Hello and welcome to The Download. I'm your host, Dave Richardson. And we are going tech-heavy today with our good friend, Marcello Montanari. Marcello, welcome back.

It's great to be back. I always have a good time on this podcast.

I'm actually changing my interview approach.

Really?

And you'll get this. So I was actually out in the public domain, and somebody had gone on to ChatGPT and said, describe the host of this, the Download, describe his interview style. And so it came up with all this stuff, which was actually quite nice. It was quite complementary in a way. So again, it's a very enjoyable experience for the guests. So I'm going to change style, so I'm going to annoy you today. We're going to be more confrontational.

Oh, my God.

We're all afraid of robots taking over, and you're probably leading the charge on that.

Well, I'm not leading the charge. I'm just observing it. Robotics is going to hit us sooner than I think most people think.

Well, one of the things we missed, and so again—just to annoy you because I know how humble you are—but you and Rob Cavallo, who we also have on the podcast, are investment executive fund manager of the year in 2023. We somehow missed that at the start of the year. We just listened to all your great ideas. So congratulations. That's a well-earned honor. All of a sudden, an email just came out today about it.

It was just referencing back to what has happened over the last year. Here at GAM, for the portfolio management team, our measurement year of how we measure our performance is basically September, or October 1st to September 30th. So that happened in this fiscal year for the sake of our performance. That's probably why it got mentioned in there.

Well, it's still a heck of an honor because you got a lot of people who are operating in this space, and this is one of the reasons we love having you on and why so many people listen to your appearances. Because this is a space that everyone is interested in. But to get that recognition for the fantastic work that you've done for several years now, well, congratulations.

Thank you.

But I knew it would annoy you because you're so humble.

I'll pass on the kudos to Rob.

Rob is pretty fantastic. I have to admit that. I don't know if we're doing a video for this one, but I thought Marcello was about 20 years younger than me, and we end up being roughly the same age. And we've been with the firm together for pretty much exactly the same time. So that was an interesting one. And Rob's the young one.

Yeah, he is. He's just a baby.

So then very quickly, because we actually have some interesting things to talk about. How long is this tech thing going to go? This just feels like nothing I've seen in a quarter century.

We're calling it a super cycle, which means it's going to last a long time. When you think of the last super cycles we're familiar with, which is the Internet, we're still in there, right? But we had the bubble, and then we had the comedown from there, and then that was followed up with mobile, which fed into Internet. So you had another big cycle there. And then you had cloud computing. So these have all been really large cycles. But artificial intelligence, don't take it from me, just listen to people like Bill Gates and Jensen Huang, people who really know this space more than just about anybody, they're calling this akin to the onset of computing, the onset of mobile and stuff like that. So this is big. So I suspect it'll take many years. And these things go through cycles. First, you have to deploy the technologies. They need to get embedded into the economy. And then once the economy starts to use them, then you get all sorts of benefits at that end. So these things last a very long time. And through each part of each stage, you get different winners, and in some cases, different losers. So we're still in the deployment phase where we're starting to deploy all these things. And every company—RBC is perfect for this—everybody's looking at it and brainstorming, what can we do with it? How can it help our operations here and there? And then we put in a little pilot, and we test it out, see if it works. And if it does, then we either devote more resources to it or we say, hey, that was great, but this other thing over here was much better. So let's do that one first. So we're in that process now where that's all happening. But in the meantime, all the infrastructure is going into place. And that's why you've got these winners like NVIDIA who are right at the infrastructure layer. It's NVIDIA, it's Broadcom. And then you got the hyperscalers who are basically putting in the infrastructure that will actually be used by end customers. And that's all going in, too. So they've been the big winners in all this. And that'll probably last for some time. And then it'll broaden out. And you're going to start to see it more at what we call it the application layer. Microsoft has a whole series of applications, like a GitHub copilot, the copilots that they're putting on every desktop. And we just saw some interesting news come out of salesforce.com and ServiceNow. So they're now starting to get embedded within applications.

Yeah. You're watching and tracking along with this every day because you have to do that, to understand where you want to invest people's money in the space and maximize the right exposures, get the right returns for the clients. But this is where it's getting to the levels of me, middle manager in a big organization, and how can AI help me day to day. We've got all this data. I've got customers. I need to understand my customers. I've got a team of people. And how can I use AI, or how can AI even jump before I even know I'm missing something and give me insights into what my people should be doing, how we can better serve our customers. And that's what you're saying is where we're really starting to get to. And that's going to be the next phase where we really, really start to feel AI more than just hopping on ChatGPT and getting them to write a letter for me, but actually get those insights to drive businesses forward.

Yeah. And it becomes symbiotic, because as more and more organizations start to implement these things, they're going to need more and more infrastructure in order to support its diffusion through the economy. So it's going to feedback and continue to help propel the hyperscalers who are putting all the GPUs and everything and the underlying chip manufacturers and the networking guys and the memory guys. So it's big. And that's why multiples have run ahead. Nobody really knows how big, what the future will really hold. Every time I look at this sector, I go, wow, I didn't see that coming. That was weird. Things happen in ways you didn't expect. That's what's happening here. I don't profess to know how everything's going to happen, but it's all symbiotic and it's all working its way through. I lost my train of thought, but coming back to the multiples, nobody really knows. What typically happens is investors say, I don't know how big this is, but it's going to be big. And so multiples will creep up to try to capture a little bit of that upside, front run it a little bit. So we've certainly seen some multiples get higher in the sector for sure.

Yeah. And then when you're in that phase of growth, you're going to see some ups and downs. It's like any market. It's never a straight lineup. Although it's felt like that more recently, but it's never a straight lineup. But when you're investing—and we talk about this all the time on this podcast with all our guests—we're not trying to guess where something is going to go over the next three days. If we're investing, we're trying to see the big picture opportunities over decades because if we're saving for retirement, even two old fellows like ourselves will still have time to grow our investments above the rate of inflation, grow our wealth, and create a better retirement for ourselves and our families. And that's what we're trying to do. And then in your space, it's just that things just evolve really, really fast, and new stuff comes up all the time. And that's what makes it particularly fascinating and interesting.

Yeah, it's a lot of fun. To that point, whenever Rob and I are doing presentations around the funds that we're running, we highlight that we take a longer-term approach. We expect that there will be drawdowns as we go along. We try to manage expectations. Nothing goes up in a straight line forever. And we always bring the example of Amazon. If you go back to 1997, Amazon every year has had a pullback, sometimes multiple pullbacks in the year, and sometimes quite significant. But if you look over the long term, it's been up and to the right. Like I said, we try to manage expectations. Could we have a pullback? Could we have a lull? Could things level out for six months and do nothing? It's all in the cone of uncertainty. Of course it could. I think we've helped our unit holders understand that that's in the cards and expect it. But we've been rewarded by having a stable base of unit holders that have stuck around with us.

You want exposure to this area. I don't think anybody would argue that. And this is part of the process of financial planning and investment planning, and you're working with an advisor, and you determine, as you're putting that plan together, what's the appropriate level of exposure for you. That's going to be different for everybody because some people are going to want to be overinvested in this sector, and some people have longer time horizons, and they accept more risks or more of the ups and downs. Then there's other people who are going to want exposure, but it's going to be because they recognize the value that it provides over the long term, but it's going to be a pretty simple exposure that they can be comfortable with. Again, that's why working with a good advisor helps. Having a great investment manager to manage the space for you, that's not too bad an idea either.

Yeah. The Life Science and Tech Fund is almost tailor-made for a Canadian investment faster. Canada has great banks. We got great railroads. We've got a lot of great sectors. But the two areas that we don't have that much exposure to is tech and health care. So it's a nice fit there. So we always tell people, don't put all your money in technology. It's something you add depending on your risk profile, obviously.

So again, we talk about this long, long cycle that we're in the midst of and AI is now this next big wave. What are these next things in AI that you're seeing coming. You reference very quickly the salesforce.com and agentic AI. What is that and where are some other places you're seeing or even just some ideas of what that might look like for, again, someone who's out there managing a business or managing a team of people within a business that's going to help them be more effective in what they're doing?

Yeah. So there's this term agentic AI, which is like laying in the weeds. And if you watch the sector very closely, you could see it coming. This has been in the lexicon for a while. But on the last quarterly conference call from salesforce.com, that just blew out. Now everybody's heard the term agentic because they actually have a product or a suite that goes across their applications called Agentforce. To really just simplify it, agentic AI is basically an AI that will actually do things independently and execute them for you. The easiest example to understand—and we're not there yet, but it's coming; you can see it coming—the perfect example is basically your iPhone. Your iPhone, because you live on it, it knows a lot about you. It has access to your emails. It has access to your Ticket Master account. It has access to your Air Canada account. So it's seen everything you do, what your preferences are. And so you go, hey, I'm going to New York. Book me a flight on this day in the morning, whatever. And then I need a hotel as well. And the AI agent, because it has access to all these things, in particular on your cell phone—which puts Apple and Google in a very interesting spot—it can basically see your history. It knows that you like to have an aisle seat when you're on an aircraft. It also knows that you like to fly it before eight o'clock or whatever. It can actually go out and independently start to execute the purchase of tickets for you to go and then come back. And then it knows that your favorite hotel in New York is the Crosby Hotel. I think it's in Soho or something like that, a little boutique place. But anyway, so that's where it's heading. Like I said, we're not there yet, but at the enterprise level, Salesforce is rolling it out to their client base. So Salesforce has all different types of applications, starting off with basically sales and marketing management, things like that. But they're putting in this agent that can actually start to execute things. And again, just to simplify things, if you're thinking about a call center and someone calls in and it might actually be a ChatGPT voice that's speaking to you. But if you're asking it to do things in your account, it'll be at a point where it could actually go in there and maybe change things or give you feedback on something. It'll start to act independently and actually start to execute things.

Or like when I punch in my account number and I actually get to the agent, there might have actually been a purpose for me to punch in all those numbers. They'll be like, hey, Dave, how's it going? Nice to talk to you. Instead of just punching it in and then having someone come on and just seemingly not know anything about me. That would be amazing, wouldn't it?

Yeah. And then they ask you for the numbers again, usually, right?

That's right. And I thought of something else that the agentic AI could probably do for people that would be very useful. So many people listen to this podcast, but they don't subscribe. The agentic AI could actually get them to subscribe to the podcast, which we would love, because then we know how many people are listening. And the agentic AI would know from the texting afterwards how much they love the podcast. They could give it a five-star review and continue to pump us up the ratings. I did that for the marketing people, Marcello. Did you notice that? They want me to do that, and I hate doing it, but I can weave it in different ways.

Love it. So that's the most recent trend that's emerged out of all this. I'd say the other thing this narrative has built is, are the models still scaling? And are scaling laws holding? The idea behind scaling laws, just to simplify things, is these models are gigantic. They use immense amounts of compute power. These chips all need to act, they call it coherently. So basically, each chip knows what the other chip is doing. They've got these clusters of tens of thousands of chips that are all training these models. They use immense power, like I said. It's very expensive to do to build the ultimate model, whether it's ChatGPT or LLaMA or Claude. So the view is that as you give more data and do more compute, i.e. spend more money on this, the models get smarter and smarter. So there's been a narrative and it's been fed by some stories that some of the engineers of some of the model companies, mostly off the record, are saying, we're training the model, and it's not given the results we expected. We thought it'd be better, and it's not. But like I said, a lot of it is off the record. If you look at where the dollars are going, the dollars are still pouring into this, whether it's Meta, Google, Microsoft. Last night, I was listening to a recent podcast with Satya Nadella from Microsoft, and he basically said he's still a big believer that the scaling laws are holding. But over time, it's basically an exponential function. As you try to get the models better and better, it's taking more and more compute and more power. So there will be a time where the economics go like, are we willing to spend a trillion dollars to train a model? But it seems like we're not there yet. But that's been one of the narratives that's out there.

But that was one of the things that you were talking about, as early as a couple of years ago when we had you on, this whole idea that all of this spending on the infrastructure of AI, at some point, will have to start to pay off in a meaningful way. If I'm Amazon or Microsoft or Google and I'm putting billions of dollars into this, I'm doing it for a good reason because I'm going to get multiple billions out on the other end. And at some point, you have to see that. Even though these are brilliant companies, they're long-term thinkers as we all should be, and we've already touched on that here on the podcast today. But at some point, it's got to pay off in a big way.

Yeah, it does. And that was the big narrative during the summer, which was, is there really an ROI on this thing? I found it a little odd because for very large projects, if we're building a hydroelectric dam, we're not expecting an ROI while we're in the middle of building it. You got to get it in place, get it running. Then you start to see revenues flow in. And then over time, you can say, okay, it's generating hopefully a positive ROI. So that was the summertime and so this narrative around the scaling laws, it's almost like we've resurrected the negative narrative from the summer. And now we're going to change the flavor of it. And it's all about scaling laws. But beyond scaling laws, there's time inference scaling. Anyway, it gets really complicated, way beyond the scope of the podcast. But coming back to ROI, Meta is a perfect example. They've been spending immense amounts of money on GPUs. But if you look at how they've used those GPUs, it's not just for generative AI. They're also using it for all their machine learning algorithms. So many things are basically fueled and supported by advertising. The way to get good advertising dollars is to make sure that the right ad shows up in front of the right person at the right time. And so Meta has used AI and machine learning in order to make sure that their ads are the most relevant at the most relevant time. That way, they generate more revenue. So if you look over time, the returns on invested capital of Meta are up and to the right. For some companies, for sure, there'll be winners and losers in all of this. But for the ones that are winning, I suspect the returns will be, like I said, just for Meta, it's already been positive.

And we're Canadians, I guess that's why I keep getting snow shovel ads pushed at me every time I'm on the web. Well, Marcello, that was always great to catch up with you. Always fascinating. Everywhere I go, I bump into people who listen to the podcast, and they say get Marcello on more often, get Marcello on more often. And I know you've got a real job, not being just a podcast guest. But thanks for coming on again. And we'll get you on early in the new year, because I know you're going to spend the holiday season thinking about looking into the future and seeing what's coming next. And you'll have some interesting things to say on 2025 if we can get together early in the new year.

Sure. Sounds like a plan.

Excellent. Well, take care and Happy Holidays to you and your family. Merry Christmas, and we'll see you in the new year.

Yeah, likewise. Sounds great.