Talking Tech and AI

Jeffrey Buchbinder (00:00):

Hello everyone, and welcome to LPL Market Signals. Jeff Buchbinder here, your host for this week under our new rotating host format. Please welcome Tom Shipp. And Tom, we are going to talk a little tech today. How are you?

Thomas Shipp (00:17):

I am doing well, Jeff. Thanks for having me. And pleased to be back on.

Jeffrey Buchbinder (00:22):

Yes, you have been on more recently than I have. But there is no doubt that the topic is going to be tech related, whether we talk about AI or we talk about earning season, or we talk about why the stock market is doing so well, no matter what you talk about, it all goes back to tech. That is certainly a topic that all of you are interested in, I am sure. Here is our agenda for today. Tom and I put together six key questions to frame the discussion. We will start with a high-level look at earning season so far, specifically, where are we in the capital cycle? what does it mean that technology is becoming less asset-light and why free cash flow deserves more focus, the capital cycle playbook, where does the economic surplus from AI flow, and lastly, how do we think about concentration risks within the ecosystem? So there are six topics. Let us quickly get your thoughts on earning season, Tom. It has just been incredibly strong. Of course we will leave the details on tech to the rest of the discussion, what do you make of this 27% earnings growth for the S&P 500? Tremendous.

Thomas Shipp (01:59):

It has been quite impressive and has been driven largely by tech. The last time we were on the podcast, I talked with Adam Turnquist about some of the interesting tidbits coming out of earning season, and particularly how the revision cycle for 2026 numbers really spiked up after the conflict in Iran started. The macro headline that was driving the markets down during March was actually coming at a time when earnings revisions were going up, and those earnings revisions were driven, as one might expect, by energy, because the underlying commodity spiked up, but it was mostly being driven by tech. And this was somewhat perplexing because the overarching story of AI and data center CapEx had not really changed.

Thomas Shipp (02:57):

It was as if the analysts on the sell side had just woken up and started revising their numbers. Perhaps this reflected their conversations with companies. The earning season really reaffirmed what was already going on. Q1 numbers looked very strong, over 20% year-over-year growth. With the 2026 revisions spiking up, it has been pretty interesting to watch. It goes to show that this market's drive upward has not just been driven by hopes and dreams. It has been driven by fundamentals.

Jeffrey Buchbinder (03:41):

It has not just been tech. The Mag Seven has driven about 55% of that earnings growth. The median S&P 500 company grew earnings 12%. That is pretty good. The easy comparisons to the first quarter of 2025, when we were dealing with tariff angst toward the end of that quarter. That certainly set us up for some really strong growth rates. Growth has been very strong. Tech is certainly the biggest story, and you mentioned revisions, the fact that earnings estimates for the next four quarters are up another 3%. That has been a continued pattern over the last several quarters. The spending numbers just keep going up and up from the big AI hyperscalers. Let us use that as our lead into these six topics. Tom, where are we in the capital cycle? When we started the year, we thought we would get maybe, what, approximately $550 billion in hyperscaler CapEx. Now it looks like that number could be well north of $700 billion. Those fundamentals are clearly very strong right now. But what should we be watching to gauge where we go next?

Thomas Shipp (05:07):

The fundamentals are real. We are shifting from a mindset of investors questioning the demand. We are well through that. We know the demand is there, the demand is real. Now we start to ask: will the returns on the incremental CapEx that everyone is spending be attractive enough to support underlying returns on equity? The big thing to watch, and this is an interesting read from individual companies that have reported, is that a lot of the CapEx being spent now is because the demand is right in front of them.

Thomas Shipp (05:56):

Supply constraints are there, the demand is right in front of them. They have data center demand lined up. So as soon as they spend the money, they can turn around and rent that out to customers. Whereas companies that were not in the business of selling compute, they are building data centers, but they are not selling that space. Those companies have been put in the penalty box and have not seen as much stock appreciation over the last three to six months. Investors are starting to look more for what the returns on AI and CapEx are, as opposed to simply asking whether there is enough CapEx and demand.

Jeffrey Buchbinder (06:48):

That next phase will be really interesting to watch. It is very early in this cycle, and we are only beginning to see companies show value. Speaking of early, a lot of people ask what inning we are in. I put together a chart of the ChatGPT era from 2022 of the Nasdaq and overlayed it against the Nasdaq from the late nineties, starting with essentially the birth of the internet in 1994 with Netscape. Back then the Nasdaq was essentially a 10-bagger. So far this period we are only about a two and a half bagger, a little more than that. Based on that simple analysis, we are in the early innings for sure. Let us go to our next topic, Tom. Tech is becoming asset heavy. What does that mean for how you think about valuations?

Thomas Shipp (07:52):

This is something that is starting to get more attention paid to it. The easiest way to think about this is to look at a simple calculation of revenue divided by CapEx. Or, set another way, how much of a company's revenue is ultimately going to CapEx? If we go back to 2023, still in the ChatGPT era, if we take the hyperscalers including meta and Oracle the companies were spending about 11% of the revenue they generated in CapEx in the same year. The estimates for 2026 show implied revenue and CapEx projections up over 40%. That is four turns of increased capital intensity, meaning they are now spending that much more money in property and plant.

Thomas Shipp (09:00):

Data centers are hard assets that companies must spend cash on and depreciate over their lifecycle to generate revenue and ultimately profits. At a high level, companies such as those that are software-based or a communication services name such as Google, these were typically asset light businesses that had high returns on the capital that they put into the business. And so they would be particularly rewarded with high valuations because you build it once and sell it many times. Now if we are getting into an era where these tech companies are more asset heavy or capital intensive, you might have to start thinking about that more in terms of the cyclical aspects of it, and also the returns you are going to get on that capital. So they would look more like a utility, with obviously higher sales growth, or an infrastructure type play. It is something we look at quite frequently and think about from the perspective of how to value these companies based on their returns on equity.

Jeffrey Buchbinder (10:22):

The question is how long they will remain capital intensive and what does the trajectory look like after peak spending? We could do a whole series of podcasts on that, but that is not even one of our six topics today. Topic three: free cash flow deserves renewed focus. It is related to what you just talked about. Essentially all the free cash flow is being spent to build out AI capabilities. Are there any other implications beyond the fact that maybe valuations should be pulled back?

Thomas Shipp (11:02):

This ties into the prior point we just made. At the end of the day, whether you are looking at a shorthand valuation using multiples or projecting out cash flows and discounting them back to today, that is ultimately what a business is worth. The $64,000 question is: what is the timeline? Is this a spend-a-lot-now period and then they go back to being asset-light, working off the asset base they have built and generating returns without having to continually reinvest?

Thomas Shipp (11:52):

Or is this a new paradigm where they will always have high spend? Because what that means is, as you are discounting out those future cash flows, if you have to continually reinvest so much in the business, you are going to have less free cash flow available back to the company, whether that is for reinvestment, share buybacks, dividends, or what have you. And that is ultimately how you value the business, which is a hugely important piece. Looking at free cash flow rather than just pure earnings matters here. Over time, they are going to start to look the same if the spend levels remain the same. We prefer to look at free cash flow multiples or free cash flow yields, as those have had better predictive power over long periods of time. And that again comes back to the question: are they spending a lot now and will not have to in five years, or are they becoming more asset-heavy? And therefore the cash flows will always be reduced because at the current moment, the hyperscalers are trading at very high free cash flow multiples, whereas their earnings multiples do not look as high because they are able to stretch the spend over a longer period of time.

Jeffrey Buchbinder (13:08):

That is right. As an example, tech is trading at 26 times 2026 estimates. That is current year, not even forward. That is a pretty low number. Even if you go to comm services and consumer discretionary where the rest of the hyperscalers are, you are still looking at 30 times and under. And these companies are growing very fast, and they are very large. If you use simple PE and earnings analysis, these things look cheap. But that is a great point, Tom, that on free cash flow, they look pretty expensive. But I always go back to the metaverse example with Meta, where the market said they were spending too much, so they pulled back and essentially recouped all of the valuation impairment they experienced.

Jeffrey Buchbinder (14:07):

The same thing could happen again, maybe not as dramatically, and maybe not in as short a time period, but we will eventually see the market push back on all the spending. Valuations will certainly move around in both directions as we figure out what this cycle looks like in phase two, phase three, phase four. So let us move on to the capital cycle playbook. As you have dubbed it, it essentially raises the question of whether we are seeing companies pull forward purchases, essentially hoarding, which would imply that you are seeing stronger growth now and weaker growth later. What is your sense of what is going on there, Tom?

Thomas Shipp (14:59):

That is an open question. We have heard management teams at the big three hyperscalers, Microsoft, Amazon, and Google, speak to the fact that the demand is there for every dollar effectively that they are spending. So that would suggest they are not pulling anything forward, but their end customers could be, whether it is still-private OpenAI, Anthropic AI lab, or even core enterprise customers, your typical Fortune 500 S&P constituent. They could be ultimately saying, we need more cloud space, whether we have something to do with it right now, we are going to buy it. So maybe it is the end customers who are pulling forward demand. We do not completely know if that is the case. At a higher level, the capital cycle framework is just about any capital-intensive business that goes through spending cycles.

Thomas Shipp (16:02):

Let us use the energy sector as an example. You see this time and again, where oil prices are high capital flows into the industry, there is a bunch of additional capacity added, whether it is rigs, whether it is offshore platforms, whether it is frack crews, et cetera. And ultimately, you overbid, you overbuild, there is too much supply, too much oil is produced, causing a glut in the market. Oil prices go down, and then that supply starts to get either retired or scrapped or what have you. And the capital cycle begins all over again. Typically, this has happened in various industries, including semiconductors. The idea is the same. Too much capital flows in, you start getting that pull-forward, bidding wars emerge for the assets you are spending on, and ultimately there is too much supply.

Thomas Shipp (17:09):

The same thing happened in the dot-com period with laying fiber in the ground. The big story that everyone always quotes is: we built so many fiber lines that did not end up being used. It took 20 years for internet bandwidth to catch up to how much was laid in that period. I am of the mind that since this is becoming a more asset intensive business, the data center build-out of compute for AI ultimately reflects a capital cycle behavior, and we must continue to watch. That is not a bad thing. It is not saying good, better, or indifferent. These are hard assets. There is capital flowing in, there is demand for it currently. And we just need to keep an eye on when we may be getting to a point of overspend and oversupply of all these assets.

Jeffrey Buchbinder (18:09):

I think we are aligned on this, Tom, that there will be overbuilding at some point. It is just a question of how dramatically overbuilt we get and how much pain the market may have to feel down the road as we recalibrate. But based on the growth of these companies so far, it is still very early innings. Just look at the Mag Seven earnings growth in Q1: 56%. You could certainly look up the private companies and see how fast they are growing. Clearly the demand is there. As the saying goes in Field of Dreams, if we build it, they will come. In the late nineties, they built it and then it took decades for them to come. This is a much different scenario, where they have already arrived. There are fans watching the game, and then they are building the field. That is how I would think about this. This demand is there. That does not mean it will always be there for these dollars that are being spent, but certainly for now it is.

Thomas Shipp (19:19):

Another important point is that bottlenecks are slowing down the demand that is there right now. We are not even able to supply the hard assets, the data centers. So some of these bottlenecks are actually slowing down the build-out, and therefore, that may actually elongate the cycle. You also brought this up in one of our strategic and tactical asset allocation committee meetings: the fact that there are these bottlenecks may actually elongate the cycle.

Jeffrey Buchbinder (19:56):

That is right. The dollars are going up partly because of shortages and short-term price inflation. This latest leg up from $680 billion in CapEx this year to around $720 or $730 billion is largely driven by higher memory prices and is not quite as dramatic as it might seem, which is important to keep in mind.

Thomas Shipp (20:31):

Yep. Good point.

Jeffrey Buchbinder (20:33):

All right. So who is going to benefit from this? What you have called, I think in our notes, an economic surplus from AI: will that flow downstream? Should we be playing the picks and shovels, the folks building this, or is the best opportunity ultimately going to be the folks adopting this and generating margins from productivity gains?

Thomas Shipp (21:02):

Right now, obviously the picks and shovels are where all the benefit is accruing in terms of earnings and profits, as you alluded to, with the memory providers, with the high-compute GPUs, et cetera, all the chips, that is very much in the picks and shovels, as well as some of the more industrial tech hardware type industries as well. Ultimately, that is where we move from the capital cycle piece: the picks and shovels into a transformative technology that increases productivity and output from the economy writ large. The way I think about this, and this is a longer-tailed event, is that when we have these transformative new technologies, whether it is railroads, electricity, or the internet, the economic value typically tends to accrue to the users, not the builders.

Thomas Shipp (22:08):

How many railroads went bust? But everybody benefited from being able to cross the country by train rather than on a horse and buckboard. I think ultimately that is what will happen, but right now we are very early. If you look at the actual monetization of AI, those numbers vary from as low as $25 billion to $70 billion annually in pure AI monetization. That is obviously a fraction of what is being spent on the hard assets of the data centers and the chips. We are very early, and right now that value is accruing to the picks and shovels. Long term, I believe it will accrue to users, whether it is knowledge work, healthcare, or whatever it may be.

Jeffrey Buchbinder (23:08):

That is where you are talking about: well beyond 2027. At this point, maybe we have visibility into 2027, but beyond that it is really hard to say. Based on what I am hearing from tech companies and people who follow this more closely, it is going to take more time before the market really becomes skeptical about the payoffs. It will happen at some point. The hype always goes a little too far, and the actual reality does not quite measure up. We will probably get there, but I would argue it is going to be much more gradual than it was in the late nineties and early two thousands, just because of the amount of capital that found its way and was essentially lit on fire.

Jeffrey Buchbinder (24:10):

We are not going to have quite the extreme cycle this time, I do not think. And that is the consensus view, not telling you anything you probably did not already think yourselves. So let us go to the last point here, Tom, before we wrap, and that is concentration. How should investors think about that? Because of course, we have a very small number of companies building out all of these capabilities and a lot of people are saying that the stock market rally is not really sitting on the strongest foundation because it has been so narrowly led. So how do you think about those things?

Thomas Shipp (24:52):

We have been talking about this for a few years now. As we mentioned, there are opportunities that arise. Whether you are playing within the Mag Seven or we are starting to see newer names participate, whether it be across semiconductors, whether it be in that more industrial and tech hardware space. The benefits are accruing beyond that hyper-concentration, but the theme is still pretty concentrated. Especially with so many people passively investing in indexes, whether it is the S&P 500, whether it is the Nasdaq 100, you want to have a finger on the pulse of what you own.

Thomas Shipp (25:44):

Over the long term, index investing has shown that it is generally, after costs, the way to go. For more tactical investors looking to make thematic ETF investments or individual securities, there are going to be those opportunities. That concentration is a risk, so just know what you own. This applies even to individual country indexes. I was looking at the South Korea index, which I believe is 50% in just two stocks and no surprise, they are semiconductor companies. As we noted, concentration can amplify risk. If the Mag Seven goes through a drawdown, it is going to bring down the entire index. That is something for people to keep an eye on.

Jeffrey Buchbinder (26:53):

This concentration issue, at least in the S&P 500, for example, has been a concern for the past couple of years. You could trace it back to 2021, coming out of COVID. The chip makers are probably going to be a bit more cyclical and have more ups and downs, but companies that adopt AI and use it profitably are not going to experience the dramatic ups and downs. There will be a broadening out, hopefully a gradual rotation, as we have seen over the last seven or eight months where the Mag Seven works and then it does not, and the rest of the market picks up the slack, and vice versa.

Jeffrey Buchbinder (27:47):

But the rest of the market will benefit from AI adoption. There is no doubt in my mind. It is just going to take time to sort out who is going to win and by how much and what the eventual earnings benefit is from that AI adoption. Listening to companies on conference calls, almost all of them are mentioning AI and saying they are going to use it to their benefit in some way, but tangible examples of benefit are a little harder to parse out. That is something we will increasingly focus on over the next several quarters. I am sure we will have more examples of the power of AI, especially in 2027. When we wrote our 2026 outlook late last year, I had 2027 in mind as the year of AI adoption. That still looks like a reasonable timetable. I think it is going to ramp up significantly.

Thomas Shipp (28:52):

There is an interesting McKinsey report that speaks exactly to this, likely talking their own book a little, since they are selling consulting services on how to rework an organization for the AI era. But it called out that entire workflows may have to change. The way we think about so much of the work that is done has to change. And that is just going to take time.

Jeffrey Buchbinder (29:23):

We will continue to track this with you. Tom, I look forward to future discussions with you on this topic because you, as head of equity research at LPL, are listening to companies probably as much as any of us. Thanks for bringing your insights from corporate America. With that, we will wrap up. Thanks everyone for listening to LPL Market Signals. We will be back next week. Take care.