David Tron, CFA - Senior Portfolio Manager, North American Equities

There is a debate raging in the U.S. equity markets right now as to whether or not we are in an AI bubble. The consensus view is that this time period is similar to that of the.com bubble in 1998 to 2000. We would argue this time period actually looks a lot more similar to that of a commodity supercycle akin to that of 2002 to 2007.

And the evidence to support this is simply that this particular cycle over the past four years has been nearly, predominantly been driven by earnings growth as opposed to multiple expansion, whereas 98 to 2000 was almost purely multiple expansion, speculative multiple expansion. We think it's really important to resegment the U.S. equity market today to better understand this.

Four years ago, the Mag 7 was popularized. The seven largest stocks in the S&P 500, drove the bulk of the earnings growth and then the 493 didn't drive as much earnings growth. We would argue that Mag 7 classification is getting a bit tired. And there's a new way of looking at the U.S. equity markets today, which is simply five very large AI CapEx spenders meta, Google, Microsoft, Amazon, Oracle. 40 AI suppliers who are beneficiaries of this CapEx think Nvidia, Broadcom but also GE Vernova, Digital Realty and some of the energy stocks and then everyone else. So 5 stocks, 40 stocks, and everyone else. As it relates to S&P 500 weight, it is really 20%, 20%, 60%. What's particularly notable about this time period is that the bulk of the earnings growth in the S&P 500 in 2026 and in the couple years prior, has been largely driven by these 20 AI supplier stocks that are benefiting immensely from the CapEx to the tune of 770 billion in 2026, and likely one trillion in 2027.

What is immensely important for us as we look forward to making money in the stock market, is where are the bottlenecks within the supply chain for this CapEx, which bottlenecks should and could be bottlenecked by the AI spenders (the CapEx providers), and which one of these AI suppliers are most likely to see durable earnings growth as opposed to de bottlenecked earnings deceleration going forward?

So that is what we're focused on immensely today, and I think that's the right way of framing today's environment.

Software stocks are in a stark drawdown right now, which started roughly a year ago. To better understand where the opportunity lies, if there is any within software stocks today, I think it's first very important to understand the AI data centre, which we're increasingly thinking of as a token factory. Think of a token factory as energy coming in the door, flowing through compute and math and other componentry around those things,

and at the other end of the factory, spits tokens. Tokens are what are being purchased by software companies today, as well as many other businesses going forward. Tokens will be an input cost for all sorts of businesses going forward. What is a token? A token is simply a syllable of a word, an output of an LLM. “Canada”: three tokens.

Software companies are one of the largest purchasers of tokens today, because they're going to have to start offering AI modules, AI products to customers, because that's what customers want - and that is likely going to be the better way that software companies are going to be doing business going forward. All to say, tokens are now an input cost to software businesses, and software businesses do not create tokens.

They have to purchase tokens from the likes of Open AI, Anthropic, Google and whatnot. So it is our view that the average software business, operative being average, the average software business is likely going to be an inferior business going forward, simply because the input costs are going up and because they're going to have new, competitive, pressures from AI native companies that are going to be trying to do the same thing that they're doing going forward.

So as we think about the software landscape, we think about it in three buckets: software businesses that need to sell tokens, software businesses that likely don't need to sell tokens because they sell to perhaps a less sophisticated end market, or they're just incredibly mission critical, and software companies that perhaps need to sell tokens but can actually ride on the back of the AI super cycle.

And because they provide the scaffolding for the agent infrastructure, which is likely to proliferate in 2027 and 2028 - unsurprisingly, we are more cautious on the group of software stocks that need to sell tokens. We are increasingly interested in this cohort of businesses that we don't actually think need to sell tokens to maintain market share and maintain margins, but have been pulled down alongside this first group.

And then the third group is the one that we own the most of today, and the one that is doing the best today because they are seeing the most revenue acceleration by virtue of providing the scaffolding - the picks and shovels - to agents that are being built today and will proliferate in 2027.

So we believe the most attractive segment of the U.S. equity market right now is not, in fact, the large or the mega-cap stocks, which have seen relatively prolific returns, both in terms of earnings and multiple over the past four years or so. But in fact, the better and more attractive opportunity, we do believe is within the mid-cap growth arena of the U.S. equity market.

Think 5 billion enterprise value at the low end, 60 billion enterprise value at the high end. The reason for this is because there is tremendous amount of CapEx still being spent by these large hyperscalers. Much of that CapEx flows down to revenue for many of the picks and shovels mid-cap stocks that we do own and continue to own.

But I think what excites us more about the U.S. mid-cap market is the ability to harness this technology that is being built today and is going to be with us for our lifetime. These companies that can harness this technology today to increase margins and inflect revenue growth. What we're finding is that the businesses that we own in the mid-cap market tend to be winning businesses and growing industries.

The market share owner in a particular profit pool that is growing. Implementing AI is tricky. Doing it very well requires a good culture and a lot of money, and we find that the largest market share owner in our profit pool is best well-suited to harness this technology and employ it better than the competitors. So what we're finding is that the most sophisticated adopters of this technology are, in fact, these winning businesses and the growing profit pools.

Many of these are mid-cap stocks that have small-cap competitors. So we do see a scenario playing out where these large mid-cap stocks can steal share from their small-cap competitors by virtue of harnessing this technology better because they have the right culture and they have the resources to spend the money to execute on this, to expand margins and to inflect revenue.