Latest / Deep Values: Value Investing Deep Dives / How Micron's Margin is Even Higher Thank Nvidia?
Transcript
- 0:00Imagine, you know, a company that basically makes its money
- 0:03by manufacturing physical factory produced metal and
- 0:07silicon. Right, like operating heavy
- 0:09machinery, managing these massive physical supply chains.
- 0:13Exactly. They ship actual hardware in
- 0:16boxes, but suddenly, like out of nowhere, they start generating
- 0:20the profit margins of a pure dominant software monopoly.
- 0:24Which I mean in the financial world is a total paradox.
- 0:27Yeah, because the laws of financial physics basically
- 0:29dictate that making physical things is expensive and writing
- 0:33code is cheap. But today, well, those laws
- 0:36appear to be completely breaking down.
- 0:38We are seeing a total deviation from historical norms.
- 0:42Honestly, it it forces anyone analyzing these markets to
- 0:45completely rethink their fundamental assumptions about
- 0:49what a hardware business can actually earn.
- 0:51So welcome back to another deep dive.
- 0:54Today we have a really specific kind of structural mission.
- 0:58We are unpacking a single financial report from September
- 1:0130th, 2026. Right, Micron Technologies
- 1:04fiscal Q4 and full year 2026 results.
- 1:07Yeah, because analyzing numbers of this magnitude and trying to
- 1:10decipher if they represent a permanent shift or just a
- 1:13temporary mirage is, well, it's arguably the ultimate exercise
- 1:18for you as a modern intelligent investor.
- 1:21To do that properly, though, we have to move past the flashy
- 1:24headlines and really look under the hood at the underlying
- 1:26mechanics of the business. Definitely.
- 1:29But before we pull that engine apart, want to quickly remind
- 1:32you that dissecting this kind of structural shift requires the
- 1:36right analytical framework. Absolutely right.
- 1:38You need the right tools. Right.
- 1:40So if you're trying to calculate the Buffett style intrinsic
- 1:43value of a complex company like Micron, you really should be
- 1:46utilizing the Deep Values app on iOS.
- 1:49And, you know, for gauging whether the broader market is
- 1:51just riding a wave of irrational hype, you can check out the
- 1:54Buffett Indicator app on iOS. Or the Bubble Watch app if
- 1:57you're an Android user. Yeah, we're also drawing heavily
- 2:00on the multi agent AI analyst at deep values dot AI for this
- 2:03research today. And as always, this deep dive is
- 2:07strictly for educational and research purposes.
- 2:09We are not giving investment advice.
- 2:11Right. Please do your own due diligence
- 2:12before making any moves. OK.
- 2:14So with that baseline established, let's frame the
- 2:17core question here. We aren't just reading a balance
- 2:20sheet today. No, we are trying to determine
- 2:23if artificial intelligence has fundamentally transformed what
- 2:26was historically a highly cyclical, often brutal commodity
- 2:30business into an unstoppable cash machine.
- 2:33Yeah. And to understand why that
- 2:35question even needs to be asked, we have to look at the raw scale
- 2:38of what Micron just accomplished in their fiscal year 2026.
- 2:43The numbers are staggering. They really are.
- 2:45They produced $133.2 billion in revenue.
- 2:50Which is a massive top line number obviously, but the true
- 2:53anomaly here is the profit. Out of that 133.2 billion in
- 2:57sales, they generated 99.34 billion in GAP operating income.
- 3:03And just to clarify for those listening, GAP stands for
- 3:06Generally Accepted Accounting Principles.
- 3:09Right. It represents the strict
- 3:11standardized accounting rules that prevent companies from, you
- 3:14know, hiding their actual operational costs.
- 3:16So this is their profit from purely running the business.
- 3:20Exactly. And the ratio between those two
- 3:23figures is where the historical comparison just completely
- 3:25breaks down. If you divide that 99.34 billion
- 3:29in operating profit by the revenue, you get a 74.6%
- 3:33operating margin. Meaning for every dollar of
- 3:36physical memory chips they manufacture and sell, they are
- 3:39keeping nearly $0.75 in pure operating profit.
- 3:43Right. And contextualizing that against
- 3:46the broader tech landscape makes it even Wilder.
- 3:48I mean take NVIDIA. The undisputed king of the AI
- 3:52Gold rush. Right, exactly.
- 3:54Nvidia's trailing 12 month operating margin has been
- 3:57hovering around 65%. But Micron, a company
- 4:01fabricating physical memory cells in massive foundries, is
- 4:05currently beating Nvidia's operating margin.
- 4:08Wow. And that 75% was just the
- 4:11average for the full year. Zooming in on their final
- 4:13quarter Q4 shows it's actually accelerating.
- 4:17Oh yeah. In Q4 alone, they reported 54.23
- 4:21billion in revenue, which produced 43.75 billion in
- 4:25operating income. That's an 80.7% margin.
- 4:30And breaking that Q4 data down by specific business units shows
- 4:34you where this immense leverage is coming from.
- 4:36Micron Score data center unit reached a 90% gross margin and
- 4:40an 85% operating margin. That's insane.
- 4:44Even their cloud memory division hit an 83% gross and a 76%
- 4:50operating margin. OK, I have to pause and
- 4:52challenge the reality of these figures for a second because my
- 4:55instinct is to assume there is some sort of accounting maneuver
- 4:59happening behind the scenes. How do you mean?
- 5:02Well, we frequently see companies like Alphabet report
- 5:05massive quarterly gains, but when you dig into the footnotes,
- 5:09it's just mark to market accounting noise.
- 5:12Right, like their private investments?
- 5:14Exactly. If a private startup Alphabet
- 5:17invested in raises new funding at a higher valuation,
- 5:20accounting rules force Alphabet to record that paper gain is
- 5:23income even though no actual cash changed hands.
- 5:26So are we certain Micron isn't benefiting from some similar
- 5:31paper gain phenomena? It's a really critical
- 5:34distinction to make. But yeah, we can verify that
- 5:36this is not an accounting mirage.
- 5:38This is genuine cash generating reality.
- 5:42So no venture capital markups padding these numbers.
- 5:45None at all. This is pure operating profit
- 5:48derived directly from the physical manufacturing and
- 5:51selling of high demand memory chips.
- 5:54OK. So if we accept that these
- 5:56margins are real and technically superior to Nvidia's right now,
- 6:01the immediate temptation is to just crown Micron is the better
- 6:04overall business. Sure.
- 6:05I mean, if they keep $0.80 on the dollar and NVIDIA keeps 65,
- 6:09the math seems pretty straightforward.
- 6:11But taking a more rigorous Warren Buffett style view
- 6:14requires us to look beyond the income statement, doesn't it?
- 6:17We have to focus entirely on owner earnings.
- 6:19Exactly. We have to follow the actual
- 6:21cash, particularly the cash required just to keep the lights
- 6:24on and the factories running. So if I'm building a financial
- 6:27model, I can't just take that 99 billion in operating profit at
- 6:30face value because running physical semiconductor factories
- 6:33bleeds cash, right? How much of that actually ends
- 6:37up in the owner's pocket after they reinvest in the business?
- 6:40Let's run those specific cash flow numbers.
- 6:42For FY20, 26 Micron generated 89.7 billion in operating cash
- 6:47flow. OK, However, fabricating modern
- 6:51memory chips requires A staggering amount of capital.
- 6:55They have to build clean rooms, construct massive facilities,
- 6:58buy ultra expensive lithography machines.
- 7:01Capital expenditures or CapEx? Right.
- 7:03And for Micron, that CapEx bill was 27.4 billion.
- 7:07So taking the 89.7 billion in operating cash and subtracting
- 7:12the 27.4 billion spent on factories and equipment leaves
- 7:16them with 62.3 billion in adjusted free cash flow.
- 7:20Which means out of that initial 99.3 billion in GAAP operating
- 7:24profit, about 63% of it converted into actual free cash
- 7:27flow. Still pretty amazing.
- 7:29Oh, make no mistake, generating 62.3 billion in free cash while
- 7:33simultaneously investing that heavily in future manufacturing
- 7:36capacity is an extraordinary feat.
- 7:39It fundamentally improves any assessment of their underlying
- 7:41economics. But, and I can hear a but
- 7:43coming, I sense a structural caveat approaching.
- 7:47You're right. The caveat is that despite this
- 7:49incredible cash generation, Micron is still not inherently a
- 7:53Better Business than NVIDIA, and the crucial distinction lies in
- 7:56capital intensity. OK, unpack that for us.
- 8:00To illustrate this, we simply have to look at Micron's
- 8:02performance exactly 1 year prior in fiscal year 2025.
- 8:06Oh, right, pulling up the 2025 numbers reveals the volatility.
- 8:10Just 12 months ago, Micron's total revenue was 37.4 billion
- 8:15and their operating income was a mere 9.8 billion.
- 8:18So they scaled from 9.8 billion in operating profit to 99.3
- 8:23billion in a single year. That is roughly A tenfold
- 8:27increase. Wow, it's like OK, if you
- 8:30observe a surfer who typically rides decent 6 foot waves
- 8:33suddenly catch 100 foot tsunami during a massive storm, you have
- 8:37to ask a foundational question. Did the surfer suddenly become a
- 8:41literal God of the ocean, or is it just the unprecedented size
- 8:44of the wave doing the work? Exactly because weather patterns
- 8:48inevitably change and waves eventually crash.
- 8:52That framing perfectly isolates the core problem for investors
- 8:55here. How much of this $99 billion
- 8:58represents a permanent upgrade in the surfer skill, like
- 9:02sustainable structural earning power, versus an extraordinarily
- 9:06favorable wave caused by a sudden spike in AI demand
- 9:09colliding with temporarily constrained supply?
- 9:11Yet deciphering the skill from the wave seems infinitely more
- 9:15important than celebrating an 80% quarterly margin.
- 9:18It is literally the only metric that matters for long term
- 9:21valuation. And this brings us back to why
- 9:23NVIDIA retains the structural crown.
- 9:25NVIDIA operates an asset light model.
- 9:28Right. They focus entirely on chip
- 9:30architecture and software design.
- 9:32They just write the blueprints. Yep.
- 9:35They hand those designs to Taiwan Semiconductor
- 9:37Manufacturing Company TSMC. TSMC acts as the foundry,
- 9:41spending 10s of billions to build the physical factories and
- 9:44assuming the manufacturing risk. Right.
- 9:46So NVIDIA reaps the massive margins without carrying the
- 9:49immense capital burden of replacing physical equipment
- 9:53every few years. Exactly.
- 9:55Whereas Micron, by contrast, has to buy the land, build the
- 9:58factory and manufacture the silicon themselves.
- 10:01That. Makes perfect sense.
- 10:04The underlying business model of NVIDIA is vastly superior from a
- 10:07capital efficiency standpoint. But that still leaves us with
- 10:11the immediate Micron dilemma. Right.
- 10:14If we are concerned this $99 billion is merely a temporary
- 10:18wave, we need leading indicators to tell us if the water is
- 10:22beginning to recede. Historically, the semiconductor
- 10:26memory market is a brutal boom and bust cycle, right?
- 10:29Oh, absolutely. Prices rise, everyone builds new
- 10:32factories, Supply floods the market, prices collapse and
- 10:35profit margins evaporate. But analyzing Micron's current
- 10:38order book suggests this specific cycle possesses
- 10:42distinct structural differences. They are not merely selling into
- 10:46a spot market anymore. They have secured strategic long
- 10:50term customer agreements backed by substantial financial
- 10:53commitments. Customers are putting cash
- 10:56upfront to guarantee access to future supply, correct?
- 10:59I mean, Reuters indicates Micron has secured customer deposits
- 11:02totaling roughly $32 billion. That's huge.
- 11:06And more importantly, they reported remaining performance
- 11:09obligations, which are essentially contracted future
- 11:12revenues of approximately 150 billion.
- 11:15Just to put that 150 billion figure into perspective, that
- 11:19isn't just an airline overbooking a single flight
- 11:21because demand is high this weekend.
- 11:22No, not at. All that is major tech companies
- 11:25signing ironclad contracts to buy every available seat on
- 11:28every light for the next 5 years, paying non refundable
- 11:32deposits just to ensure they aren't less stranded.
- 11:35And Micron's own management guidance reflects this locked in
- 11:38momentum. They are not forecasting A
- 11:40reversion to the mean for the first quarter of fiscal year
- 11:432027. They are guiding for revenue of
- 11:45approximately 61.5 billion. Wait, that implies acceleration.
- 11:51Q4 was 54.2 billion and they expect the subsequent quarter to
- 11:55jump another 7 billion? Yep.
- 11:57Along with an expected gross margin of roughly 86% on that
- 12:0161.5 billion, the financial signals point entirely to
- 12:04expansion, not a peak, OK. I have to play the skeptic here
- 12:09regarding the underlying mechanics.
- 12:11I fully grasp the demand is currently astronomical because
- 12:14every data center on earth is clamoring for high bandwidth
- 12:17memory or HBM to feed their new AI processors.
- 12:21Right, it's the bottleneck, but. Fundamentally, memory has always
- 12:25reverted to a cheap commodity. The laws of supply and demand
- 12:28dictate that competitors like Samsung and SK Hynix will
- 12:32eventually build enough HBM factories to flood the market,
- 12:35destroying this pricing power. Why is this time actually
- 12:38different? It's a great question, and the
- 12:40difference lies entirely in the physics and manufacturing
- 12:43complexity of high bandwidth memory.
- 12:46HBM is not simply a faster version of traditional memory
- 12:49chips. So what makes it so hard to
- 12:52make? Well, fabricating it requires
- 12:55vertically stacking 8 or 12 individual memory dyes on top of
- 12:58each other, and to connect them, manufacturers have to drill
- 13:01microscopic holes straight through the silicon, known as
- 13:05through silicon vias or TSVS. Which sounds incredibly
- 13:09delicate. It pushes the boundaries of
- 13:12physical engineering. You have to perfectly align
- 13:15these microscopic vias across multiple layers, bond them
- 13:18seamlessly, and then solve the immense thermal dynamics of
- 13:22dissipating heat from the middle layers of the stack.
- 13:25Wow. Because of this complexity, the
- 13:27yield rates, the percentage of chips that actually work without
- 13:30defects, are much lower than traditional memory.
- 13:33If one layer in an 8 layer stack fails, the entire component is
- 13:36often ruined. So this is an artificial
- 13:39scarcity created by a cartel. This is genuine scarcity
- 13:42dictated by the laws of physics and manufacturing yield.
- 13:45Precisely the point. This physical bottleneck
- 13:48transforms Micron from a high volume producer of an easily
- 13:52replicated commodity into a specialized manufacturer
- 13:55possessing genuine scarcity. And true scarcity is the
- 13:59foundation of sustainable pricing power.
- 14:02So, because the chips are physically harder to produce,
- 14:05the supply wave is naturally restricted, which stretches the
- 14:08duration of the cycle and permanently improves the
- 14:10economics of the business. Exactly.
- 14:13And this structural evolution demands a complete reassessment
- 14:16of where Micron sits within the broader tech hierarchy.
- 14:19OK, let's reorganize that map. When we evaluate the landscape
- 14:23of Big Tech purely on business quality and durability, ignoring
- 14:26current stock valuations for a moment, how do these Titans
- 14:29group together based on their underlying business models?
- 14:32Let's start with the Apex tier, the asset light software and
- 14:36ecosystem monopolies. Microsoft sits firmly at the top
- 14:39here. Right, they're recurring
- 14:41enterprise revenue is nearly unassailable.
- 14:43They generate roughly 155 billion in profit with a 47%
- 14:48margin without needing to pour endless capital into physical
- 14:52goods, and Apple shares this tier.
- 14:54They possess phenomenal owner earnings conversion and a
- 14:57Fortress Light consumer ecosystem.
- 14:59They generate about $155 billion in profit at a 33% margin.
- 15:04So incredible recurring cash flow without the burden of heavy
- 15:08industrial manufacturing. Who occupies the next year?
- 15:12The second tier is the high margin AI engines NVIDIA.
- 15:16As we discussed, they have extraordinary asset light
- 15:19economics generating about 198 billion in trailing 12 month
- 15:24profit at a 65% margin. But their true durability comes
- 15:28from their software mode, right? The CUDA platform.
- 15:31Exactly. Developers use CUDA to write AI
- 15:34software, and that code only runs efficiently on NVIDIA
- 15:37hardware, creating massive switching costs.
- 15:40So the hardware is protected by a software lock in.
- 15:43Yep, moving to the third tier, we have the advertising giants
- 15:46with capital problems. Alphabet and Meta both feature
- 15:49phenomenal core advertising platforms.
- 15:52However, both are becoming increasingly capital intensive
- 15:55as they race to build massive AI data centers.
- 15:58Right Alphabet pulls in around 148 billion in profit at a 33%
- 16:02margin, though as you astutely noted earlier, their true core
- 16:06search earnings are frequently obscured by the accounting noise
- 16:10of their venture investments. And Meta generates roughly $87
- 16:14billion in profit at a 38% margin, but their capital
- 16:17expenditures are rising rapidly to support their AI ambitions.
- 16:21They basically print money through ads, but they are being
- 16:24forced to spend billions on servers just to stay
- 16:26competitive. And then we have the heavy
- 16:29lifters. Right.
- 16:30Amazon occupies that capital heavy tier.
- 16:33While their cloud division, AWS, and their advertising network
- 16:36boast phenomenal margins, Amazon's consolidated business
- 16:40is weighed down by their massive retail logistics network.
- 16:43Operating thousands of warehouses and delivery fleets
- 16:46consumes extraordinary capital. Leaving them with roughly $94
- 16:50billion in profit, but at a blended margin of only 12%.
- 16:55Which brings us to the anomaly. Where does the newly upgraded
- 16:59Micron fit into this framework? The transition to HBM has
- 17:02elevated Micron beyond traditional cycle analysis.
- 17:05In terms of pure business quality, they have bypassed
- 17:08Amazon and substantially narrowed the gap with Meta.
- 17:11Synthesizing that reality is honestly jarring.
- 17:14We are watching a physical manufacturer of memory chips
- 17:18legitimately challenge the structural business quality of
- 17:21Meta's global digital advertising empire purely
- 17:24because AI created a physical bottleneck in memory
- 17:27manufacturing. It really highlights how rapidly
- 17:30structural advantages can shift. However, recognizing this
- 17:33upgrade comes with a critical warning regarding valuation.
- 17:37The normalization warning. Yes, the singular mistake an
- 17:40investor can make today is extrapolating Microns recent
- 17:43peak directly into the future. But wait, if they just achieved
- 17:468.5 billion in net income and management is guiding for an
- 17:50even stronger Q1, why is it incorrect to utilize those
- 17:53figures as the new baseline? Because it assumes 0
- 17:57cyclicality. In a market still tied to
- 17:59physical supply and demand for software monopolies like
- 18:02Microsoft or Apple, today's profits provide a highly
- 18:05reliable starting point for projecting normalized earnings.
- 18:08But memory, even structurally upgraded memory, will still
- 18:11experience shocks. OK, so an intelligent approach
- 18:15requires constructing distinct probability models for Micron's
- 18:18normalized owner earnings. Exactly.
- 18:21I'd say 3 distinct. Models.
- 18:23If I'm building those models, I have to account for both a
- 18:26massive cooling off in AI infrastructure spending and a
- 18:29world where this spending accelerates for a decade.
- 18:32How do you quantify those extremes?
- 18:34Well, we begin with the bear case.
- 18:36Assume the AI infrastructure build out decelerates, demand
- 18:39normalizes and competitors eventually overcome the HBM
- 18:42manufacturing hurdles to increase supply.
- 18:44In this scenario, Micron's normalized owner earnings might
- 18:47settle between 25 billion and 35 billion.
- 18:51OK. But even in that pessimistic
- 18:53scenario, 25 billion is still vastly superior to the 9.8
- 18:57billion they earned in 2025. The structural floor of the
- 19:01business has been permanently raised.
- 19:03The floor is undeniably higher. The second model is the base
- 19:06case. This assumes the AI transition
- 19:08continues at a steady, sustainable pace, allowing
- 19:12pricing to remain healthy but stripping away the current
- 19:14euphoria. Where does that land?
- 19:16Normalized owner earnings in this environment would track
- 19:18between 40 billion and 50 billion.
- 19:21And the bowl case requires the current bottleneck to become a
- 19:23permanent feature of the industry.
- 19:25Exactly. If structural AI demand creates
- 19:28a permanent severe deficit in HBM capacity and Micron retains
- 19:32immense unyielding pricing power, we could see normalized
- 19:36structural AI owner earnings consistently exceeding 60
- 19:40billion. So the entire investment thesis
- 19:43for Micron hinges entirely on assigning the correct
- 19:46probabilities to those 3 distinct futures.
- 19:50The bear, the base, or the bull. It isn't just about reading the
- 19:53quarterly numbers, it's about predicting the long term physics
- 19:57of the semiconductor market. Right.
- 20:00Of all the mega cap technology companies we actively monitor,
- 20:04Micron's underlying business quality assessment is shifting
- 20:07faster than any other. No other top tier entity has
- 20:10published recent financial results that necessitate such a
- 20:13profound recalibration of fundamental assumptions.
- 20:16This complexity is exactly why we conduct these deep dives to
- 20:21bring it all back to those of you listening.
- 20:23Navigating a market where intrinsic value fluctuates this
- 20:25wildly is why static analysis fails.
- 20:29Right, you need dynamic tools. Tracking these real time
- 20:32structural shifts is exactly why tools like the Deep Values app
- 20:36and the Buffett Indicator were built.
- 20:38I highly recommend heading over to Deep Values AI to review the
- 20:42full research notes and the AI multi agent analysis on this
- 20:45specific Micron report. It's definitely worth checking
- 20:48out. Additionally, if you find value
- 20:50in dissecting these structural market shifts, search for at the
- 20:54deep values on YouTube. Liking and subscribing there
- 20:57will deliver 1 deep investing insight to your feed every
- 21:00single day. Because in a market where
- 21:02physical hardware manufacturers are suddenly pulling in 80%
- 21:05margins, daily insights aren't just interesting, they are
- 21:08essential. Just keep in mind, again, this
- 21:10discussion is for your education and research, not direct
- 21:13investment advice. Always do your own digging.
- 21:16Independent analysis is paramount here, but taking a
- 21:20step back from the granular cash flow models, this entire Micron
- 21:24phenomenon poses a much broader, almost philosophical question
- 21:28for anyone allocating capital today.
- 21:30A macro implication? What's the take away?
- 21:32We just witnessed A notoriously capital intensive cyclical
- 21:35hardware producer suddenly command software like margins
- 21:38entirely because a new technological paradigm, AI,
- 21:42created a severe physical bottleneck.
- 21:45The question you have to ask is what other boring, capital heavy
- 21:49traditional hardware sectors are currently being ignored by the
- 21:52broader market, sitting on a hidden structural goldmine just
- 21:56waiting for the right technological trigger to
- 21:58fundamentally rewrite their economics?
- 22:01That reframes how you look at every industrial sector on the
- 22:04market. It echoes our opening thought.
- 22:07You assume you understand the fundamental rules of the map.
- 22:10You think you know precisely what a hardware company is
- 22:13capable of. And then you realize the
- 22:16economic ground beneath you is far more fluid than you ever
- 22:19imagined. The landscape is constantly
- 22:21evolving. The investors who recognize the
- 22:24shift first are the ones who benefit.
- 22:26On that note, we'll leave you to ponder what other dormant
- 22:29industries might be sitting on the edge of a tectonic shift.
- 22:33Thanks for joining us on this deep dive.
- 22:34We'll catch you on the next one.