TIP813: Microsoft (MSFT): Is Microsoft a Misunderstood AI Opportunity? w/ Daniel Mahncke & Shawn O’Malley

16:26Shawn O’Malley:So originally, they were basically sold as this perpetual license. I still remember in school when you had Office 97, when they didn't upgrade it for a long time, or maybe Windows XP, where basically those products were bought at a store. And then once you had them, you owned them forever. So for Microsoft, that would basically mean that you book all the revenue immediately. But also, you had no idea what the next quarter would actually look like. And when you move from selling Office for, let's say, a$400 one-time payment to only charging$12 per user per month on what is mostly like an auto-renewing subscription, revenue becomes recurring and predictable.

17:01Shawn O’Malley:And that's what Adobe wanted. That's also what Microsoft wanted. And that's why they went through this change. I think it's also important to know that, and we said it for Adobe too, this hasn't been the new holy grail when it came out. A lot of people were acting like you basically lose out on$400 that you immediately can get by hoping people would stick with your product for only$12 a month. Today, though, Microsoft has 450 million commercial seats, which is just incredible if you think about it, at an average of roughly$25 per user per month. So that's generating over about$130 billion per year and essentially pure recurring revenue.

17:39Shawn O’Malley:So that's just an incredible business. And once again, it's also high margin. So this segment alone makes about$77 billion in operating profits. And if you adjust for LinkedIn that we just talked about and some of the other smaller operations in this segment, I think there's an argument to make that probably about$70 billion comes from only the software solutions alone. So we are talking Word, we're talking PowerPoint, and we're talking Excel. And if you just think about that, it's an incredible business, especially if you think about the 20, 30 years that it's already operating. But also, I mean, that single profit pool is larger than the total revenue of most S &P 500 companies.

18:14Shawn O’Malley:and it has 60 % margins, that kind of feels right for disruption, right? Especially if you think about AI coming, you can kind of get a glimpse that this might not work for the next, let's say, 20 years. Right.

18:25Daniel Mahncke:And so then the CEO of Microsoft, Satya Nadella, has said in public on multiple occasions that software apps could just go away, vanish. And that's a pretty remarkable thing for any CEO to say, especially when that would be incredibly bearish for you as a company, right? It's not something I would have expected him to say publicly.

18:48Shawn O’Malley:On one of our last episodes, we actually talked just recently about the fact that no CEO sees their own company threatened by AI, at least not publicly. And then Satya Nadella came out and has pretty much done that. Although to be fair, he's also doing that while expecting Microsoft to become a winner in this new era of AI. So he's technically also just marketing Microsoft's new products when he does that. It's kind of similar to what Salesforce's CEO, Mark Benioff, also did, right? So basically the question for us as potential investors is how much of that$70 billion survives this AI transition?

19:22Shawn O’Malley:And also how long would that transition take? And what does Microsoft replace it with? If you don't just want to lose the customer, you have to give them something else. And we will come back to this detail. It's kind of more than a detail. It's basically the whole story about Microsoft. But for the sake of keeping it simple, I would say we first finish the segment overview, because I know we're teasing a lot today, but that's just because I fear we will kind of forget things when we're not at least going through the business segment one by one. So the second big segment is Intelligent Cloud. And that's also where Azure is basically the core of, but you also have things like SQL servers, Windows servers, and GitHub.

19:56Shawn O’Malley:And this segment generates about$125 billion in annualized revenue. So it's about the same size as the first one. However, it is growing at 29 % per year. And that's a lot more than the first segment, right? Azure itself actually has a run rate by now of$75 billion per year. And it's growing at almost 40%. Again, only 39. What a bummer. In contrast, though, to almost every other Microsoft product, Azure is not a subscription, but a consumption-based model. That's also what we've seen with a lot of the other clouds. They basically all work the same. So you pay based on how much you actually use the product.

20:30Shawn O’Malley:And this means that revenue can pretty much grow explosively in periods where you have rapid adoption. But then when customers become more efficient with using your product, which just naturally happens as they kind of master the platform, wherever you can also de-accelerate, right? And that's what you've seen, for example, in this optimization cycle that 2022 and 2023 have been. And then it started re-accelerating because you just had this new AI wave that basically created a whole new vector of demand.

20:58Daniel Mahncke:I mean, we've kind of laughed at it, but it is absurd if this premise is true that the cloud business is growing at 39%. And that is the reason, falling short of some sort of market expectations for the stock to sell off. But before we go deeper, there is one more segment we need to talk about. So what is the third segment?

21:18Shawn O’Malley:The third segment is more personal computing. And that's where you have Windows, Xbox, Surface devices. And since late 2023, You also have a gaming business because you basically bought Activision Blizzard for$75 billion. By the way, that's the biggest gaming acquisition people have ever seen. And it's not like it didn't make any impact. I mean, gaming for Microsoft is a business also generating about$43 billion in revenue. So it's not nothing. I mean, Xbox Game Pass, for example, has about 40 million monthly subscribers. But this segment is certainly the lowest margin and it's also the slowest growing of the three.

21:55Shawn O’Malley:So again, kind of like with LinkedIn, it's somewhat of a great and interesting business, but it's also not central to why you would own the stock today. I still want to shortly touch on the gaming business though, also because I know that we got some feedback when I said that I would cover Microsoft that some people are interested in. So we'll talk about it. And I would also say it's interesting that if you spend$75 billion on an acquisition, again, the biggest in gaming history, and I actually checked it roughly double what Disney paid for Lucasfilm and Marvel combined. I kind of assume that you have a good plan for the IP that you bought.

22:25Shawn O’Malley:So for example, what is that? You have Call of Duty, which has sold tens of millions of copies annually for almost two decades. You have World of Warcraft, which also still has millions of paying subscribers, and then a bit lower quality, but also more people playing it, Candy Crush, which has hundreds of millions of monthly mobile users. And I would say that the main idea has probably been to turn these games, maybe except for Candy Crush, into system sellers for Xbox and their Game Pass. And basically making it a no-brainer subscription for every gamer out there. It's kind of like what we talked about with Netflix, where they wanted to buy even more IP.

22:59Shawn O’Malley:So that subscribing to Netflix just feels like a no-brainer if you think about the competition. I think it's safe to say that for Netflix, it worked. For Microsoft, it didn't work at all. It clearly hasn't happened. I didn't ever at any point saw more people buying an Xbox or buying an Game Pass. so now Microsoft has essentially stopped trying to win this traditional console race it's also something that we talked about with Nintendo and it's kind of pivoting toward what they call an Xbox everywhere strategy so essentially that would mean that you can access the game pass across devices and cloud infrastructure and you're not dependent on actually owning an Xbox anymore I think they will not stop producing an Xbox there's this project which is codenamed Project Helix, and it's rumored to play both PC and Xbox games.

23:47Shawn O’Malley:I think to me, it sounds like a great idea on paper, but I kind of lack conviction that Microsoft can actually pull this off, just given how little they have done with the acquisition of Activision so far. Let's take a quick break and hear from today's sponsors. All right.

24:03Daniel Mahncke:I want you guys to imagine spending three days in Oslo at the height of the summer. You got long days of daylight, incredible food, floating saunas on the Oslo Fjord, and every conversation you have is with people who are actually shaping the future. That's what the Oslo Freedom Forum is. From June 1st through the 3rd, 2026, the Oslo Freedom Forum is entering its 18th year, bringing together activists, technologists, journalists, investors, and builders from all over the world, many of them operating on the front lines of history. This is where you hear firsthand stories from people using Bitcoin to survive currency collapse, using AI to expose human rights abuses, and building technology under censorship and authoritarian pressures.

24:46Daniel Mahncke:These aren't abstract ideas. These are tools real people are using right now. You'll be in the room with about 2 ,000 extraordinary individuals, dissidents, founders, philanthropists, policymakers, the kind of people you don't just listen to, but end up having dinner with. Over three days, you'll experience powerful mainstage talks, hands-on workshops on freedom tech and financial sovereignty, immersive art installations, and conversations that continue long after the sessions end. And it's all happening in Oslo in June. If this sounds like your kind of room, well, you're in luck because you can attend in person.

25:22Daniel Mahncke:Standard and patron passes are available at oslofreedomforum.com with patron passes offering deep access, private events, and small group time with the speakers. The Oslo Freedom Forum isn't just a conference. It's a place where ideas meet reality and where the future is being built by people living it. Curious about online trading, but haven't taken the first step yet? You're not alone. And Plus 500 Futures is a great place to start. The futures markets are moving fast. And with Plus 500, you can explore popular assets like oil, gold, S &P 500, Bitcoin, and more. From crypto to commodities, there's always something happening.

26:02Daniel Mahncke:The platform is super easy to use so you can trade on the go right from your phone. You can get started with just$100 and jump into the action. See something interesting? Once your account is open, you can trade it in just a couple of clicks. And if you're not quite ready yet, you can practice with a free demo account. No risk, no pressure. With 20 years of experience, Plus 500 makes trading more accessible than ever. Check it out at plus500.com. Trading and futures involves risks of loss and is not suitable for everyone. Not all applicants will qualify. Plus 500. It's trading with a plus. Every business is asking the same question right now.

26:40Daniel Mahncke:How do we make AI actually work for us? The possibilities are endless. Guessing is too risky, but sitting on the sidelines, that's not an option either because your competitors are already making their move. With NetSuite by Oracle, you can put AI to work today. NetSuite is the number one AI cloud ERP trusted by over 43 ,000 businesses. It brings your financials, inventory, commerce, HR, and CRM into one single source of truth. And that connected data is what makes your AI smarter. It doesn't guess, it knows. It automates routine tasks, delivers real insights, and helps you cut costs and make decisions with confidence.

27:16Daniel Mahncke:And now with NetSuite's AI connector, you can use the AI of your choice plugged into your actual business data. Ask any question you've ever had, key customers, cash on hand, inventory trends. Let's see your competitors do that. I use this and you should too. If your revenues are at least in the seven figures, get the free business guide, Demystifying AI at netsuite.com slash TIP. The guide is free to you at netsuite.com slash TIP. All right, back to the show. It doesn't seem like too bad of an idea. We just saw the huge success of the Switch 2. So consoles, though, are getting more and more expensive and focusing on an ecosystem around something with much cheaper and more flexibility seems like a pretty good option to me.

28:05Daniel Mahncke:But maybe the way to think about it is the Netflix of gaming of sorts. And you have this huge library of games that you can access no matter where you are, and regardless of what type of hardware you're using. That feels like really democratizing gaming in a way.

28:23Shawn O’Malley:I guess you can view this as either an upside optionality or also potential opportunity cost. But no matter how this turns out, again, it will be relatively unimportant to the overall thesis for Microsoft.

28:36Daniel Mahncke:All right, well, let's get back to the topics that are key to the thesis and talk more about the office products in particular. So I think there are two ways this segment could get disrupted going forward. The first is by facing competitors that have a new chance of disrupting the office ecosystem by leveraging AI more effectively. And so one of our mastermind members just recently posted a chart that applies Clayton Christensen's disruptive innovation theory to Google's and Microsoft's productivity suite and basically asked, why couldn't Google disrupt Microsoft Office after winning the lower end of customers.

29:14Daniel Mahncke:And so Christensen's model for context predicts that a cheaper, initially inferior competitor enters at the lower end of the market and then improves over time and eventually marches up market to displace the incumbent. And so Google launched Docs, Sheets, and Slides all the way back in 2006. And they won the low the market, you had a lot of individuals, students, and small businesses that really rely on those products, including ourselves. And then still 20 years later, Microsoft Office completely dominates at the enterprise level, right? Large corporations are not using Google Sheets. So if you assume the average time it takes the new competitor to catch up based on Christensen's model, the crossover where Google's products should have met the demand of more professional users.

30:06Daniel Mahncke:Well, that should have happened in 2018. So the interesting question to me is why has that not happened?

30:13Shawn O’Malley:I still remember that many, many years ago, before I started investing and kind of ever heard of things like a mode or switching costs, I used Google Docs for the first time that was still back at school. And I immediately asked myself, why does anyone still ever use Word? And funnily enough, I never used Google Docs again up until I started at TIP. So that kind of explained it. But anyway, now I kind of know about things like modes and like switching costs. And I would say there are multiple reasons why Google didn't succeed in kind of disrupting Microsoft's Office Suite. So the first is that Christensen's model was mainly built around hardware innovation and products not embedded in these larger ecosystems that we currently see, right?

30:52Shawn O’Malley:So software with dominant file format standards is kind of different. The Word, the Excel, and the Power Performance have become international standards, pretty much embedded in contracts, legal filings, and just all sorts of documents across millions of organizations. And this kind of reminds me of Adobe's mode regarding the PDF file format. So you as a person, or even as a company, you just can't switch unilaterally, right? Your entire professional ecosystem has to switch kind of simultaneously. And now it is relatively easy to convert a Google Sheet into an Excel file. But that's where we get to the second reason for why Google's workspace couldn't disrupt Microsoft.

31:29Shawn O’Malley:And that is that at the high end, especially for enterprises, the users are still doing fundamentally different jobs than the average users. So the Excel power user building, I don't know, a leveraged buyout model, custom VBA macros, they don't use Google Sheets because Google Sheets just doesn't offer all the same features. And even if the majority of employees don't need those features, as long as any meaningful number does. So let's say the enterprise has 5 % of their spreadsheets that are way too complex for Google Sheets. Then they still need to just stick with Excel because, first of all, those 5 % of sheets might be the most important in the company.

32:07Shawn O’Malley:And also, you would just make sure that everyone in your company can use the product, even the top 5 % of users.

32:13Daniel Mahncke:And to be fair, Microsoft has not entirely just stood still. They ran their own counter disruption effort by launching Office 365. which offered cloud delivery, automatic updates, and collaborative editing as well.

32:29Shawn O’Malley:Yeah, Microsoft did try to adjust, although I have to say that working collaboratively in a Word document is a complete nightmare, while it works great in Google Docs. So I would somewhat question how successful those adjustments actually were. And then perhaps the last reason are just that you don't only have network effects within your own company, but also with people from other companies that want to receive a file in a certain format. And also that switching software in large enterprises is just always a headache. And there's a huge bias to just stick with what works. I actually talked to Clay a couple of weeks ago for one of our last episodes.

33:04Shawn O’Malley:And he just talked about prior in his corporate life, how difficult it could be to actually change software at a big corporation. And it's such a huge headache that even if there's some value to gain and you're maybe a bit more effective, you generally just don't do it.

33:19Daniel Mahncke:So the takeaway for me is that Christensen's model was right for the bottom half of the market, but it wasn't good enough, or maybe there was too much of a change to disrupt the enterprise level. So you told me offline that AI may be able to change that dynamic. And so maybe we can reopen the door to the conversation of the innovators dilemma still being relevant and Google finding a way to disrupt Microsoft's basically legendary productivity suite? Is that something that's truly possible?

33:52Shawn O’Malley:I think we can. I think the innovators dilemma is not yet dead. So basically, Google failed to disrupt the office suite because Google was offering a competing product. And a competing product always runs into kind of the same structural barriers that we just described. What AI is doing, at least I would argue, is categorically different. So AI is not offering a competing product per se. AI is essentially reducing the need for the product. So when someone used Google Docs instead of Word, they were still spending the same amount of human time creating documents, basically just in a different environment, right?

34:27Shawn O’Malley:So the job was to create a document and then Google was a different tool for the same job. But when someone uses an LLM to generate a first draft in, let's say, 30 seconds and then spends 20 minutes editing rather rather than two hours actually creating that first draft, they fundamentally changed the human labor intensity of their task. And Microsoft's per-seat subscription-based model is kind of built on the assumption that you need one license for each human doing cognitive work. So if AI handles a meaningful share of that cognitive work, then the per-seat assumption kind of stops working. That is kind of why the$70 billion profit pool is at risk in a way that Google, without AI could have never threatened.

35:12Daniel Mahncke:So compared to Constellation software, for example, where the AI risk is mostly that there will be competing products made with AI, the AI threat here is not about switching to a different product. It's that you need fewer seats because each human is now doing the work of what several would have previously been capable of. Exactly.

35:33Shawn O’Malley:And I mean, this is not a new argument either, right? We basically discussed the same thing for almost all SaaS companies that we looked at here on the show. And I would say what makes this a bit different in my mind is that most other high quality SaaS companies that we discussed prior kind of still had a unique product. I mean, no other company can currently at least do what Adobe does. And the same might be true for Salesforce, although to a much lesser extent. So for them, and especially a company like Salesforce, I think that the main problem is that they will likely become some sort of database layer that is then used by AI agents instead of human users.

36:10Shawn O’Malley:And I would argue that this reduces pricing power, but it's something that we don't know yet and only the future will tell. And I think that could happen to Microsoft's Office products as well. So companies don't have to abandon Word. They just keep Word as the final format for the finished product. So they keep all the compliance workflows. They keep the governance infrastructure that Microsoft offers. And they also continue renewing the Microsoft subscription, which is obviously the biggest part of the profit pool. But while they're doing that, they just kind of resist further price increases on, you know, many of the price increases that Microsoft has basically come in form of new and more expensive bundles.

36:45Shawn O’Malley:So perhaps that's something that Microsoft cannot do in the future as easy as it has done that in the last two decades. So basically, that would mean that disruption not comes through this wave of cancellations, but rather as our pew. So basically the average revenue per user stagnating and, you know, companies just saying, we don't pay more for the same product that we even use less than we did five years ago. So even if the seed count told the perceived value of that seed would change. And I remember that just a couple of years ago, it was one of the most normal things to list Excel as a skill on your CV.

37:18Shawn O’Malley:So there were hundreds of courses online that, you know, taught upcoming investment bankers how to use Excel. And I'm just thinking about this now, but to me, it feels somewhat ridiculous to still list Excel as a skill on your CV today. And I mean, you can just go into Claw or OpenClaw, whatever you use as your favorite LLM, and they will create a model with at least the same quality in a matter of minutes. So this alone kind of shows you in a nutshell how much the skill advantage and the kind of embeddedness of a tool like Excel have been disrupted?

37:51Daniel Mahncke:The day up and coming investment bankers stop listing Excel as a skill on their resumes, that is the day I will declare Microsoft's office's moat as dead. Obviously joking there, but when you go just app by app, it is pretty remarkable to see the disruption, right? Word is arguably the most exposed to disruption risks. And so the activities that really represent the majority of time spent in Word, whether that's drafting from scratch, rewriting and restructuring, summarizing, reformatting bodies of text, that is precisely where large language models are already very good. And we're honestly good at that a couple of years ago.

38:35Daniel Mahncke:And so that is also probably the most used feature outside of AI, power users, those purposes. So, you know, we are guilty of that ourselves, although we certainly lack the skill or knowledge to really manipulate Microsoft Suite as much as we could using different AI tools.

38:55Shawn O’Malley:I just realized how little I know about LLMs when we recently had one of our, I would say, AI experts lead a call in the community. At least I wasn't alone because it turned into a hundred minute call just because no one wanted it to stop. And as you know, we even have a member who is part of OpenAI's team that is basically tasked with training their models in all of the financial stuff. And even for him, much of it was still new. So I certainly came out of the call a lot more bullish on AI and also a lot less bullish on software, not because it will go away, but simply because of this database agent dynamic that we discussed earlier.

39:31Shawn O’Malley:And also the fact that everything we've seen up until now is still only the beginning. So in theory, the tools that we currently use are the worst versions of LLMs and AI tools, and they will only get better. And that's kind of scary if you think about them, and especially how much better they have gotten in the last just two years. I mean, you said it, you could basically draft a Word document two years ago, but the amount of detailed work that, for example, Claude can now do, that's something that ChadGPT a year ago just couldn't do. Anyway, you're right to say, though, that Word is probably the most exposed.

40:04Shawn O’Malley:And integrating Copilot, which is kind of Microsoft's answer to the other AI models, just didn't really help either. Not only is it very limited in what it can actually do, at least with a satisfactory result, but most people also just don't start in Word anymore. So they go directly to Claude and then just make some final adjustments to the Word file that Claude has generated.

40:26Daniel Mahncke:So that's what baffles me, because why does Copilot struggle so much still with generating useful output inside of Word, right? I mean, it seems like creating tables or similar things like that are really what it struggles with most. And a lot of that stuff could really be a nightmare if you're having to do it yourself, right? I mean, formatting things in Word has always been a problem throughout my entire life. It can drive you utterly crazy. And it seems like even with large language models, Microsoft has not cracked the code on how to make Word useful to people, at least from reformatting things.

41:03Shawn O’Malley:I was talking about that kind of gives me flashbacks to my bachelor thesis. That was a nightmare. I mean, not the thesis itself, but the formatting in Word. I mean, you want to change the positioning of just one graph and your entire 100-page document is all over the place. But yeah, letting that behind me to answer your question. What was designed as a tool for human creation first? And that's something that I also got as feedback, especially from programmers lately. So the entire paradigm basically assumes that a human is doing the writing and the work in Word. And that makes it just not the best interface or just environment in general for NAI to do its work.

41:43Shawn O’Malley:So something like LaTeX, for example, is much better at that. And that's the different angle of competitive pressure where you had Google prior, and now basically reducing this skill part of using Excel helps Google Sheets. And at the other end of the spectrum, you might have tools that are just way more or way better fitting for AI like LaTeX, and they could get a boost that they have never seen before. So perhaps that's the long term risk that, you know, many people are thinking about when they think about Microsoft. So the less important Word becomes, because we use it much less, the more likely it is that a tool will take over that just works much better with AI than Word.

42:20Shawn O’Malley:Because it's AI first instead of assuming a human user as the main starting point.

42:27Daniel Mahncke:Let's just move on to the next app in Microsoft's productivity suite. PowerPoint seems like it's been similarly vulnerable to disruption. And the pace and quality at which AI can generate a complete slide deck has improved enormously just in the last four months, it feels like. I'll be the first to admit that Claude has helped me build a number of PowerPoints at this point. And I mean, I'll provide it with the information. I'll give it kind of a narrative structure. But then it does all the formatting and presentation work much better than I ever could. And honestly, like maybe a hundred times faster than it would take me to do it.

43:07Shawn O’Malley:It is incredibly helpful. Although I gotta say, just in the last couple of weeks, I got highly allergic to presentations that look like AI immediately, especially Claude, because you know, that's what everyone is using nowadays. So I think it's completely okay to feed it the data and then let it create an outline. But when I see that someone just uses the presentation that came straight out of Claude or chat gpd i kind of immediately discount the value of the presentation i know that's not the smartest thing to do it can still be your own data your own research i don't know it just kind of gives me the egg if i see that yeah i mean powerpoint's core value proposition has always been to lower the barrier to creating visually coherent presentations and that basically is what ai has now essentially collapsed so that barrier is now if it before was it like you know 20 it's now at zero it's just there is no barrier anymore you just you know give a hint or a prompt to claude and he will do whatever you want.

43:58Shawn O’Malley:And again, it feels so much easier to just ask Claude compared to Copilot in PowerPoint that it's just ridiculous. You would feel that Copilot should know Word, should know PowerPoint and all those tools better. But if you compare it to these generalist LLMs, they just tend to do a much better job. So a year or two ago, you could have still argued that AI is quite good at creating a first draft, but it has problems improving the presentation when you have things to adjust afterward, for example, with how the presentation looks. And that's a problem that's bigger than you would first imagine, because many companies have these individualized templates.

44:31Shawn O’Malley:So if AI can't use them properly, you really just can't use the presentation at all. And that's no longer the case either. I mean, if I want Claude to come up with a template for me to use, then I just upload some presentations that I have done before and tell Claude pretty much do exactly what these presentations have done and use this data. And the result is pretty good.

44:51Daniel Mahncke:None of that sounds very positive for Microsoft, right? I mean, is there anything that should make us bullish on the future of these Office products? This is a$70 billion profit pool after all. So if we're not feeling bullish about this, really the whole thesis feels impaired.

45:09Shawn O’Malley:I should emphasize that what we have discussed by now is basically the bear case. I mean, the unsettling part about that is that at least to me, it seems kind of realistic. And to some extent, it is currently playing out. Nevertheless, though, Microsoft Office is still the dominant player and even all the LLMs use it to generate their results. So I think we don't have to say that Microsoft has no chance of getting back here. And the bear thesis here could also quickly turn into a bull thesis if Microsoft were able to turn Copilot into an actual value add, which might only be a matter of time and organizing within the companies that use it.

45:45Shawn O’Malley:If that's the case, the runway itself is still huge. only 3.3 % of Microsoft's existing customers have paid subscriptions for Copilot. And if there's one thing that Microsoft excels at, it is bundling things together and just increasing the conversion of their product. And also generally, and that's something I learned recently, LLMs are about equally good. So some are better at specific things than others, but it mostly balances out. So the reason for Claude's sudden success, for example, has not only been its new model and how much better it is than, for example, ChatGPT, but mainly its ability to use user contacts in co-work or through its agents.

46:22Shawn O’Malley:And that seems to be co-pilot's struggle, although this should have been its advantage. So it's difficult to access all the necessary data because most companies, without even knowing it, have highly unsorted files. So basically files that meant to be confidential are technically accessible to people who shouldn't see them because nobody cleaned up the settings when, I don't know, project ended three years ago. So this existed for years, pretty much with no consequence, because most employees don't hunt through SharePoint folders. They're not supposed to, because they want to find access to files that they shouldn't have access to.

46:57Shawn O’Malley:When in turn, though, you turn on Copilot, and it starts reasoning across everything that your account has technical access to, that file mess quickly turns into a problem. And I've read about several enterprises that basically ran early pilots on Copilot, and kind of figured out that there was information surfacing that shouldn't have been surfacing and that employees have access to files that they shouldn't have access to. And that's obviously a huge problem if the main argument for using Copilot is that it only has access to the files and to your corporate identity that you want to give it to, right?

47:29Shawn O’Malley:So maybe one last thing I should mention about Copilot is that you have access to all kinds of models by now through Copilot. So you can also use Claude and you can also use all the other favorite LLMs that you have. And that's obviously a good way to make sure that people still access them through Microsoft and use Microsoft tools when they do. So there's certainly a future where I would say pretty much nothing changes and everyone stays subscribed to Microsoft Office. Only the only difference is fewer human hours in Word, Excel, or PowerPoint. And also if you bundle all the products together, you can also justify price increases even if the Office tools are used less than in the years before that.

48:08Daniel Mahncke:So to summarize, the major problem for Microsoft right now is that this is probably the first time in many, many years that you have even been able to come up with what is a plausible bear case against the office segment. And that just has not been the case for a long time. And if you're a company that traded at 40 times earnings, then the market is expecting really no signs of any potential risks. It's just not baked into the price of that kind of valuation. Otherwise, if uncertainty like this pops up, your stock is going to get dramatically re-rated. And well, that is a pretty quick summary of exactly what has happened to Microsoft stock in the last six months.

48:51Stig Brodersen

48:52Shawn O’Malley:I would say that's accurate. I think we've seen the same thing when we had our portfolio review episode and we talked about FICO, for example. If companies traded these premium multiples, you just can't have anything happen that just looks like a bear case.

49:03Daniel Mahncke:To your point, though, I have just seen recently that you can now directly integrate Cloud inside of Word, inside of Excel, and also inside of PowerPoint. And so at first, I think that does seem a bit odd because it is basically replacing exactly what Copilot is supposed to be there for, right? It's almost like they're acknowledging that their own model isn't good enough, maybe, or maybe it's just part of a more model agnostic approach to maybe put it more nicely.

49:33Shawn O’Malley:yeah i've seen that too i think it's still in better but i think this will ship pretty soon and i think it's not easy to say how this will turn out i mean on one hand this is obviously competing directly with copilot but on the other hand it at least ensures that people are still using you know the microsoft tools like word like excel like powerpoint so i think it kind of makes sense and also to some extent you still benefit because claude for example goes at least to some extent through Azure. So Microsoft is monetizing that through the cloud business.

50:05Daniel Mahncke:That's a good point. And perhaps that's a good time to leave Office behind us and talk more about Azure and Microsoft's cloud capabilities generally, where I think we have more leeway to be bullish responsibly, right? So to understand this segment properly, we might first want to discuss the three layers of cloud computing and where each major provider competes. Cloud has talked about a lot. It's one of those buzzwords you always hear, but understanding what that actually means is really not very straightforward. So at the bottom layer is infrastructure as a service, or you'll see this abbreviated as the acronym IAAS.

50:46Daniel Mahncke:And so this is raw compute, storage, and networking rented by the hour or the second. And you essentially get a virtual machine or container, and then you're responsible for everything that runs on it. And then AWS, Amazon Web Services, pioneered this model in 2006 and continues to lead with roughly 30 % of the global cloud infrastructure market share. Azure is at about 20 % to 22%. And then Google Cloud is at about 13%. And so the three of them control more than 60 % of a global market that crossed$100 billion in a single quarter in Q4 of last year, growing at a rate of roughly nearly 30 % year over year.

51:31Shawn O’Malley:It's pretty astonishing to me just how stable the market share has been between these three players. While all of them had fantastic growth rates, over the last few years, they didn't really gain share from one another. So it's a bit odd because technically, this is a business or a segment where there really isn't a huge moat naturally to that business. So the compute you offer is more or less commoditized. At the same time, though, it is very capital intensive to build the data centers, as we all know by now when just looking at their CapEx investments. And then when the market is dominated by the likes of Amazon, Google, and Microsoft, you can kind of imagine how much capital you would need to play in their league.

52:09Shawn O’Malley:So although it's technically difficult to differentiate your product, there are just high enough barriers to entry for this to be a profitable oligopoly. And there are also huge switching costs over time, which is why customers usually don't switch from, let's say, AWS to Azure or Google Cloud and also the other way around. So it basically comes from this factor that the more multi-part of the cloud business is the second layer, which is called platform as a service or P-A-A-S. Unfortunately, all those acronyms are not sounding as good as SaaS when you just spell them out. But this is basically where developers get pre-built services.

52:45Shawn O’Malley:So we're talking databases, machine learning platforms, especially for Google's cloud and some developer tools. And this is where Azure has a structural advantage since Microsoft owns GitHub, which is the dominant global code repository. Everybody who codes will likely know GitHub and also Visual Studio, which is the most widely used development environment in the world. So if you want to build your application on GitHub using Virtual Studio and want to deploy it, it's pretty much impossible not to use Azure as your cloud service.

53:17Daniel Mahncke:So Azure's competitive advantage is about having captured the developer relationship, which is where a lot of these cloud decisions are being made. And so that helps with onboarding new clients, especially at the enterprise level. Is that the right way to be thinking about it?

53:34Shawn O’Malley:Exactly. And it kind of goes back to this competitive advantage of just bundling everything, right? If your ecosystem is so huge that everything is interconnected, and we haven't yet mentioned this term of a flywheel today, but you can only talk about it, that where everything is going hand in hand with each other. And basically, one plus one equals three, right? And this is how Microsoft operates. And there's also the third layer of the cloud business above platform as a services, which is the actual software applications that business users interact with. And that's what we all know, even as consumers.

54:04Shawn O’Malley:So we are talking Microsoft 365. We're talking Teams. We're talking Dynamics. And Microsoft really is the only major hyperscaler simultaneously competitive at all three layers with an enormous installed base at the top. So AWS, for example, has virtually no enterprise software presence and Google Workspace competes at the application layer, but as we've discussed in pretty much 30 minutes, doesn't have Microsoft's death or enterprise penetration. So that has been a huge tailwind for Microsoft's cloud. And it's in part why they could have this second position in this incredibly profitable market.

54:39Shawn O’Malley:Let's take a quick break and hear from today's sponsors.

54:42Daniel Mahncke:Curious about online trading, but haven't taken the first step yet? You're not alone. And Plus500 Futures is a great place to start. The futures markets are moving fast. And with Plus500, you can explore popular assets like oil, gold, S &P 500, Bitcoin, and more. From crypto to commodities, there's always something happening. The platform is super easy to use, so you can trade on the go right from your phone. You can get started with just $100 and jump into the action. See something interesting? Once your account is open, you can trade it in just a couple of clicks. And if you're not quite ready yet, you can practice with a free demo account.

55:19Daniel Mahncke:No risk, no pressure. With 20 years of experience, Plus 500 makes trading more accessible than ever. Check it out at plus500.com. Trading and futures involves risks of loss and is not suitable for everyone. Not all applicants will qualify. Plus 500. It's trading with a plus. Okay, be honest. How many times have you been lying on the couch, scrolling your phone late at night, and you see that one thing you've been looking for? You tap the link, throw it in your cart, maybe browse around a little more, and then you hit checkout. And that is when it hits you. Your wallet is across the room. You have no idea what password you use for this site.

55:56Daniel Mahncke:And suddenly, buying a$30 item feels like a whole project. But then you see it. That purple button, shop pay. One tap and you're done. All your info is already saved. No fumbling, no frustration, just that purple button. That's Shopify. And if you're a business owner, that button is a game changer. Shopify has the best converting checkout on the planet, meaning fewer abandoned carts and more sales actually going through. But Shopify isn't just a checkout button. It's the commerce platform behind millions of businesses and 10 % of all e-commerce in the US. They've got hundreds of ready to use templates so you can build a beautiful online store that matches your brand.

56:35Daniel Mahncke:Plus, Shopify is packed with AI tools that write your product descriptions, create page headlines, and even enhance your product photos. It's like having a full team without the full team price tag. See fewer carts go abandoned and more sales go with Shopify and their ShopPay button. Then sign up for your$1 per month trial today at shopify.com slash tip. That's shopify.com slash tip. No, it's not your imagination. Risk and regulation are ramping up, and customers now expect proof of security just to do business. If you're a founder or business leader, you already know this. Every new deal, every new partnership, somebody's asking to see your compliance credentials.

57:20Daniel Mahncke:And getting that stuff together manually, it is a massive time sink. That's why Vanta is a game changer. Vanta automates your compliance process and brings compliance, risk, and customer trust together on one AI-powered platform. So whether you're prepping for a SOC 2 or running an enterprise GRC program, Vanta keeps you secure and keeps your deals moving. And here's a stat that really got me. Companies like Ramp and Ryder spend 82 % less time on audits with Vanta. That's not just faster compliance, that's more time you're spending on actually growing your business instead of drowning in paperwork.

57:57Daniel Mahncke:I love that over 10 ,000 companies from startups to big enterprises trust Vanta to handle this stuff so they can focus on what actually matters. Get started at vanta.com slash T-I-P. That's V-A-N-T-A dot com slash T-I-P. All right, back to the show. I don't want to be the one to keep bringing up the bear case, but in my defense, Microsoft CEO Satya Nadella is the one who's really bringing it up and beating me to it. And so he talked about what happens to the platform layer, specifically when AI agents can just bypass it and what the implications of that are. So what are your thoughts on that and what that means for the future of Microsoft's business?

58:43Shawn O’Malley:Satya Nadella is not holding back on the massive changes that might come for Microsoft. And I mean, that's also why he doubled down so massively on Copilot. And he basically didn't want to get disrupted. You know, he wants to be the first mover. He does not want to be the victim of the innovators dilemma. Generally, this is what you want to see from a CEO. But it's a fact that the execution part has really been lacking. So the comparison might be harsh, but it even somewhat reminds me of what Mark Zuckerberg did with the metaverse. So he basically rebranded the entire company and invested hundreds of billions, and it all ended up failing miserably.

59:19Shawn O’Malley:And Microsoft now has almost 80. It's 80 different products with the name Copilot in it. So their new laptops even have their own Copilot keys. And yet, it is still such a big disappointment up until now. So coming to your question, and this is kind of an introduction to that, why might platform as a service be getting disrupted? So the reason the platform layer has been so valuable and so sticky is that it basically solves a complexity problem for software developers. So if you want to build an application that needs user logins, for example, you don't have to build your own password database, your own multi-factor authentication system, or your own single sign-on infrastructure from scratch.

1:00:00Shawn O’Malley:And it all sounds pretty complex when I say it out loud, but it's just that Microsoft really takes a lot of the things that you would otherwise have to do on the backend and simplifies it. So instead of building all these password databases, You just call Microsoft Entry ID, which is this identity management service at the platform layer. And that service takes care of everything that you need. And the same exists for database management, message queuing, and also analytics. So the platform layer is a collection of these pre-built tools that developers can use. And the reason that creates stickiness is that once you build your application on top of Android ID or Azure SQL, or even Azure Machine Learning.

1:00:40Shawn O’Malley:Those services are basically integrated into your code. So the problem comes from what Nadella has described as being CRUD plus business logic. And CRUD stands for create, read, update, delete, plus all set of rules. It sounds pretty complex, but this is what most software basically is. And an AI engine can now generate its own logic. And that pretty much changes how everything works. So it does not need all these abstraction layers that we just talked about. It can go directly to the underlying database, get whatever it needs, apply its own logic, and then return the answer. So say someone wants to know the total accounts payable owe to vendors in a specific region.

1:01:24Shawn O’Malley:Before AI, you as a user would open software, which calls platform services, which then access the database, which then runs calculations, which then returns a result through the software interface. So there are just a lot of layers involved and you as a normal employee wouldn't know that, but that's the case on the backend. An AI agent though, just goes directly to the database, finds the data, does all the reasoning, and then gives you the answer. So it makes it just significantly easier. And you don't need all these platform layers because you as an employee, you just don't go through them anymore.

1:01:58Daniel Mahncke:And so that's why Microsoft is trying to turn Azure into a platform for AI agents themselves through Copilot Studio and this agent framework. And essentially, it's a race to replace the old platform layer with a new one before the old one disappears. And you have this like melting ice cube dynamic.

1:02:18Shawn O’Malley:That's the bet you're taking. I mean, the question is whether they can execute that transition, what it basically is, faster than the erosion happens. And that is not guaranteed, right? which basically means there's some sort of uncertainty. And as we all know, markets don't like uncertainty.

1:02:33Daniel Mahncke:And there is one more dimension to this that I think is worth talking about. Even if Microsoft successfully deploys AI agents at scale, the question then is who actually acquires the economics of those agents? That remains up for debate.

1:02:50Shawn O’Malley:That's a big question. And I think it really depends on what layers of the entire process you still control. I try to keep it as simple as possible, but I will still want to explain the layers. And then also please give us feedback in the comments if you're an expert on this and you kind of have a theory on how it all works out. So the first layer is the software where the agent's output lands. So that would be Word, that would be Excel, that would be PowerPoint. The second layer is the platform managing what the agent does, what data it accesses and how it's governed across the organization. What does that mean?

1:03:24Shawn O’Malley:Basically, what file can they access and what file can they not access? And then the third is the actual AI model generating the output. Okay, so that's Claude, that's ChatGPT, those kinds of things. And then fourth is the compute infrastructure. So that's the GPUs physically running the model in a data center somewhere. And Microsoft's aspiration is to own all four of those layers. And then the question is how much of that stack they can actually control. And that mostly depends on whose agent is being deployed.

1:03:55Daniel Mahncke:I think we need to dumb things down for me, Daniel, because I like tangible examples. So maybe just to go through some scenarios, what happens if an enterprise uses Microsoft's software, so Excel, Word, Teams, but then deploys a third-party agent? So they're using Cloud or ChatGBT.

1:04:16Shawn O’Malley:so in that scenario microsoft obviously keeps the office subscription because the output still lands in either word excel or powerpoint but the platform that manages the agent's behavior goes to the model provider or to no one if the enterprise builds it itself so for example that's what salesforce is aiming to do then all the value would go to salesforce the inference margin which is just the profit margin left after accounting for the operating costs of running ai models is running at about 60 % right now. And obviously all of that would go to Anthropic, OpenAI or Google. So Microsoft sees none of that.

1:04:52Shawn O’Malley:And then if the enterprise is running a GPT-based agent, then OpenAI's models run on Azure. So Microsoft captures the compute revenues on that. But Anthropic's primary cloud provider, for example, is AWS, not Azure. So if an enterprise deploys a cloud agent, most of the compute revenues would go to AWS. And of course, the same is true with Google's Gemini models, would run obviously on Google Cloud. So in scenario one, Microsoft's compute capture depends entirely on which third party model the enterprise actually chooses. So they get it for OpenAI, but they only get it for them, not if you use Cloud, Mistral, or any other model.

1:05:31Daniel Mahncke:Okay. So going to scenario two, let's say it's Microsoft software plus Microsoft's own agent being, let's say, co-pilot. What does that look like?

1:05:45Shawn O’Malley:That's the ideal outcome for Microsoft economically. So they keep the office subscription. Again, they own the agent layer and the compute runs on Azure. So the catch is that co-pilot today still runs primarily on OpenAI's models, meaning Microsoft effectively still has to pay OpenAI a cut on every single co-pilot response generated. So even in scenario two, they're not capturing the full value chain yet. That's why Microsoft is investing so heavily in building its own AI models and also its own chips. And the goal is basically to gradually shift Copilot off OpenAI's models and onto their own so they can stop writing that check and really keep the full margin to themselves.

1:06:25Daniel Mahncke:And then there's a third scenario in which Microsoft's orchestration layer wraps a third party model that then runs on Azure. Tell me about that situation.

1:06:37Shawn O’Malley:Well, this is what Azure AI Foundry is positioning for. And it might be Microsoft's most durable long-term position if they can pull it off. So in this scenario, enterprises use Cloud, Llama, or I don't know, even Mistral as the underlying model. They access it through Microsoft's managed platform. So you can imagine it as somewhat of an app store taking a cut on everything that runs through the platform, even though the intelligence itself belongs to Anthropic. Meta or whoever has the LLM. And the reason this is the most durable position is that it doesn't require Microsoft's own model to actually win that LLM race.

1:07:17Daniel Mahncke:It sounds great in theory, but what's the risk that the model providers just bypass Microsoft and go direct to enterprise customers, right? If Anthropic is good enough, why does a large enterprise need Microsoft in the middle at all.

1:07:33Shawn O’Malley:I mean, that is a possibility. But since Microsoft is already integrated in all of these large corporations on earth and has so much more experience in both the governance and also the regulatory world, I do see why handling it through Microsoft is just a sort of safety layer for enterprises. But even in the best version of Microsoft's platform strategy, the economics of the AI era might look more like Spotify than like the Office era. So Spotify is arguably be the best music distribution platform ever built. So an easy to use product. It's only for music, enormous scale and a dominant market position.

1:08:11Shawn O’Malley:We just covered it here on the show a couple of weeks ago. And still, the labels still capture roughly 70 % of the revenue, leaving Spotify with relatively thin margins compared to most other platform companies. And Microsoft software era was pretty much the opposite. They were the creator, they owned the content, the margins were extraordinary. And the AI era could shift Microsoft toward the distribution side of the ledger. And that doesn't mean it's a bad business. So Spotify at scale is still a large, important, profitable business, but it is a different margin profile. And I think justifies a different multiple.

1:08:48Shawn O’Malley:And I think that is generally part of why the market is repricing Microsoft right now. And actually, we just spent a couple of minutes also figuring out in our heads how exactly it works that you have all these circular agreements where basically you have partnerships with open ai or in amazon's case with cloud and then all of this compute is to some extent business for you and it's you know part of the growth numbers of the cloud business and at the same time a lot of it is actually paid by vouchers for example in this case so it's not actually money that you receive most of the money for example for copilot still comes from the subscription and service.

1:09:26Shawn O’Malley:So then you can open up the entire argument of who is actually benefiting the most. And I think we'll still talk about it. But if you think about that, and then the relationship between Microsoft and OpenAI, it seems more and more that OpenAI is the main beneficiary here, not really paying for the compute, at least not with real money, only to some extent starting in the last couple of years. And it all makes it so incredibly difficult to understand who exactly will make money, where the margin will be in five years. And it just increasingly feels like this just goes in your too hard pile.

1:09:57Daniel Mahncke:You took the words right out of my mouth, but we still got to push through and understand the rest of the business and think about the valuation before we make any decisions about the too hard pile. And I guess the question for investors isn't really, does Microsoft make money from AI agents? Because they almost certainly do to an extent in all three scenarios. It's more about how much of the value chain they'll own and what that will mean for the margin structure of the business over the next 10 years. The famous example of this that people like to throw around is the airlines. Think about them as creating this massive utility worldwide.

1:10:32Daniel Mahncke:A hundred years ago, I could not just get on a plane in DC and meet Daniel in Frankfurt. It did not exist. In the meantime, these companies have created a way for people to be able to travel anywhere at honestly pretty reasonable prices when you think about what is actually happening to take you from one point to another point on the other side of the globe. And yet airlines are historically very, very unprofitable businesses that have, Buffett has famously said that in aggregate, airline stocks have destroyed more shareholder value than they've created. So we were talking about capturing the value.

1:11:10Daniel Mahncke:It's one thing to make a great LLM and to be involved in this AI race, but then can you actually capture that value. And with airlines, they have not captured that value. Almost all of that surplus value has gone to consumers. We all benefit greatly from the existence of airlines and airlines have really not been able to profit from that. So that's a little tangent, but I do think it's a helpful lens to think about what it means when we say capturing value.

1:11:43Shawn O’Malley:I think it's important to understand that because we all are prone to this recency bias of thinking that the big tech companies mostly are the companies that end up with most of the value for whatever product comes out. But I do feel that to some extent, and it's only gut feeling because understanding this all entirely is pretty difficult. But when I researched Microsoft, it felt more and more like perhaps this paradigm comes to an end, at least for the type of business model that Microsoft has. And this is not a prediction. It's just that it's incredibly difficult to model or kind of understand.

1:12:13Shawn O’Malley:And we just don't know yet which scenario is the most likely to play out. And if enterprises broadly adopt Copilot, Microsoft owns the stack and compounds at perhaps software era economics on a much bigger scale or dominating a much bigger market. But if enterprises use a third party model, especially if Claude stays on AWS, and that becomes the biggest player in the field, that would be a huge hit to Microsoft. And perhaps you're kind of laying there with this format landlord status that's been systematically excluded from the high marginal layer on the AI value chain. Again, the answer is almost certainly somewhere in between those two extremes, but somewhere in between can still mean a lot of different things for a stock that has historically been trading between 30 to 40 times earnings.

1:13:01All right.

1:13:02Daniel Mahncke:Well, that's enough on that. Let's talk more about this OpenAI partnership. From the outside, it looks like this could have and perhaps should have huge implications for Microsoft. Most investors thought that it would, with Microsoft being able to use chat GBT models for co-pilot and how that would have been a major edge. And just 18 months later, OpenAI has gone from being really the anointed champion of AI that seemingly deserves to be a trillion-dollar company to now seeming like, in many ways, it's behind the competition. And we may look back and determine that OpenAI made the fatal choice of getting bogged down with consumer products, while the real money was to be made in B2B AI products, which is exactly what Anthropic has prioritized.

1:13:50Daniel Mahncke:And yeah, I don't know if you ask me, honestly, that may end up being one of the biggest oversights in 21st century business history. I mean, there will be case studies on it in colleges, you know, 10, 15, 20 years from now, if things keep moving in this direction. So why has Microsoft not benefited more from its relationship with OpenAI and leveraged that tech to make superior B2B products itself? Because now it looks like they both have massively dropped the ball.

1:14:21Shawn O’Malley:Microsoft made its first investment in OpenAI in 2019. Back then it was a billion dollars at a time when most of the world hadn't even heard of a company or even just AI. And they followed with additional investments in 2021 and then in early 2023, committing$13 billion more in a deal, giving them roughly 49 % of economic interest and an exclusive cloud partnership that we just mentioned under which OpenAI's workloads run on Azure. But again, a lot of that money are actually these redeemable vouchers. And it actually didn't go from Microsoft in terms of actual cash to OpenAI. And as you said, OpenAI is no longer seen as the best model out there, and it might have missed the best opportunity to play its AI elite.

1:15:05Shawn O’Malley:And of course, that makes people ask whether OpenAI will even be able to come up with all the money that it promised companies that they make business with, right? I mean, for Microsoft, if you just pay them with vouchers, that's totally fine. But there's so many other companies that they have basically promised hundreds of billions of dollars. And you kind of ask yourself, how are they even supposed to make that money? About 45 % of Microsoft's$625 billion backlog is tied to these OpenAI relationships. And those vouchers are not enough to cover that. Again, they already start also paying real cash to Microsoft.

1:15:36Shawn O’Malley:And OpenAI is far from being profitable. It's burning through cash so fast that it really needs these continuous large funding rounds. So I don't know, its ability to fulfill a quarter trillion dollar commitment to Azure alone. And it kind of depends on just continuing to grow revenue fast enough to then generate the cash to pay those Azure builds.

1:15:57Daniel Mahncke:A quarter trillion, Daniel. Come on. That's so much money. And then just recently, Cloud actually announced a revenue run rate of over$30 billion. That was very much in the news, topping OpenAI's run rate. So we already see competition overtaking OpenAI. I mean, that has already happened. And that does not help with their additional funding needs. And another negative from Microsoft came when OpenAI announced something called the Frontier deal with Amazon. So maybe you can tell us more about that.

1:16:29Shawn O’Malley:That's the other part of the OpenAI relationship that is just a bit odd. So OpenAI launched a new enterprise AI agent platform called Frontier and signed a$50 billion exclusive cloud deal to run it on AWS instead of Azure, arguing that this new product pretty much falls outside of the scope of the existing exclusivity agreement with Microsoft. So despite the ownership sake of Microsoft, it kind of seems relatively easy for OpenAI, at least if this holds up, to just create new partnerships that basically bypass Microsoft. And if that's the case, you kind of have to ask yourself with every new OpenAI product, will this be channeled through some competing cloud provider rather than exclusively to Microsoft and its cloud.

1:17:13Shawn O’Malley:And then there's also the problem with OpenAI's conversion from this non-profit organization, which they currently still are, to a for-profit company. And there you have all this uncertainty about whether Microsoft's 49 % stake or economic interest actually fully converts into a comparable stake in this new for-profit entity should they get there.

1:17:33Daniel Mahncke:The relationship has started to resemble those early partnerships in the smartphone era, where you'd have the platform providers, Apple and Google, they initially needed carriers like AT &T or Verizon for distribution and funding. And then once they had enough scale, you saw that power dynamic just completely reverse. And so OpenAI started as a dependent and well, look, now it's valued at almost an$800 billion valuation and is increasingly able to act in its own interest, with its own autonomy.

1:18:04Shawn O’Malley:It feels like the OpenAI partnership kind of resembles pretty well how the power dynamic generally has shifted. So in the last 20 or so years, big tech companies just bought whatever company they wanted, and either they integrated them into the ecosystem, or they just killed it off to basically just avoid competition. And no matter what they did, they certainly decided how things turned out. With OpenAI, I think it feels like there has been, for the first time in a long, long time, a company that grew so fast, and to an extent that no one could have seen coming, that they grew too big to be controlled even by companies like Microsoft or Amazon.

1:18:41Shawn O’Malley:And instead by now, it kind of looks like Microsoft lost control over them. And when they realized that, they were starting investing even more heavily into the AI model program run by themselves, which is called MAI-1, very simple name once again, developed toward GPT-4 class capabilities. Of course, at lower inference costs, running on their own Maya 200 chips. And the model agnostic founding model was also launched because of this. So Microsoft is doing a lot, but again, this also doesn't simplify the business generally. All right.

1:19:14Daniel Mahncke:So we have one more thing I want to talk about before we get to the valuation, and then we can kind of wrap up today's discussion. And that is to talk about CapEx, which is a conversation being had of many of the Mag7 right now as hundreds of billions of dollars are being invested into AI and the impact it has on financial. So what does that look like for Microsoft?

1:19:36Shawn O’Malley:As you said, as every other hyperscaler, Microsoft obviously increases CapEx spend as well. And just the six months, the first six months of the 2026 fiscal year, CapEx is more than$70 billion. And the guidance is between$120 to almost$150 billion. And for context, in fiscal 2023, CapEx was in total$28 billion. So Microsoft now spends more on CapEx in a single quarter than in an entire year, just two and a half years ago. I think looking back or kind of zooming out, you kind of realize why the market is so scared because of these investments and the CapEx spend, because even for those companies, it's so much money.

1:20:14Shawn O’Malley:And roughly two thirds of the quarterly CapEx is going towards short-lived assets, so primarily GPUs and CPUs. And they have, at least that's what it's currently saying, a useful life of about three to five years. And the depreciation charge runs through the income statement. So we already know for a fact that there will be massive depreciation charges coming in the next couple of years, certainly lowering margins.

1:20:38Daniel Mahncke:So the ROI question then would be at what revenue generation level does all of this spending pay for itself? And then on what timeline, right? When will that happen?

1:20:50Shawn O’Malley:I fear I bring up too many numbers again, but if Microsoft is spending$140 billion a year on infrastructure. And those assets need to at least earn the cost of capital. You would, let's say, want a minimum of 12 to 15 % return. That would imply that you have about$17 to$20 billion of incremental annual revenue generation from the infrastructure just to break even. Azure is already generating about$75 plus billion in trailing revenue, growing at, again, close to 40%. And at that growth rate, Azure adds roughly$28 billion in you annual earned you, which does clear the cost of capital bar, if you believe the capacity being built now will actually be utilized at the current growth rates for the foreseeable future.

1:21:35Shawn O’Malley:I think the concern is about the margin of safety here. Because if Azure growth decelerates to, let's say, 25%, while CapEx still stays at these levels, then the returns don't look that good anymore. And, And 25%, we just talked about how potentially the current growth numbers are inflated by these circular investments in OpenAI. So it does open up a lot of questions to investors. And since growth even came in slightly below expectations, again, not 40%, only 39%, that's one of the market's fears right now. And although Microsoft's CFO has said in the latest earnings call that demand for Azure is certainly there and growth could have been a couple of percentage points higher, if Microsoft wouldn't have decided to allocate compute capacity to other parts of their own businesses first.

1:22:20Daniel Mahncke:Preston Pysh, MD, To quote her, she said, the first thing we're doing is solving the increased usage in sales and the accelerating pace of M365 Co-Pilot, as well as GitHub Co-Pilot, our first party apps. Then we make sure we're investing in the long-term nature of R &D and product innovation. And much of the acceleration that I think you've seen from us and products over the past is coming because we're allocating GPUs and capacity to many of the talented AI people we've been hiring over the past years. And just continuing with her quote here, there's one more paragraph. She says, then you end up with the remainder going towards serving the Azure capacity that continues to grow in terms of demand.

1:23:03Daniel Mahncke:And a way to think about that is if I had taken the GPUs that just came online in Q1 and Q2 in terms of GPUs and allocated them all to Azure, the KPI would have been over 40. And I think the most important thing to realize is that this is about investing in all of the layers of the stack that benefit customers.

1:23:23Shawn O’Malley:But this quote alone, you might realize that if you will listen to an entire earnings call, how often they would say the word copilot, and maybe also how complex the business is and how often they talk about the GPUs, the CPUs, how that It changes the economics. And it's all pretty difficult to understand. I think if people tell you that they fully understand what's going on at these companies right now, they're kind of lying. I think it's just incredibly difficult to follow everything that's happening. And beyond what the CFO just said, all CEOs of the big tech companies saying the same thing.

1:23:53Shawn O’Malley:They would rather overshoot and invest too much than underinvest and lose their competitive position. And if Microsoft underinvests and enterprise AI demand is even greater than projected right now, they end up behind on capacity at exactly the wrong moment in time. So losing customers, at least potentially to AWS and also to Google Cloud, who just made the bet earlier. So to some extent, I feel there's peer pressure. We just have to have these investments just to not potentially lose market share to direct rivals. And I think I do agree with that strategy. If we are at such a big changing point, you would risk losing basically your entire business or at least your position by underinvesting, well, overinvestments themselves would only mean a couple percentages less in returns.

1:24:36Shawn O’Malley:I mean, we've seen that again with Mark Zuckerberg's metaverse. That was a$100 billion flop. I mean, literally no returns at all. And still, it didn't really hurt meta in the long run. So I think the risk of losing your position is much, much larger than the potential of underinvesting right now.

1:24:53Daniel Mahncke:And the metaverse was a complete moonshot. Whereas we already have seen what AI can do and how it can impact business models and profitability. And after spending a lot of this episode on the challenges that Microsoft is facing, how about you give us just a refresher on the bull case? Because I think we're already forgetting why we were initially bullish on the business.

1:25:14Shawn O’Malley:Just a brief one. I mean, we already talked about the three different scenarios that could come for Microsoft. And let's be honest, it's hard to imagine that Microsoft will not play a vital role in software or AI in the future. So the company is still positioned to potentially succeed in all parts of it, the entire value chain. So offering the software tools, running the cloud and the compute, and having a competing model, or at least offer all major ones through their co-pilot product. A second bull argument pretty much writes itself if you just take a look at the numbers. I mean, Microsoft is still generating approximately$350 billion in annualized and to a large extent recurring revenue, growing at 17 % per year.

1:25:52Shawn O’Malley:And the operating margin over the entire business is 45%, which is just insane at this scale. Even free cash flow conversion is still looking good, although it has gone down from about 80 % historically to about 50 % today, just mainly due to the CapEx investments. And the company is also returning cash to shareholders. It's about 40 billion in fiscal 2025, mostly through dividends and buybacks. So that's more than any other company in the S &P 500 in absolute dollar terms. Still, unfortunately, just due to the size of the company, you have a low dividend yield and also a low buyback yield. So they have about$90 billion in cash as well.

1:26:27Shawn O’Malley:Obviously a financially stable company, that's not surprising to anyone who listens to this podcast. So yeah, if you just summarize, this is a company that can fund an $80 billion capital spending program, return$42 billion to shareholders, and then still show a huge cash pile in the bank.

1:26:44Daniel Mahncke:And if we want to keep being bullish, there's that backlog number that we should revisit. And so just even setting aside the open AI portion,$344 billion is committed in non-open AI revenue, growing 28%. So that does not look like a dying company.

1:27:03Shawn O’Malley:And when a Fortune 500 company signs, let's say, a three-year Azure commitment, they're essentially making a statement that, hey, we are all in on Microsoft's AI infrastructure, So it's not very likely, at least to me, that they plan on leaving the Microsoft ecosystem anytime soon. Another bull argument is also that they have this custom silicon story, pretty much similar to what AWS or Amazon is also doing. So the so-called Maya 200, we already talked about them, are deployed at scale in Microsoft's own data centers. and early benchmarks, at least the one that Microsoft itself is putting out, suggest that it delivers competitive inference performance at meaningfully lower costs per token than equivalent NVIDIA GPU configurations.

1:27:46Shawn O’Malley:Let's say Microsoft can shift even 20 to 30 % of its AI inference workloads from NVIDIA GPUs, which we all know are incredibly expensive, to their own chips over the next, let's say, two to three years. I think that alone would mean margin improvements that will be very significant for the overall company. Again, Amazon is doing the same and it seems like it's working out pretty well. Finally, before we actually go to the valuation, you also have GitHub, which shouldn't be underestimated. It's 100 million registered developers. So GitHub Copilot has, I think, the clearest demonstrated ROI of any AI tool in this category, especially for Microsoft, because not only does it generate revenue for you right now, it's also developers writing code 20 to 30 times faster in ways that show up directly in the shipping times for future products.

1:28:34Shawn O’Malley:So there's a lot of debate about how much more efficient AI is actually making people. In terms of writing code, it's actually a thing that is the real deal. And this also matters because developers often make the cloud infrastructure decisions at these big enterprises. So you benefit from it in many ways. So when a developer's entire workflow is on GitHub and Visual Studio, I think the path of least resistance, and we kind of mentioned it multiple times today, is also deploying in Azure.

1:29:01Daniel Mahncke:$7.5 billion for the world's most dominant coding platform, which is also now the world's most adopted AI coding tool, and then also the primary channel through which Microsoft acquires developers who make cloud decisions. I mean, in retrospect, that just looks like maybe one of the best technology acquisitions of the last decade.

1:29:23Shawn O’Malley:It's one of the positive examples of Microsoft's acquisitions. And it's why I do give such an Adela capital allocation instincts significant credit, even while asking some hard questions about the execution in recent years, just overall. I think the perpetual license to a subscription bet was correct, which obviously has been a long time ago, but still, to some extent, we gave Adobe CEO credit for that, not only because it happened, but also because it kind of shows the skill of adapting your business model to changes in the future, which is currently happening. Then the Azure bet was obviously also correct.

1:29:57Shawn O’Malley:And then you have GitHub and you also have OpenAI, which doesn't sound like the best deal anymore, if we look at it today. But honestly, three years ago, AI was completely new. Nobody knew about it. So that is a win.

1:30:10Daniel Mahncke:All right. Well, we've teased it enough. Let's get to the valuation. is Microsoft our Google of 2026, or is it not as attractive as it might look at first glance?

1:30:21Shawn O’Malley:And I got to say, it's certainly way more complex, which is probably not a good thing. But in terms of valuation, I've divided my model into a base and a bear case. The base case assumes Microsoft kind of successfully navigates the AI transition on roughly the terms that management has described. So revenue compounds at about 17 % annually through the next five years, which would be fiscal year 2030. And that might sound aggressive, but if you consider that Azure is still growing in the high 20s to low 30s and becoming a bigger part of the pie, and also co-pilot adoption, just improving from, let's say, 3.3 % toward 10 % to 12%.

1:30:59Shawn O’Malley:I mean, for that, obviously, you wouldn't need it to be a much better product. But on the install base that Microsoft has. And I believe they will find a way to bundle that. It's not like it's a heroic assumption at all. We still talk about five years out. So the key assumption in this model is on the margin side, and basically means that free cash flow margins do recover modestly from today's depressed levels because of all the capex. So that means we go to somewhere in the mid 30s in five years time. And that basically means that the capex build out kind of peaks and that depreciation at that point stabilizes.

1:31:33Shawn O’Malley:And on that basis, I think normalized free cash flow per share could be around$30. And if you then say, hey, well, this is a company, still one of the best in the world, probably deserves a 25 times exit multiple, then I would say that you could get a fair value today of around$500 per share. The company is complex. We don't fully understand it in all parts that it has ready for us. So I would say a 20 % margin of safety for the stuff that we don't know is probably fair. And then you would end up at about$410, which is pretty much exactly where the stock is trading today. And you could say there's a bias between making the model and suddenly it's exactly the price it is today.

1:32:08Shawn O’Malley:But when I built it, the price of Microsoft was originally still $375. At that point, the return that you could have expected was about 16%. At today's prices, it's a bit closer to 14%.

1:32:21Daniel Mahncke:Well, I feel like we've already done it, but shall we discuss the bear case in your valuation?

1:32:26Shawn O’Malley:Well, we've certainly discussed the bear case and all of its qualitative factors. So in terms of numbers, I expect the top line growth to slow down slightly to about 13 to 14%. So it's not that I expect the company to just not grow anymore. But the bear case is mostly about what happens to the margin structure. If the commoditization thesis that we discussed, especially with AI, is actually playing out. So in detail, I do think that we could see margins compressed to about 35%. I'm talking about the operating margins here, in five years' time. And that mostly means you see a shift from this high-margin software company toward a more capital-intensive infrastructure provider.

1:33:04Shawn O’Malley:And I think that margin trajectory reflects pretty much a world where the AI value chain economics that we've discussed at length today favor the model providers, where Microsoft's platform as a service stickiness kind of erodes as AI agents bypass that layer, and where the infrastructure build-out never fully earns its cost of capital. If you would then assume that we have an exit multiple, maybe at about 20 times, which kind of makes sense because you have a lower margin business, it's also not growing as fast, then you would get to a fair value that's significantly lower, somewhere in the 280s.

1:33:37Daniel Mahncke:For as interesting as today's episode has been, man, I've hardly ever been so confident about a company being in my too hard pile. Thinking about good old fashioned value investing, man, this just feels complicated. And we talk a lot about using our consumer insights to shape how we invest. And honestly, I do not use Microsoft at all somehow. I'm like an anomaly. I use Google's productivity suite personally and for work. I'm not a coder. I don't use GitHub. I don't use Copilot. I'm very loyal to Claude and Gemini. I have an iPad and not a Surface tablet. And actually, I was really kind of raking my brain for this.

1:34:14Daniel Mahncke:My only connection to Microsoft that I can really directly think of is my Xbox that I've had since college. So a lot of this was all very new to me, which makes it difficult to kind of digest everything. But then when you layer in how their massive capex is going to affect the income statement, while this huge profit pool in their productivity suite seems quite ripe for disruption, it is hard for me to get really excited about the company. I'm sure it's great, but I'd probably be much more excited to just add more to some of the businesses I already know, like Amazon, if I want more Mag7 exposure.

1:34:54Daniel Mahncke:And then with Alphabet too, we already have a significant amount of exposure to great cloud businesses in our intrinsic value portfolio. And that really seems to be the most compelling part of the Microsoft story anyways. So it feels like taking on new risks, but not necessarily getting much more in return, except for perhaps a better valuation, at least on paper compared to Alphabet. And so, I don't know, maybe Microsoft will grow on me, but it's not so obviously a great opportunity to me that I need to invest in it over many of the other companies we follow for our intrinsic value portfolio.

1:35:29I agree.

1:35:30Shawn O’Malley:I think I was a bit surprised actually by how many reasonable bear cases there are for Microsoft. And we might get a bit of pushback here on this episode, because I know there are a lot of people who love Microsoft. But when I look at companies like Amazon, I just find it hard to find as many bear cases. And I do believe that when we looked at Google about a year ago, there was certainly a lot of bad press. And still, if you looked at the business model, it seemed easier to understand. It kind of seemed easier to see how this is actually completely mispriced by the market. Of course, Amazon is a company that is trading at higher multiples than Microsoft is right now.

1:36:04Shawn O’Malley:But Microsoft is also not the deep value play either. So it's not that you would buy it because of its cheap valuation. So I do think that after looking into it and getting even more into this AI field and what exactly the economics of the cloud computing are and these circular investments with OpenAI, I do feel like it's going on the too hard pile. And I feel like you might learn more over time. You might have some feedback in the comments of people who understand it way better, who think that they could explain it to us. If so, please do. But I do believe that currently it is on the too hard pile.

1:36:35Shawn O’Malley:Put it on a watch list, but not in our portfolio. And with that, I would say I close it for today with a quote by Satya Nadella, of course, who was the CEO of Microsoft. And he said, our industry does not respect tradition. What it respects is innovation. And while there's a bad case to make for Microsoft, I'm pretty confident Microsoft will not go down because it missed out on innovating and kind of disrupting itself. So the question is whether it will be able to do that without disrupting its most profitable business before replacing it with a better one. I think only the future will tell if that's the case.

1:37:11Shawn O’Malley:And with that, see you all in the next episode. Thanks for listening to TIP. Follow the Investors 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.

1:37:45Shawn O’Malley:Nothing 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