Why the Biggest AI Winners May NOT Be Nvidia or the Mag 7

Hinzugefügt: 2026-05-14

[00:00:00]Everyone is chasing the companies building AI, but what if the real winners are the companies using AI?

[00:00:05]Joining me now is Kai Wu, founder and chief investment officer at Sparkline Capital. Kai, great to have you here.

[00:00:11]Thanks for having me. So, let's set the stage here because you're bullish on AI long-term, but you're skeptical on parts of the AI trade today. So, break that down for us. Make the distinction for us. Yeah, look, I I I think if you can break down the current cycle, I'd say there's kind of two things happening at the same time. So, first is the the build-out of the infrastructure layer.

[00:00:31]That's, you know, building out the the chips and the power and the data centers, the hyperscalers, the model developers. We are far underway with that part of the cycle, right? Trillions of dollars are now flowing into that and that's driving value all the way down the chain. The second part, however, I think we're much earlier in and that's the adoption phase, right?

[00:00:50]Most surveys suggest that only 10% of businesses are actually using AI in production. So, while firms and individuals are experimenting with AI, are they actually using it in core parts of the business? Not quite yet. So, I think we're we're still we're still early with that part of the cycle. And so, the big question comes around the big risk, I guess, is around the timing mismatch there, where we're spending all this money building out this infrastructure layer, yet the demand hasn't yet fully materialized.

[00:01:15]I believe it will, you know, in the very long run, but there's some risk that, you know, over the next two, three, five years, uh it just takes a bit too long to materialize. And, you know, GPUs have a depreciation of, you know, five years or so, and so the window is not you know, super it's not forever, right? And if you look at the the history of past cycles, like the dot-com boom, the railroads, all these historical episodes, what you find is a consistent pattern of overinvestment too soon in a technology that has it's not yet mature enough to uh drive the demand, and then the overcapacity leads to um you know, falling prices and then, you know, trouble perhaps even bankruptcies in the folks like the WorldComs and many of the telecom stocks in the dot-com boom struggled.

[00:01:56]Right now, the market is assuming that this spend will amount to produce huge profits. Is that not the case?

[00:02:06]Yes, that's of course that's being priced into into the stocks. And you know, to be fair, if you look at you know, the down the value chain, there's massive like bottlenecks, massive constraint supply constraints. I will say there's there's two questions on my mind that might suggest that this data is a little bit less reliable than one might believe. First is the fact that most of these labs, AI labs, are running the kind of old Uber playbook, which is let's subsidize the price of tokens in order to drive demand to kind of capture the market with this idea, this expectation of a winner-take-all scenario. Right, so the the the price signals coming out of that are not as reliable as if they were pricing it at the correct amount. Right, it's quite possible that they say for example that 95% of ChatGPT users are not paid.

[00:02:49]Right, if it cost us money to use it, maybe we wouldn't be using it as much.

[00:02:53]Right, so that's that's an important component. Um and the second is of course what I was saying before around a lot of businesses are experimenting with AI. There's you know, initiatives at companies saying, "Hey, you actually have to as an as a CEO and as an employee use AI, otherwise, you know, it's going to go into your comp." Um but that's not that's kind of an artificial inducement and isn't necessarily something that's is sustainable over the next say 5-10 years.

[00:03:15]>> Okay, and this of course is an ETF spotlight. We'll get into your funds that help investors invest in the AI adopters versus the AI builders. But first, if I'm an investor listening here and I think, "Okay, he's sounds more bearish or more skeptical of the AI builders and more bullish on the early adopters, does that mean I need to sell the Mag 7 in my portfolio?" I mean, look, I think it's a pretty big risk.

[00:03:39]Obviously, over the past 10 years, we've seen you know, uh you know, huge bull run in the the big tech companies, the Magnificent Seven. They've become a third of the index of the S&P 500 index, let's say. And now if you add to that the other infrastructure companies like the Broadcoms of the world, the Oracles, you get to near 50% of the index is now in this one trade, right? Which has as I mentioned the two risks of, you know, valuations having increased due to the success of these companies combined with the CapEx risks that they're spending so much money. If things go south, they're on the hook for the losses. So I do think that investors who are just kind of, you know, investing passively in these ETFs, which is what we're all told to do and that's the most responsible thing, are a little bit less diversified than one might believe at at at face value. So if traditional indexes are still very much positioned for the AI build-out phase, what does that mean for the everyday investor who is just invested in the S&P 500?

[00:04:34]I think they are already, you know, exposed in a, you know, sufficient way towards being long AI. So this is just about adding early adopters to your portfolio.

[00:04:45]>> Yeah, so so what our ETFs are trying to do now is to help counter position against that, right? So Our Our review is that if you look at again the history of of these these technology paradigm shifts, in almost no cases have the builders actually reaped the rewards of their inventions.

[00:05:00]It's kind of a bit ironic that, you know, all these railroad companies sprouted up to to build the transcontinental railroad and most of them went bankrupt. Instead, who benefited was the the users, the the person who wants to go visit their relative in California, the company that wants to ship their goods across country. Same thing in the dot-com boom.

[00:05:17]We saw, you know, Netflix and Meta, um, Google came in after the dot-com collapse and were able to benefit from subsidized bandwidth, right? So there's just, you know, kind of disconnect there. And I think what's what's happened is that so much capital and so much investor interest has gone into the infrastructure layer of the stack that people are kind of forgetting that this this lesson of history that it's actually the adopters that historically been the long-term winners of the boom.

[00:05:41]What's interesting is not only do they have less CapEx risk, right? But they also have they are trading at much lower valuations. So, to the extent you you see risk in, you know, buying over elite stocks, which of course go back to the dot com boom, that was the reason why you you lost money buying at the peak of the dot com boom. Not that the internet didn't work, internet did it work and it did change the world. The problem was that you bought at inflated multiples and it took you, you know, decades to kind of dig your way out of the hole once the valuations compressed. So, that that's kind of the thesis around the early adopters and the ETFs are of course designed to try to give investors exposure to um those those themes.

[00:06:15]>> Okay, so the ETFs are I 10 and D 10, which is the international play. Give us a few examples of companies that are already seeing measurable benefits from early AI adoption that could be in your portfolio or in your funds. Yeah, it's So, it's a it's a mix of things. I'd say broadly speaking, there's kind of two categories of companies. There's kind of the more under the radar old economy plays like industrials, financials, even even pharmaceutical biotechs that could theoretically be using um you know, AI for R&D in their drug development. And [music] they're just trading at, you know, multiples that are not pricing in the potential upside from AI actually, you know, becoming a factor. Cuz again, if AI becomes what we all believe it to be, then almost by definition, you have to see an uplift in margins for the folks in for the users who are actually leveraging AI to gain advantages over their competitors. And one interesting thing we see is that pick any given industry, whether it's financials or even even software, there's a huge spread between the companies who we consider early adopters who are kind of really um open-minded and and pursuing AI as opposed to the long tail folks who are just kind of like doing nothing about that. And the market's again not pricing any difference in those two types of companies. So, obviously to the extent that AI were to be successful, that should drive, you know, pretty pretty significant separation. And historically, we've done, you know, back tests and look at all this data, that's what we see in the past.

[00:07:39]>> What actually qualifies a company as an early AI adopter. So, it comes down to a variety of metrics we use. It's all systematic, right? Because you want to try to be as objective as possible in assessing these companies. So, obviously we'll look at what the companies are saying in their earnings calls. That's a bit more subjective. We'll look at the patents, the job postings, the LinkedIn bios of the employees. You know, one of the nice ground truth things is which what where are companies actually investing their their, you know, money from a human capital standpoint. If you're spending all this extra money on AI uh trained engineers, you probably actually mean it when you say you're investing in AI. So, we look at, you know, metrics across the the board again from ground truth to um you know, third-party data to company um reports in order to try to isolate which companies are, you know, tilting towards AI versus not.

[00:08:27]What are some of your of your biggest holdings? And what are some holdings that would actually surprise us being in there? Uh you know, so there's the second category of names are the folks that are kind of bombed out, right? The folks that people actually expect to be losers in AI. Like what? We own like Accenture, for example. Accenture's stock is down big. We own Salesforce.

[00:08:44]Salesforce is down big. These stocks are kind of perceived as being losers in the AI era. And you and you and you what's really interesting is if you go back to historical episodes, the market has a really bad a really hard time actually identifying who ultimately wins. And in fact, it's quite common for them to first punish stocks thinking they're losers and then for those stocks ultimately recover and then thrive, right? Like Walmart or New York Times.

[00:09:07]You know, they all survived the the media and retail wipeouts. Um in the case of Accenture, I think it's an interesting story, right? Like, you know, folks like Dario, the the the leaders in the luminaries in AI, they they kind of like are assuming, yeah, we just get diffusion because like that's just what happens. But in reality, it takes, you know, a lot a lot of energy and a lot of organizational change to to change the battleship's direction, right? We actually saw this with OpenAI a couple days ago launching this new joint venture with a bunch of private equity companies and consulting firms with the idea they they understand, starting to understand now, actually, like, organizational change is a big deal. We don't have enough word employed engineers. We need to get if we actually want to make AI integral to enterprises, we need to start pushing this through.

[00:09:48]The case of Salesforce is another similar case. People are selling it all down assuming that because of AI coding tools, you know, the cost of the software is now zero, so therefore they have no moat. What I would ask is was was a code ever ever Salesforce's moat?

[00:10:03]I don't think it ever was. Even before AI, there were plenty of Silicon Valley startups that had, you know, flashier products and kind of more slicker user interfaces. You know, in my mind, Salesforce and companies like that, their moat is more around the network effects, the human capital, the brand, right? The switching costs of trying to switch your CRM away from, you know, Salesforce to a competitor or a live-coded alternative is not really like a a realistic thing. So, I think these companies have been unfairly punished.

[00:10:30]>> So, if we use history as a guide, what does that mean for the Mag 7 as we think 5 years, 10 years down the road? Look, I don't think they're going to go out of business. I mean, they have amazingly profitable like franchises in in social and search and and so and so forth. I do think that one area of concern is the fact that they are becoming more and more capital intensive over time. So, if you go back and ask the question of why were they so successful, the answer is that they were asset-light businesses, that they managed to be able to, you know, create huge returns on invested capital without without requiring too much money in, right? And Google famously only raised, you know, small amount of money before being able to just become this asset-light compounder.

[00:11:06]What's happening is now they're transitioning away from that business model because of the AI race and spending more and more money on capex.

[00:11:12]So, Meta, for example, is spending, I think, a third to a half, and Microsoft, too, of their sales is being now spent on capex, on building out AI data centers, right? And what we know is that, you know, two things. So, first is that when you see capital booms, these these companies that are investing heavily in building out physical infrastructure, that those companies have tended to underperform historically. Second, we know that asset-heavy businesses, right, utilities being a perfect example, have tended to be less attractive businesses to be in.

[00:11:40]So, think of it as this way, which is these these amazing asset-light businesses are now becoming asset-heavy, they're becoming more like utilities.

[00:11:47]Does that mean they're going to go out of business? No, it just means that they're diluting, let's say the profitability of their uh businesses.

[00:11:52]And so, then the risk just becomes, from the standpoint of investors, multiples and expectations being kind of predicated on a past era where they were, you know, once these amazing businesses, that that needs to eventually correct. And so, there could be some short-term pain there, but yeah, I mean, I just don't think that being a data center uh operator is like a like that attractive of a business to be in.

[00:12:12]So, just quickly, give me five names that you think five years from now will be considered AI winners.

[00:12:19]Well, obviously my ETFs, so that's two of them. No.

[00:12:22]Outside of your ETFs and outside of Accenture and Salesforce. Yeah, so, look, I I let's let's speak in categories, okay? I think I think there I think the the categories of firms that will likely be successful are, you know, the self-service software stocks that have complementary assets, like the Salesforces of the world, the beaten-down stocks that, you know, are unfairly being being thrown out with the bathwater. I think there are parts of the infrastructure uh supply chain that are actually positioned okay, and I think that the old economy winners is kind of the final category, um are, you know, a really interesting play. You're you're basically getting a free option where you can, if AI works, these guys who are invested in AI and positioned for AI will benefit, yet you're not paying anything in terms of multiples or CapEx in order to get these guys. So, an example of an old economy?

[00:13:07]I don't know, like Capital One. We own Capital One. They actually are one of the firms with the most AI patents, um so, believe it or not. Do you think the biggest AI winners, say five years down the line, could be companies that investors don't even think of as AI companies right now, then? Absolutely, and companies that don't even exist. Like I remember, like Google wasn't a public company. Um you know, I don't know Facebook wasn't even founded until what 2005 2004. Okay. All right, I think it's a good time for ETF spotlight. We don't typically do our rapid fire game of this or that, but our audience wants more of this or that, so I thought this would be a good one to do. So, this is quick questions, very quick answers. Are you ready to play with us? All right, let's play.

[00:13:47]>> AI bubble or AI revolution?

[00:13:49]Long term revolution, short term bubble.

[00:13:52]Nvidia or the next wave of AI winners?

[00:13:55]Next wave. AI builders or AI users?

[00:13:57]>> Users. Own the mag seven or look beyond it?

[00:14:00]>> Beyond.

[00:14:01]Better AI opportunity today, semis or software? Software. More attractive AI sector right now, healthcare or industrials?

[00:14:10]Healthcare. Bigger portfolio risk, too much AI exposure or not enough?

[00:14:15]Too much in the US.

[00:14:16]More crowded AI trade right now, chips or hyperscalers? Chips. Better investment today, infrastructure or applications? Applications. US or international AI winners?

[00:14:28]International. One name that we should own internationally.

[00:14:35]I think the category of firms that are your large cap international biotechs would be our interesting place to invest. Bigger opportunity from AI, productivity gains or revenue growth?

[00:14:48]Productivity gains.

[00:14:50]Better risk reward today, the AI winners or AI laggards?

[00:14:55]AI laggards. More vulnerable right now, expensive AI stocks or non-AI stocks?

[00:15:02]Non-AI stocks. Most Most vulnerable AI stock right now?

[00:15:09]Oh, um probably Micron.

[00:15:13]More important for investors right now, AI spending or AI profits?

[00:15:19]AI spending.

[00:15:20]Bigger mistake today, chasing AI momentum or avoiding AI completely?

[00:15:25]Avoiding AI completely. All right. We'll leave it there. Kai, thank you so much.

[00:15:28]Really appreciate your insights and for shedding some light on ITAN and DTAN for this ETF spotlight. That's Kai Wu, founder and chief investment officer at Sparkline Capital. Thank you. If you enjoyed this ETF spotlight, check out our full interview with Rich Pzena. He breaks down why value stocks could start to outperform.