Hi everyone. I just wanted to share my ideas about Micron Technology because I’ve been really fascinated by it lately.
This might be a bit of a long one because I want to give a fair chance to bull case and actually address the points often made. I know those aren’t often very popular. If you’re not in a for a long and relatively technical read you definitely want to skip this.
Before doing this, I want to note some important things.
I will be presenting a base case for MU predicated upon the negation of the bull arguments.
I will then summarize the bear case.
To do this, it will require **looking at the bull case and critically evaluating how reliable the arguments in favor of the bull case are**.
First, some definitions I want to share so you know exactly what I mean I say it:
* **Cyclical Commodity Industry:** A market where products are functionally identical (homogeneous) and buying decisions are driven entirely by price.
* **Oligopoly:** A market structure dominated by a very small number of sellers.
* **Oligopsony:** A market structure dominated by a very small number of buyers.
* **Monopsony:** A market structure dominated by a single buyer.
* **High-Bandwidth Memory (HBM):** A specialized, high-performance DRAM (Dynamic Random-Access Memory) architecture where memory chips are vertically stacked.
* **Strategic Customer Agreements (SCAs):** Binding, long-term supply contracts signed between chip manufacturers and their customers.
* **Backlog De-Fleshing:** The sudden removal, cancellation, or multi-quarter delaying of duplicate and phantom bookings by corporate buyers once a supply shortage ends.
* **Yield Maturation:** The predictable engineering curve where a factory transitions from early-stage, high-waste production to highly optimized efficiency.
* **Institutional distribution:** The systematic process where large, smart-money market participants quietly unload massive equity positions.
* **Operating Leverage:** The risk profile of a firm with high fixed costs.
* **Capital Expenditure (CapEx):** Funds used by a company to acquire, upgrade, and maintain physical assets.
* **13F Disclosures:** Mandatory quarterly reports filed by institutional investment managers with over $100 million in assets, detailing their equity holdings.
# Bull Case Arguments
* **The “Sold-Out through 2026/2027” Visibility:** Bulls emphasize that Micron’s entire HBM capacity for calendar years 2026 and 2027 is booked under finalized volume and pricing agreements. They argue this removes the traditional spot market volatility and guarantees record revenue visibility.
* **The 3-to-1 Wafers Cannibalization Effect:** Analysts point out that the technical reality that manufacturing High-Bandwidth Memory consumes roughly three to four times more silicon wafers per bit than standard DDR5 DRAM. The bulls claim that this will drive up average selling prices (ASPs) across their entire traditional portfolio.
* **The Mature Oligopoly Re-Rating:** The argument that the memory market is no longer a commodity sector but a disciplined, rational three-player oligopoly (Micron, Samsung, SK Hynix). Because building advanced fabs now costs $14B+, the barrier to entry is seen as an insurmountable barrier to entry, allowing the three players to preserve pricing power.
* **Strategic Customer Agreements (SCAs) as a Margin Floor:** The claim that long-term contracts signed directly with hyper-scalers for AI infrastructure act as a structural revenue cushion, preventing the catastrophic margin collapses seen in past cyclical downturns.
* **Exploding Free Cash Flow & Trailing Valuation “Cheapness”:** With trailing quarterly net income recently passing $28 billion and gross margins hitting historical peaks (guiding toward 81%), the forward P/E compresses into the single-digits or low teens. Bulls point to this forward data to argue the stock is fundamentally “cheap” relative to its earnings growth rate (PEG below 1.0).
* **The Only Sovereign U.S. Supply Chain Play:** As the sole major American-based supplier of advanced DRAM and HBM, bulls argue Micron will command a permanent premium and receive massive legislative backing (CHIPS Act subsidies/New York mega-complex) as Western hyper-scalers look to ensure a reliable supply outside of Asia.
* **HBM4E Changes The Game:** HBM4E (and assumedly future iterations of advanced memory) is NOT a pure commodity. HBM4E contains a logic die that is designed with customers for specific needs. Bulls argue that the ramp up of HBM4E and future iterations will free Micron from the cyclicality of the past because it is no longer a pure commodity manufacturer.
# Concerns with the Bull Arguments (Base Case)
* **The “Sold-Out through 2026/2027” Visibility:** Historically, in every major memory deficit, hyper-scalers panic-order double or even **triple** their actual needed capacity from **all three** suppliers. They do not just have a record of ordering the amount they actually want from one supplier, but from all three. I**n a base case, ordering double the amount they need from all three suppliers would imply a “ghost demand” of 6x the actual demand.** Based on historical data, this is a **conservative** assumption. Once supply starts exceeding the true demand by as little as **3% to 5%** an average selling price collapse historically occurs and margins rapidly compress back to \~40-50% before eventually flipping **completely negative** as supply continues to hit the market and **capex depreciation** stays on the balance sheet. This argument has been made **in every memory cycle in the last 30 years and has never protected margins.** Although Micron will certainly sell the **volume** they are producing, **there is no guarantee at what price they will be doing it at**. In the cyclical commodity industry price is everything. Building volume is worth it, but only if it doesn’t impact the margins by overbuilding. **The memory industry has never been able to achieve this.**
* **The 3-to-1 Wafers Cannibalization Effect:** HBM revenue accounts for only **2.4% of total revenue** for Micron. Assuming that this effect will even be relevant enough in the near/medium term to avoid the historical margin compression **is not rational**. Assuming that in the longer term that this **in isolation** will continue to create a supply bottleneck in the face of historic **record capex expenditure, yield maturation and backlog de-fleshing is not rational.** The increased silicon requirements increase the capital requirements of the industry. It does not increase the time it takes to actually produce the product. **This is more likely to make the cyclical busts worse as memory manufacturers have to become even more asset heavy** than in previous cycles.
* **The Mature Oligopoly Re-Rating:** An oligopoly is only able to exercise coordinated supply discipline if market shares are static. If market share is fluid the Nash equilibrium breaks down and **individual optimization forces aggressive competition.** Looking at the shifting market share amongst memory producers, a static market share **is not a reality in the memory industry.** The memory producers are locked into a prisoner’s dilemma where the first to start building to meet excess demand will have the best outcome, regardless of whether or not the other participants begin to build to meet that capacity. **This always historically leads to oversupply as all manufacturers are rationally forced into overbuilding, because the outcome is even mathematically worse if they don’t.** This doesn’t require a company to be foolish, undisciplined or short sighted, it is a structural guarantee in the market dynamics. One other point is that **the market is refusing to give memory commodity manufacturers this re-rating.** Instead of seeing this multiple expand recently, **the exact opposite is happening, the multiple is compressing.** This suggests that market participants do not currently believe that the industry has reached stability.
* **Strategic Customer Agreements (SCAs) as a Margin Floor:** If we were to assume that the SCAs guaranteed prices **as high as this year for 100% of Micron’s produced volume** this would **still not guarantee the company actually continues to grow earnings even as early as next year.** Were the revenue of next year to actually be able to continue to match the revenue of this year due to these SCAs, Micron’s earnings will actually still be lower because of the capex depreciation, which is growing at a 66% YOY clip. **This expenditure will completely eclipse any additional growth** from maturation yield volume or new fabs even if we assume prices do not decline next year, and **the SCAs in reality only cover 20% of DRAM and 33.3% of NAND** volume and prices are very likely to begin declining in early 2027 as new fabs come online. Additional problems are that we don’t even know the price floor of these SCAs, and **the historical precedent is that these contracts do not end up protecting ASP and volume from supply gluts due to backlog de-fleshing.**
* **Exploding Free Cash Flow & Trailing Valuation “Cheapness”:** The FCF displaying two things that make it look good in a short term time frame that might not look so good on a different timeline. Currently, the FCF is representing both the backwards looking sales made at peak supply shortage prices and the panic orders that were paid upfront. **This dramatically inflates** **the current FCF of the business and may not actually reflect the future FCF**, which will be heavily pressured by capex depreciation, likely loss of pricing power and the accounting miss of the products being shipped later but already paid for today. When the books don’t show the cash coming in now later, but do show that capex depreciation, FCF can be heavily impacted. **Micron is trading at a forward PE of under 7.** If you value that how you traditionally value a stock, Micron looks underpriced to an absolutely insane extent. People use this as an argument that the stock is cheap, but the issue with this is that in a cyclical commodity industry, you DO NOT use forward PE to value the stock. This is not a valid way to apply forward PE. **Historically, very low forward PE indicates the absolute peak of a stock price** in a cyclical commodity industry because two things are happening simultaneously. Firstly, we see a one time anomalous explosion in earnings & projected earnings caused by a severe supply shortage, pushing up the E in forward PE. Secondly, we see the market begin to completely refuse to assign a high valuation multiple to the stock. This is rarely a mistake or mispricing, and the market is not stupid. **The market refuses to assign a high forward multiple in some instances because it is acutely aware that the earnings growth projected are likely sustainable.** When looking at memory stocks, you’re typically meant to accumulate when the PE is high (and often negative) and sell when the forward PE is very low and the price increases of the underlying commodity start to decelerate. (Which is exactly what we’re seeing in memory).
* **The Only Sovereign U.S. Supply Chain Play:** I think this is the most solid argument from the bull case for Micron over say SK Hynix or Samsung right now. The US market undeniably has the most liquidity and capital and simply being in the US and listed on the NASDAQ commands a certain premium. The only issue with it is that that’s *already* factored into the multiple of Micron and has been for quite some time. I’m not sure I see any room for this to grow, but it’s unlikely to evaporate either. Hynix listing onto the Nasdaq might cause valuation multiples to converge, but I think that the physical location of being in the US provides it a unique geopolitical stability that Hynix still lacks.
* **HBM4E Changes The Game:** Some argue that Micron is no longer really part of a cyclical commodity industry. Discounting the revenue from standardized commodities (HBM3E and HBM4), HBM4E with logic dies custom designed for customers make up roughly $200m-$300m of Micron’s current revenue. **Currently, truly custom HBM is only making up 1% of total revenue.** Wall Street consensus models place mid 2027’s HBM total revenue at $3.5BN-$4BN. Even if we assume that custom memory grows to make up half of this volume, that is still only $2BN. Whilst an absolutely **insane** growth rate, make no mistake, this will not be sufficient to cause Micron to actually be able to grow earnings in the face of all the other structural issues facing them as early as next year, though it may alter the extent of the impact in a very minor way. Furthermore, **the only custom component that is actually part of HBM4E is not manufactured by Micron**. The majority of the stack of HBM is made of purely fungible memory chips. In addition, **Micron faces new unique problems by transitioning towards custom products.** Firstly, Micron is typically operating within an oligopsony but in the context of custom memory **it will be operating within a monopsony**. This is because memory that would not be purchased by one buyer could simply be sold to somebody else. That is no longer the case with custom memory, and **having large contracts that end up providing revenue that the manufacturer is completely dependent on whilst existing in this monopsony dynamic typically hits the ASP**, which eats into margins. Another issue with custom products is how the high expense of the custom dies produced by TSMC interacts with yield. Whilst yields reach maturation, the costs of lost yields become inflated. They have to partner with TSMC to actually manufacture the custom logic dies in the HB4ME, leaving them open to margin leakage as some of the profits they would’ve made are siphoned by TSMC, which operates at around a 50% margin themselves.
# Bear Case Arguments
* **Insider Behavior:** Over this year, **insiders have sold 23 times the amount of shares more than they have bought in dollar terms**, with zero buys after the run from the 300s and only sells. (1 buy versus 9 sales). Historically, Mehrotra would sell blocks netting $2 million to $5 million at a time. This year, the dollar volume exploded into single-day liquidations of $32.7 million and $13.57 million and $35.95 Million. These are **not** unstructured plans that sell at a fixed interval whatever the price is. **Insiders currently hold only 0.23% of the total float**. This is not necessarily a sign of abandoning a sinking ship, but it is not exactly a sign of high conviction towards the idea that the stock is an underpriced generational wealth opportunity.
* **Price Increase Deceleration:** Memory price hikes are starting to decelerate. This is difficult to reconcile with the narrative that demand is increasing at a faster rate than supply is. Typically, when demand is increasing faster than the supply, you would not expect to see a deceleration of this nature. This could be due to demand destruction, yield maturation & the double/triple ordering discussed earlier. Historically, this deceleration has called the exact apex of the last two cycles - long before underlying earnings actually hit their peaks.
* **Price Action & Technical Analysis:** In reaction to blowout earnings, the stock surged to new all time highs at around $1250 before ending lower to close on the day. This is the defining signature of a classic **institutional distribution top.** When institutions exit a cyclical stock, they do not just dump millions of shares into the open market, but instead use retail investors, momentum and trend-following quantitative funds, passive index funds and ETFs, squeezed short sellers and closet-index portfolio managers as exit liquidity to prevent the price from crashing so that they can offload further shares later at a higher price. On June 24 of this year, Micron reported a massive Q3 FY26 blowout. Management guided Q4 revenue to a staggering $50BN. Wall Street immediately flooded the (news/analyst) tape with reiterations, pushing price targets as high as $2,200. Retail FOMO reached an absolute fever pitch overnight. The next day, on June 25, the stock rocketed to its absolute historic all time high of $1,255.00 intraday. The moment it hit this peak, massive hidden institutional supply completely absorbed the incoming retail and momentum buying. The selling pressure was so intense that it entirely overwhelmed the buying surge to the point that the stock closed lower than it opened on this day. The next day on June 26, the stock continued to fall after a failed morning rally as institutional VWAP/TWAP engines slammed the bid. The stock collapsed 6.69% in a single session on massive volume, turning the previous support into an overhead ceiling of supply. The stock collapsed 6.69% in a single session on massive volume, turning previous support into an overhead ceiling of supply. Note that over the subsequent sessions, every attempt by retail to buy the dip was violently capped right at that breakdown point. In distribution, old support flipping into a rigid, algorithmic ceiling of supply is the ultimate structural confirmation that the big funds have completely shifted from “accumulation” to “inventory liquidation.” While skeptics frequently argue that fragmented retail capital lacks the structural depth to absorb large-scale institutional selling, the microstructure data from June 25 tells a different story.. In distribution, old support flipping into a rigid, algorithmic ceiling of supply is the ultimate structural confirmation that the big funds have completely shifted from “accumulation” to “inventory liquidation.” While skeptics frequently argue that fragmented retail capital lacks the structural depth to absorb large-scale institutional selling, the data from June 25 tells a different story. On June 25, an estimated 18.5 million \[±5%\] purchased shares was accounted for purely by retail. This is a approximately $22.9bn in one single session. This kind of massive selling pressure into a rip on a positive catalyst is a classic indicator of institutions entering a distribution phase for the stock. While the precise identities of the funds driving the massive June 25–26 distribution block trades remain masked behind anonymous TRF codes and institutional prime broker routers like MSCO and GSCO, they are following a well-established blueprint. Heavy smart-money distribution has been quietly accelerating all year. 13F disclosures from the primary leg of the run show that elite asset managers like Capital International Investors and Capital World Investors had already begun aggressively slicing their positions, cutting 59.5% and 27.8% of their exposure respectively. The mechanical VWAP markdown witnessed on June 26 is the continuation of this institutional distribution phase hitting its absolute volume climax. We are seeing in the chart a classic topping pattern indicating structural decay. The ultimate confirmation of this structural rollover is found in the **volume divergence**. Over the last two weeks, the volume profile has been completely asymmetric: massive, surging volume occurs strictly on down-days (like the 86.41-million-share liquidation on June 26), while the green relief sessions are characterized by thin, drying-up volume.
* **Higher Capex:** A key trap for Micron right now is its CapEx, which management recently raised to $27bn for 2026, with guidance that next year’s will climb to over $40bn. This is partly due to the higher costs of manufacturing advanced HBM. By becoming significantly more asset heavy at the peak of this pricing cycle, Micron has increased its operating leverage. This means that any future drawdowns could actually exceed the historical averages.
* **The Subsidized Supply Trap:** This structural over-expansion is entirely locked in by political and regulatory constraints. A massive portion of Micron’s domestic expansion is tied to $6.14 billion in federal CHIPS Act grants and billions more in New York State “Green CHIPS” tax credits. These state and federal subsidies are not free; they are contractually bound to rigid construction timelines, capital deployment milestones, and localized employment metrics. **In the event of a supply glut and collapsing margins, Micron will not be able to stop increasing their supply.**
* **The China Problem:** Another long-term threat to Micron is China. ChangXin Memory Technologies (CXMT) is rapidly scaling production towards 350,000 wafers per month, approaching Micron’s own volume capacity. Even if this new supply is unable to flood into the international market, Chinese scalers will begin to substitute any memory they were previously relying on from abroad with domestic supply, potentially pressuring Micron’s own ASP.
* **The HBM Cannibalization Trap:** The final blind spot is the artificial tight supply in standard DRAM, which makes up the lion’s share of Micron’s revenue. To meet aggressive HBM demand, Micron cannibalized its standard memory lines. This created a temporary shortage and inflated prices for standard PC and mobile memory. However, because this supply constraint is not structural, it could act as a coiled spring. The exact moment HBM demand cools off, Micron, Hynix & Samsung will be forced to pivot the newly expanded wafer lines back to standard DDR5. **Because of the 3-to-1 capacity multiplier, this shift will trigger a massive, overnight supply shock in the commodity market, driving a catastrophic collapse in general ASPs.**
# Summary, Conclusions & Thoughts
Personally, the structural cracks in the bull case are significant enough that I cannot justify holding Micron. This remains a binary-outcome equity: either the traditional semiconductor cycle is permanently dead and the stock is cheap, or it isn’t, and Micron is a textbook value trap facing years of deep underperformance. When confronted with binary risk, the safest default is to trust historical precedent. The cyclical nature of memory has not changed; if anything, the case for the next down-cycle being far more severe is vastly stronger than the case for the cycle being completely broken.
While Micron’s stock has likely established its macro apex for this cycle and entered a definitive institutional distribution phase, the company itself is fundamentally sound. It will undoubtedly survive and experience powerful cyclical peaks again in the future. This analysis assumes absolute flawless operational execution from management, zero product delays, and no sudden macro shocks.
Ultimately, no one can predict the future with absolute certainty, and this thesis represents my own research and market outlook. However, the most concerning element of the current market dynamic is the pervasive belief among retail investors that Micron is a low-risk, secular AI compounder at these valuations. This is incredibly dangerous. Misidentifying a high-risk asset as a low-risk asset can lead to catastrophic portfolio concentration, forcing investors to absorb maximum drawdowns because they mistake a temporary cyclical peak for a permanent secular floor.