Finding Alpha in the 2024 AI Chip Wars

[00:00:00] Welcome to Non Consensus Investing. I'm Ram Ahluwalia, your host and CIO at Lumida Wealth, where we specialize in the craft of alternative investments. At Lumida, we help guide clients through the intricacies of managing substantial wealth so they don't have to shoulder the burden alone. Through this podcast, we draw back the curtain to reveal the strategies employed by the best in the business for their high-net-worth clients so that you too can invest beyond the ordinary.

Ram: Good morning. I'm pleased to have a very special guest with us today to talk about one of my favorite topics, semiconductors. I'm joined by Dan Nysted. Dan is also maybe the first guest to have a bookshelf behind him, just like me, so I know we're going to hit it off. Dan is a former journalist. Turn financial analyst, he lives in, he has lived in the greater China region for over 20 years, and he was a Chinese language student in Taiwan, and you worked at the Wall Street Journal, [00:01:00] and you also were at a single-family office focused on Taiwan and Asia, so you had an investing angle.

So it's just a really terrific, backdrop and background to have, Dan, I'm really pleased to have you join us on Non Consensus Investing Podcast. Again, I'm Rahm Elawali, I lead Lumida Wealth. You can follow us on YouTube, Twitter, LinkedIn, Instagram. And we've got a newsletter as well. So we're gonna talk to Dan about semiconductors.

There's been so much happening, from the foundries, which is where Dan focuses, as well as Nvidia, the hyper scalers, metaverse, AI, China, and, super-regional and sovereign competition. The rival chips act around the world. There's really quite a lot to get into. So why don't we start off with, Sam Altman.

So Sam. He, it looks like he's intending to raise money in the desert, Middle East. He's in Abu Dhabi, and the report said he may be looking to raise trillions of dollars to build [00:02:00] a fabrication venture. So that would compete with TSM. It would compete also with, Global Foundries and Intel. So why is that important, Dan?

Give us some 

dan: context there. First of all, thanks a lot for having me on the podcast, Ram. I'm a big fan. I've watched some of your podcasts. I really enjoy them. You guys get a lot of good content, ask a lot of good questions, and I hope that we have that kind of discussion tonight. We're talking about my favorite subject, semiconductors, so if you need to shut me up, just, just say the word.

 Yeah, that's a good question. Sam Altman was on, he was at Intel IFS Direct Connect this past week. And he actually was able to clarify some of the things he was talking about with that 7 trillion figure that was thrown around. And he basically said that he's talking about the entire ecosystem around AI data centers and trying to get everything set up so that this is going to all work out.

And I think some of that, what you really need to understand about that is, that [00:03:00] we're not just talking about a lot of chips that are needed and a lot of foundry work and things in the semiconductor supply chain, but also, The electric grid, this is a huge amount of electricity, so when Warren Buffett's buying Occidental Petroleum, you look at that and think, now that's a smart idea, because if this gets as big as what they think it's going to be, I remember there was a, in November of last year, Samsung had a conference, and what they were talking about was, if we don't add any new, electricity or power capacity between now & 2028, By 2028, all of the power capacity we have now will be needed just for the AI data centers and whatnot that's being built.

That kind of speaks to an awful lot of, electric, the electric grid needs to be upgraded, probably nuclear power plants, they've, Bill Gates venture about the, the compact nuclear power plants and whatnot. A lot of that is going to be necessary to get this up and running because, the average data center uses an awful lot of power, about the same amount of power as a small city.

And [00:04:00] the AI data centers are going to use much, much more. Yeah, so he was talking about the whole ecosystem. And in some of the connectors So 

Ram: he's not as crazy as that number initially was called. I know Jensen took issue with it. It did feel hyperbolic. I agree with your point. There already are brownouts taking place, and it's quite interesting to see the AI semiconductor narrative and thematic idea interact with the nuclear renaissance.

And nuclear, of course, is another thing. We won't get into that now. You need cheap, efficient energy to power these data centers. We're an investor in a nuclear play that's powering some AI cloud providers in the United States as well. And we think there are multiple ways to bet on the rise of AI and the energy component and the transport.

The transition to that, is a great angle. Let's get into the foundry wars. I know this is one of your specialty areas and, you focus on TSM. How do you see TSM, which is the leader in the world? [00:05:00] Their primary customer is Apple. I think Apple is like a quarter of their revenue.

TSM chips are in virtually every mobile computing product we have around us. They are a global, systematically important institution, more than J. P. Morgan is. And how do you see them versus Intel, Global Foundries, or any other foundry competition. 

dan: That's, that's going to be one of the keys for the next few years, right?

It's, the key rivals are definitely Samsung, Intel, and TSMC, because they're all the leading edge. GlobalFoundries is a good company, but it's not doing anything at the leading edge, so it's in a little different position in that segment. Not even on 

Ram: photonics? You don't think they're doing anything bleeding edge on photonics?

dan: Oh, they're doing some very interesting things in photonics. I'm talking about the process technologies where you're getting down to 3 nanometer, 2 nanometer, 1 nanometer. That's one thing that Global Foundries a while ago said, we're not going to have the money to invest in, in those advanced processes.

Now, of course, they may change their minds, they, [00:06:00] they may try to do something different. They could get a big investment from Sam Altman. I really have no idea, that's, that would be a way for him to go. If he was going to try to boost some of that, 

Ram: and just briefly more on GFS, they just won a, multi billion dollar multi year US government contract, and it looks like Chipsack will be.

A source of funding for that. In general, what you're seeing is sovereigns are footing the CapEx bill, which is amazing because, the downside of these businesses, the capital intensity, but if you have a government paying for your CapEx bill, number 1, and you have customers paying Thank you for your funding and paying for CapEx.

It really improves the story from an investor perspective. I guess on GFS, any thoughts on that defense contract? 

dan: The interesting thing about what's happening is, ever since during COVID when the shortage hit, right? And there were automobiles that couldn't be produced because they just couldn't source a 50 cent, literally a 50 cent chip.

Would stop a $50,000 SUV from being manufactured [00:07:00] and that went on for so long and it hurt an awful lot of companies and it's really hard to say that this was, some, this company's fault or that company's fault was just one of those things, a pandemic hit, how many times has that happened in the last, 50 years and, and, the auto companies they basically thought business was going to dry up.

They canceled all their orders and then. As companies are pulling back. I think Toyota is one of the few companies that did not cancel orders. And so they kept their partners going. And then they were an early beneficiary when the chip started rolling in and when the demand rolled in, because one of the facts was that if you're in New York City, or if you're in any other big city during a pandemic, you don't want to ride the subway.

You don't want to ride the trains. You're going to buy a car because that seems safer and less chance of getting sick. It was a big mix up in the supply chains, but because it was the auto industry, that's a big industry for every government that, that the focus of a lot of governments is to make sure their auto industry is strong.

And [00:08:00] then out of that came a lot of what's happening now with the money, the sovereign money for building, trusted supply chains. Making sure that you've got plenty of manufacturing on your own. In the U. S., those defense contracts, one of the reasons they were very keen about that is because a lot of manufacturing is done in South Korea and in Taiwan, a lot of chip manufacturing.

And if you think about both areas, they're great places. They're the, wonderful people. They do an amazing job. China really wants to get its grips on Taiwan. North Korea is always in, shooting missiles off to test their missile prowess and to threaten South 

we're good. All right, let's get back into it. On the fabs, right? So on Intel, right? Intel, I believe they're standing up a three nanometer fabricator. And, it seems to be the sentiment is People don't really like Intel now. In two years, Intel might have A real shot at challenging TSM.

How do you read that? 

dan: Intel is a great company. I have no doubt that they will get [00:09:00] back up to their old kind of leader of the industry ways of doing business. And, they've always been a great manufacturer. So that's also a point wheRe. yeah, they've got that advantage and they lost track at around 10 nanometer.

They, made a few mistakes and they didn't correct them. And then Pat Gelsinger came in and really turned that around and said, no, we are definitely going to be manufacturing. We're going to definitely keep doing our own products. We don't want to outsource all of our designs. And he's, yeah, he's turned it into, he's.

Just at this recent event, he basically turned it into a product, there's an Intel product company, which all the chips that they sell to the computer industry and to the data centers and to everything else. And then there's the Foundry, which is going to be the manufacturer and of course the biggest customer for Foundry is going to be Intel's product groups.

 the, it's going to take a little bit longer, in my view than it seems like the kind of Wall Street is expecting at this point. And that's only because, Intel's kind of still catching up. They're still not back up to where they were. The five nodes [00:10:00] in four years, thing that they did, that's a bit of a marketing pitch.

dan: It's one of those things, when TSMC says that it's doing three nanometers, it's in high volume production. It's in mass production. And Intel was making announcements that were saying, we are manufacturing ready. As far as I can tell, when Intel says they're manufacturing ready, they will actually be in high volume production about a year after they say that they're manufacturing ready.

And that's a big difference. When you're talking to engineers and chip designers, you need to let them know exactly when they can expect their chips, because their customers want to know when to expect their chips. Making that distinction is pretty important, and it lets you know that as much as Intel is making a nice comeback, Intel's or a TSMC started three nanometer production about, in the middle of last year, they had revenue from their three nanometer in the second or in the third quarter.

And in the fourth quarter. So the second half of the year, they were already drawing revenue. And, yeah. And so they've got a lead in that sense. Intel's at four nanometer right now, [00:11:00] mass production. So they're not even quite yet at three nanometer mass production. So there's still It's just hard to get 

‍