Bert Hochfeld | TickerTarget | March 31, 2026
Micron Technology is in the midst of the most extraordinary financial inflection in the history of the memory semiconductor industry. The company reported Q2 FY26 revenue of $23.9 billion — nearly tripling year-over-year — with GAAP gross margins of 74.4%, GAAP operating margins of 67.6%, and free cash flow of $6.9 billion, all records. Q3 FY26 guidance of $33.5 billion at 81% gross margin implies a single quarter that exceeds Micron's full-year revenue for every year through FY24. The shares, at approximately $357 and a market cap of $403 billion, trade at a forward P/E of roughly 6.3X and a PEG ratio well below 0.5X — an extraordinary discount to both its direct peers (SK Hynix at 11.5X) and the AI infrastructure companies it enables. The HBM supercycle, structural supply constraints, and the advent of multi-year Strategic Customer Agreements (SCAs) argue persuasively that this is not a normal memory cycle. The question investors must answer is whether they believe memory has undergone a structural transformation or remains destined for mean-reversion. I believe the evidence strongly favors the former, and the shares represent a relative bargain at current levels.
I haven't covered Micron in the newsletter as a dedicated deep dive before, though I referenced the company's strong results in the December 2025 portfolio commentary when Micron's Q1 FY26 report helped drive late-week risk-on sentiment. Applying my framework to this company for the first time:
Micron is one of three companies globally capable of manufacturing advanced memory semiconductors — DRAM and NAND — alongside SK Hynix and Samsung. It is the only U.S.-based memory manufacturer of the three, a distinction that carries increasing strategic weight in the age of AI infrastructure buildout and CHIPS Act subsidies ($6.4 billion awarded to Micron).
What makes Micron exceptional in the current environment is the company's position in High Bandwidth Memory (HBM), the critical memory component for AI accelerator chips. HBM is not commoditized DRAM. It requires dedicated TSV (through-silicon via) stacking technology, thermal management expertise, 2-3 year manufacturing lead times, and hyperscaler qualification. Supply cannot respond quickly to price signals, which is precisely why Mehrotra can say with credibility that Micron can only supply "50% to two-thirds" of key customers' near-term requirements.
The competitive moat here is not permanent — Samsung has now qualified for Vera Rubin HBM4, making this a three-way race rather than the SK Hynix/Micron duopoly it was becoming. But the moat is real and durable for the next 12-18 months because of qualification cycles, yield learning curves, and the sheer magnitude of demand relative to supply. Management's introduction of the five-year SCA framework — replacing the prior one-year LTA structure — is the strongest signal that Micron believes it can lock in pricing and volume visibility across the cycle. If SCAs perform as described, they would represent a structural change in the memory business model.
The financial trajectory of this company over the past 19 quarters is, frankly, one of the most dramatic I have seen in the semiconductor space:
| Quarter | Calendar | Revenue ($B) | QoQ % | YoY % |
|---|---|---|---|---|
| Q1 FY24 | Nov-23 | $4.7 | +17.9% | +15.7% |
| Q2 FY24 | Feb-24 | $5.8 | +23.2% | +57.7% |
| Q3 FY24 | May-24 | $6.8 | +16.9% | +81.5% |
| Q4 FY24 | Aug-24 | $7.8 | +13.8% | +93.3% |
| Q1 FY25 | Nov-24 | $8.7 | +12.4% | +84.3% |
| Q2 FY25 | Feb-25 | $8.1 | -7.5% | +38.3% |
| Q3 FY25 | May-25 | $9.3 | +15.5% | +36.6% |
| Q4 FY25 | Aug-25 | $11.3 | +21.7% | +46.0% |
| Q1 FY26 | Nov-25 | $13.6 | +20.6% | +56.7% |
| Q2 FY26 | Feb-26 | $23.9 | +74.9% | +196.3% |
| Q3 FY26E | May-26 | $33.5 | +40.4% | ~260% |
What strikes me about this table is the acceleration. After appearing to decelerate in Q2-Q3 FY25 (YoY growth compressing to 36-38%), the business re-accelerated violently. The $10.2 billion sequential revenue increase in Q2 FY26 is the largest in the company's history — that single-quarter incremental revenue exceeds the total revenue of any pre-FY24 quarter. And the Q3 FY26 guide implies doing it again.
| Quarter | Calendar | GAAP GM% | GAAP OM% | Net Margin% | FCF Margin% |
|---|---|---|---|---|---|
| Q1 FY24 | Nov-23 | -0.7% | -23.9% | -26.1% | -8.4% |
| Q2 FY24 | Feb-24 | 18.5% | 3.3% | 13.6% | 0.6% |
| Q3 FY24 | May-24 | 26.9% | 10.6% | 4.9% | 6.2% |
| Q4 FY24 | Aug-24 | 35.3% | 19.6% | 11.4% | 4.2% |
| Q1 FY25 | Nov-24 | 38.4% | 25.0% | 21.5% | 0.4% |
| Q2 FY25 | Feb-25 | 36.8% | 22.0% | 19.7% | 10.6% |
| Q3 FY25 | May-25 | 37.7% | 23.3% | 20.3% | 21.0% |
| Q4 FY25 | Aug-25 | 44.7% | 32.3% | 28.3% | 7.1% |
| Q1 FY26 | Nov-25 | 56.0% | 45.0% | 38.4% | 28.6% |
| Q2 FY26 | Feb-26 | 74.4% | 67.6% | 57.8% | 28.9% |
| Q3 FY26E | May-26 | ~81% | ~77% | — | — |
From negative gross margins in Q1 FY24 to 74.4% in Q2 FY26 — a swing of 75 percentage points in 8 quarters. And the Q3 guide of 81% gross margin would put Micron above most enterprise software companies. I am a numbers guy, and these numbers are without precedent in the memory semiconductor industry.
| Quarter | FCF ($B) | FCF Margin |
|---|---|---|
| Q2 FY25 | $0.9 | 10.6% |
| Q3 FY25 | $1.9 | 21.0% |
| Q4 FY25 | $0.8 | 7.1% |
| Q1 FY26 | $3.9 | 28.6% |
| Q2 FY26 | $6.9 | 28.9% |
CFO Murphy noted on the call that Q3 FCF could "roughly double sequentially" — implying $13-14 billion in a single quarter. At that rate, the annualized FCF yield against a $403 billion market cap would approach 13-14%, which is extraordinary for a company growing revenue at triple-digit rates.
Every business unit delivered record revenue simultaneously in Q2 FY26. This is important — it means the supercycle is not confined to HBM alone.
| Business Unit | Q2 FY25 | Q2 FY26 | YoY % | GM% | OM% |
|---|---|---|---|---|---|
| Cloud Memory | $2.9B | $7.7B | +163% | 74% | 66% |
| Core Data Center | $1.8B | $5.7B | +211% | 74% | 67% |
| Mobile & Client | $2.2B | $7.7B | +245% | 79% | 76% |
| Auto & Embedded | $1.0B | $2.7B | +162% | 68% | 62% |
The Mobile & Client business unit — which many investors wrote off — delivered the highest gross margin of any segment at 79%. Core Data Center was the fastest-growing sequentially at +139% QoQ, reflecting the HBM volume surge arriving at scale. The product mix is: DRAM 79% (18.8B), NAND215.0B). DRAM bit shipments were up mid-single-digits QoQ, but prices increased in the mid-60s% range. NAND bit shipments were up low-single-digits, but prices increased in the high-70s% range. This is overwhelmingly a pricing story, not a volume story — and pricing is the more durable of the two.
I have written extensively about AI infrastructure as the defining investment wave of this era. My thesis with Nebius, Pure Storage, and Arista has been the picks-and-shovels layer: you don't need to own the ballpark to enjoy the game. Micron fits squarely in this framework. If anything, it is more fundamental — you cannot build the ballpark without memory.
Mehrotra articulated this precisely on the call: "AI has not just increased demand for memory; it has fundamentally recast memory as a defining strategic asset in the AI era." He went further in a CNBC interview the following day, calling memory "a strategic chokepoint." This is language escalation from "AI is central to our growth story" (September 2025) to "strategic chokepoint" — a CEO signaling that he believes the structural transformation is real and permanent.
The HBM franchise changes the character of Micron's revenue. Prior memory cycles were characterized by interchangeable supply — any DRAM was a DRAM, and overcapacity compressed pricing on a 3-4 year cycle. HBM requires dedicated technology, 2-3 year lead times, and customer qualification. Supply cannot respond quickly to price signals, which is why Mehrotra can say the market remains tight "beyond calendar 2026."
Management introduced the Strategic Customer Agreement (SCA) concept with considerable emphasis — 8+ mentions in the Q2 call, raised unprompted in prepared remarks before any analyst asked. These replace the prior one-year LTA structure. Key characteristics from management commentary:
Analysts pressed repeatedly on downside protection mechanisms, cancellation terms, and pricing renegotiation frequency. Management deflected on all specifics, citing confidentiality. I understand why they cannot disclose terms, but the deflection creates an analytical gap. What I can say: if the SCAs perform as described — providing pricing floors in cyclical troughs while locking in volumes during tight markets — they would represent the most significant structural change in the memory business model in the industry's history. The question is whether the "robust provisions" are genuinely robust, or whether they contain renegotiation clauses that effectively revert to spot pricing when conditions ease.
Sanjay Mehrotra has run Micron for nearly 9 years, through a brutal down-cycle (FY23: 4 consecutive quarters of GAAP losses, total net loss of $5.8 billion) and a historic recovery. His track record deserves respect:
Beat-and-raise cadence is accelerating:
| Quarter | Guide ($B) | Actual ($B) | Beat % |
|---|---|---|---|
| Q1 FY25 | $8.7 | $8.7 | +0.1% |
| Q2 FY25 | $7.9 | $8.1 | +1.9% |
| Q3 FY25 | $8.8 | $9.3 | +5.7% |
| Q1 FY26 | $12.5 | $13.6 | +9.1% |
| Q2 FY26 | $18.7 | $23.9 | +27.6% |
Five consecutive beats with escalating magnitude — the Q2 FY26 beat of $5.2 billion is larger than most semiconductor companies' quarterly revenue. The guidance pattern suggests deliberately undemanding estimates, which I view as a positive signal of management credibility. Mehrotra's public admission that Micron can only deliver 50-67% of key customers' near-term requirements is unusual and credible — most CEOs hedge on supply/demand gaps. He quantified it.
The 30% dividend increase (to $0.15/quarter), two credit upgrades to solid BBB, and $5 billion in debt reduction over three quarters further demonstrate capital allocation discipline. Share repurchases are constrained at $350 million/quarter by CHIPS Act terms, but the signaling is consistently shareholder-friendly.
Employee sentiment: Glassdoor 3.9/5 (7,784 reviews), Recommend 78%, Positive Business Outlook 75%. Named to Glassdoor's inaugural 2026 Best Places to Work in Tech & AI list. Management quality (2.9/5 on Blind) remains the pain point — hierarchy and micromanagement flagged. This is a monitoring item, not a thesis-breaker.
This is the elephant in the room. Every prior memory cycle has reverted. The market is pricing in a significant earnings reversal — at a forward P/E of 6.3X and EPS run-rate of ~$57 (NTM consensus), the market implicitly expects FY28 earnings to revert to the $10-15B range. If HBM pricing holds and SCAs perform, this is dramatically wrong. If HBM demand softens and SCA provisions permit renegotiation, the market's skepticism could prove justified. I believe the structural transformation thesis is correct, but I acknowledge this is the central risk and will not pretend otherwise.
Samsung has now qualified for NVIDIA Vera Rubin HBM4 alongside SK Hynix and Micron. Three suppliers on the platform reduces the pricing power of the SK Hynix/Micron near-duopoly. Samsung has enormous manufacturing capacity and strong incentives to regain HBM share (its share fell from roughly even to 17% in 2025). If Samsung ramps HBM4 yield successfully, HBM pricing comes under pressure in 2027+. The 16-Hi HBM4 race (where yield and packaging execution matters more than design-in share) is the next competitive battleground.
FY26 CapEx has been raised to >$25 billion (from the prior 20billionestimateatQ1).FY27CapExwillstepupby>10 billion in construction spend alone. This is the 200billionUSmanufacturingcommitmentinaction—IdahoFab1(mid − CY2027), IdahoFab2(construction2026), NewYork(brokenground), SingaporeNANDfab(2028), Tongluoacquisition(FY28production).Ifdemanddeterioratesbeforethesefabscomeonline, theCapExoutpacesFCFandthebalancesheetdeteriorates.Thecounter : atcurrentFCFrunrates(25-30B annualized if Q3 guidance delivers), the CapEx is comfortably covered. But it requires the supercycle to sustain.
Google Research published TurboQuant on March 25, 2026 — a KV cache compression algorithm achieving 6x memory reduction on H100 GPUs for LLM inference. The stock fell ~14% from its ATH of $471. Morgan Stanley explicitly labeled it "evolutionary development, with basically no surprises for memory." TrendForce concurs. The key nuance: TurboQuant only affects KV cache in inference workloads — not training, not model weights. The likely effect is that the same hardware handles more throughput, which increases aggregate memory demand over time. This is a stock narrative risk, not a fundamental risk. Panics tend to be indiscriminate.
Not included in Q3 FY26 guidance explicitly. Micron's US-only manufacturing origin provides a "security premium" with Western hyperscalers, and CHIPS Act subsidies create policy protection. But China exposure (~10% of revenue historically) and supply chain dependencies remain non-zero risks.
Here is where the analysis gets genuinely interesting. Let me work through the numbers at the updated market cap of approximately 403billion(sharesat 357, 1.14 billion diluted):
| Metric | Value |
|---|---|
| Market Cap | ~$403B |
| TTM Revenue (Q3 FY25 – Q2 FY26) | $58.1B |
| P/S (TTM) | 6.9X |
| Forward Revenue (annualize Q3 guide) | ~$134B |
| P/S (forward, annualized) | 3.0X |
| Trailing P/E (TTM GAAP EPS $21.18) | 16.9X |
| Forward P/E (NTM consensus ~$57) | 6.3X |
| Annualized Q3 Guide P/E ($19.15 x 4 = $76.60) | 4.7X |
| FCF Yield (Q2 FY26 annualized, $27.6B) | 6.9% |
| FCF Yield (if Q3 doubles per CFO, ~$52B ann.) | ~12.9% |
| PEG (forward P/E 6.3 / ~100% EPS growth) | <0.1X |
| Company | Forward P/E | P/S | Revenue Growth | GM% |
|---|---|---|---|---|
| Micron (MU) | 6.3X | 6.9X TTM | +196% YoY | 74.4% |
| SK Hynix | ~11.5X | 5.6X | ~120% YoY est. | ~65% est. |
| Samsung (Blended) | ~15X | — | — | ~50% est. |
| NVIDIA (demand driver) | 35-40X | — | ~70% YoY | ~75% |
Micron trades at a 45% discount to SK Hynix on forward P/E — its closest peer and the company with the largest HBM market share. This discount reflects the market's ingrained belief that Micron is the "commodity" memory company while SK Hynix is the "HBM premium" company. But at 74.4% gross margins (approaching NVIDIA-like territory) and with HBM4 now in volume production for Vera Rubin, the gap is not justified by fundamentals. Even if I apply a 20% cyclical risk discount to SK Hynix's multiple (recognising Micron's larger commodity DRAM/NAND exposure), the implied fair multiple for MU would be ~9X forward P/E — roughly 43% upside from here.
At a PEG well below 0.5X, a forward P/E of 6.3X, and a 38-analyst consensus with an average target of ~$515 (44% upside), the valuation case is straightforward: the shares are priced for a brutal reversion that may not come.
At $403B market cap and a Q3 annualized EPS run-rate of 76.60, themarketisimplyingthatbyFY28 − FY29, earningsrevertto 25-30B annually (10-13X trough P/E normalisation). If HBM LTAs/SCAs hold pricing through FY28 and the TAM reaches $100B as management projects, the stock is dramatically mispriced. If the cycle reverts on the old schedule — 2027 peak, 2028-2029 trough — the market's caution is prudent.
I believe the weight of evidence — SCAs, sold-out supply through 2027, structural supply constraints from HBM trade ratios and construction lead times, AI capex acceleration across hyperscalers — favours the secular transformation thesis over the mean-reversion thesis.
Micron Technology at ~$357 represents a relative bargain. The company is delivering software-like margins on semiconductor revenue, growing at triple-digit rates, generating record free cash flow, and trading at a deep discount to its nearest peer on every valuation metric. The Q3 FY26 guide — if delivered — would be one of the most profitable quarters in semiconductor industry history. Management credibility is intact, the beat-and-raise cadence is accelerating, and the first five-year SCA signals a structural evolution in the business model.
The risk is cyclical mean-reversion. I acknowledge it. But it is rare to see quite this much valuation compression for a company that is profitable, generating $7 billion in quarterly free cash flow, and which continues to grow at 196% year-over-year. The market is demanding proof that memory has structurally changed. Q2 FY26 and the Q3 guide provide that proof. If Q3 FY26 delivers at or above the guide — and the beat pattern suggests it will — the re-rating should follow.
Recommendation: Buy. My belief is that the shares will deliver significant positive alpha over the next 12 months if management executes on the Q3 guide and SCAs provide the durability they promise. The condition for being wrong: Samsung ramps HBM4 yield faster than expected, HBM pricing breaks, and SCAs prove to have renegotiation clauses that revert to spot economics. I would monitor Samsung's HBM4 production yields and hyperscaler CapEx commentary as the two most important leading indicators.
Suggested portfolio position: 3-5% in the high growth portfolio. This is not a 10%+ position because the cyclical risk, while diminished, has not been eliminated. The five-year SCA framework needs at least two more quarters of execution before I would increase conviction from the current 4/5 level.
Sources: Micron Q2 FY26 Earnings Press Release (March 18, 2026), Q2 FY26 Earnings Call Transcript (Motley Fool), GuruFocus Forward PE, TradingKey Memory Guide, SK Hynix Valuation (Simply Wall St), CNBC Q2 Earnings, Morgan Stanley TurboQuant Analysis, TrendForce TurboQuant Decoding.