Lead-Lag Live

Hype vs Reality: Kai Wu on AI CapEx Risks, Market Concentration, and the Next Phase of AI Investing

• Michael A. Gayed, CFA

Share episode

In this episode of Lead-Lag Live, I sit down with Kai Wu, Founder and CIO of Sparkline Capital, to break down how the AI investment cycle is evolving as scrutiny rises and market enthusiasm becomes more selective.

From hyperscaler spending and data center economics to market concentration and fading euphoria, Wu explains why investors need to rethink where real AI returns may come from and why the next winners may not be the obvious names.

In this episode:
– Why massive AI capex is starting to worry investors
– How market concentration amplifies downside risk
– The productivity debate behind AI adoption
– Why AI adopters may outperform AI infrastructure plays
– How to position portfolios for the next phase of AI investing

Lead-Lag Live brings you inside conversations with the financial thinkers who shape markets. Subscribe for interviews that go deeper than the noise.

#ArtificialIntelligence #AI #StockMarket #MarketRisk #TechStocks

Start your adventure with TableTalk Friday: A D&D Podcast at the link below or wherever you get your podcasts!
Youtube: https://youtube.com/playlist?list=PLgB6B-mAeWlPM9KzGJ2O4cU0-m5lO0lkr&si=W_-jLsiREjyAIgEs
Spotify: https://open.spotify.com/show/75YJ921WGQqUtwxRT71UQB?si=4R6kaAYOTtO2V

Support the show

SPEAKER_00:

Explosion innovation in this space has, if anything, only um accelerated the trend and the value of these sorts of data. So, you know, I think I think to your question, um, what the AI boo has done is actually kind of full further reinforced my thesis with regards to, you know, uh the the belief that uh you know an even greater share of corporate value now lies in these intangible assets, which of course include intellectual property, AI, and other um, you know, AI models and other like linked ideas.

SPEAKER_01:

All right, well, it's the holidays, and I'm officially going to be spoiling one of you. I'm giving away this duffel bag packed with a bunch of our signature Few Crew branded merch that has all the inside jokey slang that you only get if you actually get it. So what's inside? Well, a men's What Up Plitches hoodie, an exquisite hoodie for her, and a few other things to take you from, I think I get it, to Few Crew certified. Now, if you want in, here's the deal. You have to follow at lead leg report on X, follow me, Mela underscore Schaefer on X, subscribe to Lead Leg Media on YouTube, and like and share this video. You do that, and boom, you're entered. No gimmicks, no funnels, and no nonsense. One winner gets the whole package. The rest of you stay f until next year. Happy holidays from the Few Crew. I'm your host, Melanie Schaefer. Welcome to Lead Leg Live. Now, AI investment is still accelerating, but the conversation is starting to shift. Governments are debating rules around AI accountability and model oversight, while economists question how quickly AI actually feeds through to productivity and growth. At the same time, markets continue to reward a narrow group of companies most exposed to AI capex and data advantages. My guest today is Kai Wu, founder and CIO at Sparkline Capital. Kai, it's always great to have you here. Yeah, thanks for having me back. So if we start at sort of the high level with AI spending still rising but scrutiny increasing, how has your view of the AI opportunity evolved over the past year?

SPEAKER_00:

Yeah, well, I think it's been an interesting year. Um, for the first six to nine months, I'd say the market was more or less euphoric about the AI boom. Whenever a company announced a new AI CapEx um plan, their stock would go up. I mean, the the bellwether would of course be Oracle, whose stock went up 30% in a single day on the back of an announcement that they'd be building data centers on behalf of OpenAI. Um since then, the um fate has the fortunes have sort of changed for these names. Um Oracle is stock is down 40-50% since that point in time, um, you know, basically fully retracing um it its gains and its CDS. So the uh credit default swap, um this the credit spread between their debt and the risk-free one has blown out. So investors are starting to develop concern around the sustainability of the spending and um in particular um you know OpenAI as a uh as a credit for Oracle. Um you also saw the uh the recent news this week about um Blue Owl pulling out of a big data center deal um that that um Oracle was trying to um put together, which is you know kind of rattled markets as well. Also, Core Weave, which is you know perhaps one of the few pure play data center cloud um companies, they IPO'd earlier this year to great fanfare. Stock was up significantly, they're down over 50% as well since this point in time. Um you you look at the the hyperscalers, right? Um in the last quarter, they they announced their earnings, and some companies like Google and Amazon did quite well, but others like Meta actually fell a significant amount um after they announced earnings, which I think signals that the market is no longer uniformly um euphoric on AI spending, but it's instead being more discerning around how that spending will actually be uh what will it actually generate in terms of RY for these companies, right? So clearly the market and investors are happy with what Google's spending their money on, while they don't trust Zuckerberg as much with regards to how he plans to spend the you know tens of billions of dollars um on AI data centers.

SPEAKER_01:

Yeah, Kai, so you mentioned Oracle and Meta and Google, but we continue to see a heavy concentration in this small sort of group of AI-linked megacaps. Do you see that concentration as a risk factor overall or a feature of how AI economics uh just works?

SPEAKER_00:

Yeah, I mean, I think this is a unique time because you know, even before the AI boom, right, um, the magnificent seven stocks, um, so that's Meta, Google, Amazon, Nvidia, Tesla, um, Microsoft, Apple, um comprised 30 to 33% of the index. Today they're around 33% of the SP 500 index. So that means as an investor, one third of your money is in these names. And it just so happens to also be the case that these Magnificent Seven stocks are also, in addition to Oracle, the companies that are investing most heavily in the uh AI data center uh boom. Right. So the top four names, um Google, Amazon, Microsoft, and Meta, are collectively spending um, you know, over you know, uh$400 billion in data centers, um, which is obviously quite quite a large um expenditure and with plans to kind of increase that um over over time. So this is a huge bet that they're making. And so, yes, I mean, I think you know, as an investor in even a passive index, you should be concerned about the fact that a third of your money is um in the hands of these big tech CEOs that are basically betting the firm on this um platform shift to AI.

SPEAKER_01:

Yeah, and I mean there's there's another bet because there's debate around whether AI is delivering real productivity gains, at least currently. From the data that you look at, are we seeing measurable economic impact, or is that still sort of ahead of us?

SPEAKER_00:

So I think that's the that's the trillion dollar question. Um so with all the money that's going into building out these data centers, obviously, you know, that's that's fine as a venture bet, right, um, for now, but at some point investors will want to see a return on their capital, a return on investment in these in these things. And and the challenge is that the end users, right, of these of these AI data centers, like for example, Meta and well, Microsoft, for example, will say, yeah, you know, we're we're fully utilized. Um, in other words, when we build out these this cloud, um, we're basically unable to fulfill all the demand that the end users have with regards to um GPUs and AI um inference. Now, of course they're gonna say that for now, but the question is, is that sustainable? Like who ultimately is the end user of the Azure cloud with regards to um you know AI? And you know, obviously the biggest um uh user, the biggest uh you know demand center is OpenAI itself. Um, ChatGPT has been tremendously successful in scaling its business to you know hundreds of millions of users. Um we have anthropic and the other AI model platforms, as well as you know, a few cursor and other companies that could be considered wrappers around that. Now, the problem is that this whole complex of companies has revenues on the order of$20 to$40 billion. Um I think it was uh um one of the consultant uh companies that estimated that in order to generate a sustainable ROI on these on um a reasonable ROI on the data center investments, revenues for kind of pure AI applications would have to reach around um$2 trillion on an annual basis by 20 in five years. Right? So from now to five years from now, um we effectively get 100x the um amount of revenue that we earn um on kind of pure AI applications. Right. And that may be sustainable, but that's that's the the big question, which is will we actually be able to do that? Now we we've seen some, you know, we've seen both kind of good and bad um data on the side. To be honest, it's probably still too early to tell whether or not um this will transpire. Um, you know, ChatGPT, for example, despite not being three years into existence and having you know scaled its free service tremendously, only about five percentage points or so of uh their users are actually monetizing or actually paying for their service. Um and and you know, we also hear mixed re reports around you know, is the enterprise actually adopting and using AI and finding it to be valuable um in their processes? Now, to be fair, um enterprise adoption is slow to begin with, whether AI or or any form of technology. So again, that may be still too soon to tell. Um but I think that'll be the big guess for markets as we you know turn the page to 2026. Um, you know, are the you know Fortune 500 companies, these um uh enterprise customers actually finding value in AI and are they willing to pay for it? If the answer is yes, then perhaps the AI boom is um will continue. And if the answer is no, then um you know we may see pain ahead.

SPEAKER_01:

Yeah, and so I want to pivot just a little bit and talk about uh something that we discussed last time uh you were on the show, and that was intangible assets and how traditional signals can break down. Has the rise in general and generative AI changed how you think about valuing intangibles like data and algorithms, et cetera?

SPEAKER_00:

No, I mean I think you know, if you step back, the the reason why um, you know, I set up my firm and you know all the research I conduct is around this idea that you know traditional value investing, they tend to focus too heavily on tangible and not enough on intangible assets. In other words, the traditional value frameworks do not put enough emphasis on the value of data and AI and other technologies. Now, the explosion in innovation in this space has, if anything, only um accelerated a trend and the value of these sorts of data. So, you know, I think I think to your question, um, what the AI boom has done is actually kind of f further reinforced my thesis with regards to, you know, the the belief that you know a even greater share of corporate value now lies in these intangible assets, which of course include intellectual property, AI, and other um, you know, AI models and other like linked ideas.

SPEAKER_01:

Yeah, and we've talked a little bit about the the risk involved, but uh for investors who are trying to position across the AI adoption curve, where do you think expectate expectations are the most stretched, which you've kind of touched upon, but where might the market still be underestimating change?

SPEAKER_00:

Yeah, I mean, I think this is a really, really interesting question and something I thought a lot about. Um, you know, if you step back, I went back, you know, one of my early research papers went back to the dot-com boom, and I said, if you had the a crystal ball and you wanted to play the dot-com boom perfectly, how would you have done that? Well, what you would have done is you in the kind of mid-90s, you would have invested in the highest beta um names, the kind of infrastructure players of the time. That includes telecoms, which are basically the same as the hyperscalers. They're investing the money into the CapEx to build out fiber optic cables, which are analogous to data centers today. You would have invested in Cisco, which is basically the NVIDIA of the day, right? Selling picks and shovels to the data center builders or the fiber optic um, the telecoms, those stocks would have done the best early on. But then there would have reached a time period where these names would have would have gone up so much that their valuations became stretched, right? You know, Cisco was trading, you know, was one of the biggest stocks in the in the stock market and traded at very elevated multiples by 2000. So at that point in time, again, with a crystal ball, the correct thing to have done would have been to rotate away into less overvalued names. Now, one thing you could have done would have been to sell everything and just go into bury your head into the into non-internet related stocks, which you know may have worked tactically, but obviously would mean that you're kind of giving up on the internet revolution if you're kind of not benefiting from the technological tailwinds. So, what you could have done instead um would have been to invest in internet adopters, right? So what I would these are the companies that stood to benefit from the rise of the internet, yet weren't themselves making huge capital, risky capital expenditures into the fiber optic build-out, um, nor were they trading at excessive dot-com style evaluations, so not the pets.coms of the day. So these kind of you think of these as kind of the early at early adopters within like the old economy or you know, more traditional businesses that are just implementing a uh internet. I think that's kind of where we're starting to reach today in the AI cycle as well, where you know, we've seen a huge amount of um uh returns from NVIDIA, from the Magnificent 7, you know, in the past few years, and these stocks have done some done very well. I think now is the time to that we to start thinking about how you know the next cycle might play out, right? The the adopters of AI, which as I mentioned, it was stir early early. There have been a few companies that have reported earnings, like CH Robinson is a good example, um, that have you know done done really well, their stocks have done really well on the back of you know um alleged AI improvements in their in their businesses. I think what we'll likely see over the next year, and again, this is the big test, but I do expect it to be the case, that we'll see more and more companies um you know announcing and talking about how AI is driving out performance. And these companies, while they may be in kind of stodgier industries, they will be able to then use that improvement to outcompete their even stodgier rivals. Um so I think these companies trade at you know basically no premium to the market. So you're getting kind of free AI exposure, so to speak. And they exist not just in technology and communications, they exist in all sectors of the US, as well as internationally, right? In Germany and France and Japan, uh countries where you know maybe their valuations aren't as high. So you get more diversification, you get less capital um intensity, um, you get less um you know overvaluation while still maintaining positive exposure to AI. And so you mentioned you asked the question about what my signals are telling me. Um, that's what the signals are telling me. So I'm not just like telling you this is, I think, where we are. I mean, that's bottoms up, um, just looking at what the models are doing, is what the mod what the models are basically having us rotate out of the Magnificent 7. So at one point we were overweight Magnificent 7. We were no longer overweight. We our top position is Google, um, but only have a couple names in that category. Most of the weight's now starting to shift away into the kind of adopters away from the infrastructure plays. Um, and so that's kind of where we think that one should consider positioning um, you know, for the next year.

SPEAKER_01:

And so you mentioned that you don't have a crystal ball, but for uh investors and advisors who want to learn more about all the research that you do and uh see more of your thoughts on AI, where should they go? Where is the best place to contact you?

SPEAKER_00:

Yeah, um if you want to look at the research I write, you can just go to sparklandcapital.com. Um, and I have all my my publications posted there. There are probably 24 or so more than that now on papers. Um, and then in terms of contacting me, you can email me. My email is you know on the website, or you can contact me either the form, or I'm also on Twitter and LinkedIn with a handle CKI W U.

SPEAKER_01:

Also, Wilkai, thanks for joining me again, and thanks to everyone for watching. Be sure to like, share, and subscribe for more episodes of Lead Leg Live. Always a pleasure, thank you.

Head Shot

Michael A. Gayed, CFA

Host