Andy D’Souza: Welcome back to The Investment Conversation. I'm Andy D'Souza, Chief Marketing Officer here at Lord Abbett. Today we're talking about artificial intelligence [AI]. The scale of the investment here has been extraordinary and [there are] a lot of questions now about whether demand can keep pace and, more importantly, are we seeing returns on those investments? I'm joined today by Matt DeCicco, Lord Abbett Partner and Director of Equities. Matt, how are you doing?
Matthew DeCicco: I'm great. Thanks for having me.
D’Souza: Thanks for being on, Matt. All right, let's dig into these topics and more, Matt. So, let's kick it off with maybe the first question here. We mentioned there's a lot of investment into this space. Last time you and I spoke on this podcast you had mentioned (I think at the time it was about), the five hyper-scalers were looking to invest about $550 billion in [capital expenditure] Capex for the year. Where do we stand today on that number and is it still going forward?
DeCicco: Yeah, the number's higher now. If we look out to next year, these companies are probably going to spend closer to $1 trillion in Capex.
D’Souza: So, that sounds like a big number by any stretch of the imagination, no matter what you put into play here. It's kind of reminiscent of, in some ways, to some people I should say, what happened with the internet boom back in the late-90s, early 2000. And people would say, "Well, you know what? Those kind of numbers you're throwing around, this is a bubble." Do you think we're in a bubble here?
DeCicco: Great question and a question we hear a lot and we think about a lot. And I want to break this down into really two components. The first is let's just look at the equity markets and are they pricing like it is a bubble? And then the second is I want to come back to the actual dollars that are being spent and if that spend is sustainable. In other words, is there a bubble in that actual spend?
D’Souza: Got it. So, valuation first you're saying.
DeCicco: Exactly. And please interrupt me if I go too long. But the first part is just on the valuation. I don't think that the equity markets are pricing a bubble. I think what the equity markets are responding to is incredible earnings growth. If you look at the earnings growth for the S&P 500® [Index] this year it will be north of [estimated] 25%. And if you look out to 2027 earnings, those 2027 earnings will be over $400, most likely. That means that the S&P 500 [Index] is trading at a valuation of something like 19 times.
I'm not going to say that that is inexpensive compared to historical valuations. Historical valuations are probably 16 times. So, that's 20% rich. However, if you go back to the internet boom, the S&P 500 [Index] traded at something like 26 times. Actually over 26 times. So, if you applied that valuation to the [estimated] $400 in earnings, that would imply that the S&P should be trading at something like 10,500 [index level]. So, that would be, in my view, the equivalent of what a bubble would be today. So, if we were sitting here today talking and the S&P 500 [Index] was at 10,500, I would say, "Yeah, we're pricing in probably something similar to what was being priced in during the last bubble, the internet in early 2000."
I also think if you ground that in comparing it to interest rates, it also gives you some additional context. So, back in, again, early 2000 the S&P 500 [Index] had an earnings yield of [approximately] 3.5% and the 10-year [U.S. Treasury Bond] yield was closer to [approximately] 6.5%. So, 300 basis points less on your earnings yield versus the 10-year yield. I'm sure you know where the 10-year yield is. Today it's right around [approximately] 4.7% and the earnings yield on the S&P 500 [Index] is higher than that. It's more like [approximately] 5.25%, based on that $400 earnings number I've been giving you. So, whether you look at it just on a valuation in a stand-alone context or you bring in where interest rates are, I think it's very hard to argue today that the market is anything more than a little bit rich, compared to historical measures, not at a bubble-like valuation.
D’Souza: And what about the demand side of things though? There seems to be such insatiable appetite for all things [artificial intelligence] AI in the marketplace. We seem to be a little bit early on in the whole life cycle of this thing. How are companies and how are you seeing demand and how are you trying to gauge whether or not it's durable?
DeCicco: Yeah. I want to touch on that and remind me I want to come back to the spending that the hyper-scalers are actually making, the Capex number that I promised I'd talk to you about. So, I think looking at demand is critical and getting a sense of how demand is tracking. My view is that demand for intelligence is nearly insatiable and is running well ahead of what the supply is. There are a few reasons I come to that conclusion, and there are a few things that we can track. The first is just the revenue ramp of OpenAI and Anthropic.
Those companies right now, based upon public data, are annualizing revenue at [approximately] $105 billion per year, of which Anthropic is roughly, let's just say, [approximately] $65 billion of that and OpenAI is [approximately] $40 billion. So, this is people like you and me using Claude or ChatGPT but also of course businesses. If you put that into some context and compare that to publicly traded software companies that I'm familiar with-- so a company called Datadog does about $6 billion in annual revenue. A company called ServiceNow does [approximately] $20 billion in annual revenue. Those companies have disclosed (in the case of Datadog), that they have over 2,000 customers that spend with them, excuse me, over 4,000 customers that spend with them $100,000 per year. ServiceNow says they have over 2,000 customers that spend with them a million dollars per year.
So, if you reason through this with me I think it's probable that Anthropic, which is mostly spending by enterprises, has thousands of companies that spend $100,000 with them. And I think you have to ask yourself, "These are the biggest, the best, the brightest organizations in the world and thousands of them are choosing to spend hundreds of thousands of dollars with Anthropic, perhaps hundreds of them are spending a million dollars per year with Anthropic."
And I don't think these companies would be doing this if they weren't seeing a real return on their investment. The other thing that we can look at is this company Ramp does a survey of about 50,000 companies to measure their AI spending.1 And the median company, this is the median company in that survey, spends $12 per month per employee, so call it $150 per year, and that's been steadily increasing. But I think this also gives you an idea of how much headroom there is for the median company to increase their spending per year.
So, these are just a couple examples, quantitative, of what we can look at to see the demand for these AI tools and why so many people in so many businesses are using them. You have to believe that to spend this amount of money that these businesses would be using them because they're seeing a real return.
And then I'd say the last thing, which is much more qualitative, is just listen to the earnings calls, just listening to the companies and what they're saying. And there are examples really from across the economy that you hear from what companies are spending on where they're seeing a return. Of course it's things like software development, but it's also things like customer service, where companies are reporting that 60% to 80% of their inbound calls are being handled autonomously. You're hearing great early reports on accelerating drug discovery, which is collapsing the amount of time it takes to identify a target candidate from years down to months. And then you're hearing examples across the legal field, of course across financial services, even in places like energy exploration. So, I think the demand signals are critical, and we are going to look at both quantitative and qualitative measures just to see how this is tracking. But our base right now, our view right now, is that demand far outstrips the supply, which is why the hyper-scalers are spending so much on Capex because they are the ones who are supplying intelligence and they see that demand is running in excess of supply.
D’Souza: It's a pretty powerful story and the way you describe it, it sounds logical, it's exciting. But I guess the cynic in me a little bit says, "Okay, but what could throw this whole thing off?" Is part of it, number one, would it be ironic if a new technology came out to actually disrupt itself in terms of storage or speed, capacity, all those things we talk about? Right now, I think it feels like we're spending a lot of money on building a lot of storage-capacity computers, all those things. Could something change that end of the equation, that part of it? Or also the demand side, could you find something where maybe all these companies that have initially invested are not seeing the actual [return on investment] ROI they want to see or it's not coming to fruition and they're saying, "Yes, it's great that I have this now in my company. I maybe have seen some early productivity gains. But I'm not seeing much beyond that, and I've hit these diminishing returns"? So, are either of those things of concern to you at all?
DeCicco: Yeah. I think you've hit exactly the right things to focus on. And to oversimplify it, there's risks on the supply side and there's risks on the demand side. So, let me start with the risks on the demand side. The reason why the hyper-scalers are spending as much as they are is because they are seeing an almost immediate uptick in the compute that they bring online. So, that dollar figure we talked about earlier, the hyper-scalers have spent, let's just call it, $650 billion over 2023, excuse me 2024 and 2025. So, those two years together they spent $650 billion. If you look at the revenue of those companies in the first half of 2026 and compare it to the revenue of those companies in 2024, it has increased by [approximately] $250 billion. So, they spent $650 billion and have seen an increase in their annualized revenue of [approximately] $250 billion. And then they tell you that the operating margin, the operating profit on that is about 40%. The incremental margins are higher because their margins are increasing. So, they're spending money to build these intelligence factories, these data centers, and immediately they're seeing revenue increase from that spend. And in fact, you can even look at the rate of growth that these companies are putting up.
Back at the end of 2023, for example, Google's cloud business was growing 25%. Last quarter it was growing 80%. So, huge acceleration and (we call them hyper-scalers), on a massive scale. Google's Cloud is the smallest of these three businesses, compared to Microsoft's Azure and Amazon AWS, and it's at $100 billion per year. And it's the smallest of the three. So, I don't think that supply is at risk because as these companies are bringing on supply, they are able to immediately see demand for it, which is encouraging them to then create more supply. The other thing is they're all telling us that those growth rates that I cited, they'd be even higher if they had done more, if they had built more.
So, I think the risks to supply are more about things like a pause from pushback, regulatory, but the return for companies themselves that are doing the building is, again, nearly immediate and it's at very high levels. The demand side, I think you raise a very good risk, which is that this technology's brand new. There is a lot that is changing. And it's very hard to see around corners of what new breakthrough might occur that could change the demand from businesses or customers. Just like back in the internet, I don't think people really imagined that Netflix would exist the way it does today or that you'd be able to use your phone to call an Uber whenever you want. I think there is an unknown nature to how this technology evolves from here that could create some sort of change in the demand inflection. Those examples I gave you though are things that happened that changed the demand inflection to be greater than what was expected rather than something that changes the demand to be weaker than what it is today.
But again, going back to what we were talking about earlier, that's why we're so focused on these quantitative and qualitative ways to measure what the demand signals are because right now the demand is extremely high. It continues to increase, and the supply continues to be short of demand for intelligence.
D’Souza: Gotcha. And I guess, speaking of demand and the shift of that, the examples you gave were positive examples of Netflix and Uber and those things, but just like back in the day, to your point, when the internet first came on the scene there were a lot of different, for example, search engines that came on. And I remember all the ones we used to use back in the day and some of them are no longer here, and some weren't invented till much later and became the dominant search functions out there. So, back to the point of the last podcast we did, there'll be dispersion. There'll be winners and losers here.
DeCicco: Yes.
D’Souza: Are you seeing any signs, and not just with the hyper-scalers but as this broadens out and touches, to your point, all the other companies because it touches everything and everybody either personally and/or the companies in the world, are you starting to see any kind of a dispersion or shake-out and who's actually doing this well and who's not?
DeCicco: Yes, absolutely. The level of dispersion in the market right now is unusually high. Another way of saying that is intra-stock correlation is unusually low. And generally, in the bull market most stocks are dragged higher. We actually, over the last few years, have had a bull market where there's been more negative-returning stocks than you typically see. In every industry I'd say you have companies that are out-performing and companies that are meaningfully under-performing. And in fact, I'd say there are very few industries where you could say, "I should just be void the industry." It's much about, "Within this industry, there are stocks that are doing quite well, and in this industry, there are stocks that are under-performing."
It's very hard to trace it back to AI right now, but I think you just have to use some judgment and say, “Okay, this company, on their earnings conference call, they might be talking about new tools that they've put into place and they're seeing better customer retention.” They're seeing better leading indicators of demand. They're talking about how they're able to onboard customers faster through AI tools. They're talking about how they can launch products faster with AI tools. And then other companies are not talking about those things, and you can draw some conclusions that dispersion is happening because this company has made some good choices about how they've deployed AI in their organization and another company has made poor choices or hasn't deployed AI as effectively in their organization.
I wouldn't say that is the only cause for the dispersion that we're seeing in the market, but I think it's reasonable to assert that it is a cause of the dispersion that we're seeing in the market. I think the other thing too I want to come back to, the [return on investment] ROI question. Just think about where we started the conversation, where I said, "The growth rate in earnings right now is as high as it has been." It's 25% for the S&P 500. And also, earnings have been revised upward this entire year. So, earnings are growing fast but they're also coming in much faster than what consensus expected. The earnings growth is also much broader than it was a couple years ago. So, the earnings growth is in small companies, it's in mid-sized companies, it's in the U.S., it's outside the U.S., it's in developed markets, in emerging markets.
Again, let's reason through this. What's changed in the last three years? This brand-new technology that companies are using more and more. And I think if you wait for, "Oh, there's the ROI. I see it in the economic data now," by the time you wait for that you're going to miss the investment opportunity. So, I think you can look at these different datapoints and identify that there is an ROI that's occurring. It is not being distributed equally throughout the market. There are companies that are taking advantage of it. There are companies that aren't. On balance, though, it's lifting the entire economy. On balance, it's lifting the markets. But there is incredible opportunity for security selection right now because of this dispersion that we're seeing.
D’Souza: Gotcha. And then I guess one other way to look at this would be you have the hyper-scalers and they're providing all the infrastructure, effectively, for this all to happen. And then you're having the actual consumers, the customers, the companies and individuals actually using these tools and taking advantage of that infrastructure. And at first thought I'm like, "Oh, there's five hyper-scalers now. It seems like they have dominant positions and they're incumbent." But just like we learned from the internet before, there could be somebody we're not even thinking about yet that may come on the scene or these might have a hiccup or something. So, it's not like they're there forever.
DeCicco: Yeah, I think there are a couple things to draw from that comment. The first is if you just think about the internet or even back to the integrated circuit or the beginning of the semiconductor industry or you could go back to electricity, those technologies were foundational and they supported a growth in the economy and investment opportunities that lasted decades. But we all know markets do very well over those long periods of time. If you go back to the integrated circuit, semiconductors, 1960s, the market's done very well since then. You go back to the internet, late '90s, the market has done very well since then. But there are of course periods where expectations get ahead of where the reality is. There are companies that were believed to be leading that fall apart. And if those are the biggest companies in the market, that could be challenging for market returns.
So, you asked before about some things that could cause some risk or what are the things that would cause demand to slow. I think it's things like this. It would be if perceived market leaders were not the eventual winners and there's disruption that's caused by that. That could cause some weakness in the stock market. Or if there's a kind of pause in demand for some reason. Again, I have trouble imagining what that might be. Again, I threw out the idea that it could be maybe something regulatory related. I don't think that changes the end state of where we're ultimately going.
But could that cause some weakness in the market? For sure. So, I think the risks around where the adoption of AI goes will track similarly to the risks around the internet or the semiconductor or electricity, where you have a very long period of growth in the economy and in the markets related to that, but you'll have these periods of cyclicality or gaps in returns that could cause some periods of market under-performance.
D’Souza: Got it. So, kind of wrapping it all together if I can a little bit here, it feels like we touched on a few topics today, but the demand is real and durable and it's across the entire economy, sectors, geographies. It's everywhere. Demand is outpacing supply still. Despite all of the ROI, there's more to be needed in that space. But people are seeing the ROI increase on the large Capex spending that's going on, that it keeps going up and up and up. But overall, it's pretty exciting and I just wanted to thank you for coming on today and talking about this with us, Matt.
DeCicco: It's incredibly exciting. Thank you.
D’Souza: And thank you for listening to The Investment Conversation. Find more episodes wherever you get your podcasts or at lordabbett.com. And follow us on LinkedIn, Instagram, and YouTube for more perspectives from across the firm. Thank you.