, so they are. But there's also something called an emulator where essentially it's like the arm chip will run a mini, you know, fake x86.

And then it will by using that like, you know, emulation, it can do can move that software and use it all in that hardware. But so it is a big deal. So right now it seems like Windows, from my understanding, can run on M1. And but the problem is like when it does, it's going to be buggy. Like, have you ever used a Linux computer, for example? Like, you can't play video games on a Linux computer because they're not compatible. And so that same kind of, like, problem with the software ecosystems is going to have to be hammered out. And that's going to be a huge growing pain for everyone involved.

So I'm curious. Everyone is very excited about these specialized chips. And we were talking a little bit about Moore's law, this idea that, like, maybe it's not. completely dead, but people are having to look at new ways to make their chips more efficient rather than just making them faster. After we do this, after we do these integrated chips or whatever you want to call them, what's the next big thing that you see down the line for the industry? It's kind of like there's this like mix of things going on. We're still getting smaller, right? That's one thing. That's a huge, a huge thing. TSM just had their 5 nanometer. As we discussed,

As you guys discussed in your last episode, you know, the nanometers don't actually mean anything anymore. It's still a marketing term. But you do get benefits from getting smaller. So that's still going to continue just at a slower pace. The specialization is going to continue until it stops working. And people, there's kind of like, you know, I've read this paper that there is a theoretical limit until when that stops working. But like, we're going to keep running down that road until it stops working. After that, I really don't know.

I mean, quantum stuff, you know, I keep hearing and reading and researching and trying to figure out if this is real. Clearly, it's a little bit away, but that would still only be really useful for a very small subset of problems. And so I think the way that we're going down, which is specialization, is probably the surest way forward. And not only will it become, it will like start from most of the hardware, but like the software will probably start to have to have to be written. and better to couple with the hardware. That's something that's like a big trend where the software and the hardware will kind of specialize together because that's something that you can continue forward, right?

Like the less flexible, the faster it will be versus the more flexible, the less faster. And so if you can just like kind of make it as the least flexible as possible, in theory, you can make everything into an ASIC. And the speedups that you get between specialization and non-specialization are our magnitudes order faster. So that's the reason why it's so exciting. Because it's like you could have, if you just made a single program that only had to run one type of thing,

you could make it 100 times faster by just running it on the most specialized chip possible. Now, you have to be really, really sure that all you want to do is just this program. And a great example of this is Google's TPU, which is a tensor processing unit, which is their ecosystem called TensorFlow, which is focused on deep learning.

So Google is like, hey, we know that we're going to be running a lot of tensor flow in the future. And AI is going to become this huge exponential problem. So we're going to make a chip that only runs TensorFlow and it's really, really fast. It's compared to how it would be done on a CPU, I couldn't really tell you. But it's like 10 to 100, maybe even more. Like the magnitude, it's a lot faster. And so until the specialization stops working, I don't really see anything that's going to, I'm going to stop, stop it for now.

So basically, it sounds like sort of wrapping some of these thoughts together is that as chip architecture has migrated towards more specialized tasks, more specific tasks, as you just mentioned, Google and it's deep learning AI stuff, it really becomes more part and parcel with every technology. So if you're working on machine learning or AI, you, it's almost like you have to have a chip component to it. Or I guess maybe it makes sense to pursue a chip component rather than just sort of writing code, writing software, and then running that on generic off the shelf chip from some other company. Yeah, that's the perfect way to put it together.

And frankly, it used to be awesome. And we used to just do it. Like, think about how awesome would be. It's like, oh, you can just write the software. And every year it just gets, you know, every two years, it gets faster. like 100% faster. You didn't have to think about any of this. And it's a pain to think about all this stuff where it's like,

hey, we have to make a custom chip platform in order for our software to get faster. You know, we miss the old days of Morris Law for a reason. But now that we're slowing down, clearly like we're going to have to do some very specialized, very top to bottom structures of software and hardware that work together to make the best outcome.

I'm June Grasso, inviting you to join me for the Bloomberg Law podcast. Every weekday, we help you make sense of the legal stories that shape the nation and the world. Listen for complete analysis of the biggest court cases, the latest actions from Congress and regulators, and the legal moves driving the markets. From corporate law to constitutional law and from state courts to the Supreme Court. At Bloomberg Law, we go beyond the day's headlines. We speak with top attorneys, judges, scholars, and policy experts to break down what the rulings really. mean. We do this every weekday, then bring you the best conversations in our daily podcast. Search for Bloomberg Law on YouTube, Apple, Spotify, or anywhere else you listen.

On the East Coast, listen as you start your day. And on the West Coast, catch up in the evening. That's the Bloomberg Law podcast with me, June Grosso. Subscribe today wherever you get your podcast. So I just have a couple more questions, but one that I'm interested in is, like, you know, we've been talking about Intel, we've been talking about Apple, we've been talking about Google, Amazon, and Nvidia, and Taiwan, but Nvidia has become this monster. And I'm curious, like, their role in it, because I mostly think of them as a company that makes chips for video games, but then I always see stuff, like, about how they're also doing

stuff with self-driving cars. And sometimes I see these really cool demos or, like, other sort of machine learning image recognition types of things. What is their role in the ecosystem, what is their competitive advantage and how did they emerge to be just like such an incredible player? The stock's just unwind. Yeah, yeah. So the company is wild. Jensen Huang is awesome. Like he's he's he's an amazing CEO, but notoriously, um, notoriously stubborn. He's very like Steve Jobs-esque. GPUs think about this way. There's a gradient of like the most general to the, to the least general, GPUs are one step past the CPU and where it's like it's still pretty flexible and you can, you know, you can make it do a lot of things, but it's not like so,

but it's a lot faster, but it's not so inflexible that it's like a pain to write for. And so that's kind of their been their huge benefit. So as the original use of a GPU is essentially to make all the pixels on your screen in parallel. And that's really hard, because you have to do it all at the same time. And so whenever you hear about parallelization and that GPUs are like these super parallel engines that are really good at, calculating everything at the same time. And so the reason why that's so important is, like, AI actually happens to be that kind of that same use set, almost identically.

And so they have this huge ecosystem, and they're like, hey, well, why don't you use all your AI things? And we'll speed up with this GPU, and it's very easy to use. And so that's been their huge step is that it's much better than a CPU. It's extremely easy to program. And they've started to add a lot of software demo. And so they're kind of, in some ways, what Navidia is doing is that they're also making their own ecosystem. But instead of starting with the software and moving down, like let's say Apple, they're starting with the hardware and trying to move up.

And GPUs are, by the way, specialized. They are less specialized than A6 or an IP or all the other really, really specific things that just do one thing really well. They can do multiple things pretty well. But they are obviously more specialized than a CPU that can do everything. you know, kind of okay. So that's, that's been their big role. And, and they've really taken a lot of share as, as things have started to specialize a little bit. They're the, they're the, like, the easiest first step, if that makes sense. Right. I have one more question. So, you know, for a long time, I feel like, and correct me if I'm wrong, but I feel like the semiconductor industry for like

decades was sort of like a cyclical industry. And you'd have these sort of, uh, there's almost like manufacturing or like an industrial thing and you'd have these up cycles when there was growth. And then you'd have like an inventory correction when the foundries or the chip companies have made too many chips and then they'd have to cut prices or de-stock and they'd go into a downturn and so forth. And then it feels like in the last several years, maybe five years, 10 years, I don't know, it has stopped being a cyclical industry and become more secular as and that is as chips get put into more and more things and phones and all kinds of different stuff. So it's like less of an up and down cycle and more just like a sort of growth industry.

A, is that true? Is my perception correct? And B, is there any slowdown on the horizon for the world's overall demand for chips? Or do we just keep finding new places and new places to put them in new needs for them? So yes, historically it was a crazy. cyclical industry. If you guys have ever read the innovators, what I'm, dilemma, yeah, innovators dilemma. The case studies he uses is almost all semiconductor companies because they're so cyclical and they're, they're like constantly, you know, you know, destroying each other's modes and they're, you know,

they're innovating and then, you know, growing up and then they're dying. Like, it's chaos. Things have definitely changed a little bit since those days. And so one of the largest markets that really dictates the cyclicality of all the semiconductor and markets is memory. And there's been this thesis that it's become a lot more consolidated, meaning that there's like, I think there used to be, you know, tens of players. There's like three for DRAM, which is a type of memory, and five for NAND, five to six or whatever. And so that's, on one hand, that's, that's dampen the cyclicality. And pretty much they used to do, they used to do the race to the bottom thing where whenever things would go down, they would just add more supply and then everyone would would bankrupt together,

kind of like shale is to some extent or does. And they've done less of that. And that's been one one thing that's dampened the cyclicality. But I think going forward, it seems like there is more end markets than ever. In the past, it was really just like PCs, right? And then phones became, well, PCs and there's some like some industrial and like, you know, aerospace. And there's always been small markets, but for the most part, it was just the PC market. And then phones became a thing. And then so now became PC and mobile. And now it seems like auto is becoming a thing. So auto is this new third, third market. And so it's industrial IoT stuff, which I keep hearing about, but I'm always confused about how the semiconductors work into it. And then the AI market in itself

will probably become so demand intensive from the cloud from the cloud customers that they'll have their own end cycle. And so there's each of these markets have, their own cycles to to their own extent, but together all the demand is starting to like overlap and make it a lot more smooth, if that makes sense. To be fair, the market is still cyclical. It will always be, it will always probably be cyclical, but I think all the demand aspects are really pulling things forward. And at least in the near term, I think next year seems really well positioned. Like, for example, they're saying like this huge influx of auto demand in China, they're like, well, we can't really make it because we're, we're struggling with

semiconductor capacity. That's a headline that came out in the last few days. And I think that means, you know, that's kind of a sign of all the new things that semiconductors have become a part of, not just your PC, but your car, but your, you know, your nest thermostat, all the other things in the world that you use. So can I just ask one follow up on what you just said. You said someone came out and was talking about Chinese car production. Like, ramping up who was that and that they couldn't make all the chips needed for it it was uh Volkswagen Volkswagen okay thank you sorry I just wanted to look it up after this yeah yeah I want to say it was a reuters article but it pretty much was like hey semiconductor manufacturers can't add capacity

because it's like so I mean it's that's really interesting yeah and like like auto constantly is just called out is this market that's going to be this huge thing and it is obviously a very big market, but like if we're really going down the route of EV stuff, like hardcore, you know, we can, you know, no comment on the EVs back bubble or whatever, then there's going to be a lot of more semiconductors, right? It's like almost completely digital. Doug, that was fantastic. You're, uh, you're great at explaining this and really appreciate you making your public, uh, non-anonymous debut on odd lots. And, uh, I learned a lot. So thanks for coming up. Thanks, man. I really enjoyed it.

Thanks, Doug. So, Tracy, are you cool now to change the focus of our podcast on a permanent basis? I don't think I'm quite there yet. I admit that semiconductors are a lot interesting than perhaps I gave them credit before we've recorded these two episodes. But yeah, I don't think I want to do all thoughts, comma, semiconductor specialist just yet. Well, I was going to say maybe like when we're a gigantic media brand in our own. right, we could do like a spin-off show. The semiconductor vertical?

Yeah, the semiconductor vertical. Excellent. What I was going to say is I'm quite interested in possibly looking at one of those Apple M1 computers at some point. Yeah, I don't think we intended for this episode to be a giant advertisement for Apple's new chip. But a lot of people are excited about it. And it was really interesting listening to Doug go into some of the technical details about why exactly it's considered such a new thing. Yeah, no, I thought that was super interesting as well. I feel like the next one we do in our series, we should do like a TSM focused one.

For sure. Yeah, let's do that. Like I feel like, you know, like we've done Intel. This one is sort of a little more Apple heavy, although everyone. But it feels like TSM is just like such a fascinating. story in its own right, combining their sort of tech prowess and geopolitical significance that whenever we do our next trip episode, let's do it on that. Agreed. All right. Should we leave it there? Yeah, let's leave it there.

Okay. This has been another episode of the Odd Lots podcast. I'm Tracy Holloway. You can follow me on Twitter at Tracy Allaway. And I'm Joe Wisenthall. You can follow me on Twitter at the stalwart. Follow our guest, Doug O'Loflin on Twitter. He's under the handle at Fool All the Time. Also, be sure to check out his newsletter, Mules Musing. Follow our producer, Laura Carlson, at Laura M. Carlson. Follow the Bloomberg head of podcast, Francesca Levy, at Francesca Today. And check out all of our podcasts at Bloomberg under the handle at podcast.

Thanks for listening. Join Bloomberg in New York or via live stream on May 19th for building an AI future-ready business. This event will examine how organization, are moving beyond experimentation to scale agentic AI across the enterprise. By joining us, you'll hear from successful executives who are at the forefront of designing human plus agent operating models that are secure and outcome-driven. This event is proudly sponsored by Kindrell. Register at Bloomberg Live.com slash FutureReady.