TIVP098 (Video): Alphabet (GOOGL): The Megacap That Still Might Be Underrated w/ Kyle Grieve & Shawn O’Malley

24:43Find tickets in the full agenda at theintrinsicvalueconference.com. That's theintrinsicvalueconference.com. And if you'd rather join us as a member and get the conference plus the full weekend included, apply to the Intrinsic Value Mastermind at theinvestorspodcast.com slash mastermind dash application. All the links are in the show notes below. Hope to see you in New York. How many times has your inbox followed you home? You close the laptop, sit down to dinner, and there's still that one email you've been putting off the thread you'd have to scroll back three weeks to even make sense of. That is exactly what I use superhuman mail for.

25:24It is the most productive email app ever made. And it's part of the superhuman productivity suite, which works right inside the tools and sites you already use. So the AI comes to you. You're not opening another tab and re-explaining yourself. Two things I lean on constantly, ask AI instead of digging through a long thread. I just ask and the details I need come back in seconds and then write with voice on the mobile app. I say what I mean out loud and it comes back as a polish strap that actually sounds like me. I can clear a reply before I've left the driveway without getting pulled back into work mode.

26:00Less friction, better routine, more of your evening back. Get out of your inbox and back to what matters. Sign up and get one month of superhuman mail free with my link superhuman.com slash T-I-V-P. Yeah, the reality is that all types of searches are not equally valuable. So some questions have lots of commercial value, like maybe asking for the best Italian restaurants in your city or for product reviews of sneakers. But other searches really are not monetizable from an advertising perspective. And so the conversations that I'm having with Claude in ChatGPZ, for example, I mean, when researching these episodes, it's very technical and niche, and they don't directly relate to me looking for information about some purchase or spending that I'm hoping to do that would anchor them into having an advertising connection.

26:52So an advertiser could pay to put up a billboard that I might see during one of those LLM conversations. if I didn't have a paid subscription, but it's really just not a great model compared to Google search because you don't have the same targeting at scale, right? An Italian restaurant can pay to be the very first result that shows up in your town when you Google restaurants near me. That's very valuable digital real estate because it's linked to real economic activity that may happen, like you going out in the world and ordering food at a restaurant. Whereas, like I said, these more abstract conversations with LLMs are really not particularly appealing to advertisers who have lots of better options for more accurately targeting their ideal customer, whether that be via Google Search or Facebook, Amazon or Reddit, whatever it is.

27:40They have tons of tools at their disposal. And I don't want to make it sound like LLMs are objectively not a good way to do targeted advertising, but we certainly have to put into context just how well they can stack up against something like Google search and what the limitations are. Right. And looking at the growth in search for alphabet, it's kind of come from two main areas. So we got AI overviews and AI mode. So when I was just speaking there about finding answers to very, very simple questions, this is exactly kind of what I'm talking about. And while I've heard that Google's model Gemini three and a half isn't as good as Claude or ChatGBT, I've actually found the answers to be a pretty equal quality, just delivered in a much shorter time.

28:19But when I think about the use of AI overviews and AI mode, my mind still wanders to just how Google makes money from this because unlike a premium subscription to ChatGPT or Claude, Google isn't getting any money from me when I use these services. So, you know, I just kept going down the rabbit hole and I found some very interesting things. First, I think Google has probably done the best job of meshing the monetization of search with AI. So I remember when ChatGPT came out, I was just kind of amazed that they just weren't showing ads of some sort, you know. Google, I think, has figured that out very, very well.

28:49So whenever you do a Google query for some sort of question, you'll get an AI overview. Google has ads that will show above or below the AI overview, as well as just inside of the AI overview. Now, the benefits of the ad inside the AI overview are that the businesses can target customers to what you were just talking about, Sean, using kind of these non-traditional paths during search. If someone wants to use the AI mode to deepen their questions, then ads can be better placed to be the clear next step in trying to figure out that solution. And that exists as a complement to their existing search interface in contrast to something like ChatGPT.

29:20And with Google's distribution advantage, it's sort of their game to lose because they could immediately just roll out AI mode to over a billion monthly active users, allowing them to collect even more data on how consumers are using AI for things like shopping, which obviously is a great, great help for advertisers. And actually on the latest earnings call, Alphabet CEO Sundar Pichai really focus on just how successful this has been. He noted that Google is actually sending billions of clicks to websites each week from the AI features embedded in search. And to me, that's a pretty strong counter argument to the bearish fears that AI answers would make it unnecessary to visit underlying websites.

29:59And then therefore you'd have all these stranded publishers trapped in a sort of exploitive relationship with Google, where their sites are being scraped to feed answers into AI overviews at the top of the search results, but they're losing out on a chunk of that traffic that no longer needs to go to the underlying website. I mean, to some extent, that is a real phenomenon, but the scale of which I think could be exaggerated. And so the other detail I would probably flag is the cost side. They've dramatically reduced the cost of what it takes for them to generate a response in AI mode to solve a query down to the lowest level since their launch.

30:37And that's very promising. It is. And one of the most surprising things I learned from you about Alphabet is that the majority of searches basically carry no ads at all. So you mentioned that only about 20 % of searches have ads according to data from Google in 2020. And more recently, Google's vice president of search went on a podcast and confirmed that less than 25 percentages have ads still. So Google has spent a long time just learning how many ads it can show without losing users and bothering people or pissing them off. So I think it seems like using AI summaries and AI mode will just help them to continue to leverage these findings longer into the future.

31:10There's also another twist that I didn't anticipate when I first looked at Alphabet, because it wasn't totally clear yet that Apple would take a more passive role in the AI race and choose these other companies' models. But for context, in January, Apple announced that it would pay Google something about$1 billion per year for access to the technology that underlies Gemini to use that to power Siri for the next few years. And what that also means is that Apple is going to rely more on Google Cloud, too. And so obviously that$1 billion in cash annually is not material to Alphabet's bottom line, as crazy as that is to say.

31:47But I think it's revealing in what it shows about Apple's AI strategy and also in how the two companies are becoming even more intertwined and in different ways. Yeah, I would say it's a pretty impressive turn of event for Google. And speaking of cloud, that's maybe the most impressive and exciting part of the entire company right now. So the cloud division recently inflected into earning positive operating income, showing that while growth has been off the charts, this is actually a segment that Alphabet can make real money on because they actually lost money on cloud for nearly its entire existence that they've scaled it.

32:18And they've tried to take share from Amazon and Microsoft. We're talking about$58 billion in cloud revenue last year at a 24 % operating profit margin. And so these are huge numbers, but what's even more promising is that compared to Amazon, Alphabet's cloud business was about two thirds as profitable last year. And that doesn't necessarily sound like a good thing at face value, but Amazon has been operating in the cloud space for two years longer than Alphabet. And that may not seem long either, but that is a meaningful amount of time in cutting edge tech. And then it also does about twice as much in revenue with the point being that we can actually look to Amazon's profitability with AWS as perhaps an indicator of what Google Cloud may be able to approximately accomplish.

33:06That was something that I really talked a lot about when I first looked at the company last year. And so through that lens, Google Cloud is a business that's been compounding revenues at these impressive rates, 40 % a year over the last decade. And actually operating income growth should be more leveraged than revenue growth going forward as margins expand. That's basically an amplifier of earnings growth on top of the revenue growth that you get. And so the crazy thing is that if anything, I actually underestimated Google Cloud's growth. And so actually in the first and second quarters of this year, Cloud grew at a rate of 63 % and then 82 % year over year.

33:43And if we annualize their second quarter revenue, that run rate on that is in the$100 billion range. And that would be a double from where they were early last year. And so on top of that, Google Cloud's profitability is coming in just years ahead of schedule. And so their margins last quarter were nearly 36%, which moves them from being two thirds as profitable as AWS was in 2025 to being on par with Amazon's 2025 cloud profitability. So, you know, in pretty short order, I think we've seen Google Cloud go from being just a secondary competitor with AWS to really being almost on the same level.

34:23So, you know, perhaps this is an industry where it will be somewhat of an oligopoly, at least in North America. But, you know, Google is growing even faster than AWS is right now. So it's going to be interesting to see how that battle plays out over the next few years. But we actually do have some clues as to where revenue is going to show up specifically for Google Cloud in the next few years, because Alphabet gives us a very, very important figure. Yes, that figure is the backlog. I figured that That's what you're assuming too. Alphabet has more than$500 billion of remaining performance obligations.

34:55That's another way to say backlog for their cloud segment. And so for those unfamiliar with what backlog means, it reflects the contracts Alphabet has signed with customers for future deals where the work has not yet been delivered. And so therefore the revenue has not shown up yet on their financial statements. But we know with fairly high confidence that it will come in the future thanks to these order requests. And so it'd be sort of like a bakery, maybe having a bunch of orders for wedding cakes next year. They haven't been paid for those yet, and they haven't delivered the cakes yet either.

35:30And unless the weddings are canceled, that business will come through. And so maybe that's a crude proxy for how to think about Alphabet's order backlog. And just to give you an idea of how fast this is growing, Google's cloud backlog was$106 billion in Q2 of last year. So we're talking about five times growth in the order backlog in about 12 months. I mean, that is, that's just absurd, Kyle. It really is, Sean. So just to take the other side of this argument. So when Sean and I were talking about comfort systems, I mentioned that I personally am not the craziest about taking too much data from backlogs simply because of the revenue recognition issues that I've seen before in other businesses.

36:11So an example of that would be where backlog might be realized in, let's say, a year or two. And then maybe you try to extrapolate, okay, well, how much of that backlog is going to turn to real revenue? And then you can kind of back into a number. But from my personal experience, at least, I've kind of gotten to a little bit of trouble doing that. But in Alphabet's case, the fact that cloud revenue is exploding while the backlog grows is still a very strong signal. Alphabet says they expect about 50 % of cloud revenue growth over the next two years. So that's another$130 billion or so over the next year, which would be higher than the current run rate is showing.

36:40Then you have to layer on the fact that the demand for Google Cloud's products doesn't appear to be going away. The fact that they're investing so heavily into that area of their business is a pretty good suggestion that Alphabet believes that the demand for their cloud services is just improving with time. The other thing I probably missed in my original pitch for Alphabet is where a chunk of this demand was coming from. And that's specifically for their computer chips. And so Google designs its own AI processors called Tensor Processing Units or TPUs. And they're already on their seventh generation of the technology.

37:13And so for much of TPU history, they were used as an internal cost-saving tool, specifically for Google. And so that has allowed them to bypass buying chips from NVIDIA or to at least reduce their dependency on NVIDIA. But since the TPU has proven successful, you've had select companies that have actually been offered the opportunity to purchase Google's TPUs outright from them. And so that point that you just made about the TPUs going to other companies, I think actually kind of helps with diversifying the backlog because it's actually the concentration inside of that backlog that if I'm nitpicking, isn't really my favorite setup specifically for Google.

37:51So, you know, if I had to choose between a backlog with a diverse customer base where no one customer is making up more than 10%, let's say, of the total backlog, I take that any day over a backlog that only has a few key customers where let's say one customer is making up 50 % of that backlog. Now, it's impossible to say exactly what the structure of Google Cloud's backlog looks like. It's likely very concentrated given that they just announced a five-year$200 billion deal with Anthropic to use Google Cloud. So that implies that Anthropic makes up somewhere around 40 % of that backlog. So, you know, don't get me wrong, you know, cloud backlog growth is incredible, but it comes with this trade-off in customer concentration where, you know, let's say something were to happen to Anthropic, let's say their business falls off since, you know, after all, they're still in the very early innings of AI as an emerging industry.

38:37And so if an event like that were to happen, well, then a lot of that backlog growth would just appear to be kind of fugazi to quote Matthew McConaughey from The Wolf of Wall Street. Well, it's a great movie and a great quote. And to keep using that bakery metaphor from before, if one customer made up 40 % of the value of your wedding cake backlog, That would be very concerning, right? So it's a sort of extreme example because Anthropic is right now on path to do a multi-trillion dollar IPO and will be raising lots of fresh capital. So it's not like they're going anywhere anytime soon. But ideally, your biggest customer would not be burning billions in cash while making up such a hefty percentage of your backlog.

39:18But if we can pivot to highlighting a part of the company that I found to also be very promising, even if it's not as breathtaking as the cloud business, that would be the Google subscription segment. And so basically, Google has a bunch of different licensing and advertising fees that it earns on top of paid subscriptions like Google One that allow you to, for example, increase your storage space on Google Drive and in Gmail. Yeah. So I think when most people think of a platform, they might think of a business like Apple, which is arguably one of the best platform companies on the entire planet.

39:52But Google subscriptions are a very, very strong segment. And I use it very heavily personally. So when you're using something like Drive or Gmail and your files and emails accumulate, one day you might get a notification saying, okay, well, you're running out of space. So you have a couple of options here. You can delete things, which you can certainly do, but I've done it before. And believe me, it's time consuming and offers a lot of friction. or you could just take the simple route, which is to just pay Google a little more money for that space that you can fill up again into the future. So, you know, another example would be YouTube.

40:23You know, if you're sick of watching ads, well, then you can purchase YouTube premium and just do away with ads while also being able to do offline downloads and background playback. I personally like the background playback a bit as it allows me to watch content while answering text messages, for instance. And the cool part about the subscription business is that they're all relatively low priced. So when you think of switching costs, they're generally low enough where you just don't really give it that much thought when you think about canceling or switching. And yet this business generated over$25 billion in revenue for Alphabet in the first half of 2026.

40:52After years of resisting paying for YouTube because I had just gotten accustomed to it always being available for free, Daniel actually finally convinced me to see paying for premium as not being so different from maybe paying for any other streaming service like Netflix. So I'm now a proud YouTube premium user and actually opted for the slightly cheaper plan that doesn't include YouTube music. I don't know if most people know whether that's available, but we talked about that in our Spotify episode from a few months ago, but I use Spotify instead, so I don't need to pay for music streaming twice.

41:27And YouTube is now effectively the largest streaming service on the planet where they also pay much less for the most popular content on their platform than say maybe Netflix, where they either have to make the hit shows themselves, produce it themselves, or pay a premium to license them. And so YouTube is bigger and probably more profitable inherently with monetization being split across ads and paid subscriptions. And so YouTube ads are seeing nowhere near the explosive growth that other areas of the business are, but I don't think it requires much spending to grow this part of the business either.

42:04So as long as YouTube attracts more content creators, that in turn attracts more users, there'll be a steady stream of people to advertise to. And so that's really the beauty of the flywheel behind that YouTube business model. Now, believe me, we aren't being paid by Alphabet to push any of their products here, but I've even noticed their ability in making things like PowerPoints faster and cleaner. And that's really, really helped with just giving better presentations and preparing them in a quicker manner. So, you know, the fact that I could just create a template and edit it easily and just get some images up sounds simple, but it really saves me a lot of time and helps me just optimize my work processes.

42:39And then kind of touching to your point on YouTube there, the thing I love about YouTube is just how the business model works. It's not like Netflix or Disney Plus, which they're spending billions of dollars to create that content. YouTube has basically empowered its own content creators to handle almost all of that on their own. And then yes, okay, they do pay them about a 55 % spread on that advertising revenue, but I think it's just a really, really good lean business model. And as you said, YouTube doesn't need things like movie studios. They don't need to find capital to finance. They don't need to produce.

43:07They don't need to license movies or series. It's pretty cool. And if you look at YouTube versus public market comps like Netflix, which trade out somewhere around seven times sales, that makes YouTube worth something like$300 billion. And honestly, probably a lot more, maybe as much as$500 billion. So when you put that all into context, that's not too bad for an initial$1.7 billion investment. No, not bad at all. And YouTube is now also testing out a new feature called Ask YouTube. and so it uses Google's Gemini models to let people ask questions about individual videos and I guess the idea being you get some fast maybe takeaways or you can use it to help you find the most relevant moments in a video be able to filter through the video more quickly I actually think that is a pretty interesting value add and so Sundar Pichai noted that more than 140 million people have used that service in June of this year alone and so that doesn't obviously drive revenue directly, but in theory, it makes the user experience better.

44:06Maybe people spend more time on the YouTube platform. And so that increases monetization by allowing them to run more ads and really just illustrates the rationale behind how some of these AI investments can improve the entire business beyond what's just obviously directly attributable to AI where somebody's paying X dollars a month for a Gemini subscription. All right. Well, let's talk here about one of Alphabet's biggest moonshot bets, Waymo, which I know Sean has pretty strong feelings of. So in your original episode, you noted that Waymo just raised money at an evaluation of about$50 billion, which was actually down considerably from an earlier estimate of about$200 billion.

44:44But it actually appears Waymo is increasing in value as it's valued at nearly$130 billion in its last funding round in February of this year. So this piece of Alphabet, which at one point could have been seen as just a rounding error, albeit with a lot of optionality, has nearly tripled in value over a year and a half. Now, I know that you prefer Uber to Waymo, at least in terms of business models, but Waymo definitely has some pretty strong fundamentals. They've now driven 127 million fully autonomous miles and are reporting 90 % fewer serious injury-related crashes versus human drivers on that exact same mileage.

45:17They did 50 million rides in 2025, three times as in 2024, and they're currently running about 500 ,000 rides a week with a target of a million rides by the end of the year. So for anybody who missed it, Daniel and I just did an episode the other day revisiting Uber. And I think I said that it was pretty obscene for Waymo to have the same valuation as Uber. I mean, Uber is doing tens of millions of rides per day and generating billions of dollars in profits with a very proven and quickly scaling business model. While Waymo is still in the cash burn phase, it's losing a ton of money and its viability as a business model has not been proven.

45:55It's entirely speculative. And then to some extent, it's also dependent on whether competitors are able to make breakthroughs in AV tech that offset some of the first mover benefits that Waymo may have. So as an Alphabet shareholder, to the extent that they can capitalize on Waymo by maybe monetizing their stake partially or fully from a Waymo IPO down the road, I think that's sort of all gravy for us. As an Uber shareholder, I would say I'm pretty skeptical of Waymo being able to grow into this valuation. But objectively, the technology is really cool. I saw it for myself in Austin, Texas. So it's it's really makes you feel like you live in the future.

46:31It does. I haven't got a chance to use it yet. But Alphabet made another big bet in 2015 for about$900 million in a growing space company. Now, I bet you can guess what that company was. I think so. Yeah, I think most people know that it just IPO'd SpaceX. That's the one. So today, Alphabet's 5 % stake is worth$95 billion. So between businesses like Waymo and SpaceX, Alphabet has some monster winners in its other bet segment. But, you know, I also totally agree with your points from the original research, especially your point that it's just really difficult to evaluate some of Alphabet's more speculative technology bets.

47:06But, you know, at least with a business like SpaceX, we can get some idea of what the market thinks, which can maybe provide a little more clarity on some of its other bets, such as Waymo. There is one part of Alphabet that I don't think I fully wrapped my head around and that I still find fascinating. And to be fair, I don't think the market has really wrapped its head around this either. And that's what the ROI from their data center commitments is going to be. That is the recurring question in today's episode. And so if Alphabet is going to part ways with hundreds of billions of dollars to directly and indirectly invest in data centers, I think it's very timely to try and figure out what kind of returns this spending is likely to generate, especially since the magnitude of this estimated spending seems to just get revised higher every quarter.

47:48And so I will say maybe it's more of an exercise of trying to be directionally correct. It's a little complicated to disentangle because the data centers are not only profit centers when we look at Google Cloud's backlog, but they also are a big internal investment inside of Alphabet, helping to power nearly every aspect of the business's overall productivity. So I'm not sure that we're going to get a perfectly clear answer or that it's even possible to get one. But I'm really excited for you to kind of take a shot at doing so. Yeah, let me take a swing. So I'll double down on Sean's warning here that my answer is just an educated guess and hopefully it's a decent guess.

48:28So the first thing I want you to do, close your eyes for a second and imagine what is actually being built. Imagine a windowless box the size of about 50 American football fields. Now, the interesting thing about this data center is that it's actually the third largest in the world and it's actually owned by Google and it's actually located in Council Bluffs, Iowa, which funny enough is actually just a few miles away from the hotel that I stayed at in last year's Berkshire Hathaway's annual general meeting. While the size of this facility is clearly very, very large, that's not even the most important part.

48:55What has been a hot topic of late is the electricity usage of data centers. Back in 2024, Google data centers power use nearly equaled the entire country of Ireland. And Google's data is much more bigger than that today. It's a pretty interesting way of reframing things. And instead of looking at these data centers as just being a pure technology outlet, I think what you're describing is that they're almost like a utility. And clearly, AI data centers as a utility have some pretty massive energy consumption numbers. They really do. And, you know, I'm naturally pretty skeptical of these investments overall.

49:30So maybe let's start here with the case against Google's AI data center build out. And we'll get back to that energy consumption part shortly here. So I think there is no doubt that the investment will bear fruit for Google. as it requires this kind of growing amount of CompuPower just to operate the business of Google. Not to mention that, you know, all the capacity that they're leasing out to other businesses is obviously bearing fruit as well. But the problem is understanding, okay, well, what kind of economics is Google getting from these investments and once they're all complete? So if I'm looking at Alphabet's return on invested capital, that number is kind of trending in the wrong direction.

50:04So if we look here at fiscal.ai, which we like to use very, very often, in their last 12 months, their returns on invested capital is now the second lowest that it's been since the business went public. But the interesting part about this equation is that Alphabet is still making a ton of money. You know, their operating profit margin has actually gone up. But the reason their capital efficiency numbers are trending downwards is actually this increase in invested capital. Given that Google's cap extra this year is estimated to be six times more than what they spent in 2023. I guess it's not all that surprising that the incremental returns on some of that capital are not yet manifesting or clearly higher than in the past, but for better or worse, we still really don't know yet.

50:44And so Alphabet is going from a company built around the digital world with its software to increasingly being anchored in the physical world to an extent. And so rather than just purely being algorithms and software that are creating most of the company's value and maybe the human knowledge workers that produce those technologies, more and more of what they do is just simply tied tangibly to the physical world, computer chips, electricity consumption, data center construction and maintenance. And so the implications of that are pretty profound, right? Software has driven the market's returns for two decades now, and the economics of the biggest software companies are changing.

51:22And I'm not saying that because of SaaS apocalypse concerns, even though that is a narrative in markets right now, but really just simply that the businesses are becoming a whole lot less digital. So I think it's really important to be humble at this moment and recognize that this is a paradigm shift. And no one knows with much confidence what that is going to mean going forward. Exactly. And another thing that kind of obfuscates Alphabet's numbers and cloud in particular is that Alphabet is carrying some of its assets on the balance sheets that are yet to depreciate in value. So this actually serves to inflate Google Cloud's margins.

51:58And as of their latest quarter, Alphabet carries assets not yet in service valued at about$122 billion. Now, I'm not saying that they're trying to do anything that's not about board, but I think if you really want to dig into their financials in depth, you definitely have to account for this to some extent. You know, you can make the argument that these assets will be in service at some point. And then at that point, once they're in service, they'll be added to the depreciation schedule. And perhaps that means that Google Cloud's margins at the current levels are not really sustainable over the long term.

52:24To me, this is probably the biggest black box at Alphabet. And Michael Burry, who listeners will, I'm sure, know from the big short, has said that depreciation accounting for computer chips in particular is the most common form of fraud and maybe shareholder deception in the 21st century. And so accounting is based on assumptions and assumptions like how long the quote unquote useful life of an asset is. So whether you have to replace the servers in your data center every three or six years is a very consequential accounting decision to make. And so if the answer is that they need to turn over every three years, because that's how quickly they get outdated, but you depreciate those costs over six years, then you're going to be hugely underestimating the real economic costs of those chips and artificially inflating earnings on the short term.

53:12So we talked about this in one of our YouTube live streams the other day. And Michael Burry has stated that hyperscalers are collectively understanding depreciation by as much as$175 billion over the next few years. And no matter how much reading I do on the topic, I will never know enough about semiconductors to tell you what the appropriate depreciation schedule is with any confidence. But what you're doing is you're putting a lot of trust into management. And clearly you have some smart people like Michael Burry, at least raising questions about those assumptions. Yeah. And you know, I'd also be lying if I said that I'd put that much time and effort into figuring out depreciation schedules for data center lives.

53:48So, you know, I think I'll probably leave that argument to people who are more knowledgeable in that area than I am. But I still think it's great that you brought it up. You know, perhaps it means when you're modeling this business, you have to make a few different assumptions on different depreciation scales, which could affect gap profits. But, you know, I would just add a few other concerns with the data center here as well. So I already discussed how Anthropic appears to be taking up a large share of Google's backlog, but there's also this kind of circular nature of the revenue here, which has been widely, widely shared.

54:15So just to give you an example here, a real example of Anthropics deal with Alphabet. So Alphabet committed up to about$40 billion into Anthropic as part of the deal. So, you know, to put it simply, Google invests in the customer, the customer buys Google Compute, then Google then books the backlog. So when looking at Alphabet, they've traditionally been a business that has compounded their per share value. And part of this was completed through buybacks, which ran between say$45 billion to about$60 billion between the years of 2023 until 2025. But as of the first half of 2026, they're literally zero.

54:46And additionally, CapEx has eaten up almost all of Alphabet's free cash flow. And in Q2 of 2026, they actually had their first negative free cash flow quarter in a very, very long time. Well, now we're bashing Alphabet a little bit here. But I do think that it's a productive exercise to do that because after all, we're pretty bullish on Google. Otherwise, it wouldn't be the largest holding in our portfolio. And so it's very important to look at your businesses critically. But just as maybe some more context on the composition of this AI spending, Alphabet's CFO, Anand Hashkanasi, has said the mix of the AI infrastructure investments that they're undertaking is approximately 60 % in servers and then about 40 % into the data centers and networking equipment.

55:31Yeah, and that number seems to track quite well. But I'd like to get back to the power consumption side of things. So an important metric that I've seen mentioned for data centers is something called power usage effectiveness or PUE. It's a standard ratio that measures how energy efficient a data center is by comparing the total energy used by the whole facility to the energy used strictly by the IT equipment. So in Google's case, it has a 1.09 PUE. And this is supposedly much lower than the industry average of around 1.3, meaning that Google is just just basically running much more efficiently than the average data center is.

56:04Now, it's great that Google is more efficient here, which gives them a number of scale advantages. These are advantages including cost savings, greater computing capacity per megawatt, and a reduced carbon footprint. And for anyone wondering what a megawatt is, it's enough electricity to power about 600 homes or so. And the interesting thing is that even though power makes up only about 6 % of the annual cost of this AI capacity, it is the largest bottleneck to scaling up. And so Alphabet is supply constrained with power and land and supply chain, but everything else is irrelevant if the energy needs cannot be met.

56:35And so it does not matter how many data centers exist, obviously, if they can't be powered. Exactly. And that's why I think, you know, SpaceX and Tesla Bulls are just so excited about these AI data centers in space, which we discussed on my SpaceX pitch. So, you know, renewable energy is clearly a massive competitive advantage and having data centers in space also takes care of much of the cooling that must be very, very carefully managed. But let's look at some of the economics of Alphabet's competitors. in the AI data center space. So there are a few businesses that compete in the value chain.

57:04The two best ones that I came over were CoreWeave and Nebius, simply because they're kind of pure play and don't have financials that are obfuscated by other business lines, such as, you know, an Oracle, an Amazon or a Microsoft. Now, the problem with both businesses is that they just today remain unprofitable. So, you know, even if you are making a complaint about Google's appreciation being artificially low, they're still probably much more profitable than either Nebius or CoreWeave is today. Some really interesting data points there for sure, but something I'm still reflecting on personally is how come Alphabet does have this large profitability edge over some of those comps?

57:38How do you think about that, Kyle? Yeah, I think the biggest one is probably the vertical integration part. So you mentioned here earlier, Sean, that Alphabet has developed many of its own chips, including the Tensor processing units. Now, because of this, it skips to some degree this kind of NVIDIA tax that pretty much all of these other companies are forced to pay. Now, keep in mind, NVIDIA is a very, a very good company. It has 65 % operating margins as well as pricing power. So, you know, that's great for NVIDIA, but obviously it's not so great for its customers. Now, Google doesn't necessarily have that same issue.

58:08So Google's tensor processing units come in at about a 40 % discount to the NVIDIA equivalent. And as we've seen with Anthropics deal with Google, it just makes more sense for many of these AI businesses just to rent compute power from Google rather than building it out on their own. Another simple advantage that Alphabet has as well is they have the ability to fill up their capacity themselves core weave for instance rents by the hour to their customers and alphabet can allocate compute capacity toward google search and youtube and gemini as needed on top of the baseline of customers that they have just naturally coming in so alphabet really has very little capacity being wasted they have a lot of flexibility and if there's a brief shortfall in customer demand that compute can be redirected internally and it's sort of like a hotel that doesn't have any room occupancies.

58:57All the rooms are filled and that makes a huge difference for the incremental profitability of the hotel. And that really is something of a perhaps a competitive advantage. And as of now, this probably is not a problem for neoclouds like CoreWeave and Nebius. But if Alphabet's advantages continue to grow, there may come a time when these businesses find it difficult to keep the lights on with too much unused capacity. But that is a speculative tangent for us to go on, perhaps on the other side of this AI cycle or bubble or whatever you want to call it. But just getting back to the angle of data centers and their impact on capital efficiency, it all really comes back down to this one question that we have been touching on throughout the entire course of today's episode.

59:42And that is, how much does that$200 billion need to earn for those investments to have made any sense in hindsight? it. Yeah, I mean, I think that's really all it boils down to. And I'm going to do my best here to break it down as simply as possible. So let's think of the$200 billion investment like buying a giant apartment building. Before you guys argue about whether that's a good buy, you obviously have to work out a few things such as what it costs you to own it every year. And then you need to also ask, okay, well, what rent do you need to charge in order to make it worth doing? Now for Alphabet, owning one year's worth of data centers runs about$30 billion in expenses once it's all switched on.

1:00:18And that composes of about 25 billion in chips that are slowly wearing out and the rest in power and upkeep. But just covering your costs obviously is not the goal. You also want a nice profit as well. So if we work backwards from a decent profit, you land somewhere around here, which is one year spending has to bring in somewhere between 70 and$100 billion of brand new sales every single year. If you simplify that, the rule is about 40 cents of new annual revenue for every dollar spent. That's kind of where the bar sits. But then, okay, you want to know whether that bar is high or low because, you know, 40 cents by itself just doesn't mean that much.

1:00:49And here, I have to be pretty honest because I went looking for a somewhat decent comparison and couldn't really find one. Obviously, there's a marketplace for this stuff. Nebius and CoreWeave, you know, they do things, they're buying AI Compute, they're renting it out, and then they're disclosing what those contracts are worth. But there's one kind of flag here that makes it really, really hard to really understand. The trouble is they quote it per megawatt and a figure like 20 million per megawatt can mean two very, very different things. If that's what the customer pays every year, it's$20 million a year.

1:01:15But if that's the value of the entire contract, and these deals, unfortunately, can run four or five years, and it's more like four or$5 million a year. So you're getting the same kind of headline number, but you don't actually get too much clarity. And I actually, unfortunately, found it described in both ways by different sources. So it's just not a comparison that I can really put any conviction into. But we can still use some information from these two businesses to at least develop a number where we are hopefully directionally correct. So what we do know is that Nebius' pricing has roughly doubled in the last six months.

1:01:46And Core, we've raised prices by about 25 % alone in July. And they actually said that their near-term capacity is effectively sold out. So its payback period or how long it takes until a deal repays, what it costs to build has actually dropped to under two years. On top of that, Nebius says that it could basically settle its entire 2027 capacity today if they wanted to. Now, you know, you certainly don't get to raise prices this much into a market with spare capacity. So as of now, I think Alphabet is far from guaranteed to making good returns here. But, you know, with all the numbers that we're getting here from Nebius and Corrive, I think it's telling us something useful, which is that demand is definitely running well ahead of supply.

1:02:22And that's obviously a condition that Google feels it makes sense for them to justify this amount of spend. It makes sense. And clearly, as an Alphabet shareholder, it's great to see this. But there is another crucial thing to be aware of here. And we have to look at the durability of those price rises. And since this is a product that I think at the end of the day could be mostly commodified, it doesn't mean we can necessarily extrapolate it very far into the future in terms of just linearly drawing price hikes up and to the right. Clearly, we're at this moment in time now where capacity is very scarce.

1:02:55And because demand exceeds supply so dramatically, the owners of Compute Power can charge more because there's a willing buyer out there. but supply and demand won't remain this imbalanced forever because, well, we've been talking all afternoon here about the data center investments that are going into building out the supply side of this equation. And so I've also been pretty uneasy about that 40 to$50 million number you mentioned from Nebius. And so we talked about this uncertainty with comfort systems in that episode of how, where we sit in the AI investment cycle that may be leading to excess optimism about how much more spending will continue to increase by and how profitable it'll be.

1:03:38And that is a company for context that specializes in the maintenance of servicing data centers. So the stock has gone off in a straight line, but we really don't know how good the business will be when things normalize. And to a lesser extent, the same is true for Alphabet. Yeah. And as Alphabet has become more capital intensive, it really actually surprises me that Warren Buffett, who's kind known for preferring businesses that don't require much capital to generate these cash flows, has now chosen now to be the time to invest in Alphabet after he first discovered the business, you know, 20 years or so ago while running search ads for Geico.

1:04:12Now, this is another area where the thesis may have changed a little bit since you first covered it. I remember you highlighting in the original episode how Alphabet just had so much cash that they didn't really know what to do with it. They had so much cash that they were just returning it to shareholders through buybacks, which I think felt right for a company at that time with excess capital. To your point on Buffett and Berkshire there, not only have they used their own cash piles, but they've increasingly tapped the debt markets for more financing. And now they're turning to selling equity to raise enough capital.

1:04:41Yeah, that's right. So they said they were going to spend about 200 billion on AI data centers this year. Then they said, we need more money. So they went out and raised many, many billions of dollars. And I think this shows that Alphabet isn't focused on just managing their excess cash anymore, but maybe they found something worth deploying all that excess cash into and even more. So here's what I find noteworthy about all this though. So the share count has declined pretty steadily since 2018 because of those buybacks that I just discussed. But as of Q2 2026, they have 12.3 billion shares outstanding.

1:05:12And that's actually the highest number since 2024. That kind of shift shows you how much they're prioritizing this data center spend. If you think about it, assuming Alphabet is going to continue to need funding for its AI data centers, there is a good chance that they'll use all avenues available to them, which likely means more net share issuance that increases the number of shares outstanding. And whether that is technically dilutive depends on how effectively the capital raised from selling more shares is deployed. But at a really simple level, you've spread the business's intrinsic value across more shares and shareholders.

1:05:48And that is a minimum going to put pressure on the stock in the short term because you're increasing the supply of shares to the market. And then you have to be able to justify that raising of capital longer term. And so while Berkshire invested$10 billion into Alphabet directly, Alphabet's total equity raise earlier this year was for$85 billion. And so the market was happy to fund that evidently, but it's the first major equity raise we've seen for them since they IPO'd. And in theory, Alphabet should be at a scale and a maturity where they don't need to rely on selling stock to raise money, right?

1:06:24That's thing you associate with startups and unprofitable tech companies because it's a very costly form of financing. That's right. And the fact they haven't had to issue equity for two decades, I think is a pretty obvious signal of just how good of a business alphabet is and speaks volumes about their ability to generate cash. But the share of issuance is actually just the beginning of their capital raising effort. So as of Q2, 2026, long-term debt is now$98 billion or about nine times since the fiscal 2024 year ended. Yeah. With how conservatively financed Alphabet has been, even I was a little surprised at how much this figure has grown.

1:07:00Yeah. Part of the increase in debt has been from raising money in both domestic and global bond markets. So I guess, you know, if you can't raise it all from home, why not look elsewhere, right? And the appetite for these bonds is very, very high. So in February, they issued about 20 billion of US dollar denominated bonds ranging from three to 40 years. And this was upsized from 15 billion. And I read the offer drew more than 100 billion in orders. And then outside of that, they've also just looked around the world. They've raised over$50 billion in other currencies, such as the sterling, Swiss franc, euro, Canadian dollar, Japanese yen, and even Australian dollars.

1:07:31So between the equity raise and all the bond issuance that we're seeing, maybe you can paint some color for the audience where they sit with the cash position now. Yeah. So they currently have$242 billion in cash and cash equivalents. So even with all this new debt, there's still obviously net cash, but that$242 billion is probably going to get depleted very quickly, given their current spending rate. And based on what management has said, they're not planning on slowing that down anytime soon. The Wall Street Journal had a really interesting article the other day about hidden liabilities for big tech companies where they've made these contractual agreements to build data centers or lease data centers or buy XYZ number of computer chips and so on.

1:08:10But these commitments are for 2028, 2029, 2030 and beyond. So we know they're coming and they're massive, sort of like the backlog. And with Alphabet alone, we're talking about more than$800 billion in off balance sheet liabilities over the next few years. And I'm not misspeaking when I say that. Literally almost a trillion dollars in liabilities that are not at present reflected on the balance sheet that weren't even conceivable a few years ago, right? They weren't on anybody's radar. So of course, the business is riskier today than it was in the recent past, because there's just no way to get around the fact that a tremendous amount of financial uncertainty has been injected into the business and their future financial prospects because of these different liabilities.

1:08:59But also, I think you could argue that if AI is even remotely as revolutionary of a technology as Silicon Valley thinks it is, then Alphabet's ability to do all this spending is buying them maybe several more decades of dominance and being one of the world's biggest tech companies. That would be sort of like the ultra bull narrative. And so just kind of looking at the facts, plain and simple, the risks have increased. They've changed also, right? A little bit less of a regulatory discussion and a competitive discussion and more about how the returns on this capital will look like. But on the flip side of that, Alphabet's corporate life cycle has potentially been reset pretty dramatically.

1:09:40And so all of a sudden, Alphabet looks to be a much younger company with more dramatic growth possibilities ahead instead of being some kind of stagnant mega conglomerate that was just slowly in the process of hardening. Yeah, that's a great point. And the problem with these stagnant mega conglomerates is what you just said. Sometimes it gets to a point where it's nearly impossible to find a new growth lever. And then you turn to basically this kind of cash flow generator that just returns 100 % of its cash flow back to shareholders. And while those can make a decent business, in terms of an interesting investment case, at least those that we look for to put into our intrinsic value portfolio, it's just not that interesting.

1:10:16We want businesses that have hopefully some sort of growth aspect to it because we're all looking for businesses that have a decent upside and we're not really particularly concerned with businesses that are just going to grow at the same rate as GDP growth. But I also kind of want to look at some of your points here on the risk angle that you just discussed here. So I think the first risk that I'd want to dig a little more deeper into is probably that you shouldn't give the income statement too much credit right now. So Alphabet reported$112 billion in profits in its second quarter. But when you dig in,$99 billion of that was a non-cash gain from marking up its equity stake in private companies like SpaceX and Anthropic.

1:10:53So think of it this way. If I told you you could make$100 in profit, you'd probably be pretty impressed, Sean. But then what if I said that $99 of that wasn't actual cash I could spend. It was just unrealized gains on, let's say, a stock that I own that I might be able to sell someday. You'd probably say, well, that's not really a profit. And that single gain added over$6 in EPS, but it's zero actual dollars coming into the business. And with Anthropic planning to IPO at some point in 2026, we're probably going to continue to keep seeing these large non-cash gains on the income statement. It all just kind of serves to add complexity when analyzing this business.

1:11:24So I think the flag here simple. Don't get too excited by the headline numbers. Focus on operating income or cash flows when you're evaluating Alphabet. Yeah, of course, the headlines on CNBC are always going to lead with that net income number. But as you said, accounting gains can make a business look a lot better on paper than it's actually performing. And that's why we talk about quality of earnings. That's why that's such an important concept for listeners to know. But beyond the income statement and depreciation accounting, I would come back to the reality that the cloud backlog sort of epitomizes both the upside here as well as the risks facing Alphabet.

1:11:59So half a trillion dollars in revenue being committed into the backlog is really great. But Anthropic and OpenAI have to actually be able to afford to pay that money in cash to Alphabet, regardless of whatever they promise today. And so the cash has to come from somewhere. And I'm nowhere near optimistic enough about either of those companies to say with conviction that they will be solid customers of Alphabet for many, many years to come. Right, I agree with you. And we're not here saying that these businesses are at the edge of bankruptcy by any means. There's clearly a ton of investor interest in this space.

1:12:36So my guess is that they'll find financing for a while. And with some of these other hyperscalers in there too, it kind of just raises the floor for committed revenue quality. Now, the next risk that I'd raise here is on the margins. So the margins obviously look really, really good right now. Operating margins is sitting in the low 30s. But here's where I think many people miss about this massive CapEx. Every dollar that Alphabet spends on these data centers today is going to become depreciation expenses next year or many years from now. So even if their revenue just stays flat, even if they don't grow at all, margins automatically will compress from that added depreciation expense.

1:13:09And when you're spending$200 billion a year, well, that depreciation hit is going to be quite substantial. Now, the final risk that I'd flag right now is that Alphabet doesn't really seem to me to have a variant perception as in the stock, not the actual business. You know, analysts currently love the stock and are issuing strong buy recommendations with an average target price around$430 versus the current price of about$340. Now, I think Sean did the right thing adding this business very, very heavily when it was completely out of favor. You know, this business right now is not out of favor at all at this time.

1:13:41So without getting into the weeds of valuation too much, I will say it was a really useful exercise for me to go through your model from last year on Alphabet, update it a little bit, and think about whether the market is offering us a really attractive bargain with that stock. Well, in short, I think we'd both say that we see this as an exceptional business, but one that is much closer to being fairly valued today than probably it was a year ago. And that's partly because of the run-up that the stock has had, but also because of the changes, these new uncertainties that are facing the company that personally I have less confidence in.

1:14:14I felt more comfortable saying that I disagreed with the market about chat GBT being a real disruptor of Google search. I'm really not sure that I have a strong opinion that differs from the market where I feel strongly one way or another about whether these massive investments that they're making are going to underperform or outperform expectations. And so if you're searching for something on the internet at the end of the day, there is a very good chance that you're going to touch one of Google's products. I use YouTube pretty much every day myself, and I just can't think of any substitute that would turn my attention elsewhere.

1:14:52And so Alphabet's not going anywhere. This is an incredible company. We're very happy to kind of let our winners run here. And I don't want to overthink things, but yeah, I don't see as compelling of an argument that I can make in good faith that Alphabet's stock is as undervalued as it was this time last year, despite ironically having the same PE ratio? Yeah, I really don't have any arguments here. I think my stance on Google is quite simple. I just don't think it's the right time to add to the position. If we were speaking strictly rationally, then selling is actually probably the right decision due to the fact that we already made nearly a double.

1:15:32But we are long-term investors. So sometimes the most optically rational decision isn't actually the correct move. And then speaking of maybe potentially trimming, well, we aren't really into market timing. So I think selling pieces of the stake, assuming that we can buy more for later at an even cheaper price, just doesn't really make that much sense. So in my view, I think the best move with Google, given that we already own it, is to just do nothing and let the position play out. Well, I completely agree with you there. And yeah, it's sort of a weird concept. I remember when I first pitched Alphabet on the show, I concluded that it was about fairly valued, actually.

1:16:09I didn't even make the case for being super undervalued. So when I say that it was really undervalued, I say that in hindsight. And we had a listener write to me and ask me why we would decide to invest in Alphabet if the conclusion was that it was fairly valued. And back then, mostly it was because I knew I'd been very, very conservative in the modeling when I said that it was fairly valued. But again, it kind of goes into this idea of just wanting to own wonderful businesses because they will surprise you to the upside typically and just letting them compound over time and just sitting on your hands and not doing anything about it.

1:16:47So I felt like if I could get a wonderful business at a fair price, that was a great arrangement. Alphabet has surprised me to the upside a number of times in the last year. And that's why I sort of take this benign and maybe more like tranquil view on the run-up in the stock in the last year where I wouldn't want to bet on where the stock price is going to go in the next 30 to 60 days. but looking at this with a 10-year, 20-year perspective, if you can even fathom thinking that far out, I think these are rounding errors in hindsight. And the simplest thing is to just sit back and watch because you'll drive yourself crazy if you're trying to time every single swing in the market and buy and sell and buy and sell.

1:17:27It's just not a winning formula. It's not a recipe for success. Exactly. So that's all we have for you today. But as usual, I think I'd like to leave you here with a quote. And this one is coming from the Oracle. of Omaha himself, Warren Buffett, back in 2012. So he said the chances of being way wrong in IBM are probably less, at least for us, than being way wrong with Google or Apple. But that doesn't mean that those, the latter two companies, aren't going to do, say, far better than IBM. Now, this is just such a great quote because I think it showcases Buffett's thinking process. In just a few sentences, you can see how he thinks about opportunity cost and why upside isn't the only thing that matters.

1:18:03And of course, he was completely correct that both of those businesses provided much better returns than IBM, but he just didn't have the competence in his understanding of Apple and Google at that time to make an investment. But now he's invested in both. And perhaps you can argue that maybe he feels like he's gotten competent enough or that at the time that he bought them, the businesses just had been de-risked to some extent and that downside was well-protected. That's all we have for you today and see you next time. Just a quick note before you go, this episode would not be possible without our friends at Fiscal AI.

1:18:37It is the complete stock research terminal that Daniel, Kyle, and I use on every single episode. With every company we dig into, pulling 20 years worth of financials, digging into segment level data, and grabbing quotes from the latest earnings calls, real-time institutional grade data, all in one place. And now with our new AI connector, you can plug that same data, financials, transcripts, fund letters, news, filings, and more straight into Cloud ChatGVT or whatever AI you use for your own research. If you want to try it yourself, head to fiscal.ai slash T-I-V-P to get 15 % off. The link is in the show notes.

1:19:15That's fiscal.ai slash T-I-V-P. Thanks for listening. Thanks for listening to T-I-V-P. Follow the Intrinsic Value Podcast on your favorite podcast app and visit theinvestorspodcast.com for show notes and educational resources. This podcast is for informational and entertainment purposes only and does not provide financial, investment, tax, or legal advice. The content is impersonal and does not consider your objectives, financial situation, or needs. Investing involves risk, including possible loss of principle, and past performance is not a guarantee of future results. Listeners should do their own research and consult a qualified professional before making any financial decisions.

1:19:51Nothing on this show is a recommendation or solicitation to buy or sell any security or other financial product. Hosts, guests, and the Investors Podcast Network may hold positions in securities discussed and may change those positions at any time without notice. References to any third-party products, services, or advertisers do not constitute endorsements, and the Investors Podcast Network is not responsible for any claims made by them. Copyright by the Investors Podcast Network. All rights reserved.

1:20:32Thank you.