Latest / Deep Values: Value Investing Deep Dives / [Q3] Micron Super Cycle Breaks Old Rules
Transcript
- 0:00Welcome to the Deep Dive. Today we're taking a huge stack
- 0:03of forensic analysis on a semiconductor giant, and we're
- 0:06going to boil it down to the insights that you really need to
- 0:09understand. And that giant is Micron
- 0:11Technology, ticker MU. Our mission today is a full deep
- 0:15dive on Micron. This company is really the
- 0:19poster child for cyclical volatility.
- 0:22It's the classic boom and bust stock.
- 0:24It absolutely is, and that's what makes this so critical.
- 0:27Right now the stock is trading near its all time highs.
- 0:30We're looking at a price of around what, $236.88 and the.
- 0:34Fear is palpable. You can hear everyone on the
- 0:36street screaming peak earnings, get out now.
- 0:39But that's the hook, isn't it? The forensic report we're
- 0:41breaking down today makes the exact opposite case.
- 0:44A pretty bold case. It argues that this isn't the
- 0:47peak of a cycle at all, it's the start of a multi year super
- 0:50cycle. Right.
- 0:50One driven by AI demand and, surprisingly, by the fundamental
- 0:54laws of physics. So the core challenge for us and
- 0:56for you listening, is to figure out if that old historical
- 0:59pattern has finally been broken. Exactly that mechanism that
- 1:03always crushed memory prices. You know the supply response,
- 1:07where high profits lead to this immediate, overwhelming flood of
- 1:11new production. Is that cycle well and truly
- 1:14broken? Because if it is all the old
- 1:16rules for investing in a company like Micron, they just don't
- 1:21apply anymore. OK, but before we get into this
- 1:23thesis, which the report calls a fat pitch opportunity.
- 1:28We have to cover the ground rules.
- 1:29All the analysis you're about to hear is based on publicly
- 1:32available sources. It could be subject to revisions
- 1:34or different interpretations. And of course, this is all for
- 1:36educational purposes only. It's not investment advice.
- 1:39You absolutely have to do your own due diligence before making
- 1:42any decisions. All right, let's unpack this.
- 1:44The report central idea is that Micron is changing.
- 1:47Yeah, transforming from just a supplier of dumb bits.
- 1:49You know, basic interchangeable memory chips.
- 1:52Into something more like a specialized utility provider.
- 1:54Yeah. So what's driving this massive
- 1:56cycle breaking change? It's the perfect storm really, a
- 1:59collision of two huge independent forces.
- 2:02First, you have this massive secular demand shock from
- 2:06artificial intelligence. The LMS, the data centers, all
- 2:09of it. All of it.
- 2:11And second, you have a structural supply problem.
- 2:13We're calling it the complexity wall, and it's where Moore's Law
- 2:17is basically hitting its physical limits specifically for
- 2:20memory manufacturing. OK, let's stick with the physics
- 2:22for a second, the report says. The key to this whole thesis is
- 2:26high bandwidth memory, or HBM. Why is HBM so important?
- 2:31HBM is designed to solve the single biggest bottleneck in AI
- 2:35processing the memory wall. The memory wall.
- 2:38Yeah, think about it this way. You have an incredibly powerful
- 2:42AI accelerator like Nvidia's Blackwell GPU.
- 2:46It can do quadrillions of calculations a second.
- 2:49Just a staggering number. But if the memory can't feed
- 2:51data to that chip fast enough, the processor just sits there
- 2:55idle. It's waiting for the bits to
- 2:56arrive. HBM solves that problem.
- 2:58So how's it different? I mean, how's its architecture
- 3:00different from the standard DRAM chips we're all used to?
- 3:03Standard DRAM is flat. It's laid out on a circuit board
- 3:06with these long slow connections.
- 3:07HBM just completely flips that on its head.
- 3:10It stacks the DRAM dies vertically like a skyscraper of
- 3:14chips, maybe 8 or even 12 dies high, and they're connected by
- 3:18thousands of microscopic channels drilled right through
- 3:21the silicon itself. They're called through silicon
- 3:24vias or TSV's. That sounds incredibly difficult
- 3:28to manufacture, not to mention risky.
- 3:29And that is the absolute core of the thesis.
- 3:32It's exponentially harder. Imagine drilling thousands of
- 3:36perfectly aligned microscopic holes through a stack of what is
- 3:40essentially fragile glass. One mistake and the whole thing
- 3:44is useless. Exactly.
- 3:45If one channel is off, if the wafer cracks, if the thermal
- 3:48bonding isn't perfect, yeah, the entire very expensive chip
- 3:52fails, the report says. This process quote decimates
- 3:55yields. So a huge percentage of what
- 3:57they make is just scrap. A huge percentage.
- 3:59HBM is fundamentally an engineering marvel that's a
- 4:02nightmare to manufacture at scale.
- 4:04And that difficulty, that's what creates the scarcity dynamic,
- 4:07right? This cannibalization effect.
- 4:08Walk me through the economics of that.
- 4:10It puts a hard ceiling on supply because the process is so
- 4:14complex and the yield loss is so high.
- 4:17The analysis shows that for everyone wafer Micron uses for
- 4:21HBM. The really high margin stuff.
- 4:23Right. For every one of those, they
- 4:24remove about 3 wafers worth of standard DRAM capacity from the
- 4:28global market. You just can't physically flood
- 4:31the market with both at the same time anymore.
- 4:33Hold. On, I have to push back on that.
- 4:35If HBM margins are that high, why not just build more fabs?
- 4:39Isn't that the old supply response we were just talking?
- 4:41About that's the perfect question and it's what the
- 4:43market is missing. The complexity wall we're
- 4:45talking about is happening right now.
- 4:46Building a new $20 billion fab takes four years minimum.
- 4:51So until that new capacity comes online way down the road.
- 4:54HBM production has to steal from standard DRAM supply, and that's
- 4:58what breaks the historical overproduction cycle.
- 5:00Which means the unit economics for Micron are totally
- 5:04different. The old gross margin ceiling has
- 5:06to be gone. It is memory.
- 5:07Gross margins were always trapped, you know, in that 30 to
- 5:1040% range. It was a commodity.
- 5:13If prices went up, competitors would just flood the market.
- 5:16But not anymore. Not with HBM.
- 5:18It commands a five to seven times price premium over
- 5:22standard DRAM and that is pushing Microns gross margins up
- 5:26towards 60. Percent. 60%.
- 5:29That's a logic designer number like NVIDIA, not a manufacturer.
- 5:33It's completely transformational.
- 5:34OK, this is where it gets really good.
- 5:36Let's move from the physics to the financials.
- 5:39If this scarcity is real, then the Q1 fiscal 2026 earnings call
- 5:44from last December, it should show some hard evidence.
- 5:48What signals did we get from management?
- 5:50The language was it was a complete shift.
- 5:54They went from being cautiously optimistic in prior quarters to
- 5:57just confirming physical scarcity.
- 5:59The phrase they used over and over was supply constrained.
- 6:01Meaning the problem isn't finding buyers, it's making
- 6:04enough product. They're selling absolutely
- 6:06everything they can possibly make.
- 6:07But the single most important data point, the one that really
- 6:10confirms the multi year thesis. Was that they had completed
- 6:13agreements on price and volume for our entire calendar 2026 HBM
- 6:17supply. Sold out.
- 6:19Completely sold out, locked in on price and volume for the next
- 6:22two calendar years on their most profitable product that de risks
- 6:27their revenue in a way we just never seen in this industry.
- 6:30But wait, sold out for two years sounds almost too good to be
- 6:34true. Are these contracts firm?
- 6:36What's the catch? That's the right skeptical
- 6:38question to ask. And while there are always
- 6:40clauses, management use the term agreements and specifically
- 6:44quantified the full supply, the commitment is exceptionally
- 6:47strong. And there was more, wasn't
- 6:49there? Yes, management even expressed
- 6:51disappointment that they were quote, unable to meet demand
- 6:54from other customers. They are actively turning away
- 6:57business. The bottleneck isn't sales, it's
- 6:59manufacturing. That's the very definition of
- 7:01pricing power. And we see that scarcity in the
- 7:04numbers, Q1 revenue $13.6 billion, but the guidance for Q2
- 7:08fiscal 26. It shot up to $18.7 billion,
- 7:12give or take $400 million. That kind of sequential growth
- 7:15is, well, it's basically unheard of for a company of this size
- 7:18and maturity. OK, but we have to be sure this
- 7:20growth isn't just accounting tricks.
- 7:22The forensic report applies A shenanigans test looking at
- 7:26working capital. Right, You're looking for red
- 7:27flags, channel stuffing, inventory bloat, anything like
- 7:31that. So let's look at the Holy Grail
- 7:33metric for a memory company inventory velocity that's
- 7:37measured by day's inventory outstanding, or DIO.
- 7:40Why is that the crucial 1? Because for a memory company,
- 7:44inventory is a melting ice cube. It's their biggest risk.
- 7:47When prices crash, all those chips in the warehouse are
- 7:50suddenly worth a lot less. Leading to huge write downs.
- 7:52Exactly. Back in the 2023 bust, Dio
- 7:55spiked to 182 days, a huge rotting pile of inventory.
- 8:00Today it's fallen way down into the 126 to 140 day range.
- 8:04And this is happening while revenue is skyrocketing.
- 8:07Which is the key point. The inventory is literally
- 8:09flying out the door. It completely confirms the
- 8:11shortage that management is talking about.
- 8:13And the other side of that coin day sales outstanding DSO.
- 8:17Any sign of them forcing sales by giving customers easy payment
- 8:20term? Nope.
- 8:21DSO is stable right around 57 days.
- 8:24The receivables are growing right in line with revenue.
- 8:27There's just no evidence of channel stuffing.
- 8:29The financials back up the story perfectly.
- 8:32All right, let's shift gears to the structural mode.
- 8:34What stops the other two members of the big three, Samsung and SK
- 8:38Hynix, from just building more capacity and killing the Super
- 8:42cycle? Well, the barriers to entry are
- 8:44just becoming insane. To build a modern fab with the
- 8:48latest EUV equipment, you're looking at 15 to $20 billion.
- 8:51And it's just for one. For one.
- 8:53And then it takes three to four years minimum just to build it
- 8:56and install the tools. You can't just react to high
- 8:58prices anymore. Plus, there's a global shortage
- 9:00of the specialized clean room construction crews you need.
- 9:03But we have to talk about the anti mode, the China threat
- 9:07companies like CXMT and YMPC are building a ton of capacity for
- 9:11older legacy DRAM. And that's the strategic
- 9:14imperative for Micron. They have to get out of that
- 9:16market. They have to abandon the low end
- 9:18commodity game to survive. So this whole shift to high end
- 9:21HBM isn't just a growth play. It's a survival strategy.
- 9:25It's about avoiding the red ocean of commoditized chips
- 9:28where they'll inevitably get crushed by state subsidized
- 9:31Chinese competition. OK.
- 9:33Let's tackle the biggest fear hanging over this stock CapEx,
- 9:37capital expenditure. Micron guided to around $20
- 9:40billion for fiscal 26. Historically, that's the signal
- 9:44for the top of the cycle. And that fear is totally
- 9:47rational, based on decades of history.
- 9:49But the forensic analysis shows a structural inversion of that
- 9:52cycle. We're seeing rising capital
- 9:54intensity. Can you define that in this
- 9:56context? Rising capital intensity.
- 9:58It means the industry is spending more and more money to
- 10:00get fewer incremental bits of memory.
- 10:02More money for less growth. Think of it like a Formula One
- 10:05engine versus a regular car engine.
- 10:07The F1 engine cost 10 times more to build, but it only produces a
- 10:11little more power because the engineering is so extreme.
- 10:14Micron is spending $20 billion not to flood the market, but
- 10:17just to keep up, just to get modest bit growth because HBM is
- 10:20so complex. So that huge CapEx number is
- 10:22actually defensive spending, not offensive.
- 10:26The supply response is dampened because the tech is just too
- 10:28hard to scale cheaply. Exactly.
- 10:30It breaks the old historical signal and it supports the
- 10:33scarcity thesis. OK, so let's bring this all
- 10:35together. As Charlie Munger would say,
- 10:37invert. Always invert.
- 10:39If the Super cycle thesis is so strong, how does it fail?
- 10:43What's the pre mortem? There are three big risks.
- 10:46The number one risk is an AI winter.
- 10:48The entire bull case depends on this insatiable demand for AI.
- 10:52If LLMS hit a scaling wall or if the hyperscalers.
- 10:55Google, Amazon, Meta, Microsoft. Right.
- 10:58If they fail to monetize their huge AI spending, they will slam
- 11:02on the brakes. And memory is at the very end of
- 11:04that supply chain. Whip a small cut in a
- 11:07hyperscaler spending turns into a massive price crash for
- 11:10Micron. Absolutely.
- 11:11If AI spending growth slows from 50% to 20%, memory prices don't
- 11:16just flatten out, they could collapse by 50%.
- 11:18What's the second big risk? Technological displacement.
- 11:21New tech like Compute Express, Link or CXL is all about
- 11:24efficiency. It lets servers pool and share
- 11:27memory. If you get more efficient, you
- 11:28just don't need to buy as many physical bits.
- 11:30Efficiency is the enemy of volume.
- 11:32In the third, risk loops back to that massive CapEx number it
- 11:36could and become balance sheet suicide.
- 11:38It's a huge bet if Micron commits to $20 billion a year
- 11:41through 2028 and the cycle turns hard in 2027.
- 11:45Say because of that AI winter. They'll be caught with brand
- 11:48new, very expensive fabs coming online just as demand
- 11:52disappears. That leads to catastrophic cash
- 11:54burn and probably shareholder dilution.
- 11:56We should stress though that they do have a defense.
- 11:58Management has shown a very Klarmanesque trade, a focus on
- 12:04having a net cash position. They enter the 2023 bust with
- 12:07net cash, prioritizing survival. That's a huge defense.
- 12:11And don't forget the hidden asset, the $6.1 billion CHEPS
- 12:15Act grant from the government that's basically non dilutive
- 12:18equity that lowers their future spending needs.
- 12:20OK, let's wrap this up with valuation.
- 12:22How do you even value a company? This Volvo?
- 12:24It's swung from almost $12.00 of EPS in 2018 to a $5 loss in
- 12:292023. The report uses three different
- 12:31methods. Right, you have to.
- 12:33We start with the historical view, what the report calls a
- 12:35Buffett style normalized mid cycle valuation.
- 12:38This assumes the pass will repeat itself.
- 12:40And how do you get that normalized EPS?
- 12:42You average the boom and bust cycles.
- 12:44You take an estimated peak revenue of around $53 billion
- 12:47and the bus revenue of $15 billion.
- 12:49That gives you a normalized EPS of about $8.90.
- 12:53Imply a standard multiple. A 15X multiple on that gives you
- 12:56a valuation of $133.50, which means at today's price of $236
- 13:02the stock is massively overvalued if you believe the
- 13:06old cycle will just repeat. So the whole thesis rests on
- 13:09rejecting that valuation. If the structure has really
- 13:12changed, we have to look forward.
- 13:13What does the Super cycle forward PE model say?
- 13:16This model accepts the new reality of scarcity and high
- 13:19margins. It's based on the fiscal year
- 13:212026 EPS estimate of roughly $16. $16.00 and you'd apply the
- 13:26same 15X multiple. You would because the quality of
- 13:28earnings from HBM is so much higher and more predictable than
- 13:31the old commodity earnings. OK.
- 13:32So $16.00 * 15 that gives a target price of $240.
- 13:37Which is surprisingly close to where it's trading now.
- 13:40It suggests the stock is is actually fairly valued based on
- 13:42your term fundamentals. Assuming the Super cycle is
- 13:45real, the fair value zone is probably 240 to 260 dollars.
- 13:49And what's the bowl case? If the Super cycle extends
- 13:52through fiscal 2027 and EPS hits say 18 dollars, then the target
- 13:57could reach 324 dollars. The alpha opportunity isn't
- 14:01buying a cheap stock anymore. The alpha is in the duration of
- 14:05this cycle. So the recommendation here isn't
- 14:07to chase it, but to accumulate on pull backs.
- 14:09Absolutely. You're looking for pull backs to
- 14:11that 202 hundred $10 level with the goal of riding out this
- 14:15scarcity that we're seeing confirmed by those sold out
- 14:17contracts. And that brings us to our final
- 14:20thought for you to take away. Micron's entire investment case
- 14:23really boils down to one question.
- 14:26Have AI and HBM created a permanent structural scarcity
- 14:29that finally breaks the old supply cycles?
- 14:31All the evidence we've seen, the gross margin expansion, the
- 14:34language from management, the inventory velocity, it all
- 14:37points to yes, the Super cycle appears to be real.
- 14:40So here's the long term implication to consider.
- 14:43Because of this complexity wall, the entire memory industry now
- 14:46has to spend dramatically more capital to produce less growth.
- 14:49So will this permanently transform these companies into
- 14:53high margin, low volume specialist?
- 14:56Will they be forced to completely abandoned the old
- 14:58commoditized markets? That would be a fundamental
- 15:01business model shift unlike anything we've seen in decades.
- 15:04A phenomenal pivot point for the entire industry.
- 15:07Thank you for joining us for this deep dive into Micron
- 15:10Technology. And remember, all this analysis
- 15:12is based on publicly available sources and is for educational
- 15:16purposes only. It doesn't constitute investment
- 15:18advice, so please conduct your own due diligence.
- 15:21If you want to access more advanced content or listen to
- 15:23our full deep dives, you can join our Deep Value Club
- 15:25membership or subscribe on Apple and Spotify podcasts at the
- 15:28channel Deep Value Investing. Thank you.