Hello and welcome to The Download. I'm your host, Dave Richardson. And we are joined by David Tron, Senior Portfolio Manager, North American Equities with RBC Global Asset Management for the David and David podcast. You had your big successful appearance last time. People loved it. And so it's great to have you back again.

Yeah. Pleasure to be here. Always fun.

So David, you were recently in San Francisco doing, I guess, some due diligence work and that’s a big part of what you do as a portfolio manager. When you were down in San Francisco last week, what did you see and what does it say about the way you're thinking about managing your portfolio going forward?

Thanks. It was a remarkably insightful trip. I think as everyone knows, we as active investors spend a lot of time sitting down with companies, asking them questions, understanding their strategy, how they're thinking about the world, so on and so forth. It was a very valuable trip to be able to sit down with some of the CEOs and CFOs of some of the largest tech companies on the planet and the ones that are the largest in the S&P 500. I would say before I went down there, my bias was to test the hypothesis as to whether or not we are in some AI bubble. Are we spending too much CapEx on something where a prize is too uncertain, I guess, would have been the hypothesis. To be completely candid, I was expecting to be validated by my initial hypothesis. However, the opposite turned out to be true. I heard a lot of optimism, of course, unsurprisingly, which is not necessarily thesis-changing, but we saw a ton of use cases, real-world use cases today of AI being applied to generate productivity gains inside knowledge worker organizations. That was insightful to see. But then I would say the most valuable insight was actually sitting down with the companies that are spending all of these CapEx dollars on these AI data centers to better understand the motivations behind doing so. Dave, as you know, there are roughly eight companies or stakeholders spending all of this capital. Four of them are public, four of them are private. You have Meta, Google, Microsoft, Amazon, on the public side, and xAI, OpenAI, Anthropic, and China on the private side. We would meet with the public companies, although I did hear OpenAI speak. What was new to me this time was the idea that a lot of this CapEx on the public company side is more defensive in nature as opposed to offensive. Keep in mind, these are some of the best business models ever created. It's a little funny to hear some of these management teams discuss being somewhat scared of other companies. We've never really heard it before. But we heard more than once that what OpenAI is doing, what they're building, who they're hiring, is a scary proposition for certain businesses. Assuming I haven't talked too much already on this one topic, but if I can just share one quick example—Meta is an advertising business, but at the heart of that business is its one big recommendation engine. This is their secret sauce. They recommend organic posts to us from family, friends or accounts that we follow, but they also recommend ads to us. In fact, the platform has gotten so good that the ads and the organic posts are looking one and the same now. In some instances, when you open up your Instagram app, the very first thing you see is an ad. That was not something that was ever even considered not long ago, but it is now. Okay, so it's a very successful recommendation engine, perhaps the best one ever created. But what worries them is a company like OpenAI using a superior recommendation engine than they have to also show ads. What do I mean by that? Here's a simple example. If I'm on Instagram and I see a US Open video and I hit the «like» button or I watch the whole video, Instagram Meta now knows I like tennis, probably. What they don't know right now is that if that video has Carlos Alcaraz hitting a winner, maybe that means I'm a Carlos Alcaraz fan. They can't see that that's necessarily Carlos Alcaraz in that video today. But with LLM recommendation engines, these image generation recommendation engines, they can watch the video. The algorithm can watch the video and see that it is Carlos Alcaraz. Then they can now have a more surgical recommendation engine and start suggesting more Alcaraz related posts or Alcaraz apparel. It is just the next frontier for advertising. OpenAI has this capability today. Meta does not necessarily have it today. In a funny way, they're playing a little bit of catch-up. I do think they'll get there. But I say all this because every single investor on the planet is trying to understand the durability, the duration, the magnitude of this AI CapEx, and the fact that it is proving to be a little bit more defensive in nature—there's existential risk to the core businesses that are spending this—that tells me that there is more durability, more duration, because if they don't spend it, then they worry that their core business is at risk. It actually made me more optimistic, more bullish on this CapEx cycle. I realized there are elements of this cycle that suggests we may be in a bubble and we can talk about that, but I think that would be my biggest takeaway from the entire week.

You walk away with a base that you're expecting these companies to continue to spend at the rate they're spending. Any perspective around how much they're spending right now?

Those four public companies I mentioned—Meta, Google, Microsoft, Amazon—are likely going to spend $350 billion this year. And next year, those four companies are going to spend even more, in excess of $400 billion. In fact, all four of them are very likely going to spend more than $100 billion each on CapEx. I think this is particularly important because these companies used to be capital-like businesses. They were internet businesses. They didn't build these monster data centers. They had tremendously large free cash flow margins, and they grew incredibly fast. Today, they are more capital-intensive. They are extremely large. One thing we always like to remind ourselves is that advertising is a cyclical industry. Historically, digital advertising ate legacy advertising—print, media, you name it, TV—because it was just a better form of targeting. It's eaten most of legacy now in as much as digital advertising is now just advertising. We do wonder, and this does give us worry, that these are cyclical businesses now. They're cyclical and they're capital-intensive. Typically, not a particularly healthy combination for something that you’d want to pay a large multiple for. The good news is the multiples aren't overly, strenuously high right now, but it is something in the back of our minds that we do think about.

How critical is it that you as an investment manager pick the winner? Typically, these businesses have been «winner take most» or «winner take all». Is that something you're thinking about or how do you manage through that process to make sure that you're going to get the best for the people investing in your portfolio?

In this particular instance, as it relates to AI, picking the not-loser is going to be tremendously important. Certainly, picking the winner is going to be valuable also. One of the CFOs of one of those large four companies that I mentioned, whom we spoke with, we asked, do you really need to build this or can you not just license it from, say, OpenAI or Anthropic? License the frontier model, GPT-5, if you will. Why do you need to spend tens, hundreds of billions of dollars? To which they replied, well, we're not sure if in the future, A, they will sell it to us at a price we’ll be willing to pay, or B, sell it to us at all. They really need to do this themselves, and it's entirely unclear who the winner is going to be, I should say. The race changes in terms of who's the front runner on a quarterly basis, I might say. I do worry, though, that this race is going to become so expensive that the returns for finishing first—in as much as you'll have three or four first-place finishers—might not actually justify all the capital being spent. This is another thing we think about a lot. It's entirely unclear. But back to your original question, we're always thinking about who's going to win this? What are the ingredients required for winning? But again, picking the not-loser is more important than picking the winner, if that makes sense.

Yeah. Do you notice that in the culture when you're out there, the idea that they're really trying to beat each other up or be the winner or drive someone else under almost? Or is it more of, we're building the next big technology that's going to move the human race forward and we're all in this together and everything's going to work out for everyone?

These are some of the most intensely competitive people on the planet, and I've never seen them more competitive than I do today. I'm not sure if people have seen this in the headlines, but Mark Zuckerberg has hired a team of somewhere in the neighborhood of 50 to 100 people, and he's poached them from the likes of OpenAI, Google, Anthropic, and he pays them sports team salaries. So the average person on that team is probably in their late 20s, early 30s, and gets paid on average $50 million a year. Yes, that's 5-0-million-dollars a year. These guys are all in on this, and they're willing to pay aggressive prices for the best of the best talent. It's as competitive as I've ever seen it in the stock market. Yeah, it is crazy out there right now.

That's at the builder side, but who's supplying everything that they're buying? Has that been a major focus of what you've been investing in the portfolio?

Just to level set everyone's knowledge base here, there's two forms of modeling, if you will. There's something called training and there's something called inference. Just to give a simple analogy as to how to think about these two things, if anyone knows the London taxi cab drivers, historically they have spent long periods of time studying the maps, studying the shortcut, studying the landmarks inside of London and just burying all of that content into their brain such that they can then apply it to the real world. This is what we would call the training portion of building a large language model. It's time-intensive, takes a lot of capital to do so. This is historically where NVIDIA has sold all their GPUs to. On the inference side, I think of this as I go into the back seat of a London taxi cab, I tell the driver I'd like to go somewhere in London, and he'll know exactly how to get there, take a couple of shortcuts, and get there as quickly as he can. That is the practical application of the training portion. There's two companies that are effectively providing the brains of the data center, the GPUs or the XPUs to this process. On the training side, studying the map, it's been basically one company thus far, and that's the one everyone knows, which is NVIDIA. They have pretty much a lock on the one chip that everyone wants, which is their GPU, the latest and greatest iteration of their GPU. As we shift more towards an inference world, which is basically the monetization of all of that training CapEx, there is a couple of other companies that do have a right to win there, one of which is this company called Broadcom. Broadcom is actually one of the largest companies in S&P 500 right now. What they do is they do something called Custom Silicon, XPUs. I'm just going to use Meta consistently here. If Meta wants to build their own chip, perhaps they don't like paying NVIDIA for all the chips because NVIDIA charges a lot of money. They would call up Broadcom and say, we'd like to build our own custom ASIC, our own custom chip to use for inference. Broadcom works with them, designs it, and then it gets built, and then they can put it into their data centers also. You've seen Broadcom's share price go up a tremendous amount. In fact, it's more expensive than it is NVIDIA stock right now on a multiple basis. A lot of people are betting that this inference market is going to be orders of magnitude larger than the trading market. Therefore, perhaps, Broadcom is a better stock going forward than NVIDIA. Of course, NVIDIA is also going to want to try to win some of that inference market, and they do own some of that market today. We do see a bit of a two-horse race between those two, call it picks-and-shovels providers for the chips into the data centers, which is important because these data centers, as expensive as they are, anywhere from 30 to 50% of the data center cost is just the chips. These are very expensive items to put into a data center. Those are the two that we've been focusing on. We have had a bias towards Broadcom. We're more interested in the monetization phase of this AI cycle than we are in the training phase, and we've had more of a negative bias, if you will, towards NVIDIA, over the past little while.

Are we really seeing AI pay off in other companies in terms of using the technology and monetizing and making their businesses more efficient, more effective, cost-cutting, generating higher revenues, more sales? Are there any particular areas that have seen the most progress?

There's a couple pockets that everyone points to that we've seen thus far. The first one would be contact centers or customer service. Rather than calling up your phone company and speaking with an individual, perhaps you’d speak with an artificial intelligence agent and have a conversation with that agent. We are seeing real-world applications of that, sadly displacing some of that human labor. That would be a good use case. Then the second, perhaps even better use case, is advertising. I think I've talked ad nauseam about Meta already, but there's a couple of other advertising companies, one of which we own in our US mid-cap growth product called AppLovin. And they show ads inside of mobile games, which hasn't historically been a sexy place to show ads per se, but they own the majority of that market, and they use AI or a machine learning, I should say, to match the right eyeballs with the right ad inside of mobile games. You might think, mobile games? What the heck's up with that? Well, a billion people play mobile games around the world. The people that do play mobile games tend to spend more time in those games than they do on social media, and the users skew more female, of which advertisers have historically had a tricky time targeting. This is a company that we like that is showing leadership in AI and machine learning to better match eyeballs with ads. We're seeing real-world use cases for these recommendation engines, and advertising is the poster child for this thus far.

Let's shift gears a little bit because we've talked a lot about AI and tech, and maybe you'll give me another area of tech that's a response to this question. But it seems like AI is driving the real excitement and a lot of the performance in stocks that are driving the major indices. Are there any other areas of the market that you think are fairly exciting? Again, it could be another area of technology or financial services. Any other area of the market you're particularly excited with?

There are pockets of industrials that myself and my co-managers would suggest are right for an inflection in earnings with some room in the multiples. This tax bill that was signed into law in July provides extreme incentives for US corporations to reshore some of their production. There's an accelerated depreciation schedule that allows these businesses, when building, say, a factory in the US, to depreciate 100% of the cost. Very early on in that life cycle, which shields some of the taxes being paid, the payback period for one of these factories shifts from six and a half years to five years, which doesn't sound like a whole lot, but for a business in a competitive industry, it can mean a whole heck of a lot. We are very intrigued by a number of industrials' businesses that can facilitate some of that construction, engineering, pics and shovels associated with actually bringing reshoring back to the US. It's an area that we're spending a lot of time focusing on. That would be the first one. The second one would be, there is a perception that AI is almost certainly going to kill certain businesses. As a result, some of these businesses, the multiples of these businesses, have been compressed quite meaningfully. Software is smack dab in the middle as disruption worry right now. A lot of investors are thinking, why would I need to buy some software from, say, Salesforce when my company could now use an LLM to build my own Salesforce software or something along those lines, or perhaps as an AI native company that's going to steal incremental share from Salesforce. What I would say to that is this is an area we're doing a tremendous amount of work on. I do believe there is some truth to this thesis that AI will deprecate some of the future growth from some software companies. But I think we've seen a beta sell off in software. What we're going to see going forward is a dispersion in the best ones moving away from the worst ones. I think it's going to be a big alpha opportunity. We're spending a whole lot of time trying to figure out which software companies not only going to be resilient from an attack from AI native or just in-housing software, but also optionality from using AI to sell incremental products and services to their existing customer base. That would be the two pockets of the benchmark that we feel are somewhat mispriced from an earnings multiple perspective.

I'm listening to everything you're saying, and we've always talked about portfolio management, one of the things, it's almost more important to avoid losers than to be precise in exactly the winners that you get. Avoid the Nortels of the world and just do okay with everything else. You're probably going to outperform your peers. But it seems like this AI world is creating even more of an extreme, that there's the potential that we're going to see more of these big losers. Almost like talking to an individual. My kids were home from university this weekend, and some of their friends came over, they're worried about the degrees they're taking because what jobs are going to be the winners or losers when they graduate, after they get their education, with AI in the background. Do you feel that way as a portfolio manager that you're almost looking as much for what are going to be the companies that disappear as much as the ones that are going to be the big winners?

I do worry about that a lot. Yes, I do. It's funny going back full circle to the San Francisco conversation, we spoke with and heard from a lot of venture capitalists who, again, keep in mind, are some of the more optimistic people you'll ever meet. Nonetheless, what they told me is, for the first time in history, they're adding zeros to the total addressable market of some of the markets they historically invested in, say, software. They're adding a zero—and this is a bit of a scary statement and I'm not sure I believe it just yet—they're adding a zero because for the first time in history, some of this technology is able to displace knowledge work and human labor. This is what they believe, and it is a scary proposition. Historically, the economy has been remarkably resilient and has found new jobs for those people that have been displaced. But just to give you a sense as to how people are thinking about it, when you add a zero to a number that already has a lot of zeros, that obviously makes it a much larger number. But this is the level of optimism that currently resides in Silicon Valley and San Francisco today.

Yeah, and you can see it. We were talking just on the side before we started recording. And many Canadians, for a number of reasons, have not been traveling to the US as much as they might have in previous years. I know I was in Phoenix earlier this year. You were talking about San Francisco and the driverless taxi cabs that are zipping around the way, all over the place. And if you don't go down there, you see, oh, yeah, okay, there's a couple of cabs driving around, and the regular cabs are there. But the sheer number of them, how much you run into them, it's just incredible. You think of that going to trucking and all kinds of other areas of industry. It's quite amazing.

Absolutely. Yeah, it's freaky to see a self-driving Waymo with a passenger in the front seat and no driver in the driver seat.

Or a whole group of people in the thing. It's something else. So David, let's just get back and finish off with, just at the core of what you're thinking about what's going on in the US economy and how that flows through to the markets. I know in the spring when we spoke, you thought that 2025, first year of a new administration, you get a reset, a readjustment. There's new policies being put in place. In this case, your thinking was the Trump administration is doing some of the heavy lifting, some of the harder work, maybe even some of the more negative stuff at the front end with the anticipation where Congress is up, the Senate and House of Representatives are up in 2026 for the balance of power in the US. And so you want to have everything humming along come election time next November. And that was setting up perhaps a better 2026. Is your view still aligned with that, or have you shifted a little bit because we've come so far so fast market-wise in the last six months?

No, the view still does align with that. We try to be consistent here. Obviously, as the facts change, we want to change our mind but the facts haven't changed a tremendous amount. In fact, this tax bill is remarkably stimulative to the largest companies and broadly, most consumers, particularly at the low-end will receive some form of stimulus over the next 12 months. The employment picture is relatively healthy. The consumer is relatively healthy. It should be a relatively good holiday season for a lot of retailers. The AI CapEx trade is obviously still in full force. Then the CapEx xAI dynamic is one that is inflecting, and we do believe is continuing, is just starting. The only thing, I would say we're below trend-ish growth right now for the economy, generally speaking. But as we head into 2026, the headwinds from the tariffs will turn into tailwinds because we're going to lap that period of time, but the benefits from the tax bill will remain, and we're likely going to have some rate cuts that will be stimulative. I'm not even sure if we need these rate cuts, but it is what it is. The only thing standing in our way is likely going to be the multiple of the broader benchmark, which is already starting to price in some of this good news, but I do think there is some more room to run. I think it's a good time for active management right now because there's going to be a lot of capital moving underneath the surface within the benchmark to provide new leadership going forward. I do believe that. We're always hunting for the mispriced opportunities, the inflection points that are just around the corner. I think the latter half of '25 and certainly '26 is going to be a remarkably healthy time for active management. We love these periods of time.

Yeah, the whole discussion we've had seems to point to that. Again, that whole making sure that you're avoiding those losers as being as important or more important in finding the big winners, and then making sure that you're identifying what becomes the next leadership as this one trade can't go on forever. I'm sure people are listening and thinking this, having someone like you managing the money for you is perhaps is a better way to go about it.

Yes, I do think so. Just to level set my view on AI, I do think we'll get an overbuild eventually, certainly on the trading side. I do think if we're not already in a bubble, we might go into a bubble. What that means exactly to the forecaster, I'm not sure, because bubbles can grow and get bigger for long periods of time, and it's difficult to know when they pop. But we're level setting our view, the optimism and the real-world application associated with AI with there is likely going to be an overbuild. So our head is on a swivel and we're standing on our toes at all times when we talk to these companies and we look at incremental news. So that's how we're thinking about these things.

Well, fantastic. David, fascinating conversation. I didn't even think we were going to get off in some of these directions, but I know the listeners must have enjoyed your insights. I learned a lot, as I always do, talking to you. So thanks for joining us, and hopefully we'll be able to get you back on again. Maybe let me know the next time you go on an interesting trip, and we'll get you on for an update after that.

Absolutely. Thanks so much for having me.