What If AI Growth Is Just Debt In Disguise

A $500 billion AI funding headline can sound like destiny, but we’re more interested in the mechanics and the weak links. Steve Davenport and Clem Miller trace the AI investment chain from money-losing model builders to the very real world of data centers, semiconductors, and the energy required to keep it all running. Then we ask the uncomfortable question that almost nobody wants to say out loud: what happens when a major backer finally says no?

We dig into why calling AI “infrastructure” is both persuasive and misleading. Data centers can be long-lasting capital assets, but chips are on a short replacement cycle, more like maintenance than a bridge. That difference matters when you’re trying to value the AI capex cycle, understand who benefits when spending accelerates, and spot the risks when it slows. We also talk about debt, private credit, and how higher interest rates can turn an optimistic buildout into a fragile refinancing story.

From there, we move into market signals. Forward earnings and forward PEG ratios can look precise while the assumptions get stretched by hype, so we explore a more grounded cross-check: short interest and what it may reveal about informed skepticism. If you’re investing in AI stocks, semiconductor stocks, or data center infrastructure, this conversation is built to help you think clearly about funding, cash flow, valuation, and what “growth” really costs. 
Subscribe for more skeptical market breakdowns, share this with a friend who’s all-in on AI, and leave a review with your take: slow deflation or a sharper unwind?

Straight Talk for All - Nonsense for None

Please check out our other podcasts:

https://skepticsguidetoinvesting.buzzsprout.com

Disclaimer - These podcasts are not intended as investment advice. Individuals please consult your own investment, tax and legal advisors. They provide these insights for educational purposes only.

Transcript

Steve Davenport 0:28

Hello, everyone. Welcome. This is Steve Davenport from Skeptics Guide, and I'm here with Clem Miller. And today we're going to take a little bit of a hard look at what's happening with the AI trade, what's going on with the capital needs, and how does this future really lay out as a continuous operation versus something where there will be a stopping

Welcome And The First No

Steve Davenport 0:52

go and a refunding. And someday, someday, someone might actually say no. Um, Clem, is anybody ever going to say no to an investment in AI? Or is it always going to be a yes?

Clem Miller 1:08

Oh, there will there will definitely be no's, no question about it.

Steve Davenport 1:13

We haven't seen any yet, though, right? I mean, everybody's patting each other on the back. I mean, what do you think about today's situation?

Clem Miller 1:20

No, I think I think what's gonna happen, what we're seeing right now is we're seeing a situation where the AI companies, that is the model builders, okay, right, the software model builders, are operating at huge losses that need to be financed.

Losses Fund The AI Supply Chain

Clem Miller 1:44

Those companies, once they get the financing, are dependent on data centers, and those data centers are filled with various kinds of equipment, , namely semiconductors, but other types of equipment too, and in turn are also dependent on energy resources to keep those data centers nice and cool so that the chips all don't burn out. So, what's happening is you've got a very profitable, very fast-growing semiconductor and hardware sector, , technology hardware sector, that is supplying stuff, stuff, physical stuff, to these data centers for companies building models that are making huge losses and need to be funded. So essentially, if you run into, and this is what you're alluding to, Steve, if you have a situation where you know, somebody financing OpenAI or Anthropic or somebody, those are the model builders, if they don't get their money, if they don't, if they somebody says no to them on financing, that's going to impact their ability to build models, to build data centers, to buy chips, and so on. And it's gonna spread, , that problem is going to spread into the rest of the technology sector, into parts of the energy sector. Uh it's going to be widespread. So I think the real question, you know, it's not going to go on. This boom is not going to go on forever. Right? It's just not. It's not going to go on forever. And the question is, yeah, I mean, two questions. When will it end? And I don't know. I don't have an answer to that when when it will end. Is it going to end this year? Probably not. Is it going to end next year? Could end next year, could end 2028. Um, but what happens when it ends is the second question. Are we going to see it unwind as a sort of a gradual deflation? I don't mean deflation in an economic sense, but I mean the bubble will deflate, right? Gradually, right? Or are we gonna see the bubble burst and cause widespread damage to the economy? Uh and I don't know the answer to that. Um, I would hope that it's a gradual deflation rather than the bubble bursting, but we don't know.

Steve Davenport 4:45

So yesterday, last night on CNBC at five o'clock, there was a round table with Jensen Wang was surrounded by six of the most powerful people in finance. Um, and they were announcing a $500 billion funding for various parts of the AI ecosystem. And I thought it was a commercial

The CNBC Hype And The $500B Pitch

Steve Davenport 5:09

that Jensen Wang had created and paid for, but instead it was something that CNBC reported as news. And when I looked at it and I listened to it, and I heard that we need to stop thinking about technology and start thinking of it as infrastructure, I started to put my skeptics' hat on and I said, wait a minute, isn't infrastructure something that's long-lasting? And if the chips have to be replaced every two to four years, and I'm hearing it's more like two, then can that really be considered infrastructure? Can chips be infrastructure or are they a short-term asset? And I'm of course dismayed by a lot of things, Clem. So that's why I ask you is are chips a part of the US infrastructure? Or am I missing something? And did somebody immediately turn AI into I heard about the great American system of financing innovation, and I heard about how great America was at giving people money for things to make a difference in their lives, and I almost stood up because I thought they were gonna play the national anthem. Um, is this really what's happening? That we're gonna start to twist this as part of American greatness, and we should just give them money because it helps make America great.

Clem Miller 6:41

Okay, so Steve. I being cynical, yeah, yeah. Yes,but appropriately, okay, in some regards. So let's let's talk about infrastructure, right? That is a very misused

Data Centers vs Chips As Infrastructure

Clem Miller 7:06

term, right? And I would say that part of the problem here is that is that infrastructure, the term infrastructure now applies, you know, when you're in the investment world, right? And you know, deep into the investment world, infrastructure has a specific it's a specific term of art. I think you know this, Steve. That you've got firms that have created infrastructure financing products, or you've got you know, infrastructure of investments, like ETFs and whatnot, right? BlackRock is out there doing infrastructure financing, for example, or Apollo or whomever is doing infrastructure. And and infrastructure in those contexts is referred to things like you know, electrical distribution, water projects, you know, sanitation projects, roads, right? Toll roads, bridges, energy, power plants. That's infrastructure. You know, infrastructure meaning long, let's call it what it is, long-lasting capital assets. That's what infrastructure is supposed to be. And so, you know, data centers, , yes, data centers I would call infrastructure by that kind of definition. But that's like the shell of the building, right?

Steve Davenport 8:48

But if the if the data center needs new chips in order to run semiconductors, are not infrastructure, to your point.

Clem Miller 8:55

Chips are not infrastructure. Chips are like you know, three years, two years. I mean, NVIDIA goes through different rounds of chips like every year or two, right? So what is each new investment in additional chips? Are we should we be considering that to be infrastructure? No, that's not infrastructure, right? Do you consider the paint on a bridge to be infrastructure? Do you consider the repaving of a road to be infrastructure, the repaving itself? No, no, right? The the data centers are the infrastructure, the chips are not the infrastructure. And I think I think what Jensen Wang, who is a cons mate pro promoter, with his leather jacket, his cool leather jacket, and I'm sure he was wearing that, right? His cool leather jacket and and and I don't know, his demeanor. He's just he's just he's a fantastic promoter of his company. And obviously he went to be at the center of things. He's trying to make maintain Nvidia's relevance in this world. And so he he not just is relevance, he wants to create the impression, the image that he's at the center of this AI infrastructure, infrastructure, quote unquote universe. And so the one well last thing I would say about that, about that meaning and the 500 million is that 500 million dollars 500 billion, sorry, 500 billion dollars even 500 billion dollars, well depending on did they say what time frame? Whatever that is, right? Right five hundred billion dollars spread out across a lot of countr companies. Um depending on what you count in that is maybe not a lot of money, actually. Okay.

Steve Davenport 11:43

Uh well my whole point is is it is is you and people always exaggerate too.

Clem Miller 11:49

So 500, you have to count 500 billion dollars as an exaggeration, too.

Steve Davenport 11:55

Yeah, there's six, you know, there's six leaders there, less than 100 billion a leader. Um, I I could see those institutions over some period of time providing funding for these types of things. My only point was at some point when you're going on national TV and talking about this as an American, you know, America's future is dependent on it and telling everybody it needs to happen. Um, if you need to tell people, then it's probably, you know, the investment's being sold to you, not given to you, right? Yeah, exactly. You're not given an opportunity, you're sold an idea. Yeah. And I feel like I'm being sold this idea on AI. So if they're starting to sell you on this idea and you don't arrive at it on your own, I think that something's changed. And then to turn around today and see, you know, Intel is going to go to the market for equity for another 16 billion in funding for its AI capabilities. I sit there and I say, you know, the the company was needing investment so that it could survive six to nine months ago. Now all of a sudden, we're going to market at a much higher level because people have bought into the idea that they're part of this ecosystem.

Debt Funding Rates And Recession Risk

Steve Davenport 13:19

And I think that what it's telling me is we're not in Kansas anymore. Cash flow is not doing this. It's going to be built the same way a lot of things are built, which is by debt. And when you build things with debt, you build in riskiness. And therefore, you build in the possibility that someone doesn't pay, and you build in the possibility that you know there is , and and the part of this that bothers me is the part of this in the back of my mind that is Warsh, talking about the fact that we may need to increase rates once or twice this year. So if we need more debt, and rates are probably going up in this short three to six month time frame, and the reason rates would probably get lowered quickly is if we went into a recession. So it just feels to me, Clem, like we're pushing ourselves towards something that is a typical recession where we've overextended, debt becomes too much of a factor, and we then have higher rates that make the payments on that debt higher. And as a country, I think we've all been reminded we're over a trillion in debt spending, and that's the interest, not the 30 you know, trillion or so that we're adding to the overall debt. So, to me, this future opportunity is looking a lot to me like overspending that we've seen in previous recessions.

Clem Miller 14:57

So, Steve, I was just as you were talking, I was thinking just that. Uh, and so let me ask you do you think that we're heading for a TARP two in the AI sector?

Steve Davenport 15:12

I don't know.

Clem Miller 15:13

That's that would be I mean explain what TARP means to people, our listeners who may not know.

Steve Davenport 15:21

TARP was the government funding of banks during the financial crisis, where they gave them the money and

TARP Two And Slowing Capex

Steve Davenport 15:29

said, You need to pay this back with these restrictions, and we're gonna own some stock in you until you pay it back. And a lot of banks didn't want to take it because by taking it, you were getting painted with the brush that I'm an injured bank or an unsafe bank. And by the fact they gave it to everybody, they said, Well, everybody got it, so there was no picking of winners and losers. And so I agree that Intel doesn't want to be painted with this brush as a loser, but they're also not exactly showing us by finding the cash flow to do what they need to do. I don't think that we will have TARP, I don't think we will get to TARP. I think that all these companies, given their own desires, could easily stop or slow down the capex. But then climb the question is what's the forward earnings? If we're not gonna have this exponential growth by the build-out, then we're gonna go back to lower growth rates. If we go back to lower growth rates, as a growth investor, tell me, does lower growth rates mean higher stock prices?

Clem Miller 16:48

Of course not.

Steve Davenport 16:50

No, of course not. That's why so but so you should be saying to me, Steve, please don't tell me my growth rates are gonna slow because I'll go from 50% cash to 75. Do you want, you know, do you see this in the horizon, Clem, or are you still living in the la la land of well every every street corner should have a server farm on it?

Clem Miller 17:18

I think there's a lot of there's a lot of variation among companies in terms of tech companies in terms of their short ratios that I've been observing. And so I've tended to go for those. Um well, I'm exclusively going for those that have relatively low short ratios.

unknown 17:42

Okay.

Clem Miller 17:43

Because honestly, like you were implying, I think you were implying this, , the earnings-growth

Why Forward Earnings Feel Unreliable

Clem Miller 17:50

ratios, forward earnings might not be as reliable as they've been in the past. And if that's what you're saying, I totally agree with that. Uh, because when I saw you know what convinced me about that, and I I you know I like the idea of PEG, of forward peg. I like that idea because first of all, it was forward-looking, right? Wasn't look, wasn't past looking. Secondly, it took into account growth when you had you know PE ratio not looking at that at all, right? So obviously you should look at growth. So I liked conceptually the idea of forward peg. It's just that you've got you know people who are so analysts who who make these forecasts be so enamored of AI and what these companies are going to do that they're they're sort of captured by the all the excitement. When you see Micron and Sandisk with like low like low sub one percent, like 0.5 percent or whatever forward peg ratios, that's just insane, especially as their stocks were soaring. And then obviously they hit a they hit some kind of a ceiling and then they started pl meting. And and so you know that's my question.

Steve Davenport 19:17

Is it is it isn't that a perfect example of how I mean I love your idea of looking into the future, it's not based on the past, but the past kind of is real n bers of real generated profits. So I'm not sure I can always go to that future expected n ber that now is is realistically not as attainable. If you know, if the past quarter you grew at 25 and the forward quarter you're gonna grow at 45. I don't know how the same company does the same thing unless they're underinvesting or over-leveraging, or you know, there has to be something to the financial structure that changes that creates that extra value.

Clem Miller 19:59

Yeah, it's almost like you have to take take the forward earnings estimates and divide them by two or divide them by three or something and recalculate your forward peg. I think it has to get that complicated, right? It's when you have short interest that you can use instead, right?

Steve Davenport 20:19

And then I look at the capital that's being raised and I look at the people who are investing, and the idea that you're you're finally getting being able to be included with the beautiful people and private equity and private credit and private real estate, I mean, it's a wonderful thing to be, you know, a sophisticated investor. But I also look at it and say, we're not in a place where sophistication and we know we know what we're doing. This is a whole new area. So just because somebody was really good at bioscience, you know, private equity, are they necessarily going to be the ones that give you the access to the right funds for build out of you know server farms? I don't think they are, and I think that because this is a new space, because this is a space that I just look at it as a chain. And once you break the chain, whether it's the future estimates for memory are too high, maybe the future estimates for you know CPUs are too high. Maybe we, you know, maybe we need to wait and see a little more. I'm I'm I'm just a little bit confused by how quickly I saw it with the SpaceX IPO. The media just became this enabler, and they believe that Musk can be trusted. He came up with this price on the back of an envelope, and he believes this is the price we should go public at. And so all of those spreadsheets by those investment bankers are worthless. Musk says 133 or 143,

Narrative Pricing And The SpaceX Effect

Steve Davenport 22:06

that's the n ber, and I just feel like I I thought there was some analysis.

Clem Miller 22:12

I thought it sounds like Tr p and his 10 or 25 tariffs.

Steve Davenport 22:17

I thought we it feels right just arbitrary n bers, Cl p. I didn't think we just picked the n ber out of the air and said, I've got it here, it's 135. I I didn't think that's the way we were supposed to do things. Do we like sit in a Ouija board and we wait for it to move towards the n ber? I mean, tell me that we're doing some type of analysis. Like you Said the the Apollo yesterday at this meeting was talking about what a great opportunity it was. That you said the CFO of Apollo was doing an analysis that said it's it's going to eventually slow down. And when it slows down, it's gonna mean a correction, right?

Clem Miller 22:59

Well, they may be having a little talk this morning.

Steve Davenport 23:01

Right. I'm just I I I am a silly old man. I believe that n bers still matter, models matter.

Clem Miller 23:12

You're not old, you're just logical.

Steve Davenport 23:14

How you construct your models and your future investments should be based on some series of data and inf you know influenced by your experience. If we haven't had companies in the memory space growing by 45%, because it's mostly commodity-like, what changed? What's different? Tell me what's different this time, and I will I will gladly accept your analysis. Just show me some analysis. That's all I'm asking for. I'm not asking for a lot of analysis because I realize we're going to be simplistic to you know speak down to the investor level now that we're going, you know, we're almost at retail. And so we've got to tell people, you know, probably a lot more than we would rather not tell people in the private equity and private debt space. But just give me something, Clint. Tell me who is doing things in a way that really does get backed up by the n bers and and has a potential for a great return. Are you gonna tell me?

Clem Miller 24:23

Oh, I thought you were just being I didn't know you were asking me a question, I thought you were just raising a generic question.

Steve Davenport 24:34

No, I I I I depend on you to bring me back to the fundamentals because you're a fundamentals guy. I'm you know, I I look at quantitative models and I'm a little too obtuse.

Clem Miller 24:46

Look, I I I think that I think that there are a lot of questions about growth in AI space. I think funding is a huge issue. Uh, and I think that we need to pay careful attention. I mean, there there are debt n bers that are available for these companies, and I think as we go forward, we need to be much more careful in looking at how these companies fund themselves. Uh, I think that it's gonna be difficult because you see

Short Interest As A Reality Check

Clem Miller 25:19

the private privately owned companies funding, and we're not gonna have data for that. Uh so that's gonna be challenging. Uh, I think I think that that what we need to do is, you know, there's you and me, and then there are people out there who bet against these companies. And so I don't think that's not a bad yeah, and so I want to know how much they're betting against these companies, and I don't want to be fighting these short sellers, right? If somebody's gonna be short selling a lot of a particular AI-related company, I don't want to be, and they've done their research in order to do that, because short sellers can have unlimited liability when they short sell. Uh I don't want to fight those well-informed short sellers. I'd rather take the knowledge that I gain from them, that is the fact that they that they're doing active short selling with a stock and and stay away from that stock. So that's why that's why I'm emphasizing you know, one percent, one point four percent.

Steve Davenport 26:40

You know, well, I have a I have a little just fly to throw in that alignment. All right, it's not meant to be, you know, too upsetting to you, but let's think about this. You've set your n ber at one or two percent, right? Yeah, and what happens when the price of all securities goes up, yeah, goes up 20, 20, and then another 13 this year. Yes, and then we have IPOs that add

Do Ratios Change In Inflated Markets

Steve Davenport 27:13

a trillion dollars to the market, but we didn't open up any new short sellers. The short sellers have larger positions to short on, but the biggest positions are harder to short on because you know one percent of Broadcom now to short one percent of Broadcom is a bigger n ber than it was three years ago, right? One percent of NVIDIA is a bigger n ber than it was three years ago. As we get bigger and bigger, don't the short sellers shouldn't your n ber be going down as to what you accept? Because there's less and less there's the same amount of short sellers, but there's there's so much more equity.

Clem Miller 28:03

I mean, shouldn't I accept like shouldn't I require something that's like below or right around one? Right. If I if I look at the world as having you know, and I don't know what the total well, you know, one of the I'll tell you what I found interesting is for a while there, short for some of these technological stocks, especially, but for quite a few stocks, I saw short interest creeping up. Uh and now short interest seems to be coming back down again.

Steve Davenport 28:37

Uh so yeah, I mean, is it because their courage is less, or is it because their their universe is less? That's what I I just wonder if you get burned on shorting you know Broadcom and NVIDIA, maybe you try to, you know, everybody can't be in an Oracle, can they?

Clem Miller 28:59

No, no. Um so yeah, I'd rather be in a 1.5 than in a in a four or five percent one, and , and so that's why I'm not in micron or sand disk. I was at one point when I was looking at forward peg, but now I'm relying much more on short interest, and I'm not in micron, not in sand disc. Uh I do have some NVIDIA, , but you know, maybe not for long, right? Depends on what happens to their short interest. Right.

Steve Davenport 29:37

My only question is is that there's the universe is getting bigger for these names. The hyperscalers are getting bigger and bigger. And therefore, you know, if you have the same people with about the same portfolio of shorts, shorts get burned by an upmarket. Yeah, so we got an up 13 this year. Probably then they're losers if they were going against the general market. Let's just say, in general, that's how we look at it.

Clem Miller 30:04

Do you think they have a lower pool of money and smaller pool of money to apply against a larger larger trading vol e?

unknown 30:13

Yeah.

Clem Miller 30:13

And therefore, they should have automatically lower, you know, sort of mathematically lower ratios. Yeah. In which case, my range should be quite tight. Right.

Steve Davenport 30:27

That's my only thing.

Clem Miller 30:29

Yeah, I think my range is pretty tight. I mean, if you you know, I think the the conventional wisdom is that it's okay to have this is the conventional wisdom, that it's okay to be long invested in stocks up to like two, three percent even short interest. And I'm a lot tighter than that. Okay. I don't think, I mean, if something rises above two, I'm not gonna sell it right away. But you know, if you look at my list of stocks, they're , you know, I would say the average short ratio is something, you know, short interest is like 4% or 1.4%, 1.5, something like that. Okay, I got some that are down around one. I've got a few that are like 1.8, 1.9, but not nothing up around two.

Steve Davenport 31:23

Okay. I'm just saying that I've been thinking about your short interest ratio. I'm thinking about how inflated stock values could make yeah, your n ber deceiving, and you think I'm fine at 1.8. And in reality, two years ago, that n ber of short against the company would have been three. You know what I mean? Yeah, yeah, yeah. And so therefore, you kind of have to say, hmm, maybe I should look at this differently now that the the values are inflated.

Clem Miller 31:53

Yeah, there's something, I mean, I understand what you're saying, but I have to believe that markets are a bit more dynamic than that, that higher trading vol es would attract more short interest as well. And thus the the ratios wouldn't automatic wouldn't adjust so mechanically.

Steve Davenport 32:17

Yeah, okay. So something to think about. Yep. Anything else on AI before we close out? Oh, that's it.

Clem Miller 32:25

You good? I'm good. Thank you everybody for listening to us.

Steve Davenport 32:30

Yep. Be good, everyone. Send us your comments and ideas, and , we love having you listen and we love to have your input. So have a good day, everyone. Enjoy. Bye.