Title: Software vs. Semis: The Next Battle in AI | Protect the Pile Episode 26 Show: Protect the Pile — the podcast of Hedgeye Asset Management ("Free-Form Friday" episode) Hosts/panel: Patrick Kent ("RPK", Hedgeye Asset Management) and Sam Rahman (portfolio manager; surname spelling unconfirmed — auto-transcript "Ramen"; X handle hedgeye_hgrow) Date: 2026-09-19 (YouTube publishDate); recorded Friday 2026-09-18 ("it is September 18th") URL: https://youtu.be/gLouWXez2OQ Length: 57:53 Note: YouTube auto-transcript pasted by Stephen; fillers (um/uh, "you know" interjections, contentless "like"/"right?") and stutters/false starts removed; wording otherwise verbatim; (mm:ss) cues kept in place. ">>" marks a speaker change; the auto-transcript does not label speakers — RPK opens and hosts, and attributions on the analysis page follow the conversational turns. Auto-transcript garbles fixed: "Hedgei/HedgeI/hedgi"=Hedgeye, "brethren"=breadth, "moving monkey"=their nickname for the 50-day moving average (as spoken), "Russell 20000"=Russell 2000, "my index"=as spoken (a liquidity index), "GM"=as spoken (likely GS — Goldman Sachs — positioning data), "meter"=METR (AI-evaluation group), "Aaron Sorcin"=Aaron Sorkin, "social reckoning"=The Social Reckoning (film), "Bentley"=as spoken (likely "bet"), "leg reg leg reg…" stutter collapsed to "legislative, regulatory". Figures (S&P ~7,620–7,625, 10-year ~5%, CCC spread 920, diesel ~$6.50) are as spoken.
00:00 All right, welcome back to another episode of Protect the Pile, the official podcast of Hedgeye Asset Management, the only no-nonsense investment show that cuts through the market noise with a panel of practitioners who play this game every day. Or at least a pair of us today. We'll dissect the landscape, debate the opportunities and risks, and help you navigate the markets using the Hedgeye risk framework and lots of real-world experience.
00:33 All with the goal of protecting and growing your pile. I am Patrick Kent, or RPK, of Hedgeye Asset Management. With me is Sam Rahman, colleague, portfolio manager, co-host most of the time. And Sam, it's Free-Form Friday. It'll be you and I batting around whatever topics we feel like taking on.
00:55 So it is September 18th and it is episode 26, I believe, if my count is correct. The S&P is around 7620, 7625, wherever I last checked it. Just a little over 2% off the all-time high. Does it feel like that? >> I don't know. Not really. I don't know. We'll see what Sam says. But breadth has deteriorated under the covers.
01:22 So there's roughly 30 to 35% of names above the 50-day moving average, or the moving monkey as we lovingly call it around here. You can see this, it picks up in the diverging signals between the bearish Russell 2000 and the bullish-for-now signal on the S&P. I don't have it in front of me, but maybe Sam recalls exactly where the trend line is on the S&P 500.
01:46 10-year's at 5% despite a hawkish Fed meeting on Wednesday. So we'll get to that, talk about that. Dollar's back on bullish on the Hedgeye signal, mostly off the back of that. I guess I'd say surprise — I don't know if it was a surprise. The market seemed to react like it was a surprise that they raised rates, even though the odds markets had this at a 90% chance.
02:09 Which is bad for global dollar liquidity. My index kind of slows down on the back of that. Also bad for liquidity: a higher cost to make things go. Diesel is at $6.50 a gallon roughly. Some places it's $10, like California. We've seen some numbers like that. Some places you're getting spot shortages. Oil's over 100.
02:32 Physical markets are closer to 120. Maybe we'll talk about that a little bit. That might be at least in part what is happening in credit. High yield is now bearish trend. And the double-B spread to Treasury is at like 273 basis points over, something like that. It's up about 20 basis points in the last month.
02:52 But I think the more interesting story is in triple-C credit, where the spread is at 920. So for those following at home, that's a 14% cost to borrow on the crappiest companies, and those crappy companies are what your PE credit looks like. So Sam, how are you, man? >> I'm good, man. How are you?
03:15 >> I'm all right. This is a strange market to navigate. We say that every time — I feel like we say that every single time we're on here now, because we live in strange days. But it does feel like we're waiting on a direction. And as my early look this morning would sort of imply, I worry that the direction in the next three, four, five months is maybe going to be a little bit of some things we don't want to trend on.
03:48 [laughter] Let's say that. >> Sure. >> How about you? >> Yeah, I think it's been tough, man. The last month and a half, two months has been tough, because it seems like every day or every week we're churning within the market. The market levels per se, as you said at the top of the party, we're like 2% off the high, but underneath it all there's been tremendous rotation and churn.
04:15 I see it in the signal strength list: if you look back even a month ago to now, just how much change there's been, names that came on, names that came off. I think that's just the nature of what we've been seeing in this market, that there's a lot of rotation, a lot of churn, and also, particularly the last few weeks, definitely a narrowing of the market. As yields declined and oil prices have climbed, it's really kicked out sectors that were favorite longs for a lot of the investment community, like
04:54 financials, industrials — very long consensus longs over the last few months that got taken a shot the last few weeks. Consumer's been weak, and you can look at the relative strength chart of consumer versus the market over the last year and it's been weakening since spring, but — >> Yeah, I looked at that this morning, and my check on all the consumer names — I could find only a handful. I have my proxy for the signal; it's one I can do in large batches. It's not accurate, doesn't get the edge cases, but it generally gets the direction roughly correct, especially when I'm looking across lots of names. And
05:29 boy, the consumer, it's just a bloodbath. There are very few names left that are bullish trend. I have a couple that I actually was going to ask you about. Maybe we'll do it offline. [laughter] >> Sure. >> But —
05:48 >> Yes, I think that that's what's made it really hard and a tricky market to navigate. But I think the one thing that might be a sort of aligning or positive — and there's no way to time this — is that if you look at positioning data in terms of hedge fund, long-only stuff that you and I see from GS and Morgan Stanley, positioning is pretty light.
06:25 Particularly in momentum. >> In the momentum basket. So — >> And there's a lot of cash on the sidelines. Nobody wants to take a big bet. So I don't know what it'll take for that to change, but you get the sense that there's a lot of dry powder sitting waiting for some direction, some signal that — not that things are resolved per se, but that maybe at the margin things aren't getting worse, and that may be as good of a positive catalyst as the clouds part and the sun — we're not going to get, most likely we wouldn't get, a grand bargain next week at the UN or anything like that, but if we get less bad headlines, that takes the pressure off crude oil. I mean, if you really boil it down, this market is a one-factor-model market.
07:09 >> Yeah. Or maybe just — well, I guess that's probably true, right? Because that's what's pressuring, at least in part, the 10-year, right? The 10-year rate is up because we're still concerned about inflation
07:41 pressures, because I think the consensus would be that the Fed can raise rates, but that doesn't necessarily open the strait, right? It doesn't have any impact on that. So that's effectively trying to curb demand enough to slow things down, which sounds like a great solution until you see what that really means.
08:04 And then people usually start to be not so happy about the second-derivative slowing, which it would need to fix that. So I think you're right. The market reacts to it, certainly. Futures are up when oil's down, and you can kind of see that in the way this is playing out.
08:25 Now what'll be interesting is just to see the timing of all this and how this is going to show up in fractal patterns of the market, because — we talked about it before we got started — the physical markets are at 120, right? The forward market for Brent is like 105, call it, somewhere around there, and that contract at 104, 105 rolls in like a week or so, and that has to converge.
08:59 So either the physical market has to come down to this price, or the price is going to grind higher. What usually ends up happening is something in between, but that is sort of just like a law of gravity: they have to end up being kind of the same price, at least briefly, right? So that puts a sort of bullish cushion under the oil markets to inflate it, which is a headwind, I think. But to your point, if news is going to try to impact that very quickly, it
09:32 would roll. And if the physical markets are weakening, then that will actually really start to change things. Anyone listening to the show may have already heard this, because it was out on Twitter even if they're not getting JPMorgan research, but there was a report that just went around the world about JPMorgan's commodities team basically just throwing up their hands and
09:58 being like, we have no forecast, because I don't know how to model this. Which is the right answer. As much as people want to dunk on them for it — and we talk about that a lot around here — why do you feel so confident in a complex multivariate system where there's all kinds of cascading effects? Why would you think you can have a firm forecast that you're going to stick with, when even the people who are involved in all this don't even know how
10:32 this is going to turn out. So in the secondary and tertiary positions, why would we have a firm view of how this is going to go? So it does feel like this is an example where we do have to just watch the signals, and those will tell us a lot more about what's happening here than I think —
11:04 >> Correct me if I'm wrong, I think the signal on WTI, the trend line, is down in probably the mid-90s. So there's a lot of room between now and then where the price could come down but still be signal-bullish. So I think that's the — to allude to what you're saying — that's kind of the dynamics that we're going to have to navigate: even a preponderance of headlines that are viewed bullishly, that bring futures prices down to the mid to high 90s,
11:36 it's going to feel like relief, even though the physical markets are tight. There is supply tightness everywhere, particularly in diesel. So even if you drop to the mid-90s next week, it's not like some buyer in Asia for diesel fuel is going to get that diesel fuel that they want, or in Europe, or even in some states here, where there are pictures now on X where you can snap a picture of some gas station somewhere in the middle of wherever where they're sold out of
12:12 diesel. So that's not going to change for several days or weeks. >> No. And as someone who used to cover energy more full-time, we're in hurricane season. So — >> Yeah. >> That's not even — now, I'm not saying that there's going to be a hurricane that'll impact the refining complex, but that happens.
12:34 That's a nonzero probability that could actually end up causing some problems. And in fact, Exxon just brought a diesel refinery down, for other reasons, because we're running these things full out, right? We're not shutting them down, because we have product shortages — we usually do turnarounds on these things to do maintenance, and when you don't maintain a refinery, it blows up.
12:57 Because they're giant pressure cookers, basically, right? So with that kind of taken offline, that's another headwind to the market. If we get some kind of hurricane or something that impacts that, that's going to be another problem. These are just — when a system is really tight, all it takes is a couple small things to go wrong that can have, again, cascading effects.
13:22 You probably saw the headline this week where the administration was talking about maybe looking into banning exports of refined product just to deal with what they're seeing. It doesn't solve the problem. >> No, it just does not solve the problem. So it's like —
13:53 >> It might have an immediate-term impact on taking some pressure off the diesel price — goes from 6.50 to 6 and we go yay — and that's it. That's what you get. >> Yeah. But we were talking about this before we came online. I think the thing that's going to be interesting is how long this fuel issue is going to last. One, you've got tight supplies, but also you've got elevated prices. You've got two problems. So even if suddenly supply got a little bit better, elevated prices are still going to be elevated, even if it's down a few cents from where they are right now.
14:21 So we were talking about this earlier: how oil and the progression of oil has gone from this inflationary force that's been in play for several months now,
15:05 to now we are at the point where you really are at demand destruction and potentially, if you look forward, a disinflationary force, given how high these prices are and how companies and people are going to look to manage their spending, not just on fuel but also other things in the wallet, which ultimately have this disinflationary pressure on prices, which I think we will start to potentially see as we go into 2027. But how do you think about that in terms of the macro lens that we could keep an eye on, in terms of trying to triangulate growth, inflation, and to your point earlier about Fed futures expectations — the '27 expectations for hikes, I've got to believe those are in fake pencil that could be easily erased, because I don't think that's going to happen.
15:38 >> Yeah. If things — we're hiking rates now into what I think is going to be kind of a cycle peak for rate of change, right? Because, just for the reasons you just said, diesel prices at $10 are going to slow things down, right? You are going to slow things down.
15:58 Even — we'll get to talk data center capex and that kind of thing — but the only thing I bring up is that that is the biggest driver of industrial activity at the margin. And at some point there is a question mark of, does this go in at any price? Because steel price is up, copper price is up, diesel price is up, memory prices are up. So you're not only building the capex, but you're building it at peak prices for everything, and so it really becomes like, wow, you're really taking a big
16:30 gamble that even at an elevated cost of capex you're still comfortable underwriting this. So even if — again, all it takes is a pause, right? Like, hey, we're going to slow down a little bit because diesel prices are high, it's getting very hard to find some of the inputs. That alone is a second-derivative change from where things were, and when you look at it from a year-over-year or six-month or three-month rate of change, if that starts to slow down —
17:02 I know Keith, in his ETF Pro, has had the industrial momentum on as a short, right? Because when the second derivative changes, the thing that gets hit the most is the stuff that was moving the most.
17:22 Actually, I want to come back to your point about momentum, because there's an interesting thing happening, I think, under the covers too, which is that momentum is not a sector, right? It's a basket of companies. And so momentum changes. It was one thing, it becomes another thing. And so it's interesting to see the tug-of-war that's going on between semiconductors and software in the technology space. It's a fascinating one, because it was
17:53 this "AI is going to put software out of business," and that was the trade that led to that huge gap between the two for a while, and now that narrative has shifted: okay, there are losers, but there are also potentially some winners, there are people who have a little better moats. So we started to see the software recovery within that, and then obviously there are questions being raised about maybe the long-term profitability of the closed-weight models, all that kind of stuff.
18:16 That's fine. But it's bringing this back in, right? But now that tug-of-war is: when you look at three-month momentum, software is dominating it. But on the 12-month-minus-one-month, that long-wave momentum, you still have a bunch of the semiconductor stuff that's at the top of that basket, but it's falling down the basket on the three-month basis.
18:38 So depending on how you're defining momentum, there's this tug-of-war between short-wave or longer-wave momentum. >> So just on that question: these momentum baskets, how do they weight the three-month and 12-month? Is it different for each one, or is there a —
19:04 >> Depends. It basically depends on the model, right? If you're running a quant model, depending on what else you're putting in it, you might have more of a weight to three-month momentum. But I know classically the way people like to look at it is the naive model of momentum, which is: stocks that have done the best over 12 months
19:22 do the best going forward. And so that tends to be what they want to buy, and so that's the one they tend to look at for long-wave momentum. >> Yeah. >> And they usually take the first month out. So it's the 12-month minus the one month, because you want to screen those stocks. You actually don't want to capture it if they just had a pullback; you don't want to penalize them for that, because that could be your best opportunity to be buying those stocks.
19:41 So that's why they want to rank them on that 12 minus one. And so I just find that that composition is shifting: healthcare, software, some of this stuff has migrated into the shorter-term momentum, three-month, six-month, where the 12-month is still dominated a lot by the stuff that peaked back in July. >> Yeah.
20:04 And that is a very, very important observation, because if you overlay positioning with that too, I think you get a dynamic that's interesting as well, in the sense that on the AI semi stuff, the positioning is being cleared out. So now you have this confluence: okay, if people are migrating to the short-term momentum, that's at the expense of the 12-month AI stuff that was dominating a few months ago, but the positioning is somewhat cleansed. On the software stuff,
20:42 it's killing it on the short-term momentum. Positioning has improved, but it's improved from rock bottom, extinction levels. So on a longer-term positioning map, they still could go quite a bit higher, if these software names hold momentum and hold strength and continue to grind higher: they go from three months to four months and they start becoming bigger pieces of —
21:18 >> Oh yeah, yeah. And then think back to when software was the dominant one within that 12-minus-one-month momentum basket. So take yourself back in time to four or five years ago and remember what these stocks were doing, right? '21. I remember, I think it was '21, when we got to late in the year and I was looking at the multiples people were putting on a lot of these companies from an EV-to-sales perspective, and I was like, look, I know we don't care about valuation necessarily with these names, but if the
21:52 second derivative shifts on you, this is going to be bad. And I just remember the setup there being: wait a minute, if you reopen the economy and everyone was on Zoom and doing everything on it, at the margin it's just going to be an uncompable comp, and you're sitting here at peak-on-peak multiple. That's an ugly setup, right? Those types of setups are the ones that always feel the worst when you don't want to be involved at the end when they go parabolic, but then they
22:21 just turn into absolute disaster once the rate of change actually does show up, because you get the mean reversion of the multiples on top of the slowing revisions. And but this is kind of —
22:59 >> So just on that, it's going to be interesting to see going forward — it was one day, but we saw this yesterday: software and semis were up together. This goes back to this question we're going to see and have to answer in the next few days and weeks: can these two groups work together, if the market wants or decides to rally and break out? Can these groups work together, or is it going to be either/or? And I think —
23:29 >> I think the answer is yes, probably. Yeah. Well, what do you think the answer is? I'll give you mine if it's different than yours. >> There's a good chance it could be both, because software's got that three-month momentum tailwind, and I think underlying the fundamental picture of it, clearly a good number of these companies are going to be around for a long time. So the extinction factor is potentially behind you and you've got the momentum with you. And then on the semi stuff, because positioning is light and there's beta in there, people want to chase beta later this year. That could be a basket that recovers, because positioning is light,
24:05 as long as we don't get some real negative headlines on this AI stuff. We could talk about it later, but the whole AI-safety, pacing-the-frontier stuff — we'll see what happens, but it clearly felt like a DeepSeek moment part three, or whatever part we're on. >> Yeah. But —
24:33 >> But it could happen where both work. We'll see. But the elements are there in different ways for both those baskets to work together. >> Yeah. So I think my answer would be very similar to what you just said. The only thing I could add to it is that it would be effectively a way of resolving the two buckets of momentum,
25:04 right? I think the have/have-not discussion just shifts from being semis versus software to haves in semis and software and have-nots in semis and software. Because no matter what we think of memory right now — look, memory is a commodity business that the Chinese are expanding capacity in. >> And "oh, we're not going to buy Chinese capacity, Apple's not going to" — it doesn't matter. We also don't buy Chinese steel, but that doesn't mean it doesn't impact the global price of steel. It's a global market, so whether we buy it or not doesn't matter.
25:26 So if they're expanding capacity, they're expanding capacity, and they are, in both DRAM and NAND. And so you've probably passed the peak of — now, maybe famous last words, maybe we haven't — but we've probably passed the peak of the tightest pricing on that stuff. So at some point those start just acting worse.
25:46 Meanwhile, you look at an Intel — I don't know where it is on the signal right now, but it's certainly looking better, I think — and it sort of remains true that it is probably the number one best company for building out a US manufacturing footprint in semiconductors. They have a fab already created that they need to fill up.
26:11 So there's that dynamic where you could have some of, say, the opticals, which look a little better, that might still look okay and stay within the momentum basket, where other things fall out of the momentum basket. Similarly in software: Adobe still sucks, but others — we're learning Snowflake, other things in that software basket, not only are not hurt, they're actually probably in really good competitive positions — or Twilio, where the narrative has
26:39 totally changed and actually the valuation's not that crazy. So those types of things can migrate back into that momentum bucket and stay there, and then we'll cleanse out some things that were just getting the all-in bid that are now going to start to be scrubbed a little bit.
26:57 >> Yeah, another great point. I agree 100%. I think once you get past this year, the questions on the earnings power of some of the semi space, particularly, as you point out, memory — I think the rate-of-change comp that you're going to have to face is so big that I don't think any of those stocks will survive that, in terms of just comping
27:30 these massive revisions, regardless of what multiple you're at, on top of the supply things you mentioned. I think it's going to be hard. But there are going to be other places. I think we'll probably net out, as we go into next year, to a broader basket of software names than we probably would have thought at the start of the year that's in the momentum basket, and a narrower basket of semi/AI hardware names than we would have thought at the start of the year that are
28:06 going to continue on surviving in '27. >> Yeah. No, I think that's right. So maybe let's talk about the safety narrative for a second, because I'll give you my hot take and then I want to hear your hot take.
28:28 So I'm not going to go with scam, but I will say PR. I'm going to go with that. It's kind of a PR onslaught. And there are a couple things it's sort of solving for. One is sort of a classic salvation narrative, which is that it could kill us all, so you really need to regulate us and support it,
28:51 create a monopoly of just a few of us that are trusted to do this thing, and then your salvation lies here with us. That's one. Two, I heard — checking into some comments from a distributor that sells into the data center space, the utility capex space —
29:12 they're kind of just like, look, demand's huge, but we cannot keep accelerating at this pace, and we're running into regulatory problems of just not being able to move that quickly. And anyone who's ever tried to build anything — you can try to push these things at 20, 30% growth rates, but you still have to get the inspector to come out and sign off on the thing, and it needs to go through the town meeting in this area.
29:38 These things don't move at 30% growth rates. They're so slow to do anything. So as soon as all the low-hanging stuff is gone, you're up against just the physics of not being able to grow this any quicker. And so it's almost better to create a narrative like "we need to slow down for these other reasons, because we're so cool," not because we just can't grow at that pace anymore and we're running up against barriers.
30:00 Not to mention the growing liabilities of these things. So that's my third one: every breakout, every story — and now there was the story this morning, Anthropic guys hacking OpenAI's emails — this is a massive open liability that's just beginning to get going.
30:23 Anytime you get a rogue agent doing anything, you're going to be like, ah. Because this whole "AI is going to kill us" — well, we probably cause ourselves more problems with AI long before AI gives us problems. That's my hot take. What's yours? >> Yeah, I kind of — again — agree.
30:40 I think broadly the whole strategy behind it was, on the micro level, regulatory capture: trying to get the government in to kind of secure or ring-fence these companies and provide regulations that give them an additional moat as other models come onto the scene, be it open source or others.
31:13 Clearly, I think also, if you look at the coordination between the key players in the model companies when they first made this announcement, and how quickly the political classes came running to support this narrative of "hey, we need to regulate this" — >> Dude, the lobbying dollars.
31:36 Come on. >> Yeah, they should have done a better job pacing out the commentary and the tweets or whatever, because it was such a tight window: Dario, Sam Altman, the New York Times, METR — which is this monitoring body that's affiliated with Anthropic and effective altruism — which is this —
32:08 >> Now, it's also part of the circular funding of all this stuff. >> Yeah. This group that's suddenly getting a spotlight put on it, that's full of activist players that are deeply embedded in the liberal political machine — Obama, Bernie Sanders, you name it. It was in a space of three days. It was impressive, to say the least. So there's a lot of that going on.
32:35 I think the great thing was you saw a pretty good, coordinated pushback to say, look, there's enough regulatory apparatus out there to protect consumers — product liability, product oversight, the FTC — there's a bunch of different layers of legislative and regulatory bodies that have been regulating companies, overseeing companies, for like a century.
33:09 So it's not like we need to really reinvent the wheel to do this. So you had these titanic forces of closed-source models and open-source democratization of technology battling out for share of voice, but I think Dario saw the risk-factor section of his [IPO filing] and literally lost his —
33:40 >> Yeah. He probably threw up in the trash basket next to the desk. Right. >> But you know, it's like, get your stuff together. If you're going to put out a model or product, make sure it's safe. And so — >> Yeah, but they've been in such a hyper-growth mode, right? They've been in such a hyper-growth development mode.
33:55 And here's a controversial thought I had the other day — you can push back on it if you think I'm totally off base. I wrote a piece three weeks ago, right, on "I'm not an AI skeptic, I'm an AI claim skeptic." Because my point was that the claims expand to serve the sellers' ambitions, not necessarily the buyer's needs. And so this idea of a regulated, closed-weight AGI digital god is like a solution looking for a problem, because I actually think 99% of the — I'm picking a number, I'm totally making this up — but if I just think real world, 99% of the productivity gains you will actually get are from a lightweight model running on proprietary data that you can customize solutions for your own business or your own enterprise or your own industry, like law or whatever, and then that's it.
35:11 That got the job done. We don't need the next model that's even the highest one. That's not the thing that's going to decisively make us better. It's just: well, look, we're trying to cut a lot of work. I can cut three paralegals if I have this Claude-for-legal that doesn't need to be updated again in two months.
35:31 It's good enough. And especially if you're training it on your own work — if you're a large law firm and you have a whole bundle of expertise in a certain area, I don't need you to look at every legal thing across the internet. I need you to just learn this huge history of stuff we have, right? So I'm a skeptic on why we even need this until some later date.
35:55 Maybe there's something down the road, when we've learned how to better model, better train these things, better contain them, all the interim steps before you get there. What are your thoughts on that? >> Yeah, I don't disagree. I agree with you. I think we're seeing this already in terms of token consumption, in terms of share. I don't know what the number is, but it is a large number, like 90-plus percent of token consumption goes to really basic, semi-basic queries, and
36:27 it's that 10% that is using the cutting-edge models that are doing a lot of heavy lifting on research, using the power of those tools. And as you can imagine, the margins and profitability of those two pies are similarly skewed to that really high-end heavy usage, where you're consuming millions and millions of tokens to build something super high-end. And I think that's going to be the way it's going to be, with model usage of all types, open and closed.
37:10 Once we start getting these applications it'll be that way too. A similar analogy, although not as skewed in terms of share as this example: let's compare iPhone and Android, iOS phones and Android. Android, in terms of unit share, is the majority share. >> Yeah.
37:34 >> Of the phone market all over the world, by — >> Right. Isn't it something like 80/20, and then the gross-margin dollars are flipped? Apple's like 80% of the gross margin but 20% of the units. >> Yeah. The profit pie is like 60–70% Apple; the units are almost 60–70% Android.
37:55 So I think that's kind of a good framework of where — maybe it's a little bit more skewed — that's kind of where we're going. And I think what we've been missing is: we've been seeing all the spend, the cost, the price inflation, and the dodgy financing deals that are all kind of going on there.
38:23 What we haven't seen really yet is a real application. We've seen it on enterprise with Codex and Claude Code; those were really big light-bulb moments for the enterprise solution. So that's what's been driving token consumption: coding using AI has been a massive driver.
38:47 What we haven't seen yet is applications for the average user, you and me and everybody else, where it really tangibly makes a difference to our lives. And I was kind of looking for this, waiting for this to come along. Have you started using Muse yet? Meta's? >> I haven't yet.
39:10 >> I just started to use it the last few days, and I'm still kind of deciding whether I really want Meta to have access to everything on my — >> I think that's probably going to be my biggest hurdle. >> On my worst days, Sam, I'm a little bit close to just unplugging everything and going into the woods sometimes.
39:31 >> So I'm not using it yet to that extent, but the little that I've used it for, I'm like, okay, wow, this is the first real look at a consumer application that has the potential to be super useful, high utility, really, really easy to use, no necessary coding or extra steps required.
39:56 You basically just download the app, you follow a couple of steps, and Meta is really good at this, as we all know. >> Well, they're good at — here's an interesting point, because look at Threads. Threads is kind of a joke, right, a little bit. >> Yeah. >> They've got enough usage on there.
40:16 >> It's a joke now, but when it first came out, the virality — >> That was going to be my point. At the time everyone was really sour on X and looking for another place to go, and they launched Threads at the right time and it ramped very quickly. The problem is X kind of recovered, and so people were like, well, we're already doing this, and now it's a distant number two.
40:38 There's still plenty of people on it, and they continue to drive some traffic there by crossing it with Instagram. People who are scrolling Instagram will tap on a thing that takes them to Threads and then you end up in a conversation on Threads. I've gotten into it; I know my wife's done it too.
40:50 That will happen. But I think this is the point, and this is where you were going with this: they've got a consumer-facing footprint, and so if they come up with something that's truly useful to the end consumer, there's almost — I don't want to say nobody better, but almost nobody better.
41:10 They're putting it in front of consumers very quickly. So if it turns out to be something that's worthwhile, it will take off and it will work. Apple will be the other one, right? Apple's the other one where they have the consumer, and we've talked about this ad nauseam: fine, fight it out over who has a great model; we're going to be the one they're using it with; they're going to be talking to us or using it through our devices. So have at it.
41:38 >> Yeah. Ultimately most people, you and I, are going to be more interested in the app layer. Whatever happens downstream in terms of the models, nobody's going to care, for most of us. It's more: can you give me an application that really is useful, that I'm going to use a lot more, that leverages the power of these models? And I think Meta is clearly super experienced and very, very good at doing this.
42:13 Apple, as you said, is also very good at doing this. They're not always the first to do it, but they usually have something next in line that really has a high utility value. And to some extent Google too, I think, has experience of putting out applications that run on the super complicated, powerful compute but actually provide a real utility.
42:40 So I think that's what we're missing, or maybe we're going to start to see, and I think that solves for this mental equation — it doesn't resolve it, but it partially resolves: what are we going to do with all the spending? What actually is going to come of it? What are we going to do with it?
43:14 >> And I think that's where we may start to see at least some signs. It doesn't mean that we don't have a bubble that ends like we saw with the dot-com experience. You saw those websites that had tremendous utility value and they were great: Amazon, eBay, Priceline, whatever they were. >> Oh yeah. In fact, I already know of a feature application that I think is going to take up a ton of this usage, and they'd better not screw it up in developing it.
43:33 I'll get to that in a second. Finishing that: yeah, but the thing ended. The thing blew up and it ended, but it seeded the consumer and enterprise behavior for the following decades to come. >> 100%. I think that's totally right. The experiment that might end up just coming to an end here will be just that: these closed-weight models, the platform to end all platforms, the one app that does it all, might end up — looking back — being like, yeah, that wasn't the way this was going to go.
44:10 In fact, that was the old internet model. We talked about this in the past. Take us back to '99, 2000: everyone was like, this is going to be the ultimate democratization; there's no longer going to be any power accumulated in just a few companies; it's going to be everywhere. 25 years later we have the Mag 7, and it's all concentrated in a few companies, and you're like, wow, that went in totally the opposite way that people expected. Wouldn't it be fascinating if now we think that's the same model and we're trying to replicate it with these two titans, OpenAI and Anthropic, the horse race to be the platform to end all platforms — the Mag 1 — and it ends up just being like, no, actually this was the fragmentation that was about to happen,
45:01 and it just took this long to get there. So, all right, the killer application that I hope they don't screw up: have you demoed a Vision Pro? >> No, not in person. I haven't. I've seen videos of people using an older feature. I've never done it myself.
45:23 >> And I say this as a person who hates virtual reality stuff generally. Personally I wouldn't use it a lot, but I've demoed it. If you launched the Vision Pro with an NBA courtside package, for — pick a number — $2,000 a year subscription, it would sell millions. And that's just that.
45:53 How about a concert package with Live Nation? How about a — once you've demoed and seen what it can do, sitting courtside at an NBA game, watching a baseball game from the best seats in the house — you can sell it anywhere to anyone globally. It is a mad market potential. If they screw up Vision Pro — how did you guys miss this? Because Steve Jobs never would have missed it. Anyway. >> It's a really good point.
46:26 I think it's this chicken-and-egg thing. Does Apple try to build — because you've got to build out the ecosystem, as you said. You've got to partner with the NBA, Live Nation. I was at the US Open a few weeks ago. Can you imagine virtually having courtside seats for the US Open finals, which would cost like —
46:58 >> That was the — but this is the point. Steve Jobs knew this type of stuff. He's like, there's no point introducing a device that doesn't have a use case. >> Yeah. >> When they dropped the iPod, they were not the first music player. You could buy any number of MP3 players out there. That was not the point. The bigger deal was the iTunes store. That was the actual thing. The device to play it on was whatever.
47:18 You had to have something that people want for it. So the Vision Pro hasn't sold that many because there's just not that much to do with it. Man, you come out and do that — you can start naming the things that they could do: IMAX films at home, partner with IMAX. You could do Live Nation for concerts, front-row seats at concerts, courtside seats at the NBA, on the glass at the NHL. It's just crazy what they could do with this, and the fact that they haven't means they either can't or they're just missing this.
47:56 >> I think there's also a lot of hardware-related improvements they need to make. It's a little bulky, and they need to improve the storage and memory of it, power. So there's a few things they've got to go through. But to your point, it is something that, after hopefully years of development both on the content side and the hardware side, they'll be able to come out with — hardware plus a subscription to all these different things.
48:23 It would be an amazing device. >> Yeah. Anyway, I didn't mean to get sidetracked on it. Not a topic I've been talking about recently, but I had demoed one a couple months ago. >> You've seen the light. >> I just saw, once they had the NBA footage in there, concert footage in there — people will pay a lot of money for this type of thing.
48:40 Anyway. All right, I think that's — I had maybe one other topic, but we can either touch on it or save it for later. We'll save it for another week. I think we're going to try to have JT Taylor on next week. For those who don't know him, he's the Washington policy analyst at Hedgeye.
48:59 That'll be a fun one. We definitely wanted to get him on before the midterms anyway, and maybe we'll have him on more than once before the midterms. But maybe we'll talk about — I was going to say, do deficits matter, with the 10-year here at 5%?
49:17 Now granted, I think we decided that was a one-variable model that's really more related to oil at the moment, but it does raise a question — I brought this up in my Early Look this morning — a 6% deficit as far as the eye can see, assuming no slowdown, without any recession, without any exogenous COVID-like event, really is kind of eyebrow-raising when you think about that growing pie of interest payments as part of the budget. And it's every gold bug's dream, right, when you start looking at those numbers. Everyone who's been bullish gold for decades — that's the stuff that they worry about: that there's just no way out of the math on some of these things if you don't make a change.
50:07 And as I think we learned, hiring some 20-something under the nickname Big Balls to cut the deficit doesn't quite do it. So we're going to have to come up with another solution. [laughter] I don't know. >> Good point. We should definitely have a bigger conversation, because we'll see what happens the next two years.
50:29 As I say, if the Dems win the House, the purse is essentially closed. I doubt much will get done the next two years, other than emergency things like maybe defense. I suspect you won't see a lot of spending. But there's also — related to what we were talking about earlier — a competition for credit with all these massive credit requirements from this
51:08 AI boom, not just in the US; it's going to move globally. >> Yeah. >> You, buyers of credit, are going to compare — whether it's right or not — sovereign credit for a developed country versus credit from Amazon or Google. >> Right, right. >> And it's going to be an interesting question in those investment committee meetings: which one is riskier? Is it going to be Amazon, or is it going to be — >> That's a great — have you seen the Social Reckoning preview? >> Yeah, yeah.
51:45 >> There's the ominous line that Aaron Sorkin's already written into the script, because there's going to be plenty of good lines — he's a zippy writer — and it's like, we're post-government around here, some line like that. And I was just like, yeah, if you want to believe in one of those dystopian futures where these corporate technology companies are the new nation-states, well, that'll be the indicator: when they can borrow at better rates.
52:08 >> Right, right. I'd love to — I don't know what it is right now — but Johnson & Johnson's credit versus the US government, I'd love to know how close those are.
52:36 >> They trade at a spread, but those spreads have been relatively tight, right? For AAA-rated credits in the corporate space the spread's pretty tight. I'd have to look at where it is exactly right now. But — >> So I think that's going to be an issue that I don't think goes away in the next two years. I think there's going to be a moment at the end of '28, particularly with the presidential election — whatever the outcome is, that could change the equation for all these spenders. They're going to want to front-load whatever spending they can, albeit with all the issues they're dealing with — as you said earlier, regulatory, local, federal, power, electricians, all those variables combined, those rate-limiting steps — they're going to try to front-load
53:35 as much as they can before the end of '28, because whatever happens after that, nobody knows. >> Well, and what's going to be great for them is they'll end up owning a bunch of this capacity that's been built on spec by lower-credit-quality players, right?
53:53 >> These neoclouds that are out there. >> Yeah. We saw this with the internet build-out. You saw these independent ones that came out and then they went through bankruptcy or close to it, or ended up on the pink sheets, and then got bought out. AT&T bought them, Verizon bought them. >> Texas has done a really good thing, I think: they had multiple gigawatts of requests for power versus what Texas has in terms of power
54:24 generation. And they said, okay, this is ridiculous. So now they're parsing out this massive backlog of power requests: okay, who is legit, and whoever is legit and further along the process of getting plugged in, they're putting them in different categories within "batch zero."
54:47 They call it batch zero, but they're categorizing these demand requests: legitimate players with real financing, that are through most of the local approval process, with backup power, everything lined up. They're now categorizing who's in the front of the line and who's in the back of the line.
55:08 And I think what you're seeing is that there is a very long tail in the back of the line of mom-and-pop cowboy developers that are never going to see the light of day, because they don't — >> I think I had sent you that chart from Bernstein that showed how much of the incremental announcements were all from fly-by-night spec builders.
55:30 And so anyone who has permitting — even if you can't get [power] — you'll be in a position where they're buying up permits. Amazon or somebody will step in and just buy — do you have a permit? All right, we'll just take that project off your hands, because we'll do it down the road. We're not in a hurry.
55:46 We didn't borrow money to put it in the ground. So we're just going to sit on this and do it later. There'll be a buyer's market for some of that stuff at some point. >> And it's the same with contractors too: suppliers, contractors, they're going to prioritize the guys that are coming in with real capital, a real name brand, real financing. Real everything is going to get preferential treatment in terms of contracts and supply. Some guy that walks off the
56:18 street says, "Hey, I'm building a 50-megawatt data center down the road, I need whatever" — okay, show me what you've got. Amazon, Blackstone, Brookfield, KKR, all these guys are going to gobble up that supply and contract capacity. So you're not going to be able to do anything.
56:41 >> I think that's right. All right, that's a good place to close. We're up on the hour. This is awesome, Sam. As usual, when we do something like this, I think, are we going to have enough to talk about? Maybe we'll make it a short episode. And here we are. We could have probably gone on for another half hour.
56:54 So I love it. These will be some Free-Form Fridays that we do in the future. So hopefully you enjoyed that if you were listening. Tune in next week for another episode; as I previewed, I think we're going to try to have JT on. We'll get some politics from him and how the midterms are shaping up.
57:11 And otherwise, in the meantime, you can follow me on X at Hedgeye_Hells. Sam is Hedgeye_HGrow. If that was too hard to follow, you can find all our bios and other information on the HAM website, hedgeyeam.com. Otherwise, have a great weekend and best of luck out there.