AI Could Turn Software Into “Dumb Data Pipes” | Dom Rizzo

28:33You could go look, it's public. I've owned ARM for a while. Look, ARM has had an incredible run. So all of this, let's just be very clear about how we're thinking about things. The semiconductor index is up 90 % year to date and software is flat. I'm not saying that you can't see a software balance. In fact, you're probably likely to see some sort of software balance and trying to make sure that we have enough portfolio construction to make sure that we're well positioned for some sort of bounce. That's okay. And stocks go up and stocks go down and stocks go up too much. And you got to think about portfolio construction and risk management and be able to trim when stocks go up.

29:08All of that's true. But why is ARM so well positioned for a few reasons? One, because they were the architecture for smartphones, they really figured out low power, low power CPU processing. And guess what's one of the most important things to do in the data center, low power CPU processing. The second thing is their architecture is very well positioned for this agentic, being able to do tasks on the behalf of the large, using the large language model to do tasks on your behalf. And so I think ARM will have its place in the data center and it'll be a hybrid model. Sometimes, you know, they'll give the IP to, they won't give, NVIDIA will pay for the IP.

29:49And sometimes they'll help someone like Meta design the CPU, which is one of the great moves that Rene has made is changing the business model from just a pure IP royalty to hey, let us help you design this chip. We really understand how this architecture works. He was fantastic when he joined us on talking about all of those things and refer people back to that episode as well.

30:18This episode is brought to you by LSEG, the leading global financial markets, infrastructure, data and analytics provider. To learn more about how LSEG connects businesses, investors and markets worldwide, visit lseg.com. This episode of the Master Investor Podcast with Wilfrid Frost is sponsored by BNY Investments, a trusted partner for many delivering financial solutions to investors and institutions worldwide. This sponsorship does not constitute financial advice.

30:58You mentioned there, to sort of round out some semi-stuff, the amazing performance here today. Talk to me about memory. Where do you stand in that and the, frankly, absurd performance you've seen of late? Yeah. So memory, I love when people say memory is a commodity. Okay. Okay, memory is a commodity in the sense it doesn't matter if you get a Samsung chip, a Hynix chip, or a Micron chip. Okay, they're all the same. That's not true for GPUs or CPUs or ASICs. That's not a true statement. So you can literally go pull out a Samsung chip, plug in a Micron chip, and the system will work perfectly. Okay, this commodity is the hardest commodity in the world to make.

31:45I mean, tens of billions of dollars at CapEx. Only three companies in the world could do it. You know, Samsung used to have this thing called the golden price where they knew the exact price of memory where all their competitors would go out of business. I mean, this is a ruthlessly competitive, difficult business. And now you're down to effectively three people who can make it. Being Samsung, Hynex, Micron. That's right. And that's on the DRAM side. I don't want to get too technical. There's the NAND side as well. There's more players who could do that. And the Chinese are coming quickly. And we could talk about that.

32:18What's happened is in an agentic system, again, this agent being the big takeoff, memory consumption is like 5 to 10x. And so you're writing all this code and guess what? You're hitting the DRAM a zillion times. And what happened is these companies, because 22 was such a weird downturn, they really slowed down the CapEx investment. And memory is one of those businesses where there's a price of a bit, just like there's a price of an oil. There's a price of oil. There's a price of a bit. And if there's a lot of demand for the bit, prices are going up. And so what you've seen is a market that was way oversupplied go to a market that's way undersupplied with an increasing cost curve and a consolidated industry.

33:13And that's why you've seen such incredible performance out of the stocks. But I think it's really because going back to the framework, linchpin technologies, there's only three companies that can do this. That's very essential. Innovating and secular growth markets. Memory continues to proliferate all around the world. With improving fundamentals, what happened the past 12 months? You went from 0 % growth to 500 % growth, right? And then reasonable valuations. These stocks still trade at four times earnings. Now, the tough part today as we sit here. They're four times P. Hynex and Samsung are four or five.

33:46Micron's eight or nine. Because the market's so smart, right? The market's this amazing thing. The market's daring you to buy them. That's what's happening. And why is it daring you to buy it? Because fundamentals are about to decelerate. I was going to say, it's not going to be 400 % growth forever. It can't. And so going back to the fund, so the tension right now on the team and with my framework is three of the four look really good and we have decelerating fundamentals. So, you know, Dave, if you're listening, I'm quoting you again. I'm memory curious. I'm interested. I have some positioning in it, but we did just go through this major acceleration and we have to be humble about the fact that pricing probably normalizes and revenue growth has to decelerate mathematically from here.

34:36So just add onto that, I mean, I didn't realize the P's were quite that low, particularly of the Korean names. But add to me how you factor in thinking about this explosion that we've seen in ETFs in Korea, 3X levered, various ETCs and ETNs, where clearly there was just a technical market dynamic of forced buying that has played into this. I think there's a lot of technical market drivers all over the place. I mean, just, you know, the VU is now a trillion dollar ETF. Yeah. Right. What's going to happen with these IPOs, how quickly they get placed into the passive will drive not just demand, but how people think about risk.

35:20Right. Because your risk is relative to a benchmark. And then you have these levered ETS. You have higher participation from retail, which I think is net net a good thing in the world. Right. Lower cost, higher participation in capital markets is great. You know, if one thing you can do to address wealth inequality is make sure everyone's an owner, right? You want everyone to be an owner. All that being said, of course, there's these technical dynamics. But what I've found and where can we be, you know, the goal is always to be the best in the world, right? That is the goal. You have to have that goal.

35:58Focus on improving fundamentals. Focus on the technology excellence. Don't focus on the management teams. Don't focus on the technical dynamics of the market. But I guess you don't think that has contributed to the share price performance. I mean, you know, one of your four factors is obviously the price you're paying. But I guess, I mean, it comes back to the fact that P is at four. That's mad. Well, I think, well, why is the PE at four? And why does NVIDIA trade at, you know, 17 times, 15 times, 14 times? It doesn't really matter. the market is debating, is the CapEx boom sustainable? And then what percentage of that CapEx boom goes to memory?

36:37What percentage goes to GPUs? What percentage goes to CPUs? What percentage goes to optics and networking? And the area that it has more confidence that you're going to take share in the CapEx boom, trade at higher multiples. Right now that's optical and networking, but at different points, it's been GPUs or CPUs or memory. And then you have to balance that with what's the game theory at the hyperscaler level. And I think that's why this Google event of raising it. This is the third most profitable company in the world raising equity capital. This is not some small cap. This is the third most profitable company in the world raising equity capital.

37:21Which tells you what? that the capex boom's continuing yeah i think so does it also tell you for a company that's bought back tons of shares that i mean look you could argue they've been a good hedge fund they bought they bought back shares cheap and they're raising capital at an expensive price but that's also not why we buy shares in them right i mean Because I guess the other point to pair that with is their trend has been to buy back shares, not to issue shares. And their trend has been to fund either out of cash flow or out of debt. Does the fact that they have to turn to the equity markets warn something else, which is the debt markets are drying up or their cash flow is already stretched?

38:07I spent a lot of time with our fixed income team talking to this. This is the benefit of a T-Row, right? I have Mark Stodd and our fixed income analyst is wonderful. I do not think we're at a point yet where the debt markets have dried up. I think what it says is the size, scale, and scope of the spend is bigger than people think. And I don't know how much of it was wanting to get in front of pending IPOs. I don't know how much of it was, hey, we're trading at a reasonable valuation now 27 to 30 times. Like you said, they bought back stock lower. But I do think we're at this unique period of time where the game theory is who's going to spend the capital.

38:49And why is that the case? Because we've learned time and time again, compute equals revenue. I'm going to say it again because it's so important. Compute equals revenue. Compute equals revenue. And it was a great pod, so I'll reference it again. The one place I really disagreed with Dan was that OpenAI was writing checks that it couldn't cash. I think Sarah, Sam, and the entire OpenAI team were very, very thoughtful in locking up compute capacity early. And guess what? Go listen to what CC Wei said last night. 26, 27, you know, potentially out to 28. It's very hard to get capacity right now. And compute equals revenue.

39:37And if that is the case, I think it leads to a more sustained CapEx boom. One final question on the Google capital race. Step back with a perfect benefit of hindsight type world as the fund manager investing in tech companies for what tech companies do. Would it have been more ideal if all of these companies, before even issuing equity, which obviously is a decent trade, as we said, based on the prices, but they've all had to issue tons of debt. Should they all have not bought back stock for years? I don't know. I think the world changes. You have to play the game on the field. And the game on the field three years ago was not capital-intense business models, right?

40:21Why did the MAG7 outperform the rest of the market? And why did passive gain so much share, right? Why did passive gain share? It was because MAG7 outperformed the rest of the market, I think. I think what happened before November 30th, 2022, to was that these companies had effectively perfect business models, local dominance, and were the perfect aggregators, right? They controlled demand, they commodified supply, and they were free cash flow machines. Everything changed with AI. And why is that? And it's because the scaling laws, going back to that original Dario meeting, effectively. You know, people can quibble about the exact relationship.

41:05But if you spend 10x the money, you get 2x the intelligence. And that's held. And as long as that holds, and you believe that there's a very high demand for leading edge intelligence, it leads you to spend more money. And it leads you to compete with your competitors. because if you're spending the money, you've got to get an ROIC on that. And guess what? Now you have to go into search more if you're Microsoft or defend your turf from ChatGPT and Claude. A few other little areas before we get into the upcoming IPOs that are just coming to mind. Is everything that we've been discussing based on that thesis that you said that compute equals revenue?

41:51I mean, I guess I just want to test therefore that. i'm trying to think through the question but i can't but have you thought about how that could be wrong like yeah it's it's the it's the that is the central thesis right now today on i keep looking at my watch to find out the date i don't know what today's date is but june uh i don't know guys shout out the date 10th or something fourth june the fourth june the fourth 12 30 p.m uh british summertime i get mocked at sky because i think in quarters i know we're late q2 but my My life is a bunch of quarters back to back in a good way. So compute equals revenue, right?

42:29So what's a great counter case? How would I steel man the argument against compute equals revenue? It's that, Dom, you were overestimating the return to leading edge intelligence. And these N minus one models, the models that are almost as good, are dramatically cheaper. And if you have intelligence that's almost as good and 90 % cheaper, why wouldn't you use that instead of the leading edge intelligence? And so these are the rise of these open source models in China, which I actually think there's some really interesting ones. Every software vendor in the world will tell you this. I've never met a CEO of a software company who doesn't tell you, yeah, we do all of our hard work at the leading edge and then we go to N minus one the second we can.

43:21And they so desperately want this to be true because they don't want to live in a capital intense world. They don't want to move to Sparta. They want to live in Athens where there's diminishing costs. They have very high marginal profitability. And that's not the type of technology AI is, right? As long as the scaling laws hold and you can keep doubling your intelligence for every 10x compute, again, roughly speaking, I think this is a dynamic. And then people say, well, isn't intelligence enough at some point, right? And I think if you just think about IQ, that may be the case, right? What's the difference between a 160 and a 170 IQ?

44:05I don't know. You don't know. No one knows. if you think about intelligence as task completion that's that's very different and and even today as much as i love my codex there's a bunch of tasks it still gets wrong and in six months it'll be better and a year from now will be better yeah i guess it still does rest on there being an ongoing acceleration in terms of the intelligence development and that we might reach the top of that one point another little offset question i mean you might be right at the theme as a whole but clearly there'll be some losers from this and maybe it won't be winner takes all but maybe there'll only be one or two winners and and there's a lot of market cap that sits across you know whatever it is 38 of the s &p 500 now represented by this sector as a whole yeah what's the risk that half of that is not profitable like it is today in a few years time i you've got to get the stock selection right you have to get individual stock selection right and that comes down to the framework understand so linchpin let's start with the first part of the so i so i i get that and i guess my question on that is if half of those stocks have a big correction that is hugely relevant for our listeners my is my point i mean i think you're going to see the S &P 500 falls hard in that scenario.

45:27Yeah. Look, I think that this is the beauty of active management, right? And what is my job? My job is to pick relative winners and relative losers by following the framework. So another way you're asking the question on my framework is whose linchpin status is getting stronger and whose is getting weaker? Is AI positive to your business or is it negative? Great question for Apple right now. Right now, we haven't seen any negative effects to Apple's business at all because of AI. If anything, they've probably had a modest acceleration. People are probably upgrading their hardware a little bit faster because, you know, their DRAM's a little light in their old ones.

46:06They don't even know it's light, but, you know. Going forward, if Apple can't make the shift to agentic operating systems, it's going to be a major problem for them. And I think we're going to start seeing the first agentic operating systems come out in 27. Now they're partnering with Google. Let's see what happens there. But, you know, Apple's service line is primarily driven by one big payment from Google. It's$25 billion a year or whatever the number is. Does that balance a power change as Google provides more AI? These are all the questions we're asking ourselves internally right now. yes, active management really matters right now, whose linchpin status is getting stronger or weaker.

46:50And then you better have the revenue to justify the spend. And so one reason I think we've seen such incredible stock performance year to date is because before Claude Code, people didn't see the revenue, right? People didn't see agentic taking off, but Claude Code and OpenAI Codex have opened everyone's eyes to, oh, wow, there's a huge enterprise opportunity here. That's really interesting. I mean, clearly that has been one of the big shifts as we referenced at the start. We're getting squeezed on time, so we're going to have to drop a couple of topics, but definitely want to hit these IPOs. Now, obviously, as you said, in the private market, you have some positions in open AI and in Anthropik.

47:33So I really want to come at this question via SpaceX. A hard one. But we've touched on one of these issues on the technical market dynamics. And as we've discussed on some recent past episodes, the indexes, rules are changing, which is perhaps an amber flag to some people. We'll be forced to buy them quite quickly. SpaceX has not had that revenue uplift that Anthropic has had. I mean, we've got the pricing now. 1.8 trillion market cap, 20 billion revenue. Is that mad? Obviously, I'm not going to kind of say what our plan is for a live deal, right? But let me go through kind of how I think about the different businesses and the different puts and takes.

48:18And then how I kind of compare and contrast their positioning in the AI world versus Open AI, Anthropic, and Gemini. So number one, I think launch is an incredible business, right? They're, I mean, they're the only ones in the world that can launch rockets and land them. My three-year-old has never lived in a world where we can't catch rockets. Isn't that an amazing thing? That's a good way of saying it. Every Saturday, we get up or we watch rocket videos. And then there's connectivity, right? Starlink, I think, really interesting questions. What does that mean for communications all around the world?

48:59Comcast stocks, it's not been great. Well, so I think if you had a lot of fixed assets into areas that can now be addressed by satellite communication, which is really strong, I think it does kind of beg questions. I don't know the answer, but it's something to think about. Okay. Then on the XAI side, they have unique data in the form of X. I do think there's this trope going around the internet now. Now, Elon is the best at turning atoms to electrons. I mean, he's really good at building stuff. He's great at building data centers. I think that's true. And I think it's very interesting, this cursor call option they've done, right?

49:44Because they need a harness. So the model is very important, the intelligence of the model. But the harness that surrounds it is also really important. And actually, that's the takeoff in Claude Code was you figured out how to harness the model really well. And so I think the question is, can they create a strong digital enterprise solution between a leading edge model and a harness? And I don't know the answer, but if they do, I know it's a big number, but I don't know the answer there. So those are the puts and takes on SpaceX. Compare that and contrast that with open AI and Anthropic. So like you said, Anthropic very focused on enterprise, has done a really strong job there.

50:29OpenAI focused on both, 900 million monthly active users, whatever the number is, and a very strong enterprise business as well. combined those companies you know based on the public reporting look to be on a path to to to something like 200 billion dollars of run rate revenue by the end of the year so so who knows a lot could change that's the public numbers and combined they're like two trillion dollars of value right based on the the public numbers um that doesn't sound insane to me, you know, by any means. So I think you have to think about all these differently and you have to see where the prices come and then you got to play the game on the field.

51:19The question around SpaceX is can they take a leading edge model positioning, which they have a very strong leading edge model and put a harness around it and generate strong enterprise revenue. I mean, the interesting thing there, though, is OpenAI and Anthropic, based on your numbers just there, is still 10 times revenue, which is a lot. Fine if you're growing like a weed, but it's not 10 times earnings. It's 10 times revenue. But to my point, SpaceX, roughly$2 trillion, just below that same valuation, but for$20 billion of revenue. Let's see is kind of my answer. I think you're raising a really good point.

52:01And I think the question is, is there latent revenue growth in XAI and Kershaw? And frankly, I haven't decided yet. So we'll see. Do you think these IPOs as a whole are the, I mean, by their size they are, but are just unbelievably important for the market as a whole? Yeah. You know, yes, unequivocally. You know, Ray Dalio has been talking a lot recently about what makes bubbles pop. And it's when you need to turn wealth into, you know, equity wealth or assets into income. Right. And effectively, the S &P 500 has been buying back stock for a long time as a whole. And now we're in a world where supply is coming on in the form of new securities, in the form of, you know, companies like Google doing equity raises, right?

52:57So I do think they're very important. I think the question for me is how much does this passive dynamic change the typical pop and then bleed down that you see combined with the magnitude of these IPOs being so large, right? I mean, they're insanely large. What's so interesting, though, is as large as these IPOs have been, the capacity for the U.S. companies to raise in the private markets has been as large, if not larger. and so yes these ipos really matter but i also i also think there's this unique statement on the depth breadth and strength of the u.s capital markets that it's been able to support companies getting this large and and and then to to this phase i mean it really i think talks a lot about the u.s system it does i mean the u.s capital markets are just unbelievably deep aren't they?

53:59And liquid. It's been highlighted, as you say, from this is going to be a fascinating couple of weeks with these IPOs. We, Dom, we really are out of time annoyingly. So we have to jump to - We didn't even talk about valuation. To our final topic, which is, well, dwell on that quickly then for me, valuation as a whole. But I guess, you know, the companies you have, you're very comfortable with. I'll very quickly, because I think it's such an important topic, but I wanted to make sure we hit it. So first, let's just do semiconductors versus software. Both are trading at like mid-20s, earnings multiples, the semiconductor sales multiple is now closer to nine times, the software multiple is closer to six times.

54:34Historically, that would be flipped, right? That would be the other way. Part of that is the margins have gone up a lot at semis. And part of that's the questioning around the long-term software business model. What I find so interesting right now is that the Mag7 actually trade at a cheaper peg ratio than the rest of the market. So P.E. to growth, right? So they trade at roughly a peg ratio of one times, you know, 22 times and 22 % earnings growth. Rest of the market trades at like a 1.25 times peg. And then so many of these companies we just said, the large semiconductor companies, trade at peg ratios of 0.4, 0.5 times.

55:14So you're talking incredible growth, 30, 40, 50 % growth. But then you're talking, you know, 15, 20, 30 times earnings. Thanks. So I look at valuations today. I think they're fine. That doesn't mean that you can't have momentum reversals, right? We've had a very heavy momentum market year to date, just from a factor basis. You probably do have a momentum reversal at some point. Like we said earlier, software is flat, semis up 90%. Again, you can totally have a reversal. But if I look at the underlying valuations, I think that they're relatively healthy. That's really interesting. And actually, Howard Marks also said that the Mag7 are great companies and justified.

55:56It's the other 493 he worries about when he came on with us. So final question, Dom, for you. As we flag before the conversation, we ask everyone this. What is your closing, overriding piece of investment advice for our listeners? So I don't have one. I have three. I know we're well over. Bring it. So number one is find a framework that matches your personality. So I hope you could tell through today. My framework and I are aligned, right? Linchpin technologies, innovating and secular growth markets, improving fundamentals, reasonable valuations. I once listened to a framework talk where the PM said, I'm a really boring guy and I like pouring stocks.

56:37And I thought that that was a great framework for him and it wouldn't work for me. So find a framework that meets your personality. Number two, work just relentlessly hard. One of my mentors once said, this is a business where people have won Nobel Prizes saying what you do for a living is impossible. That's the efficient market hypothesis. This is impossible according to that. So you have to work relentlessly hard. And then number three, you got to get a little lucky. You got to get really lucky. You know, the Romans had this phrase Felix, and it was a title they would give to people. And so Sulla was famously Sulla Felix, Sulla the fortunate one, Sulla the lucky.

57:24And it sounds cliche, but I just feel so lucky. You know, my coverage coming in, getting small cap semis in 2015, moving at the right time, taking over the strategy the day after Chachi Patti. You have to get lucky. And hopefully that extends to personal life, right? Lucky in terms of a wonderful wife, kids, parents, sisters. So get lucky would be the great advice at the end. Well, it's a lovely way to end it, Dom. It's been a real pleasure. Thank you so much for joining us here on the Master Investor Podcast. Thanks for having me, Wolf. And next week on the Master Investor Podcast, we will be joined by my great friend and former colleague, Becky Quick, the legendary CNBC anchor.

58:03So make sure to hit follow or subscribe on your podcast app if you haven't done so already. The Master Investor Podcast is sponsored by LSEG, Interactive Brokers, the World Gold Council and BNY Investments. Please do remember the views expressed in this podcast are for general information purposes only. Nothing in the podcast constitutes a financial promotion, investment advice or a personal recommendation. More on that in the show notes. this podcast is produced by Paradigm Productions and Master Investor Limited in association with Birdline Media if you've enjoyed the show please do subscribe on YouTube or click follow on your podcast platform and you'll be automatically notified each time a new episode drops