Latest / Deep Values: Value Investing Deep Dives / Microsoft: Navigating the AI Transition and Market Valuation
Transcript
- 0:00What if the very AI tools that you are, you know, using right
- 0:04now to work faster are quietly destroying a $70 billion profit
- 0:09pool? It is a wild thought, right?
- 0:11It really is O Welcome to the deep dive.
- 0:14Today we are unpacking this fascinating paradox involving,
- 0:18well, one of the most dominant tech giants of the last 30
- 0:22years, which is Microsoft. Yeah, Microsoft.
- 0:25Right, and before we dive into that, just a quick housekeeping
- 0:27note to access more advanced researches, applications and our
- 0:31latest published personal intelligent multi agent deep
- 0:34value intelligence. Welcome to explore and join
- 0:37Deepvalues dot AI community. Absolutely.
- 0:40And to track total market temperature.
- 0:42Welcome to search and download iOS app Buffett indicator from
- 0:45Apple App Store. Also, quick disclaimer, all
- 0:48analysis is based on publicly available sources and maybe
- 0:51subject to revisions or differing interpretations.
- 0:54All contents from this channel is for information and research
- 0:56purposes only and does not constitute investment advice.
- 0:59Right. Definitely conduct your own due
- 1:00diligence before making investment decisions.
- 1:03Exactly. And hey, if you like to access
- 1:05wider topics including life and investing philosophy, please
- 1:08subscribe to our sister channel Deep Values.
- 1:10And don't forget to like this video.
- 1:12Much appreciated. Always appreciated.
- 1:14OK, so back to Microsoft. If you just glance at their top
- 1:18line numbers, they look completely invincible.
- 1:21I mean, we are talking massive revenue, staggering profits.
- 1:24Oh, totally. But the sources reveal that when
- 1:26you dig into the recent data, you see a completely different
- 1:29story playing out. Over a recent six month period,
- 1:33their stock actually plummeted 35%.
- 1:36Wow. Yeah, 35%.
- 1:38And that drop dragged their forward PE, their price to
- 1:42earnings ratio, down from a premium 40X to just 20X.
- 1:47Which is a massive haircut. Right.
- 1:49So the mission for this deep dive is to figure out, you know,
- 1:51is Microsoft a fundamentally misunderstood AI opportunity
- 1:55right now, or is it turning into a massive value trap?
- 1:58And to really get to the bottom of that, we need to synthesize
- 2:00incident from both their financial analysis and some tech
- 2:04trends happening under the hood. We're going to walk through
- 2:06Microsoft's 3 core business pillars, and what we find is
- 2:12pretty ironic, honestly. How so?
- 2:14Well, we're going to explore why the very AI revolution that
- 2:17Microsoft helped kickstart might actually be its biggest
- 2:21existential threat. OK, let's unpack this, because
- 2:23to understand why that thread is so real, we have to look at
- 2:27Microsoft's absolute crown jewel.
- 2:29The productivity and business processes segment.
- 2:31Exactly, for you listening, this is basically the software you
- 2:34probably stare at for 8 hours a day at work.
- 2:36Right, the stuff you can't escape.
- 2:39This segment generates around $120 billion a year.
- 2:42Which is just a crazy number. It is, but the staggering part
- 2:45isn't even the revenue, it's the profitability.
- 2:48They run at roughly 60% operating margin. 60%.
- 2:51Yeah, 60%, just the office suite alone.
- 2:53So Word, Excel, PowerPoint, that represents a $70 billion profit
- 2:57pool. Wow.
- 2:59Think about the scale of that. They have 450 million commercial
- 3:02seats, paying an average of $25 per user per month.
- 3:06Just clockwork every single month.
- 3:08Exactly. It's arguably one of the
- 3:10greatest recurring revenue models in the history of
- 3:12capitalism. And you know, we shouldn't
- 3:14overlook LinkedIn in this either.
- 3:16Oh, right, LinkedIn. Because the numbers show
- 3:18LinkedIn is a surprisingly robust $18 billion business on
- 3:23its own. Yeah, they bought it for what,
- 3:2425 billion? Back in 2016, yeah.
- 3:27And critics thought they wildly overpaid at the time, but they
- 3:30turned it into a massive engine. They are pulling in incredibly
- 3:35high CPM B2B ads. Right businesses paying a
- 3:38premium to target professionals. Exactly.
- 3:40Plus those recruiter seats that cost up to like $900 a month.
- 3:44It's a total powerhouse, yeah, but that core $70 billion office
- 3:49suite profit pool is suddenly looking, you know, highly
- 3:53vulnerable. OK, so to understand that
- 3:55vulnerability, I want to pose a historical question.
- 3:57Let's think back to 2000. Six, OK, going back a bit, yeah.
- 4:00Google launches Google Docs and Google Sheets, and the entire
- 4:04tech narrative back then was that Google's free cloud native
- 4:08tools were going to kill Microsoft Office I.
- 4:11Remember that? Yeah.
- 4:12And Google did win the low end market.
- 4:14You know, students, freelancers, lean startups, they all jumped
- 4:18ship, but they completely failed at the enterprise level.
- 4:22Why did Google fail back then? And why is AI suddenly the real
- 4:26threat now? Well, what's fascinating here is
- 4:28how perfectly this maps onto Clayton Christensen's concept of
- 4:32the Innovator's Dilemma. OK, how so?
- 4:34So Google failed because massive enterprises rely on deeply
- 4:38complex features like let's take an investment bank.
- 4:42They don't just use Excel to make simple tables.
- 4:44Yeah, they have crazy financial models.
- 4:45Exactly. They rely on incredibly
- 4:48complicated VBA macros, custom code that automates all these
- 4:51financial tasks built up over literally decades.
- 4:53Or like a major law firm with all their formatting locked into
- 4:56Word. Exactly.
- 4:57Google Sheets and Docs were fantastic for, you know, basic
- 5:00collaboration, but they just couldn't replicate that top 5%
- 5:03of hyper complex functionality. Makes sense?
- 5:06So the enterprises stayed put. But AI is an entirely different
- 5:09kind of threat. It isn't just our for a
- 5:11competing workspace, it changes the actual labor intensity.
- 5:14Oh I see, because if an AI can draft a like a comprehensive
- 5:18document in 30 seconds instead of a human taking two hours.
- 5:22Right. The sheer volume of human effort
- 5:24required just drops off a Cliff. Which strikes right at the heart
- 5:28of Microsoft's business model. Exactly because their model is
- 5:31based entirely on human seats, the disruption won't be mass
- 5:35cancellations where a company just unplugs Microsoft
- 5:38overnight. It'll be a slow bleed.
- 5:39Yeah, it's going to look like average revenue per user or ARPU
- 5:42stagnating. If an AI can do the cognitive
- 5:45work of three people, companies just won't need as many seats.
- 5:49That structural shift is massive.
- 5:51I was reading a detail in the sources that really stuck with
- 5:54me. Listing XLS a skill on your CV
- 5:57is rapidly becoming obsolete. Oh completely.
- 5:59Because five years ago being the Excel wizard was a badge of
- 6:02honor. Now LLMS can build complex
- 6:04models in literally minutes. The technical skill barrier has
- 6:08just collapsed. You don't need to know the tool
- 6:10syntax, you just need to know what business question as the
- 6:12AI. So if human seats are at risk,
- 6:15Microsoft's defense mechanism is obviously its own AI, right?
- 6:19Enter Copilot. Yeah, Copilot.
- 6:21The sources reveal they're putting the Copilot brand on
- 6:23like 80 different products. They're even putting dedicated
- 6:27Copilot keys on physical laptop. They are heavily forcing the
- 6:31distribution, Yeah. They're relying on their
- 6:32historical advantage of being the default software.
- 6:35But there is a glaring execution gap happening.
- 6:38Despite all this forced integration, Copilot is
- 6:41struggling with basic tasks. Right, like just formatting
- 6:43tables in Word? Exactly, and users are
- 6:47increasingly starting their workflows in better generalist
- 6:50LLMS like Clod or ChatGPT, and they only open Microsoft Word at
- 6:55the very end just to satisfy their company's formatting
- 6:58requirements. Reduces Word from being a
- 7:00primary creation tool to just a dumb terminal.
- 7:04Essentially A formatting wrapper.
- 7:05It's like Microsoft built an incredibly complex airplane
- 7:09cockpit, right? But AI is a pilot that doesn't
- 7:12want to use their dials and switches.
- 7:14That's a great analogy. It just wants to control the
- 7:16plane directly. Exactly.
- 7:17But wait, let me push back a bit.
- 7:19Isn't Copilot supposed to have the ultimate advantage because
- 7:21it has access to a company's internal files?
- 7:24What's fascinating here is that the internal data advantage is
- 7:28actually backfiring in a lot of real world corporate
- 7:30environments. Really.
- 7:32Why? Because companies historically
- 7:34have terrible file management. Oh yeah, that is true.
- 7:38Unsorted SharePoint folders everywhere.
- 7:40It's a disaster. Outdated permissions, old
- 7:43project drives. In the past no one cared because
- 7:47a human employee isn't going to spend a week manually hunting
- 7:50through thousands of folders. But when you turn on copilot it
- 7:53instantly indexes everything. Suddenly a mid level manager
- 7:57asked for AQ3 summary and the AI surfaces.
- 8:00Confidential HR documents and unannounced layoff plans.
- 8:03Oh wow, just because someone forgot to uncheck a permissions
- 8:06box 3 years ago? Exactly.
- 8:08Word was built with the assumption of a human creator,
- 8:10making it a really clunky, dangerous environment for an AI
- 8:14agent to operate. That is terrifying for an IT
- 8:17department. It's a governance nightmare.
- 8:19So if the application layer like Office is vulnerable, what about
- 8:23the back end infrastructure we should transition to the
- 8:25intelligent cloud segment? This is where the scale gets
- 8:28hard to even comprehend. Azure is the core here.
- 8:31Should say the $75 billion run rate growing at 39%.
- 8:35Which ironically disappointed the market because they expected
- 8:3840%. The expectations are just
- 8:40astronomical. So how does this cloud
- 8:42infrastructure actually work? We have to look at the cloud
- 8:45layer cake. The cloud layer cake, yeah. 3
- 8:47layers IAS which is infrastructure as a service,
- 8:51that's the raw servers and cooling.
- 8:53Then Pavis Platform as a service, that's the plumbing,
- 8:57databases, login systems, and finally Sauce software as a
- 9:02service like Office apps. Microsoft is the only
- 9:05hyperscaler competing at all three levels.
- 9:07And GitHub seems to be their secret weapon here.
- 9:09They bought it for 7.5 billion, which looks incredibly cheap.
- 9:13Now it's. The deal of this entry?
- 9:14It gives them a direct pipeline to developers who ultimately
- 9:17choose which cloud to use. And GitHub Copilot has a massive
- 9:21proven ROI right? Letting developers code 20 to 30
- 9:25times faster. Because code is the perfect use
- 9:27case for LLMS. But here's where it gets really interesting.
- 9:31Microsoft's own CEO Satya Nadella is publicly warning
- 9:36about AI agents bypassing the platform layer.
- 9:39He is, yeah. Why would the CEO sound the
- 9:42alarm on his own highly profitable platform?
- 9:44Because AI threatens to completely melt it.
- 9:47Historically, developers needed Microsoft's Pace S layer to
- 9:51manage complexity. Like login systems.
- 9:53Exactly. There's this concept called CRUD
- 9:56plus business logic. Create, Read, update, delete.
- 9:59To build an app, developers relied on Microsoft's middleware
- 10:03like Intra ID for secure logins. So it's a very sticky middleman.
- 10:07Very sticky. But now AI agents can bypass
- 10:09these middle layers entirely. They go directly to the
- 10:12underlying database, apply their own reasoning, and return the
- 10:15answer. Wow.
- 10:16It threatens to melt that sticky platform layer Microsoft relies
- 10:18on. So if AI agents bypass the
- 10:20traditional software platforms, the burning question for you
- 10:23listening becomes who actually captures the economic value in
- 10:26this new era? That is the multibillion dollar
- 10:29question. Think about the airline
- 10:30industry. Airlines revolutionized global
- 10:33travel, right? But they are notoriously
- 10:36unprofitable businesses. The surplus value went to the
- 10:39consumer, not the airlines, right?
- 10:41Is Microsoft building the AI planes but failing to capture
- 10:45the profits? Let's breakdown the value chain.
- 10:47Say an enterprise uses Microsoft Word, but powers it with an
- 10:51Anthropic Claude agent running on Amazon.
- 10:54AWS. OK, so they still pay for word?
- 10:56Microsoft keeps the $25 Office subscription, sure, but they
- 10:59lose the massive compute revenue which flows directly to Amazon.
- 11:03That's a. Huge loss.
- 11:05But what if they use copilot even?
- 11:06Then Copilot runs on Open AI's models, which means Microsoft
- 11:10has to pay Open AIA cut of the profits.
- 11:13And that open AI relationship is getting really messy, isn't it?
- 11:16Very. Messy.
- 11:17Microsoft committed $13 billion, largely in compute vouchers, but
- 11:21Open AI is burning cash fast and.
- 11:24Open AI is facing fierce competition from Anthropic, who
- 11:27just signed a massive $50 billion Frontier deal to run
- 11:30tools on Amazon AWS, bypassing Azure entirely.
- 11:34Exactly. What's fascinating here is the
- 11:36power dynamic has shifted. Big tech used to buy and control
- 11:39upstarts. Now, Open AI has grown too big
- 11:43with an $800 billion valuation to be controlled like early.
- 11:47Smartphone carriers losing leverage to Apple?
- 11:49Precisely. AT&T thought they controlled
- 11:52everything until the iPhone turned them into dumb pipes.
- 11:54Open AI is dictating terms now Which?
- 11:56Leads us to the $140 billion CapEx question.
- 12:01To stay in this race, Microsoft is spending unprecedented
- 12:04amounts of money, which terrifies Wall Street, rightly.
- 12:06So their. CapEx guidance is 120 to $150
- 12:10billion a year, up from just 28 billion in 2023 and most of.
- 12:15That goes to GPU's primarily from NVIDIA, which have a really
- 12:18short lifespan, like three to five years, so massive.
- 12:20Depreciation hits are coming. Let's do a quick math check.
- 12:23To earn a 12 to 15% return on 140 billion, they need about 20
- 12:27billion in new annual revenue. Azure is currently adding 28
- 12:31billion. So they clear the bar, but only
- 12:33if that growth holds. What's the bull case here?
- 12:36The Bull. Case is that Microsoft generates
- 12:38350 billion in recurring revenue, they have 90 billion in
- 12:42cash, and they are developing their own custom silicon.
- 12:46The Maya 200 ships to replace expensive NVIDIA GPU's which
- 12:51could. Dramatically improve margins?
- 12:53Exactly. If we connect this to the bigger
- 12:55picture, all Big Tech CEO's share the same philosophy.
- 12:58Right now they'd. Rather overshoot and waste
- 13:01billions than under invest and lose their market position
- 13:03forever. Exactly.
- 13:04Like Meta's metaverse survival over short term margins, so.
- 13:08Where does this leave the valuation The.
- 13:10Models show a base case fair value around $410, assuming
- 13:15margins recover as they build their own chips and the.
- 13:17Bare case. The bare case.
- 13:18Puts fair value at $280, assuming margins permanently
- 13:21compressed to 35% due to AI commoditization for now.
- 13:25The sheer complexity puts Microsoft in the too hard pile
- 13:28for a lot of investors. We don't fully know if they'll
- 13:31remain a high margin software creator or become a lower margin
- 13:35distribution pipe. Spotify dominating music
- 13:38distribution but handing 70% of revenue back to the labels.
- 13:41Exactly. So I want to leave you, the
- 13:43listener, with a final provocative thought to Mull
- 13:45over. OK.
- 13:46Let's hear it. As we watch tech giants pour
- 13:50hundreds of billions into data centers and ships, is it
- 13:54possible that the ultimate financial winner of the AI
- 13:57revolution isn't one of the Magnificent 7 at all?
- 14:02What if it's a completely unknown company that simply
- 14:04waits for this infrastructure to be built and then rides it for
- 14:07free? It's a.
- 14:08Great point. Historically the companies
- 14:10laying the literal cables rarely captured the biggest upside.
- 14:14It was the agile startups built on top of that cheap
- 14:16infrastructure. Exactly.
- 14:18Well, that is all the time we have for today.
- 14:20Thanks for taking this deep dive with us.
- 14:21We will catch you on the next one.
- 14:23See you next time.