Analyzing AI Limitations and Stock Earnings – Amazon vs. Google, Uber, and More

All right, let's get going. Thanks for being with us today. I'm excited to talk to you guys about some really interesting earnings reports this week, interesting news in general, and I'm going to jump into it here. I got a few things I want to talk about in particular. I'm going to Create a portfolio here for us today. Earnings watch 2626. Drop the tickers in here. I'm showing you our institutional membership here, so you're going to see a lot of buttons and bells and whistles you won't see in other memberships. These are the stocks we're going to talk about today. Before I jump into that though, I am interested in showing you guys. Something that was really pretty fascinating to me that I came across this week. And it's a guy from a firm called Zenista. I just met him this week. You can see down at the bottom, Armen Khalidian, Zenista AI, does some cool stuff. But really in particular, what I loved about what I saw Armen do is really talk about AI, how it works and what does and doesn't really work. And I think he has the ability here to really shine some light on what's real and not real about AI. And this chart, first of all, just jumped off the page at me. And I love how it points out here that, look, at menial tasks, AI performance is pretty perfect, does pretty good stuff here. But as task complexity increases, you're seeing a pretty Big drop-off. And, and by the way, perfect execution is for only the simplest of tasks today. This blue line really being current AI performance, the orange line being AI potential, and this orange wall here being we don't know if AI ever really gets past that. So these lines really probably just don't exist here. And I think this is great. This is an awesome depiction. And when I was talking with Armin about this, he pointed out, he said, listen, David, We don't have the ability, we don't have the data to give to machines to perform these really complex tasks because the few human experts that can do these things, it's in their head. Even if they've written a book on it, people aren't able to pick up the book and do what they do. Can you throw passes like Tom Brady? Well, he could tell you how to do it all day long. That doesn't mean you can do it. And it brings up another point that I think is important, and that is we underestimate how difficult it is to teach machines to do things. We think, oh my God, we just figured out AI is going to make it easy for us to teach machines to do things. We've been building robots for decades that can't do what humans do. Can barely fold laundry. Yeah, the Boston Dynamics robots can do a lot of cool things. They can do parkour. They've really gotten good at the agility stuff only. The mental stuff, we've been trying to do this for a long time. Absolutely, technology is better. Things are moving faster. And we're making great progress. But I wanna take another chart here that is my new favorite chart from Armin, which I think is brilliant. Guys, good in so many ways. So I, I like here on this part, it shows the number of words processed by LLMs during their training. Right? And we're pretty close to them going through all human text. Right? And then we're looking at the human brain number of connections versus the number of parameters. That the AI has. And you see it's gone up a whole lot. This is the ant brain down here. It's the human brain. And then this is where AI is. Notice the decline. And then the image of the parrot is perfect. How many of y'all have ever been around a really good parrot? Raise a hand. Put it in the chat. I wanna hear from you in the chat and in the Q&A. I see John. He's been around a really good parrot. They're amazing. It's fascinating sometimes how good they can be. But we know there's not really any real consciousness or sentience around that. And I think that's where AI is right now, right? It's able to pare back a lot of stuff that it knows to pare back based on programming, training data. Here's how you answer this question. Here's a book that can give you the answers to all these kinds of questions. And yes, there's some natural language processing that's made it better at understanding how humans speak so that it can respond to questions and do that well. But does that mean that the brain is any better than a parrot brain? Point being, not so much that AI won't get better, but to recognize where it is now and how far it has to go to really replicate a real human expert brain. And as I was saying a second ago, the one thing that we overestimate our human ability to do is to teach machines how to do really complex things, how to replicate complex human behaviors. It's a different language. We're not talking about teaching other humans that have the benefit of sentience. Like babies, just, they grow up, they just learn stuff. Like, they just absorb it. Human machines? No, it has to be programmed specifically. It's a different interface. It's a different kind of mind. We've been talking about that for 20 years at New Constructs. I've always said, listen, we're gonna teach our machines to parse like we do. We have to endow them with all the context that we bring to going through a filing. It's all kinds of natural things that expert humans just take for granted when they're going through the filing, like their knowledge of what the income statement looks like before they go to the balance sheet, their knowledge of what the income statement and balance sheet look like before they go to the cash flow statement, their knowledge of what all 3 of those look like before they go to footnotes, the knowledge of what to expect when you get into page 250 of the filing versus page 2. The knowledge of what to expect when you see certain phrases or certain footnote headers. All these things is context altogether that's cumulative, not just within the filing, but over time. That's extremely difficult to teach. Pattern recognition, right? The term I like to use for this is what I call the virtuoso effect. Stroke of genius. We see these things all the time through society. Right? Stephen Hawking, Mozart, Elon Musk. We gonna get machines to do these things? We don't need machines to be super genius creative, but that's what people are saying. They don't need to be super geniuses to do a lot more things that'll be useful to us. And I think they'll get to, but I wanna point out here that there is a big gap between— Armin says, mind the gap, understanding the limitations of, of AI. And we at New Constructs have a particularly I think unique knowledge of this because of what we have been able to teach a machine and what we've been able to create in terms of an AI that can really answer questions. And let me tell you, there is no trick to teaching a machine how to know how to pick stocks. It took a lot of really difficult and complex thinking about, okay, what's the ontology that we're going to use to pick stocks? The ontology Is this rating system, right? We're going to say these are the things we need to measure. This is how we're going to put meaning around the data. Then I had to create a taxonomy that would organize all the data and filings into very specific categories that would contribute to each element of the ontology. Then we got to go collect it. We got to collect it faithfully, perfectly, right? Data's got to go in the right spot because if it doesn't, Then it's going to be assumed to mean something it's not supposed to mean. And that screws the whole thing up. All this has to be done exactly right. And then over time, so well that we can get a machine to do it on its own. That's super powerful. And there was no way to teach the machine to do that except for us to do it as humans so many times that we had so many scenarios that we can say, machine, just do what the human's done. As of yet, even our machine, has not parsed and interpreted a footnote that a human has not already interpreted. Now we've done a ton of them, right? We've got probably 250,000 filings that human experts have perfectly marked up. That is our raw training data. And then we've got millions more where the machine has said, ah, I broke on this. How do you help me fix it? Boom. Fix that one. That's how we've operated is over time the machines do more and more. And the humans are able to focus more and more on the exceptions. There's a real turning point there when all of a sudden the machines are doing more and we as humans are just focusing on a narrow set of exceptions, a more and more sophisticated set of disclosures. And we basically just keep checking off the scenarios that we've taught the machine to handle. And that's our AI, and it's one of the best in the world. But is it an AI that's sentient enough to figure something out that we haven't already told it? No, that doesn't exist as far as I know. I don't think it will exist for a long time. Now, will one of the places where it maybe gets it right be in financial statements? Because you've got somewhat limited range of results. Yeah, limited. I don't know. We've been doing this for a long time and we have a joke like, never say never. Anything that can happen at filing has pretty much happened. Balance sheets don't add up. Income statements don't add up. Like gross stuff. Entire disclosures just missing. Clyde, disclosures off by a factor of a billion or a thousand and just wrong data errors. It happens. Nobody at the SEC is reviewing these things at scale. I know because I try to give them tools to do that. We're a long way off from something that's going to do all this work for us. Okay. Let's take a look at our earnings reports this week. So some pretty interesting stuff. I think it's super interesting. Let's start at the bottom here with Amazon. All right, Amazon. What's happened to Amazon this week? Let's go to see. Here's FinSites. We're going to switch off FinSites as good news and bad news on FinSites. FinSites is about to be going public right now. It doesn't work. We got enormous amount of plumbing to go through the Google system to get it from the development environment where they built it in and out through to the surface. It's pretty interesting. There is a particular firm that was doing nothing. As a consultant to help us with that. They're great, and they're getting paid a lot of money to do nothing but take it from the development environment to the production environment. So that's pretty interesting. All right, I got my watch list down here. Earnings watch today. Okay, Amazon down 7% today, down 13%, down 14% here recently. You know, those of you who follow our research, you know that we put Amazon down as one of the losers in the AI race. And we— not long after, we saw them start to cut thousands of jobs because they got to get their cash flow right, folks. People don't realize Amazon's cash flow, it's been negative for a long time. Let's go take a look. See a negative enterprise value, super expensive valuation, great profitability, right? Look, accounting earnings going up faster than economic earnings, but they're both going up, which is good. Why are they issuing all these shares? Why is debt going up? Why would a company super profitable need to be doing that over all these years? Because they're not profitable, my friends. Free cash flow negative. Negative $40 billion. Negative $46 billion. Negative $49 billion. Negative $13 billion. Negative $28 billion. Negative $42, $41.6 billion. And we're going to get a new 10-K here, or new 10-Q or 10-K. for the end of 2024. And I bet it's going to be probably a really big negative number because we've been talking about the spending going. So, can Amazon keep up if they haven't been able to spend or haven't been able to build up a big cash hoard like Google? What did Google come out and say this week? We're doubling down to like $175, $185 billion in CapEx. Can Amazon keep up with that? I don't think so. They're losing money before AI came on, you know, and this is what it's about. This is how companies compete. Google went out and body slammed all the other companies that it's wrestling with for the AI wrestling tournament. We got a big wrestling tournament this weekend, sectionals here in Tennessee. So figure out who gets to go to state. So the wrestling's on the mind, but the idea is the same, right? Google's, hey, I'm gonna go do $175 billion. Who can keep up? I don't think many can. So Amazon, I don't even really care what their earnings are. I don't think it honestly matters that much what they report. What did they say? $200 billion AI spending plan spooks investors. Why did it spook investors when Google said they're going to do $175 to $185? No, because they got the cash flow. Amazon doesn't have it. This is a game, folks, of Put it out there. Say it. It doesn't really matter whether or not it's what you're going to do. It's perception. I don't know how they're going to do it. Let's take another look here on Amazon too. Let's see. Got the debt. Take a look at the debt. We saw shares going up. Debt, boom, it's going up a lot. $128, $129 billion 5 years ago, $220 billion. So added nearly $100 billion in debt. Over the last 5 years and selling shares. Why are they doing that? Because they're losing money. Amazon's always lost money. It's been okay. Now they're going to have to figure out a way to fund serious spending. And that's what the— again, that's what the— this is what this competition in business is about. I'm going to go do spending that you can't keep up with to get ahead of you because I got a better business model. And Google has a better business model. All right, we got to get some other stocks. Speaking of Google, Suspended because we're getting a filing any day now. You guys have heard me talk about Google. I think it's the best positioned of all the AI stocks. It's the only company that has its own cloud and storage and compute, has its own semiconductors, has its own AI, right? Amazon doesn't have an AI. Apple doesn't have an AI. I think Amazon makes its own semis. But then fourth, Google's also got cash flow. Huge cash flow. Let's take a look at cash flow for Google. By the way, Google shares going down, but they have very little debt. You haven't even talked about a debt issuance for Google. How much cash flow has Google had? $32 billion, $49 billion, $27 billion, $54 billion, $48 billion, $31 billion positive. That's with the spending. So what I'm seeing here with Google Is that they've been spending consistently at pretty high rates. Because even with the AI spend, you don't see that much of a drop-off in cash flow. So what does that mean? That means the spending they have been doing in the past is generating great returns. And by the way, this negative cash flow thing, this is where you also see the underestimated depreciation rates that Michael Burry talks about, right? This is where it manifests right away. It's like when you are actually having to spend more to make the same amount of money. Before it emerges in an accelerated depreciation, it emerges in negative cash flow. And that's why I'm focusing on this today. Google's got it. They got it in spades. I think they had great earnings too, making tons of money, great peak. Look, this is right. What's Amazon doing? They're cutting staff and cutting jobs. Google's just growing, getting better, and hitting on all cylinders. So that kind of speaks to my point. Uber. What's up with Uber here? Not doing so well. Look, this is like a super overpriced stock, has been forever. It's been a danger zone stock forever. Look, we got in the news for these guys. I don't really know how this business model like ever really makes that much money long term. I've been saying that forever. Yeah, it's growing. They're going to do rides. Do they have that big of a moat? I don't know. The driverless cars is going to be lower cost for them, but that's going to accrue to the consumer. You think Uber's going to be able to charge the same rate for a driverless cab when other people have driverless cabs and they're going to just charge a lower price? And that price is going to go down to basically where the margins are zero. It's a commodity service, always has been, always will be, you know, unless you're in a fancy car and you pay more for the black car, the black SUV. There's some margin in those. But below that, there isn't. It's gotten profitable, which is great, and they've done a good job of that. That's impressive. I didn't think they'd do that, but the valuation's still way, way too high. And I think that's part of why you're going to see the stock continue to get crushed. I think what we're seeing with Warsh coming into office, Kevin Warsh, the new Fed chair, is like that we may not see a big drawdown in liquidity, but we're seeing, I think, a cap on liquidity, which is almost as good. Because you're really crazy when you go to Vegas and you just got endless money. But when you know you got a limit, all of a sudden you get a little smarter. Wait a second, I'm maybe not going to play roulette. Maybe I'm going to play the pass line in craps, or maybe not so much blackjack, but I'm going to go to something that gives me better odds. Right? And I think that's what we're seeing is a rotation right now in the stock market away from super expensive stocks into more cash flow generating stocks, which is why I think our strategies will continue to perform extremely well. We've done really well, as you all know. The Core Earnings Leaders Index last year was up 27% versus 18% for the S&P last year. I think this tougher environment is going to be perfectly suited for new constructs. Palantir, another good company, bad stock, just super expensive. Absolutely super expensive. Overpriced. Overpriced for sure. And again, I don't really know. I don't really even care that much about how well Did here. It's up a lot with kind of the bounce, but yeah, 24% down in the last month. Year to date, down 24%. I think they had great earnings, great earnings, and didn't matter. Stock's too expensive. Eli Lilly, they had a good week, I think. I know they did at one point. I know it was going well. Uh, yeah, today up a little bit. Flat for the month. Huge day the other day. This week, big day, down, up and down all over the place. Interesting. This is good company, not so bad stock, so not super expensive. And I think those who've been following us know that we've been bullish on Eli Lilly back when it was around, I don't know, $20, $30 a share. So it's up to $1,000. We can actually go in here and I think see our rating history. Going back a long time. So we were back here, $200 a share, 2021. If you go to the rating history CSV, this will take us back to the beginning of time. This is '06. We first got attractive in '06, $150 a share. That's our history. Cool feature on the institutional side, giving this whole download rating history. All right, Pepsi. You know, a lot of people have been predicting that Pepsi might be a huge stock this year because we're going to see maybe a fade on some of the AI and software stocks. We've so expensive and there's such a run-up and people are going to go back to cash flow generating investments and Pepsi and the consumer cyclicals are going to have a good one. Just got a new 10-K for these guys. Pepsi has been one that has been on our focus list at some point, but it went up a lot, so we took it off. So I think back here, attractive and got really great. We let it go and it's come back a bit here. So that's a neutral. Not much to say about that. Qualcomm. So Qualcomm's up today. How's it done the last 5 days? All right, mostly down. Yeah, this is, this is one of these, I would say, good company, not so good stock. So pretty expensive. Doesn't look as bad in terms of free cash flow and price-to-economic book value, but the market applied growth appreciation period, pretty strong. New 10-Q and forecast, and we're expecting they'll beat next quarter. You can see here we got a new— we've got both the equity rating, the earnings distortion score, and the credit rating in the portfolio. We also have this in the screening tools now for professional members. You can also screen on economic versus reported earnings, which is a new capability. So screen on this and screen on GAAP earnings distortion scores are pretty cool. The screeners are unbelievable, in my opinion. To screen on 3,400 stocks based on the precise detailed models that we've built is pretty amazing. All right, we're running out of time here. Let's knock out the rest of these. Qualcomm, Pfizer— how did Pfizer do this week? Yeah, better, right? This is a stock that's got a much more reasonable valuation. Quite reasonable. Doesn't look that profitable, but super cheap, right? The current stock price is implying the company's profit is going to permanently decline by 40%. We'll be getting a new filing a little bit later. Interesting. That's a long time. Maybe they reported a little bit earlier because it's a pretty big, pretty big difference between the amount of time between today and when they report. So they did well in earnings. I didn't realize there was this tab here for us to look at this on these other stocks. And maybe we can pop back through Lilly earnings, strong results. Merck had great results too. EPS and revenue beat expectations. Pepsi beat EPS and revenue estimates also. Palantir, they beat on the top and bottom. Qualcomm, yeah, record revenues. All these guys look great, my friends, they all look great, but their valuation started to matter, starting to matter at some point. Spending has got to be reached some limit. All right, how about Shell? Suspended because we're going to have to get a new filing. Attractive rating, cash flow generator, friends. Look, 10% free cash flow yield, good cheap rating. Look, energy demand is only going up. And these guys are not just in the business of exploration and production of fossil fuels. They're also in the business of liquefied natural gas as well as other alternative energy sources. They have a tremendous distribution. I just don't underestimate what these guys could do. Whatever the source of energy, I think they can definitely profit from and deliver a lot of great— a lot of energy. Shares going down, that's a good thing. They're doing good stuff for their cash flow, not wasting it. Returns on capital, decent margins, doing all right. Free cash flow, $26 billion, $29 billion, $31 billion, $30.6 billion, $43 billion, $26 billion through the 2 quarters. Companies that make money. Can create value in your portfolio. The key is to make sure you know that they're making money. Can anybody name another research firm in the world that has said consistently that Amazon doesn't make money besides New Constructs? I've never heard anyone say that. Everyone's always surprised. What? Amazon doesn't make money? I thought they made money. Earnings are so great. Look at the returns on capital. No, they don't. No one looks at the balance sheet. Can anybody put that into the chat for me? If it's true, I'll give you a free Most Attractive Stocks report. So throw that into the chat. And if we got anything in the Q&A, let me check Q&A. Looks like we got one. Oh yeah, we got one here. You guys ever offer analysis on a company's moat? Yes, Eric. The analysis of a company's moat is exactly what in this market implied growth appreciation period. It's not just analysis. This is the market's valuation of the moat. So when the moat is small according to the market, because that's what matters, right? Like, I don't really care what my thought on the moat is. Same thing I've said to you guys many times and same thing that we say to all of our new analysts when they come in for training. Would you rather be a fortune teller or a critic of a fortune teller? Fortunately for us, luckily for us, Mr. Market is our fortune teller every day for all stocks. So someone else has got to be a fortune teller. Tough job for him. Fortunately, he's got the makeup for it because he's overly exuberant about some stocks and overly depressed about other stocks. And that's where we focus. We go look and see where Mr. Market, the fortune teller, is most wacky. And then we look long-short for that. So when it comes to analyzing a moat, yes, we look at the market implied competitive advantage period or growth appreciation period. This is also referred to as expectations investing, where we don't try to predict, we quantify what the market's prediction is. Price implied expectations for cash flows, profits, and the duration of cash flow growth. That's the definition of competitive advantage. How long can you grow your profits? Because when you can grow profits, you have a competitive advantage. When you lose your competitive advantage, you're no longer able to grow profits. Not saying profits go to zero, but when you can no longer grow profits, that's the end of the growth appreciation period or competitive advantage period. We model The market's implied growth appreciation period for every single company. And then we give them a rating on it. That's what's on the page here. So yes, we analyze moats. I think we do this better than anybody in the world. Again, please put into the chat or the Q&A if you know anyone else that does reverse discounted cash flow modeling, price implied expectations, expectations investing better than New Constructs. Anywhere close to what we do. This is the best fundamental analysis in the world with the best fundamental data in the world. That's why we have indices and portfolios that outperform. We just got the latest from SumZero, the professional buy-side network. 56 months in a row, number 1 rated stock investor. Number 1, 56 months in a row. Name another firm that does that. Again, free reports. Giving out free reports, freebies. People who can name another research firm that comes close to what we do. There isn't one that I know of, but I'm happy to be proven wrong. And by the way, this AI thing, when it comes to new constructs, this is devastating for our competitors. When you want to talk about firms that have gotten blown up, let's look at some of these data firms. Let's see, good way to do this. We're out of time, but let's just take one of them. SPGI, and we can compare to MCO, we'll compare to FDS, and let's compare to LSEG, right? These are the big publicly traded competitors with respect to financial data, down 15-plus percent in the last month. They've got a commodity product. The LLMs are coming for them. Some of our other competitors, a lot of new firms out there that focus on data, they use LLMs to go get data, financial data. They don't do any analytics. They don't really do footnotes analysis. They're not into that. Why? Because it's super freaking hard. Yeah, I just pull the data, give it to you. There you go. Your job. You go to work. I'm just giving you the ingredients. You be the chef. Nobody wants to do the work to calculate return on invested capital with integrity. They'd love to have that data, but to do the work, to go through the footnotes— this is a below-average 10-K in terms of length. It's a lot of work. Maybe 25 pages between risk factors, management discussion and analysis, and financial statements. It's 175 pages of footnotes after that. That's what it takes. And you're not going to be able to teach a machine to do it very often. Alrighty. That's all I got for today. I don't see any more questions. Thank you for the comments in the chat. I hope you guys enjoyed this one. And I think we probably got one more of these next week before earnings are over. Maybe we got a couple. We're also probably going to be doing a special AI podcast. And we are, we're also in our next monthly podcast, February 20th, we're going to have a special guest. You guys have asked me a lot about our index and things that we do in terms of our money management, asset management stuff. And we're going to have a special guest on that's going to talk just about that. Founding investor is a clue. So anyway, thank you. Have a great weekend. Happy Friday. And I appreciate you being with me today.