Latest / Deep Values: Value Investing Deep Dives / Microsoft's 2026 Four Trillion Dollar AI Gamble Exposed
Transcript
- 0:00Imagine reporting over $81 billion in revenue, beating
- 0:03every single Wall Street earnings expectation, and then
- 0:09just sitting back to watch your stock instantly tank 10%.
- 0:13Yeah, it sounds completely backwards.
- 0:14Right, but that is exactly what happened to Microsoft in April
- 0:182026. Welcome to the deep dive.
- 0:20Glad. To be here.
- 0:21So for you listening, we are looking at a really massive
- 0:25equity valuation report on Microsoft today.
- 0:27This is a company currently sitting at a staggering $4
- 0:31trillion valuation Which? Is just a massive number to even
- 0:34wrap your head around. It really is, and our mission
- 0:37today is to look past the sensational headlines.
- 0:40We need to figure out if this historical spending spree on
- 0:42artificial intelligence is actually building an
- 0:45impenetrable economic Moat. Or, you know, if it's just a
- 0:48bottomless money pit. Because the answer to that
- 0:50question fundamentally alters the risk profile of almost every
- 0:54retirement account out. There exactly.
- 0:55I mean, if you hold a standard S&P 500 index fund, you are
- 0:58heavily invested in this transition, whether you
- 1:00constantly chose to be or not. You're in it.
- 1:03You are definitely in it. OK, let's unpack this tension.
- 1:07Microsoft stock is down nearly 12% year to date and the entire
- 1:12market is basically holding its breath for the upcoming April
- 1:1529th Q3 earnings report. Right, they want proof that this
- 1:19entire AI narrative is financially sustainable.
- 1:22But if they just beat expectations, why is the market
- 1:25punishing them? I mean, 81 billion in revenue is
- 1:27no joke. Well, the fear isn't coming from
- 1:30a lack of revenue. If we look at the previous
- 1:32quarter, so Q2 of fiscal year 2026, the operational
- 1:36performance was spectacular. Yeah.
- 1:38I mean, they brought in 81.3 billion dollars.
- 1:41That is a 17% increase year of year.
- 1:44The Intelligent Cloud division alone crossed $50 billion in
- 1:47revenue for a single quarter, and earnings per share beat
- 1:50estimates by over 7%. Hold on though, for a company
- 1:53that was founded in the 1970s to still be growing top line
- 1:56revenue at 17%, wild, That basically breaks the law of
- 2:00large numbers. So how on earth does the stock
- 2:02drop after a performance like that?
- 2:04What's fascinating here is that it's all about the cost required
- 2:08to generate that growth. The market is experiencing what
- 2:11the research calls a decoupling of capital expenditures from
- 2:14revenue growth. OK, a decoupling.
- 2:17Yeah. Yes, the revenue is breaking
- 2:18records, but the underlying infrastructure required to
- 2:21generate that revenue is just bleeding cash.
- 2:25In that same 90 day quarter, Microsoft spent $37.5 billion on
- 2:31capital expenditures. Wait, 37.5 billion in 1/4?
- 2:35Just one quarter. Let me do the math on that
- 2:37really quick. That is, that's just under $416
- 2:41million a day. Exactly every single day.
- 2:43I'm trying to wrap my head around that scale because this
- 2:46feels like buying a massive beautiful mansion, but suddenly
- 2:48panicking when the first month's electricity and maintenance bill
- 2:51arrives. That is a perfect analogy.
- 2:53You don't. Spend $400 million a day typing
- 2:55code. What is that money actually
- 2:57buying? It is buying concrete, steel,
- 3:00cooling systems, and highly specialized silica Institutional
- 3:04analysts like Keith Weiss and Morgan Stanley.
- 3:07They're pointing out that Microsoft is basically being
- 3:09forced to pay a massive AI tax just to maintain his leadership
- 3:13position. And AL tax.
- 3:15Right. Investors are staring at that
- 3:17daily $416 million bill and they are demanding an ROI timeline.
- 3:22Like when does that upfront capital expenditure turn into
- 3:25pure profit margins? Right, because heading into this
- 3:28Q3 report, the consensus estimates demand $81.43 billion
- 3:33in revenue and an EPS of 4.06, and they expect Azure to grow by
- 3:3837 to 38%. But hitting targets isn't enough
- 3:41anymore. They need to prove this bending
- 3:43plateau is insight. Which shifts our focus from the
- 3:46financial statements to the physical reality of what they
- 3:48are constructing. The research mentions they're
- 3:50operating over 400 data centers across 70 global regions.
- 3:53It's a truly planet scale deployment, right?
- 3:56Here's where it gets really interesting to me.
- 3:58We're talking about planet scale, but the research mentions
- 4:01the Fairwater Data Center and these massive power needs.
- 4:04At what point does the physical world, like literal electricity
- 4:08and supply chains, put a hard ceiling on this digital growth?
- 4:13If we connect this to the bigger picture, the short term
- 4:15bottleneck isn't just money. I mean, they have the money,
- 4:18sure. The bottleneck is the physical
- 4:20scarcity of the actual hardware required to run these models.
- 4:24Are we talking about the advanced AI chips like the
- 4:26GPU's? Yes, specifically the H-100 and
- 4:30B200 GPU's from NVIDIA. There simply aren't enough of
- 4:34them in existence to meet the global demand right now.
- 4:37But why can't they just make more?
- 4:38I mean, if Microsoft is waving a $37 billion checkbook around,
- 4:42shouldn't the supply chain just speed up to take their money?
- 4:44You would think so, but the semiconductor supply chain
- 4:47doesn't care how big your checkbook is.
- 4:49Building AB2 under chip isn't just about printing silicon.
- 4:52It requires incredibly complex advanced packaging techniques
- 4:57and the foundries that actually assemble these chips, primarily
- 5:01TSMC in Taiwan, they only have so much physical capacity.
- 5:05It takes years and billions of dollars to build a new
- 5:08fabrication plant. So the supply of these critical
- 5:11chips is capped by the laws of physics and manufacturing
- 5:14timelines. Exactly.
- 5:16Which means Microsoft is caught in a 0 sum game internally.
- 5:20If the supply of chips is capped, they have to ration
- 5:22them. Wow, so they're fighting
- 5:24themselves. Yeah, that internal war for
- 5:26resources is causing in genuine friction.
- 5:29Management has to play a delicate balancing act every
- 5:32single day. Do they allocate their scarce
- 5:34TPU computing power to their external Azure cloud customers?
- 5:38Right, these are massive enterprise clients paying top
- 5:40dollar to rent that computing power.
- 5:42Exactly. Or do they redirect those
- 5:44precious chips internally to power Microsoft's own consumer
- 5:48facing AI products like Copilot? The research notes that this
- 5:52balancing act actually slightly moderated Azure's headline
- 5:55growth rate recently. Which obviously contributed to
- 5:57the stock volatility. They almost have to prioritize
- 6:00copilot though right, because the user adoption numbers in
- 6:03this report are just insane over 100 million monthly active users
- 6:08on copilot. Oh, it's.
- 6:09Staggering. Yeah, and 80% of the Fortune 500
- 6:12using their Azure AI Foundry platform, plus 20 million
- 6:16developers relying on GitHub Copilot.
- 6:18As an AI peer programmer, if you have that many people actively
- 6:22using the tool, you can't just let the servers crash because
- 6:25you rented the chips to an outside client.
- 6:27You really can't, and there's a deeper strategic reason they
- 6:30prioritize those internal users. The research outlined something
- 6:34called the AI Feedback Loop. OK, how does that work?
- 6:37By being the first to deeply integrate Open AI's massive
- 6:40models into everyday enterprise software, Microsoft created a
- 6:44data flywheel so every time those 100 million users ask
- 6:48Copilot to draft an e-mail or summarize a meeting, that data
- 6:51helps refine the models performance.
- 6:53Right. And better performance naturally
- 6:55attracts more users. Exactly, which generates even
- 6:57more behavioral data. So the flywheel spins faster,
- 7:00creating a wider mode against competitors.
- 7:03But a flywheel of that size requires immense amounts of
- 7:06energy. Like we mentioned earlier at
- 7:08that Fairwater data center, At a certain point don't you run out
- 7:11of literal electricity? Energy grid limitations are a
- 7:15huge structural long term risk. You can't just plug 100 MW data
- 7:20center into a suburban power grid.
- 7:22Yeah, that would blow out the whole neighborhood.
- 7:24Exactly. They're having to invest
- 7:26directly in energy infrastructure.
- 7:28They're exploring nuclear options, revamping cooling
- 7:31systems just to keep these intelligence refineries from
- 7:34literally melting down. Add to that the constant
- 7:37regulatory scrutiny. Antitrust regulators in the EU
- 7:40and the US are hovering. Right, they are analyzing how
- 7:43Microsoft bundles its software and questioning the exclusive
- 7:46ties with Open AI. If regulators decide that
- 7:49ecosystem is too dominant and force Microsoft to unbundle its
- 7:52services, that feedback loop could be severely disrupted.
- 7:56So the structural risks are everywhere.
- 7:58OK, so building 400 global data centers, fighting over scarce
- 8:02chips, and fending off global regulators is historically
- 8:05expensive. How are they actually paying for
- 8:07it? That's the billion dollar
- 8:09question. Right, because if I look at a
- 8:11telecom company that builds a massive 5G network, they usually
- 8:15drown in debt for a decade. But Microsoft doesn't seem to
- 8:18have that problem. That is where the operational
- 8:21brilliance of this company really shines.
- 8:24They are funding this historically massive
- 8:26technological transition using the cash flow from a completely
- 8:30different part of the business. Wait, let me guess, is it the
- 8:32productivity and business processes segment?
- 8:35You get it. So basically Microsoft 365,
- 8:38LinkedIn, Dynamics, Word and Excel.
- 8:41That is the cash engine. That segment alone still pulls
- 8:44in over 27% of their total revenue.
- 8:46It feels so counterintuitive. We are talking about cutting
- 8:49edge artificial intelligence, but the whole thing is being
- 8:51financed by the boring legacy software suite.
- 8:54It's like this incredibly wealthy sugar daddy funding all
- 8:57the sexy expensive AI token factories.
- 9:00That's a. Funny way to put it, but yeah,
- 9:02legacy software has incredibly high margins.
- 9:05You write the code once and you sell infinite digital copies
- 9:08through a SOAP description model.
- 9:10It is an asset light cash generating machine.
- 9:13But wait, isn't the whole point of AI to eventually replace
- 9:15those legacy apps? If they're building AI agents
- 9:18that can do our work for us, aren't they basically funding
- 9:20their own obsolescence? Well, this is where Satya
- 9:23Nadella's management strategy is so crucial.
- 9:26There are not using AI to replace the legacy apps, they're
- 9:29using it to supercharge them. Oh, I see.
- 9:31It's a software as a service or sauce expansion strategy.
- 9:36By bolting an AI assistant like Copilot directly onto existing
- 9:40software like Word or Excel, they aren't cannibalizing their
- 9:43user base, they are simply increasing the average revenue
- 9:46per user or ARPU. OK so if I'm a business using
- 9:49Microsoft 365 I still need Excel but now Microsoft comes to me
- 9:54and says hey for an extra $30.00 a month per employee we will add
- 9:57an AI assistant that builds your spreadsheets for.
- 10:00You exactly. The core product remains, but
- 10:02the revenue they extract from you goes way up.
- 10:04That is a brilliant upsell mechanism.
- 10:06And while that segment provides the steady cash, the intelligent
- 10:10cloud segment is now the biggest driver of revenue.
- 10:13Overall, it accounts for nearly 35.6% of the total pie.
- 10:18Azure alone crossed $75 billion in annual revenue for fiscal
- 10:23year 2025. But this transition to cloud and
- 10:26AI fundamentally alters what kind of company Microsoft is,
- 10:30right? A decade ago they were an asset
- 10:32light software company. Today they are a capital
- 10:35intensive infrastructure behemoth.
- 10:37Totally different beast now. Let's look at the actual costs
- 10:39because they are mind boggling. The report shows they're fixed
- 10:42costs, meaning property and equipment.
- 10:44The concrete and servers hit $64.6 billion in FY20 25.
- 10:49Right. And once you pour the concrete
- 10:50and install the servers, that cost is fixed.
- 10:52But the variable cost, the things that fluctuate based on
- 10:55usage, are also skyrocketing. They have massive bills for
- 10:58electricity, cooling water and human labor.
- 11:01Labor is a huge one in this report.
- 11:03There's a stat in here that they have 34,000 full time engineers
- 11:06dedicated entirely to the Secure Future initiative.
- 11:09Why do you need the population of a small city just doing
- 11:11cybersecurity? Think about what they are
- 11:13hosting. When you are the centralized hub
- 11:16for the world's corporate data, AI models, and intellectual
- 11:20property, a security breach isn't just APR nightmare.
- 11:24It's catastrophic. Exactly.
- 11:25It is a structural threat to the entire business model.
- 11:29If Fortune 500 companies don't trust the Azure cloud, the
- 11:32intelligence refinery shuts down.
- 11:34Those 34,000 engineers are basically the private army
- 11:38guarding the Moat. And they are spending
- 11:40aggressively to make sure everyone knows they own that
- 11:43Moat. Advertising spend jumped 23.5%
- 11:46in 2025, hitting $2.1 billion to maintain that AI frontrunner
- 11:51status. They are operating with the
- 11:53aggression of a hyper growth startup, but they are doing it
- 11:56at a $4 trillion scale. And we know the company is an
- 11:59operational juggernaut. They have brilliant management,
- 12:01a massive Moat, and an incredible cash engine.
- 12:04But here is the critical pivot for anyone listening today.
- 12:07Today, a great company does not automatically equal a great
- 12:10stock purchase. No, it certainly does not.
- 12:12The price you pay for the asset determines your return.
- 12:15Which brings us to the cold hard valuation math, the deep value
- 12:20investors dilemma. This is where things get really
- 12:22confusing for a traditional investor.
- 12:24I'm looking at the value metrics in the report.
- 12:26Right now Microsoft has a price to earnings or PE ratio of 26.5
- 12:32and their price to book ratio is 8.1.
- 12:35Now, correct me if I'm wrong, but book value is basically what
- 12:38the company would be worth if you sold off all its physical
- 12:40assets, paid off the debts and liquidated it, right?
- 12:43Essentially, yes, it is the net asset value on the balance
- 12:46sheet. OK, so trading at 8 times its
- 12:49book value means you are paying an 800% premium over the actual
- 12:53tangible assets. By strict traditional value
- 12:56investing standards, the stock looks horribly, dangerously
- 12:59overpriced. On paper, yeah, looks
- 13:00terrifying. And yet, the research notes that
- 13:02Michael Burry, the famous value investor from the Big Short who
- 13:06is notorious for betting against bubbles, he recently bought into
- 13:09Microsoft. What is a deep value guy like
- 13:12Burry? Seeing that these traditional
- 13:13metrics are completely missing, why touch a stock with an 8X
- 13:16book value? This raises an important
- 13:18question. He is recognizing that
- 13:21traditional metrics like Priced a Book were designed in the 20th
- 13:24century for industrial companies, railroads, steel
- 13:27mills, car manufacturers. Physical stuff.
- 13:30Right. They fall apart when you apply
- 13:32them to a digital behemoth where the most valuable assets are
- 13:35software, ecosystems and developer networks which just
- 13:38don't show up neatly on a balance sheet.
- 13:40That makes sense. When investors like Brewery look
- 13:42at Microsoft today, they ignore the book value and focus
- 13:46intensely on the quality of the cash flow.
- 13:48Meaning they want to see how efficiently Microsoft is turning
- 13:52that massive $37.5 billion quarterly CapEx spend into
- 13:57actual profit. Exactly.
- 13:59The critical metric to watch here is return on invested
- 14:01capital, or ROIC. Despite spending 10s of billions
- 14:05of dollars every quarter on new data centers, Microsoft's ROIC
- 14:09has remained remarkably stable at around 21 to 22%.
- 14:13So for every dollar they pour into this AI transition, they
- 14:16are generating over $0.20 in return.
- 14:19That proves the CapEx isn't just vanishing into a money pit,
- 14:22they're actually finding highly profitable places to deploy that
- 14:25cash. Furthermore, you have to look at
- 14:26the PEG ratio. That's the price to earnings to
- 14:28growth ratio. This is a metric that factors in
- 14:31the company's expected growth rate.
- 14:33A PEG ratio under one point O generally suggests that a stock
- 14:36is undervalued relative to how fast its earnings are growing.
- 14:41Microsoft's PEG ratio currently sits at approximately .93.
- 14:44Wait, so even with APE ratio of 26.5 the PEG ratio is under 1?
- 14:49Mathematically validating the growth.
- 14:51Yeah, that implies their earnings are growing so fast in
- 14:53the high teens that it justifies the high multiple.
- 14:56It does, and the risk of that multiple contracting is buffered
- 14:59by an absolute fortress of a balance sheet.
- 15:02Their debt to equity ratio is a microscopic .10.
- 15:05Wow, so almost zero debt? They use almost no leverage.
- 15:09Their free cash flow margin, the cash leftover after all
- 15:12operating expenses and capital expenditures are paid, has
- 15:14remained above 25% for a decade. Over the trailing 12 months,
- 15:18their free cash flow was $75.3 billion.
- 15:21That is just staggering. And what are they doing with
- 15:23that cash? The report says they've returned
- 15:26over $368 billion to shareholders over the last 10
- 15:29years. In just the 12 months ending
- 15:32September 2025 alone, they bought back $51.7 billion of
- 15:38their own stock. Stock buybacks of that magnitude
- 15:41send a very clear signal to the market.
- 15:44Management is looking at their own company trading near record
- 15:46high valuations and concluding that buying back their own
- 15:49shares is still the most accretive place to park $51
- 15:53billion. They are definitely putting
- 15:55their money where their mouth is regarding the AI pivot.
- 15:58But let's pause here because there's a reality check in the
- 16:00report that we cannot ignore. The researchers built a
- 16:03discounted cash flow model, or DCF, for a listener who doesn't
- 16:07build financial models all day. Let's break this down.
- 16:09Why are we discounting cash? Isn't a dollar just a?
- 16:12Dollar. Well, not in finance.
- 16:13ADCF model relies on the concept of the time value of money.
- 16:17Simply put, a dollar Microsoft promises to make 10 years from
- 16:20now is worth less to you today because of inflation and because
- 16:24you could have invested that dollar somewhere else in the
- 16:26meantime, say in a Treasury bond.
- 16:29OK, that makes sense. So analysts project out all the
- 16:31cash the company will make in the future, and they discount it
- 16:34back to its present day value to figure out what the stock is
- 16:36actually worth. Right now we call that the
- 16:39intrinsic value. OK, so the DCF reveals the
- 16:41actual worth of the company based on its future cash
- 16:44generation. The report models a few
- 16:47different scenarios. The first is a zero growth
- 16:50scenario. Imagine the AI bubble bursts,
- 16:54the 37 billion in CapEx only maintains their current market
- 16:57share and their cash flow never grows again.
- 17:00What happens to the stock? In that zero growth scenario,
- 17:03the intrinsic value of the stock drops to $155.04 a. $155 The
- 17:08stock is currently trading around $424.
- 17:11If growth stops, the floor falls out completely.
- 17:14It is a massive downside risk, but let's look at the nominal
- 17:17case, which is the baseline expectation If Microsoft
- 17:21continues its historical trend and grows free cash flow at
- 17:24about 15% every year for the next decade.
- 17:27The intrinsic value of the stock calculates out to $399.19.
- 17:32Hold on. So if they execute perfectly,
- 17:35navigate the physical hardware shortages, fend off global
- 17:38regulators, and maintain a blistering 15% growth rate for
- 17:43an entire decade, the stock is worth roughly $400.00.
- 17:46That's. Based on those cash flows, yes.
- 17:48But it's trading at $424.00, which means at current prices
- 17:52there is absolutely 0 margin of safety right now.
- 17:55You aren't getting a discount, you are paying a premium price
- 17:58that fundamentally demands perfection.
- 18:00If they stumble, or if that massive AI tax doesn't quickly
- 18:03translate into profit margins, this valuation has a very long
- 18:06way to fall. That's the reality of it.
- 18:08The current share price fully bakes in a decade of
- 18:10uninterrupted mid teens growth. It is the ultimate tension for
- 18:13an investor. So to emphasize this entire deep
- 18:16dive for you, Microsoft has essentially defied the economic
- 18:20law of large numbers. They are managing the most
- 18:22capital intensive technological transition in human history,
- 18:26building these planet scale intelligence refineries, and
- 18:29they are funding the whole thing with an absolute fortress of a
- 18:32legacy software business. Yeah.
- 18:34From a deep value perspective, the stock does not offer a
- 18:36traditional margin of safety. However, it represents what many
- 18:40consider the ultimate Buffett style compounder.
- 18:44A great business at a fair price.
- 18:46Exactly, rather than a mediocre business trading at a great
- 18:49price. Now before we wrap up today, I
- 18:51need to read a mandatory disclaimer source directly from
- 18:54the research text disclaimer. All financial data and analysis
- 18:57discussed today is based on publicly available sources and
- 19:01may be subject to revisions or differing interpretations.
- 19:04This deep dive is for informational purposes only and
- 19:07does not constitute investment advice.
- 19:09Please conduct your own due diligence with a licensed
- 19:11financial advisor before making any investment decisions.
- 19:14Always vital to do your own homework, particularly when
- 19:18assessing valuations in unchartered technological waters
- 19:21like this. Absolutely.
- 19:23And Speaking of uncharted waters, we want to leave you
- 19:25with a final, provocative thought to Mull over.
- 19:28Yeah, we've spent this entire discussion focusing on the
- 19:30billions of dollars Microsoft is deploying to build these massive
- 19:34AI infrastructure modes. But consider this alternative
- 19:37future What happens if AI agents become entirely commoditized
- 19:42over the next five years? Commoditized like cheap and
- 19:46everywhere. Right.
- 19:47If open source models improve to the point where every tech
- 19:50company on Earth can produce cheap, highly effective AI, does
- 19:54Microsoft's multibillion dollar infrastructure mode suddenly
- 19:58transform into a permanent margin crushing tax?
- 20:01Wow. Because if intelligence becomes
- 20:03as cheap and abundant as tap water, being the world's most
- 20:06expensive water utility is suddenly a terrible business
- 20:09model. Exactly.
- 20:10It is the ultimate tension between massive innovation and
- 20:13rapid commoditization. Definitely something to think
- 20:15about the next time you use an AI assistant to draft an e-mail.
- 20:18Thanks for taking the deep dive with us.