The AI CAPEX Illusion with Thomas Chua

0:00There is a lot going on right now with AI cloud, all of these things, and it's easy to get excited, but sometimes we need to remember what the basics are and remember why we're doing this. And remember that steady compounding is the way to sustainable wealth. And so we're going to have a great conversation for you. Go ahead and buckle in and get ready for a good show. Okay, so it's time for some real talk. I have a serious problem with shoes, like legitimate, like my wife has opinions about it type of a problem. So when I find a pair of shoes that I absolutely love and they're$300 or$400, I don't just buy them outright.

0:35I always try to find them cheaper first, you know, to keep my wife happy. That's exactly what dupe.com is for. It's an AI-powered shopping tool that finds cheaper alternatives to the expensive stuff that we want to buy. Not knockoffs. They're not counterfeits. They're the same manufacturers, just different branding and way lower prices. Let's be honest. The white label game is real and dupe is blowing it out of the water and their brand new research for me tool is next level. Just describe what you're looking for. Type something like running shoes for trail running under$100 or workout gear that doesn't fall apart after three washes and it pulls from real sources, cuts out all that sponsored garbage and just tells you what to buy and why.

1:18Straight answers. Done. Be prepared to save yourself a ton of time and money. Just go to dupe.com, that's D-U-P-E dot com, and tell it what you're looking to buy. That's D-U-P-E dot com to finally feel confident about what to buy. Support comes from Wise, the smart way to manage the currencies you need around the globe. Fed up with losing out to hidden fees when you send money abroad with your everyday bank? Choose the smart way, Wise. You can count on the exchange rate you'd usually find on Google. No unwelcome surprises. Plus, ditch that where's my money feeling. Most transfers arrive in under 20 seconds.

1:58Join millions saving billions on hidden fees. Be smart. Get wise. Download the Wise app today. T's and C's apply. You're tuned in to the Investing for Beginners podcast. Investing for Beginners podcast. The show for the long-term investor. We cut through the noise to focus on what works. Compounding, discipline, and the conviction to buy wonderful businesses and stick with them. Your path to financial freedom. Start now. Welcome to the Investing for Beginners podcast. I am your host for today, Andrew Sather, and I have Thomas Chua from Steady Compounding with us. He is returning after an almost two-year absence.

2:47So clap, clap, clap. It is so great to have you back, Thomas. Thank you for joining us. And we've got a lot of interesting things that have just happened. We've got Amazon just reported crazy numbers. Google Alphabet reported crazy numbers. Meta reported some numbers and the market didn't like it. So my first question, and this is actually why I emailed you, because you had a great email talking about Alphabet's annual report. So can you start with what you saw in there, or it might have even been a quarterly filing, but what you saw in there that is interesting to you in the context of everything that's going on with the data centers and all that?

3:34Thomas Chua:yeah thanks for having me andrew looking forward to having this conversation yeah so i thought like the newsletter that i just put out on google's quarter two that part was interesting because usually every single quarter google will go into contractual agreements with a couple content houses and all that whatnot but this quarter there was suddenly a spike up right usually it's a couple billion dollars this quarter is back up to 700 over billion dollars and these are contracts um that are you know deliverable within over the next few decades really really long-term contracts and these are largely for you know um electricity for data centers and all that whatnot and it comes up to over 700 billion dollars um they can't really get out of it unless they pay a heavy penalty for it.

4:26Thomas Chua:So I thought like, you know, while we always fixate on the capex number, which is frequently subjected to change, there are something that is less likely to change. And these kind of numbers usually hide in the footnote itself. So in note 10 of Alphabet Q2, you'll be able to find, you know, what exactly they go into. And you know, they are being guarantors for a lot of these companies also when it comes to building out all this power. Because as we all know, when it comes to this AI race, it's not just the chips, but I think the main bottleneck right now for a lot of these huge companies is actually power.

5:02Thomas Chua:They're not getting enough power to fire up a lot of these data centers right now. And if we were to look at, you know, they have a lot of, there's this item called construction in progress, like CIP, whether it's meta, whether it's alphabet or any of the hyperscalers, you'll be able to see this line item, which is stuck on the balance sheet right now. So it just means that they have the raw ingredients to create all these data centers. But right now, it's either delayed by construction or it's delayed by the lack of power. They are not able to do so. And what happens when these items are still in this line item called CIP is that it's not subjected to depreciation, which is why we still see margins increasing tremendously for a lot of these hyperscalers because a lot of whatever they've bought, whatever inventory, whatever not inventory, whatever assets they have bought is still not subjected to depreciation yet.

5:58And so is there kind of like a double whammy too? Because if not enough data centers are being built, then the price of compute goes higher.

6:07Thomas Chua:Yeah, so I think for the price of compute, right, they are all racing right now to deliver as much compute to the market as much as possible. because there's this severe shortage right now. And I think the most telling one is not just from Google. It's like all their bad logs is increasing at an insane rate. But even for Meta itself, you know, the CFO Susan Lee of Meta just recently came out to say that, you know, they are receiving crazy offers to buy over their compute. And the offers, she gave a hint, right? It is multiples of what they paid for. so there's this insatiable demand for compute at the moment i thought this was interesting because meta predominantly bought all this compute for its own use right for two users actually one is for the meta super intelligence labs um the one that's hated by alexander wang and the other one is to power up its um its core businesses right so we are able to see a lot of returns being generated at the side of the business.

7:13Thomas Chua:Whether it's using LOM to deliver ads, to read all the content, to deliver more relevant content to users and really see the user's time on it. Whether it's Facebook, whether it's Instagram, really, really go up this quarter. Yeah, so I'm going to ask the dumb question because I've been pretty AI skeptic for a while um i told you off the air i'm i'm a i'm a shareholder of alphabet i have asml i have tsmc like i'm i'm all big on technology but i've for a while been very skeptical about ai where is and the thing that kind of like opened my eyes a little bit is like wow these backlog numbers are getting absolutely insane um and you mentioned like 700 billion that they're signing enough for and then the backlog that alphabet has is also hundreds of billions and increasing by hundreds of billions so like what the dumb question i have for you is like what is driving all this compute demand because it can't just be open ai yeah so i think um one i mean of course the big players i saw a report out there so like google's uh most recent growth for its um gcp business is google cloud business is like over 80 but if we were to take out entropic it would be roughly around 60 so obviously yes um you know big players like entropic and open ai are driving a lot of demand for this compute but even within like um you know all this growth it is coming from a lot of other aspects of their business now because if we were to look at um all sorts of business whether it's tech or non-tech, a lot of them are trying to figure out how to integrate AI into their daily processes.

9:09Thomas Chua:And even for myself as a consumer, right now, if I were to book an appointment, you know, with my dentist or whatever, a lot of it is happening on WhatsApp now. And it is via a chat interface, which I thought is pretty powerful. Like I can just say whatever treatment I want, whether it's just cleaning or something, and then it will propose to me the kind of dates that are available. And then, you know, I just chat with it and then it will just fix an appointment up for me like this. Or if I would need technical help when it comes to booking my flights or anything, I'm not going to name the airline, but if I were to, in the past, if I were to use all this chatbot, it would frustrate the hell out of me.

9:51Thomas Chua:But right now I'm realizing that these chatbots actually can solve the problem faster than a human could. I don't need to wait. And they are actually able to, good enough to execute on whatever problems that I'm facing. So we definitely are seeing returns, especially on the consumer-facing front. And when I read the transcript of a lot of these consumer-facing companies, it is helping them to bring down the cost of their consumer-facing, the help there's that support this and it's also increasing satisfaction and utility actually right so right so if there's any chance whereby you can go and try like um any of this ai chatbot is it's very very different from the experience just one or two years ago and you know there's just so many companies out there trying to figure out how to deploy um ai into their daily process and i I thought the one by Meta is one of the most innovative one I've seen.

10:49Thomas Chua:Because in the past, when Meta do recommendation for their reels for you, it's largely based on the caption the creator writes and maybe hashtags and how they tag their content. And they are already able to do a very good job, right? Based on what we are able to see. Right now, the LLM is reading through every single one of the content. So that will be, I assume there will be a transcript at the back and then they'll understand what the content is about and is able to better allocate this content to the users. And likewise, they're deploying it at their advertising site also. So there are a lot of use cases, new use cases that's happening right now outside of whatever the Frontier Labs is generating the demand for.

11:33Thomas Chua:So it is not just the Frontier Labs that's driving up the demand right now, but I think across all businesses as a whole, they are finding out the benefits of AI and deploying to their business whether it is to give them an edge over their competitors or to drive down costs but I think AI overall is making businesses way more efficient than before That's super fascinating Do you think, because if we look at like even if you're an S &P 500 investor and we look at all the big companies the S &P, NVIDIA, Apple, Alphabet So you're in the AI trade, whether you like it or not. Should investors be worried when you say like companies are trying it out?

12:21Should investors be worried that maybe they're just going to try it out and then decide, oh, we don't need to invest in this anymore?

12:30Thomas Chua:So, I mean, I thought this is an interesting one because like there are companies who are out there who previously budgeted huge amount, you know, for all these AI projects, but they subsequently, they scale back. Uber is one of them. You know, they busted their entire years of AI expenditure budget in just, I think, a couple of months. And the CEO started to reel back. And then, you know, we have to rethink how we actually want to deploy this budget. So I think the amount they are going to spend on AI is going to remain consistent or rising. but what's going to happen is that previously a lot of people were using the best model to run every single thing and it is very costly so if you were to look at models like Fable Opus, Sonnet or even for ChatGPT's model or Kimi's model they all take up different costs to run so everything Entropic is generally the more expensive one so these Frontier Labs will have to figure out how to quickly bring down the amount of compute for every single from that users generate or every single if agentic is going to happen right they will have to solve this problem and so far if you were to look at the older models we can see that the costs are actually coming down extremely quickly and with every iteration of the new gpus by nvidia the cost of generating all this compute is actually coming down pretty quickly.

14:08Thomas Chua:But the companies that are burning the most capital right now, I don't think it is all the end users. Like you and me, we probably use a lot of AI. But frankly speaking, it is heavily subsidized by VC money right now. So for me, I use Cloud, I use Open, I use TARGPT, I use everything. But the amount of value I'm getting is way larger than whatever... it takes to provide me with all these tokens or answers. That's interesting. Are there winners and losers starting to kind of like, are there some companies in this space that you think are more risky? It seems like the other thing that gets me excited about what's happening right now is Wall Street seems really, they're just not as hyped as you would think on like alphabet you know they punish meta after the the earnings and and they're so hyper focused on this capex um but it seems like some companies have the capex where maybe they might be doing it like like wall street maybe shouldn't paint with such a broad brush maybe some companies are allocating the capital and it's going for good causes and others are being a little bit more wasteful with the capital.

15:31Are there companies or categories of the AI space that you feel like are maybe a little more risky and then there are some that are a little bit more safer?

15:44Thomas Chua:Yeah, I thought you raised a very interesting observation because in the most latest earnings calls, we see the market actually reward Microsoft and Amazon after the earnings and then it punished Meta severely after the earnings. And I think a lot got to do with the way the CEO is communicating, right? Because when it comes to anti-Jesse of Amazon or Satya Nadella of Microsoft, both of them were very, very clear in communicating about the economics of their cloud business, of providing and selling all this compute. And, you know, when they are able to provide this kind of clarity and the market has less confusion, generally the market would reward these stocks.

16:26Thomas Chua:But what's happening with Meta right now is I think they are doing very well, right? The ad business grew 28%. Impression growth, engagement growth, both still double digit. Despite being such a mature business, I think they are absolutely crushing it on their core business. but what is less certain is the amount of compute they are spending on their super intelligence lab or what they call MSL, meta super intelligence lab. That's where it's a little bit of a question mark. And I mean, I understand why it's a question mark because they are still in that initial, I wouldn't say initial investing phase, the heavy investing phase, whereby he spent a lot to bring in this team and also to provide them with a lot of compute to try and figure out what the frontier, to try and develop a world-class model as well.

17:18Thomas Chua:And I can see why he wants to do that because, you know, Zuck, I think he has always been scarred by other people being a layer on top of him. So what I mean by this is, you know, throughout 2022, when you see Apple wrote the iOS 14 update, which severely impeded Meta's business by reducing the amount of targeting they are able to do, That's where Meta was held hostage for a while. So I think Zuck, he don't want to rely fully on OpenAI or Entropic or even Google for that matter when it comes to using all these models. And, you know, these models are actually generating very, very real returns for their core business right now.

17:59Thomas Chua:We start to see their advertising revenue as re-accelerating up to above 21%, right? In what Zuck was, what he said is that, you know, they are the strongest advertising player with the strongest growth right now. And that part is true. Yeah. But I felt like he could have done a better job when it comes to communicating what kind of returns or what are the key deliverables we can expect out of all this compute spend. Because there's a huge opportunity cost right now. And that's largely because there are a lot of people who are willing to pay a lot for Meta's compute, right? So in the very brief note is that they could have a chance to become a hyperscaler, but Zark feels that if we were to go through that route, we will be thinking short term.

18:45Thomas Chua:Yes, we will make a lot of money in the short run, but over the longer run, what we want to do is we want to make sure that Meta remains a vertically integrated company, meaning they run on their own models, they have their own data centers, and they want to prioritize the core business. and I kind of think that that's the right move, just very, very poorly communicated. Another interesting one you raised just now is that, you know, which are the companies that are more risky? Now, when it comes to the biggest hyperscalers, right, whether it's Amazon, Microsoft, you know, Alphabet, or even Meta, who's not a hyperscaler, but they are spending a lot of capex, financially, they are still very, very sound, especially for companies like Meta, Microsoft, or alphabet, they do have their core business as well, which can absorb this compute in the event, you know, the demand, external demand for their compute isn't as high as what people think or what they expect forecast it to be.

19:48Thomas Chua:So, for example, if GCP suddenly run into a speed bump because for whatever reason, demand for compute decline, they do have their core business, whether it's the Google search, um ai overview gemini or youtube that's able to accept uh absorb all this excess compute but for other um compute providers especially the new clouds i think those are at a significantly riskier position because their core business is just purely selling compute and if demand were to falter and these new clouds they actually do have agreements with companies like microsoft whereby Microsoft actually buy compute from all these new clouds and then they resell to their big clients in the S &P 500 at a very, very big markup.

20:36Thomas Chua:So I thought that's very, very smart of Microsoft because they don't want to assume that much risk, but yet they are still able to enjoy this very, very strong compute demand. So there's this markup because companies are more willing to pay Microsoft for the Microsoft brand, right? For the Microsoft name because it signifies reliability. whereas for the new cloud if demand were to falter I would think that Microsoft would just drop their contracts with the new cloud or Alphabet they also went into this contract with SpaceX to buy compute from them that also can be dropped after a certain period so if there's any speed bumps along the way I think the biggest hyperscalers whether it's Alphabet or Microsoft they do have levers they can pull to drop off all this excess compute or to absorb this excess compute through their core businesses.

21:31Thomas Chua:But for NeoClouds or maybe Oracle, these companies are leveled up to their eyeballs. They took on an insane amount of debt to really try and soak up all this demand for compute. So if there were speed bumps, they do not exactly have a core business. All these NeoClouds, they don't have a core business like YouTube or Search or anything that can absorb all this excess compute. Yet the thing that is going to stay constant is their interest payment. They will still have to pay down their debt. So I think when it comes to this whole AI industry, I think they are at the most risk at the moment. And then you mentioned you own TSMC.

22:14Thomas Chua:I think that is a company that is just very well run at the moment because they are not overly aggressive. and I like what the management is actually doing. He's not doing this price gouging, whereby you see some players within the chip industry are doing, they're still making it consistent, not charging their customers a whole lot more just because there's extremely strong demand. And management has always been very clear that they are only going to invest heavily if they have very, very firm demand, meaning companies are willing to put up the money. for it and establish onto all these long-term contracts yeah so i mean the ai industry or a lot of players a lot of stakeholders but there i would think that there are some that are you know quote unquote more well behaved than the other and the riskier one i i think you know investors should exercise caution august is national wellness month but most health trends equal things like buying random gadgets and guessing at what actually works based on whatever's trendy at the time and i wanted to stop guessing at things like that and actually look at the data behind my body.

23:22I've mentioned it before, but lately I've been taking time in the gym much more seriously, not just to build a bunch of aesthetic muscles, but to build a good, sustainable, long-term health plan for my future. Your daily resilience leaves a clear data trail in your body and function tracks the exact markers behind your energy and immunity. Not a generic overview. They look at core biomarkers like white blood cell count, which maps your frontline defense against invaders, HSCRP, which catches hidden energy draining inflammation, vitamin D and zinc, which are essential immune anchors and commonly low in a lot of people.

23:55Plus, there are secondary metrics that they cross-reference, things like ferritin, which is iron levels behind your energy, MMA, which is your active B12 for energy and nerves, and cortisol, which is how stress is actually impacting your body. I use Function and you should too. Check your health the way I do. Function provides 160 plus lab tests for$1 a day and member pricing on advanced imaging. Join at functionhealth.com slash beginners and use gift code beginners25. Bitcoin is one of those really divisive topics. And depending on where you stand on it, either you ignore it or you can see its future utility and the things it powers.

24:29And so in a time like today, we're all wondering how much exposure makes sense and how can I get that exposure in a simple and easy way. Where you aren't subject to countless fees, endless passwords, and constant micromanaging with too much time spent on websites and apps. When you're buying Bitcoin over time, the small costs add up. That's why I like that Cash App offers zero fees and zero spread on automatic Bitcoin purchases like auto-invest, roundups, and paid in Bitcoin. So if you're building a position over time, more of each purchase goes into Bitcoin. It's a straightforward way to add Bitcoin exposure without overcomplicating the process.

25:05For a limited time, new customers can get$10 added to their balance. Just use code BITCOIN10 when you sign up. And don't forget this part. send at least$5 to a friend in the first two weeks. Terms apply. Bitcoin services by Block Inc. See the Bitcoin disclosures at cash.app slash legal slash podcast. What's the best way to get started in the market? Download my ebook for free at stockmarketpdf.com. Yeah, super good points. I agree with you on those too. Let's say an investor feels overwhelmed about, man, there's like all these companies in this industry and all these things to learn that I just feel like I don't have the time for.

25:47You know, maybe they have a full-time job, hobbies, and all of that. What kind of advice would you give to somebody who wants to learn about maybe, you know, great companies like the hyperscalers, but feels like they don't have a lot of time in their day or in their week?

26:06Thomas Chua:Yeah, I thought that's a great one because I spent the past two years writing about this. So like my book, The Lunch Break Investor is going to come out on August 18. And it's exactly targeted at people who have a nine to five job. So when we are extremely busy, when we have a nine to five, I think the way we look at companies need to be extremely, very highly filtered, meaning like we shouldn't be looking at companies without a mode. We shouldn't be looking at companies that are way too complicated, too complicated for us to understand. so in this book i'll be talking about you know like how we can actually filter businesses that we should be interested in the first place like for example i always hear a lot of people talking about memory chip company like sk hynix or micron so and so forth but they have no idea no clue about what these businesses done does and i think for anybody with a nine to five the companies we should be looking at first are companies that we are familiar with whether we are consumer or whether our job involves interacting with them um that's where we should start and you know one of the key things um that i always tell investors is that we must be able to answer this question when it comes to investing right um why does this business deserve to be bigger five to ten years from today if we are not able to answer this question is either we don't understand this business enough or this industry is so rapidly changing that we cannot um you know have any form of reliability when it comes to forecasting this company's future then i think those are not great candidates for us to actually delve into for um you know additional researching right so when it comes to investing i always tell my readers that you know we are not being rewarded for style points a lot of times the boring companies, companies that are all around us, maybe Netflix even.

28:07Thomas Chua:These are companies that are extremely easy to understand. And it's easy to see why these companies would have an economic mode over its competition, just from a consumer point of view. And I think that's where we should start researching and looking for a lot of these companies. I love that you brought up Netflix. a great example, which by the way, we may own securities. A lot of the securities is discussed in this episode. I just added some Netflix, but, um, you know, what about to the investor who looks at Netflix is like, yeah, I understand Netflix. I watch Netflix, but they look at the stock.

28:41It's down 40 % in such a short time period. And maybe that spooks some people. How do, how do, how do we approach a stock that's been beaten up like that? which can kind of be its own skill set outside of like investing in other companies.

29:01Thomas Chua:Yeah, so I think you touched on a very good point because when it comes to massive drawdowns like this, the first thing we want to identify is whether, you know, is the business going through a circular decline or just a speed bump, right? If it's a speed bump, generally, if the valuation makes sense, it's always a great time to buy. But if it's going through a circular decline, you know, I mean, the classic example is always a lot of people like to talk about Kodak. Right. That's one of them. Or Xerox. Right. Or even the newspapers companies going through a circular decline because they are either facing disruption or consumer tastes changing.

29:43Thomas Chua:Right. So when it comes to Netflix, we'll have to answer this question. Is the world moving away from streaming to another form of entertainment? or is competition offering more value than Netflix at the moment? And then, you know, for a lot of this question, not exactly, right? The world is still shifting from cable TV to streaming. I think that's the preferred mode of consuming a lot of this entertainment. And when it comes to all the streaming companies, the entertainment dollar value provided over the hours, I think Netflix is still the strongest proposition. And, you know, the key thing about Netflix's mode is that it is the biggest player right now.

30:30Thomas Chua:If you look at a lot of its competition, right, they are not nowhere near as profitable as Netflix. Yet Netflix is spending the most amount on content at the moment. And they are able to do that, generating the most value for its subscribers because they have this huge pool of subscribers that they can divide all this content production costs over, which really, really juice up the economics of Netflix. And I think that is Netflix's strength. And I wrote about this since exactly a year back, actually. I think the challenge for Netflix back then is that the valuation was just extremely high. It swung from a low in 2022 all the way to a record high last year.

31:18Thomas Chua:and the valuation was becoming very, very demanding, right? And as the share price draw down 40%, I see a lot of people try to put a lot of narrative around it, but that's just one of the things we investors need to take note of, that narrative always follows price movement. When a share price is going up, a lot of times the Wall Street or mainstream, if you were to go on X or anything like that, people will be justifying why it deserves this insane valuation and vice versa right when when it goes down people are saying oh you know it was trying to buy one of brothers because it cannot come up with great content again and so on and i think that's a very dangerous slippery slope investors could get into if we were to allow ourselves to go down that route so one of the tips that you know i i always tell people is that whenever we invest in companies we always want to write down our original thesis and also pre-mortem.