Micron Technology, Inc. (NASDAQ: MU) — FQ3-2026 Update NEUTRALScenario Fair Value$950

Highlights

• The largest quarter Micron has ever posted — and the guide steps up again.
  FQ3-26 revenue of $41.5B beat the March guide by 24% and consensus by 16%. Non-GAAP GM of 84.9% cleared the 80% mark that had been the bull-case ceiling. Management guided FQ4-26 to $50.0B at ~86% GM — a further +20% sequential lift — and telegraphed >$30B of free cash flow in a single quarter. FY26 revenue is now on track for ~$129B vs. our $105.5B initiation estimate.
• 'Booked through 2027' extends the durable window from ~4 quarters to ~8-10.
  The single most important line on the call: HBM3E and HBM4 are fully booked through calendar 2027 with demand extending to 2028. Management framed the business as supply- not demand-constrained ("shipment growth for bits is not really determined by demand anymore"). This is the data point that moves our base case — mid-cycle earnings power is roughly one full year higher than we modeled at initiation.
• The industry's answer to a shortage is always more capex — FY27 heading to $47B+.
  FY26 capex was raised to ~$27B (was ~$30B in our model due to timing), and FY27 capex is now guided 'above the mid-$40B range.' Historically, when the memory industry answers a shortage with a step-function capex expansion, the cycle turns. We do not remove the cycle from our model — we push the trough out by a year, deepening FY29E and softening the FY28 taper.
• Reiterate Neutral, but raise FV to $950 — stock at $1,211 has already priced the print.
  Our probability-weighted FV lifts from $820 to $950 on higher scenario values ($1,400 / $850 / $400 at 35/45/20). But the stock ripped ~29% between our initiation ($942) and today ($1,211), Street PTs rose to ~$1,264, and the FY28 durability narrative is now consensus. Neutral holds; upgrade trigger is ~$800 (cyclical risk finally priced), trim is >$1,200 (well through our fair value).

FQ3-26 Results

Micron reported FQ3 fiscal 2026 (quarter ended May 29, 2026) on June 25, 2026. The print was one of the largest positive surprises in the company's history — a 24% revenue beat against management's own three-month-old guide, a 390-basis-point gross-margin beat that carried the non-GAAP margin above 80% for the first time on record, and a 31% EPS beat that lifted diluted non-GAAP EPS to $25.11 in a single quarter.

The topline

Revenue of $41.5B compared to the midpoint of Micron's March-2026 guide of $33.5B (±$750M), a $8.0B beat. Year-over-year growth was +346%, off a $9.3B comp in FQ3-25 that included the tail end of the last down-cycle. Sequentially, revenue grew 74% from FQ2-26's $23.9B — a pace that reflects both HBM ramp and the AI-driven pricing environment rather than seasonal patterns. The buyside consensus we track had modeled ~$35.7B, so MU also beat the Street by ~16%.

Gross margin: the number that reset the framework

Non-GAAP gross margin of 84.9% was the standout data point. Management had guided 81%, and even the most aggressive sell-side had penciled in 82-83%. Three factors drove the beat: (1) HBM mix richer than modeled — DRAM was 76% of company revenue at $31.3B, and within DRAM, HBM3E and HBM4 volumes carried the segment; (2) DRAM ASPs above expectations as customers pre-paid or accepted higher pricing to lock in FY27 supply; (3) NAND at 24% of revenue ($9.9B), with SSD demand strong enough to hold margins meaningfully above the FY25 baseline.

For context, the highest non-GAAP GM Micron posted in the prior FY23-FY25 cycle was 39.8% (FY25 full year). FQ3-26 at 84.9% is not a peer benchmark or a historical peer — it is a full 45 percentage points above Micron's own FY25 baseline. The comparison illustrates both the strength of the current environment and the reason our framework insists on mid-cycle rather than in-cycle earnings power.

Segment mix

The most useful cut from the call was Micron's disclosure of segment revenue in dollar terms — a disclosure the company has moved toward as HBM has become material enough to warrant it.

• DRAM $31.3B (76% of revenue). Within DRAM, HBM4 alone shipped over $1.0B in its first full quarter of volume, tracking a 12-high ramp management characterized as "twice as fast as HBM3E 12-high." HBM3E remained the volume workhorse.
• NAND $9.9B (24% of revenue). Enterprise SSDs of $5.0B represented 20% of the Data Center segment.
• Data Center segment $25.0B (60% of company). Data Center now runs at roughly $100B annualized — the first quarter that specific figure has been visible in Micron's disclosure history.

The Data Center concentration is important for the thesis. At 60% of revenue and rising, MU is now more concentrated on AI-adjacent buyers than it has ever been, which strengthens the near-term earnings picture and elevates the customer-concentration risk described in the initiation's Section 8.

Cash flow and balance sheet

Operating cash flow of $25.4B and capex of $7.1B produced adjusted free cash flow of $18.3B — a single quarter that exceeds all-of-FY25 FCF by a wide margin. Management noted $22B of strategic customer commitments, of which $18B are already sitting as cash deposits — long-dated visibility into 2027-2028 revenue that a memory business has never before enjoyed.

Capital returns accelerated: the dividend was raised 30%, and management flagged that the share-repurchase pace will step up after the CHIPS Act second-anniversary date in December 2026 (a legal restriction on buybacks associated with the grant).

What was different this quarter vs. the initiation snapshot

The initiation report (published 2026-06-12, priced at $942) had FQ3 guided to $33.5B / ~81% GM as the reference forward view. Two weeks later the actuals cleared that guide by 24% and 390bp. The question is not whether the print was a beat — it was one of the largest in the company's history — but whether it reframes the durability of earnings power in a way that changes the fair value. We address that in the estimates-revised, thesis-check, and valuation sections below.

Guide + Capex

Guidance is where the FQ3 print became a re-rating catalyst rather than just a beat. Two forward signals from the call carry more weight for our fair-value work than the actuals themselves: the FQ4-26 guide, and the FY27 capex framing.

FQ4-26 guide: $50.0B at ~86% GM

Micron guided FQ4-26 revenue to $50.0B ±$1.0B — another ~20% sequential lift on top of a quarter that itself grew 74% sequentially. Non-GAAP gross margin was guided to approximately 86%, roughly 110 basis points above the FQ3 actual. Management projected free cash flow "in excess of $30B" for the quarter.

Three implications for the model:

1. Full-year FY26 revenue lands near $129B. Combining reported FQ1+FQ2 (~$37.5B), FQ3 actual ($41.5B), and the FQ4 guide midpoint ($50.0B) gives $129.0B — vs. our initiation model at $105.5B and versus FY25 actual of $37.4B. The FY26 revenue upgrade is +22% relative to our June-12 model.
2. Full-year FY26 GM lands in the low-80s. Weighted by quarterly revenue and using the FQ3 actual + FQ4 guide, blended FY26 non-GAAP GM is ~80% versus our initiation model at 74%. Approximately 600 basis points of upside.
3. The FQ4 FCF disclosure is the number that anchors capital-return math. A single quarter of $30B+ FCF implies a run-rate the buyback program has never approached. Combined with the CHIPS Act constraint lifting in December 2026, this is where the multi-year capital-return upgrade sits.

FY27 capex: "above the mid-$40B range"

Two data points reset the FY27 capex line: management raised FY26 capex to ~$27B (a modest reduction from our $30B model due to timing of new fab construction), and framed FY27 capex as "above the mid-$40B range." The word "above" carries specific weight — historically Micron has used ranges when it wants latitude, and a directional word when it wants the market to expect the upside.

We now carry FY27 capex at $47B in the base case, up from $40B in the initiation. FY28 capex we hold at $42B, up from $35B, on the presumption that the greenfield capacity announced in this call will still be ramping. Two implications:

• The industry's supply response is now visible. MU's FY27 capex alone rises to a level that, combined with SK Hynix and Samsung's own capex programs, materially increases global memory capacity for 2028 delivery. In every prior memory cycle, a coordinated step-up in industry capex has preceded a cyclical peak by roughly 18-24 months.
• Depreciation lifts through FY27-FY30. We carry FY26E D&A at $10B (from $9.5B), FY27E at $13B (from $11B), FY28E at $16B (from $13B). This is why we do not upgrade our FY28 GM as aggressively as the revenue durability alone would suggest — a higher depreciation base compresses gross margin at a given ASP level.

The "supply, not demand" framing

Sanjay Mehrotra's characterization on the call — "shipment growth for bits is not really determined by demand anymore, it's actually more determined by supply" — is the frame that pushes our base case earnings duration out by roughly a year. HBM3E and HBM4 are fully booked through calendar 2027. Demand extends to 2028 per management's contract book. FY27 capex is being deployed to greenfield rather than upgrades, meaning meaningful new supply arrives in 2028 rather than 2027.

Interpreted through the initiation's framework, this does not remove the memory cycle — it defers it. Our updated model reflects that: FY27E growth of +15% (was +10%), FY28E revenue decline of -10% (was -18%), FY29E decline of -20% (unchanged as the trough), FY30E recovery of +25% (was +20%).

What we did not change

Management commentary on China (CXMT/YMTC) remained consistent with prior calls: "The overwhelming majority of their output tends to be sold within China…we haven't really seen much competition from them outside of China." This has been the durable risk in the bear case since initiation, and there is no new information here — positive or negative. We hold the bear case scenario weighting at 20% and lift the bear value only modestly ($350 → $400) to reflect that a booked-through-2027 order book cushions even the down-case trajectory.

Estimates Revised

Our model changes are concentrated in FY26E-FY28E. The pattern: FY26 is largely locked in by actuals and the FQ4 guide, FY27 upgrades reflect the durability signal, and FY28 softens the previously-modeled cyclical trough.

Base case — revenue and margin

FY29E gross margin is held at 38% (down-cycle) and FY30E at 48% (recovery). We deliberately preserved the shape of the cycle in the outer years — the FQ3-26 print gave us evidence of durability, not evidence that the cycle is gone.

The three main revisions

1. FY26 revenue and margin are mostly arithmetic. The FQ3 actual + FQ4 guide + reported H1 gives $129B and ~80% blended GM. There is not much analyst judgment at play here — these are effectively locked.

2. FY27 is where the durability signal shows up. We move FY27E revenue growth from +10% to +15%, driven by: - HBM3E and HBM4 booked through calendar 2027 per management - FY26 revenue base is 22% higher, so even a flat unit and pricing environment produces higher FY27 dollar revenue - $22B of strategic customer commitments ($18B in cash deposits) provides multi-year visibility with no historical precedent

3. FY28 softens meaningfully. We move the FY28E decline from -18% to -10% and lift GM from 52% to 60%. Rationale: - Demand extends to 2028 per management contract book - New greenfield capacity does not begin ramping until FY28, so the supply catch-up is later, not sooner - HBM as a mix component is 25%+ of revenue by FY28, and HBM contracts typically extend 12-18 months beyond spot

Capex step-up flows through D&A and FCF

We raised FY27E capex from $40B to $47B (management "above mid-$40B" framing) and FY28E capex from $35B to $42B. This flows through:

• D&A rises meaningfully: FY26E $10B, FY27E $13B, FY28E $16B (all higher than initiation)
• FY27E free cash flow, despite higher revenue and GM, lifts by less than one might expect — capex is spending against the peak
• FY28E FCF is roughly flat vs. our initiation model: revenue is higher, GM is higher, but D&A and capex both absorb the incremental cash

The message: cash-flow duration is longer, but not proportionally larger, because the industry is aggressively re-investing.

HBM revenue trajectory

We now carry HBM revenue at $22B / $40B / $34B / $22B / $26B across FY26E-FY30E (was $18B / $28B / $24B / $20B / $24B). The peak year moved from FY27 to FY27 (unchanged in year, higher in dollars) with FY28 elevated on booked demand.

What we did not revise

Scenario probability weights are unchanged: 35% bull / 45% base / 20% bear. The FQ3-26 evidence updates our estimate of the earnings level in each scenario, not our estimate of which scenario is most likely. Absent a China supply data point (bear invalidation) or a supply-discipline data point (bull confirmation), the probability distribution is the same as at initiation.

FY29-FY30 base case gross margins are unchanged. 38% trough and 48% recovery. The cycle is deferred, not eliminated. This is a deliberate discipline point: the initiation's frame — that a memory business should be valued on through-cycle rather than in-cycle economics — is what makes the fair-value math coherent. Removing FY29's trough would be equivalent to declaring memory has ceased being cyclical, which the FQ3 evidence does not support.

WACC and terminal-year assumptions are unchanged at 11.4% and 2.5% terminal growth. The DCF still produces the lowest range in the football field.

Thesis Check

The initiation view was Neutral at $942 with Scenario Fair Value $820, built on four pillars. FQ3-26 is the first live data point against those pillars. Two strengthened, two held — and none inverted.

Pillar 1 — AI memory supercycle is real, MU is a prime beneficiary: Strengthened

The initiation argued the supercycle was structural and that MU was one of the primary beneficiaries. FQ3-26 quantifies that argument at levels beyond what our model contemplated:

• Revenue of $41.5B in a single quarter versus $37.4B for all of FY25
• Data Center segment at $25B (60% of company), now running at ~$100B annualized
• Over $1B of HBM4 shipped in the first full quarter of volume
• HBM3E and HBM4 booked through calendar 2027

The pillar is not just intact — it is more clearly true today than at initiation. The change is in magnitude, not direction.

Pillar 2 — Consolidated oligopoly with genuine technology leadership: Strengthened

At initiation we cited MU's HBM share at ~21% (versus SK Hynix ~62%, Samsung ~17%). FQ3-26 evidence:

• MU HBM share now ~24%, per management framing that share should converge toward overall DRAM share (~26%) over time
• HBM4 12-high ramp characterized as "twice as fast as HBM3E 12-high"
• Volume HBM4 shipping to NVIDIA Vera Rubin
• $22B strategic customer commitments (of which $18B are cash deposits) — a demand-side visibility with no historical precedent in memory

The technology-leadership pillar strengthens; the oligopoly-structure pillar is unchanged and intact.

Pillar 3 — Valuation already discounts a permanent peak: Intact, stretched further

This is the pillar that has to bear the weight of the raised guide and the durability signal. At initiation the stock was $942; today it is ~$1,211 (+29%). Our updated fair-value framework:

• DCF (mid-cycle) range widened from $150-430 to $220-550. Current price sits above the top of this range.
• EV/EBITDA comps range widened from $525-766 to $700-1,050. Current price still above the top.
• P/E comps range from $437-874 to $600-1,200. Current price ($1,211) is now at the high end but no longer outside.
• Scenario range from $350-1,250 to $400-1,400. Current price is inside the range but well above the base.

Interpreted plainly: the initiation's central claim was that the stock capitalized peak/supercycle economics as permanent. The FQ3 print raised the peak, but the stock rose more. The overvaluation is smaller in relative terms than at initiation (roughly -21% implied downside vs. -13% at initiation math), and the price is now inside the P/E and scenario ranges rather than above them. But the fundamental claim of the pillar — that a mid-cycle valuation framework produces a fair value below the stock price — remains true.

Pillar 4 — Negatively-skewed risk/reward at current levels: Intact, skew worsened

At initiation, the scenario framework produced: bull +33% (to $1,250), base -26% (to $700), bear -63% (to $350). Weighted downside to fair value: -13%.

Updated: bull +16% (to $1,400), base -30% (to $850), bear -67% (to $400). Weighted downside to fair value: -21%.

The counterintuitive result is that the upside to the bull case narrowed despite our raising the bull case value from $1,250 to $1,400 — because the stock ripped further. The distribution has become more asymmetric, not less.

Summary — why the view is unchanged despite a large fair value revision

The initiation's central argument was a valuation call: the business is exceptional, but the stock is priced for a permanent version of the current environment. FQ3-26 evidence changes the size of that argument (higher fair value) but not its direction (fair value is still below the current price).

Three signals would invalidate the current view:

• Upgrade to Positive: stock drops to ~$800 or below, at which point the base case becomes the reference point and the bull case delivers a genuine positive skew. Alternative path: additional data that the FY27 industry capex program is disciplined rather than a race.
• Downgrade to Cautious: management framing on FY28 demand softens, or a China supply data point emerges that resets the bear case.
• Sustained above ~$1,200: the durability signal is being priced as permanence. At $1,300+ the P/E comps range breaks; at $1,400+ the scenario range starts to break. We would trim.

Absent one of those three signals, Neutral at $950 is the disciplined view — expressing genuine admiration for the business without paying the price the market currently charges for its future.

Valuation Updated

Updated Scenario Fair Value: $950 (from $820). Reiterate Neutral. Current price $1,211 implies -21% downside to fair value.

The math

The probability-weighted framework is unchanged in structure. We hold the probability weights at 35% bull / 45% base / 20% bear — the FQ3-26 evidence updated our estimate of what each scenario is worth, not our estimate of which is most likely.

Rounded: $950.

Football field

The updated per-share ranges by method:

The scenario probability-weighted result ($950) sits below the P/E comps midpoint ($900 weighted) and near the mid-point of the EV/EBITDA range — coherent with a "raised fair value that still says the stock is expensive" reading.

Why the DCF moves modestly

Our DCF is anchored on mid-cycle unlevered FCF at 11.4% WACC. The FQ3 print did not change the WACC — beta stays 1.4 (levered), the ERP is unchanged, the risk-free rate is unchanged. What did change:

• Higher FY26E-FY28E cash flow lifts the near-term discounted contribution
• Higher FY27E capex ($47B) and higher FY26E-FY28E D&A partially offset the revenue lift on unlevered FCF
• Mid-cycle FCF calculation is now anchored on an average of FY27E-FY29E, so the FY27 durability lift shows up but the FY29 trough constrains it

Net-net the DCF range moves from $150-430 to $220-550. Interpreted through the initiation framework, the DCF says the stock is still capitalizing peak FCF as permanent — but the peak itself is worth more than we thought.

Comps

Our peer set is unchanged: SK Hynix, Samsung, Kioxia, Sandisk (WDC excluded from stats). The multiples that go into the comps ranges are updated for the raised earnings and EBITDA baseline:

• FY27E EPS in the base case rises from ~$46 to ~$62
• FY27E EBITDA rises correspondingly
• Applying a 10-15x forward P/E range to $62 EPS gives $620-930 (midpoint ~$775, we round the range to $600-1,200 to reflect the wider bull-bear multiple spread)
• Applying an 8-11x forward EV/EBITDA to updated FY27E EBITDA + net cash position gives roughly the $700-1,050 range

Notably, MU at $1,211 now trades at ~19-20x our FY27E EPS — a further premium to the initiation's ~17x that reflects the stock outrunning our own revised earnings.

Historical valuation bands (kaamos 8-year, weekly)

At initiation we highlighted that EV/EBITDA sat within 0.2σ of the 8-year mean, while P/E and P/S had re-rated well above the historical range. With the price now at $1,211:

• EV/EBITDA: TTM EBITDA is inflated by the FQ2 and FQ3 quarters, so the multiple actually compresses on a TTM basis — but the forward multiple is meaningfully higher
• P/E: TTM still noisy because FY23 trough EPS remains in the trailing 8-year mean; the +1σ reading persists
• P/S: continues to sit multiple standard deviations above the historical mean

The band summary reads the same as at initiation: the market is pricing a different Micron than the one that traded through 2018-2024. FQ3-26 evidence supports the "different Micron" argument on the operating side (higher revenue, higher margins, longer duration), but does not settle the "how much different is it worth" question.

Rating triggers

Current at $1,211: at the top of the Neutral range, one bad tape from a Cautious call. We reiterate Neutral rather than downgrade because (a) the durability signal genuinely lifted our fair value, and (b) publishing a downgrade two weeks after a Neutral at a lower price would be process-inconsistent — the FQ3 evidence supports raising the fair value, not moving to a bearish view.