Phase Space AI

Micron Valuation Analysis 2026-07-27

Micron Technology [MU]

Micron Technology, Inc. [MU] — Valuation Analysis

Investing Hub Research | July 27, 2026 | Task 3 of 9

Prerequisite: Task 1 (Company Research) and Task 2 (Financial Model, Micron_Financial_Model_2026-07-27.xlsx, verified in Excel) complete.

Price $900.20 (Alpaca SIP, 2026-07-27 close; revised from $884.75 intraday under the v1.4.0 re-run). All model outputs below were read back from the live workbook after recalculation in Excel — not from the Python that wrote it.

---

1. WHERE THE VALUE QUESTION ACTUALLY SITS

These last two lines are the entire analytical problem, and they are not contradictory — they are the signature of a cyclical at a peak. A memory company prints its lowest P/E at the top of the cycle, because the denominator is peak earnings, and its highest P/E at the bottom. Price-to-book is the metric that has historically worked on this industry precisely because book value doesn't swing with the cycle the way earnings do. At 10.1x book, Micron trades at a level with no precedent in its own history — at the FY2018 cycle peak, with a then-record 58.9% gross margin, the company was worth roughly 1.5x book.

2. METHOD 1 (PRIMARY) — NORMALISED EARNINGS POWER

For a deep cyclical printing a record margin, mid-cycle earnings power is the correct primary anchor. Assumptions (all flagged as assumptions, all in blue in the workbook):

Note how generous this already is: it assumes Micron permanently sustains revenue 3.5x its FY2025 level at a gross margin 16 points above its historical median. It still produces a value 47% below the current price.

Sensitivity — the honest picture (normalised revenue × normalised gross margin, at 13x)

(Values computed live in the workbook's DCF tab; the four corners are shown rounded here.)

Only the top-right corner of this table — $160–190B of permanent revenue at a 55–62% permanent gross margin — justifies today's price. That is the regime-change bet stated numerically: revenue permanently 4–5x FY2025 at a margin permanently above the company's all-time peak annual margin. It is not impossible. It is simply not the base case that history supports, and the memo should not pretend otherwise in either direction.

3. METHOD 2 — DISCOUNTED CASH FLOW (cross-check)

WACC 14.7% (risk-free 4.7% + beta 2.0 × ERP 5.0%; the measured 2-year beta vs. SPY is 2.41 and the 1-year is 3.25 — using 2.0 is a deliberate haircut, since a 12x melt-up inflates spot beta). Explicit period FY2027–FY2031, terminal value at 9x exit EV/EBIT. Scenario drivers are on the Scenarios tab.

Probability-weighted DCF value: $479.21 (−46.8% vs. $900.20), at 30/48/22 probabilities, now derived from the explicit two-tranche split rather than a blended margin assumption — see §9.

(All three fall vs. v1.4.0's $194.85 / $439.56 / $1,028.07. Nothing about the contractual evidence weakened — what changed is that the contract benefit is no longer allowed to run past contract expiry into FY2031 and the terminal value. See §9 for the full attribution.)

The single most important line in this report: the Bull case above already matches consensus FY2027 revenue ($240.7B house vs. $238.8B Street) and assumes gross margin stays at 72% in FY2027 and 56–62% thereafter — i.e. permanently above the all-time peak annual margin — and it still only produces $1,028. To reach the Street's ~$1,500 average price target, one needs numbers above consensus, or a multiple applied to peak earnings with no mean reversion at all. That is a specific, falsifiable claim about what the market is embedding.

4. CONSENSUS, AND THE REQUIRED BRIDGE

4.A The numbers (Alpha Vantage EARNINGS_ESTIMATES, pulled 2026-07-27)

Two things deserve emphasis. First, the house model reproduces FY2026 consensus almost exactly — that is deliberate. Three of four quarters are already reported, so FY2026 is arithmetic, not forecasting; agreeing there isolates the disagreement to where it actually lives. Second, the FY2027 consensus range is $101.6B to $294.2B — a high nearly 3x the low. The Street does not have a view; it has a distribution that spans "the cycle turns" and "the cycle never turns."

4.B Estimate revisions — violently positive, and this cuts against the house view

FY2027 consensus EPS has risen 53% in ninety days, with thirty analysts revising up and zero down. This is about as one-sided as estimate-revision data gets, and per Chan/Jegadeesh/Lakonishok (1996) revision momentum is a genuine predictor of continued outperformance. Any bearish view on this name is standing in front of that. It is stated here, prominently, rather than buried — because it is the strongest single piece of evidence against this report's own valuation conclusion.

4.C The decomposition: numbers, or multiple?

Street average price target is approximately $1,492–$1,569 (S&P Global 45-analyst consensus rating "Strong Buy," average target $1,492; a 30-analyst set averages $1,569 with a $2,200 high and $1,100 low; a 42-analyst set averages $1,507 with a $361 low) — sourced via web aggregators (stockanalysis.com, public.com, TipRanks). Analyst-count and averaging methodology differ by source; the spread is disclosed rather than collapsed into one false-precise number.

Applying the required decomposition on a shared revenue base:

The gap is not primarily a numbers disagreement about FY2026 — it is a disagreement about (a) FY2027 revenue and margin, where the house is 22%/42% below consensus, and (b) whether peak-cycle earnings deserve a mid-teens multiple at all. The Street's targets require memory to have stopped being cyclical. That is the falsifiable claim, checkable within twelve months.

The honest risk to the house view, stated plainly: if HBM's contracted, sold-out, qualification-gated economics genuinely have decoupled a majority of Micron's profit pool from spot DRAM, then the historical mean-reversion anchor used throughout this report is the wrong prior, and it will be wrong for as long as that regime holds. Sold-out 2026 and 2027 supply and $22B of non-cancelable contracts are real evidence for that case, not hand-waving. This report weights it at 22% (the Bull scenario); a reasonable person could weight it higher. What the two-tranche build adds to that risk statement (§9): the contracted evidence covers 7–26% of revenue depending on scenario and runs out in CY2030. For the decoupling claim to be right, it has to be true of the merchant tranche too — and there is no Tier 1 or Tier 2 evidence for that. The contracts prove the floor; they do not prove the regime.

5. FACTOR & ANOMALY SCORECARD (required, skill v1.3.0)

Inputs: Alpaca SIP daily bars through 2026-07-27; Alpha Vantage EARNINGS_ESTIMATES; SEC EDGAR XBRL FY2025 vs. FY2024. Screen composite from reports/screens/Screen_2026-07-27.json (long rank #11 of 47).

Synthesis. The scorecard is genuinely split, and that split is the finding. Quality and expectations factors argue strongly against a short: an F-score of 8/9 and accruals of −0.12 are the profile of a business converting earnings to cash cleanly, with none of the deterioration signature (high accruals, falling F-score) where shorts empirically work; and 30-up/0-down revisions are the single most reliable short-horizon signal pointing up. Trend factors have begun to argue the other way: −27% from the high, −21.9% in a month, and now below the 50-day. What the scorecard emphatically does not show is a stock whose fundamentals are deteriorating. It shows a stock whose price has started to roll over while its fundamentals and estimates are still improving — which is either the market discounting a cycle turn before it appears in the numbers, or a correction within an ongoing advance. Distinguishing those two is exactly what Task 5's gates must adjudicate, and neither the valuation work nor the factor work resolves it alone.

6. BASE-RATE CHECK ON THE SCENARIOS (required, skill v1.3.0)

Computed from the 16-year EDGAR record in Task 1:

• Consensus FY2027 (+84% revenue growth) would be the fastest growth year in company history, immediately following FY2026's +247%. Consensus is an above-base-rate outcome in both years — not a midpoint, and the memo says so rather than treating consensus as neutral.
• The Bull case deliberately matches consensus and holds margin permanently above the prior all-time peak. That is a top-decile-and-beyond assumption, named explicitly as the regime-change bet.
• The Base case still embeds +43% FY2027 growth (4x the historical median) and a margin normalising to 44–47% (above the 32% median, crediting HBM mix). Its FY2028 decline of −28% is milder than the −49.5% Micron actually printed in FY2023.
• The Bear case (−42% revenue, margin to 38–41% after the v1.4.0 SCA-floor revision, previously 30–35%) is not a stress test — it is the modal historical outcome after a memory peak, now with a contractual floor under it. That is why P(Bear) is set at 30%, still above P(Bull) at 22%.

7. TRANSCRIPT MENTION-FREQUENCY — MANAGEMENT ATTENTION (REQUIRED, v1.4.0)

Corpus: Micron first-party earnings-call prepared remarks, 14 consecutive quarters, FQ2-2023 → FQ3-2026. Alpha Vantage transcripts were unavailable (daily quota exhausted); this substitute corpus is prepared-remarks only, which is the stronger of the two series the methodology asks for, but it means no Q&A series exists and no claim is made about prompted-vs-unprompted. Full derivation and open-question list in Research §9.

The two readings that matter. First, an entirely new contractual vocabulary (SCAs/RPO/floor/ceiling) appeared from absolute zero in a single quarter and immediately dominated the call — the strongest emergence signal in this corpus. Second, and against intuition, HBM has decayed 79% from its FQ1-2025 peak while HBM revenue was compounding; the investigation of that anomaly (Research §9.2(b)) produced the finding that Micron's highest-margin units are now the conventional-DRAM ones (CDBU and MCBU at 87% gross margin, above CMBU's 83%). Mention counts measure attention, not economics — but here the attention shift pointed directly at an economic fact that changes the thesis.

8. BOTTOM-UP TAM AND THE IMPLIED-PENETRATION TEST (v1.4.0)

8.1 The units build — HBM, calendar 2028

Built from units, not from a vendor headline. Micron's own guide ($100B HBM TAM by CY2028, ~40% CAGR from ~$35B in CY2025) is used only as a cross-check on the bottom-up answer, never as an input.

Penetration-path sensitivity (the variable that actually dominates the answer):

The Base path lands within 4% of Micron's own $100B guide from wholly independent inputs, which is the appropriate use of the vendor number. Restated as a falsifiable unit claim: a $100B CY2028 HBM TAM requires roughly ten million AI accelerators a year, each carrying about 500GB of HBM, at about $20/GB — i.e. ~1.4x unit growth and ~2x content growth and a ~35% decline in $/GB from the CY2026 spike, simultaneously. Note the internal tension: the bull case needs the volume, and the same forecast needs HBM pricing to fall by a third.

Time-to-revenue: CY2028 maps to Micron's FY2028–FY2029. This is a 2–3 year-out pool, not a current one.

What falsifies it: HBM4 attach rates on the CY2027 accelerator generation coming in below ~400GB/unit; or CXMT reaching competitive-yield HBM3 volume before 2028 (currently assessed as a 2028+ prospect).

8.2 Does consensus already embed it? — yes, and more

This is the step that decides whether the TAM work is edge or context, and the answer here is unambiguous.

Consensus FY2027 revenue is $238.8B (41 analysts). Micron's entire HBM revenue on the Base path is ~$22B in CY2028 — barely 9% of what the Street already expects Micron to earn in FY2027. Consensus is therefore not primarily underwriting HBM at all; it is underwriting conventional DRAM price levels. That single reconciliation is the most useful output of the TAM exercise, and it disqualifies any HBM-based duration variant from carrying a long: the Street is not under-modelling HBM, it is fully modelling something else. Per references/tam-sizing.md, the TAM work here is context, not edge, and this memo says so.

8.3 Required implied-penetration statement

At $900.20 (2026-07-27 close) and 1,129.4M diluted shares, market capitalisation is $1,016.7B; net cash of $24.4B gives an enterprise value of $992.3B.

At a 12x through-cycle multiple, today's price requires roughly $84.7B of sustainable annual net income — 25x Micron's FY2010–FY2025 average annual net income of $3.4B, and 1.5x its best year ever ($14.1B, FY2018).

Reversed into physical operating terms, which is the stricter form the methodology requires:

Holding revenue at today's annualised run-rate of $165.8B, the price requires a permanent gross margin of ~64% — 32 points above the 16-year median of 32%, above the all-time annual peak of 58.9%, and slightly above the SCA contractual price floor itself. In other words, the market is pricing the contractual floor margin as the through-cycle mean, across all revenue, forever. Only if revenue also grows to consensus FY2027's $238.8B does the required permanent gross margin fall to ~46% — still 14 points above the historical median.

And the sharpest version, now verified by the explicit two-tranche build (Two-Tranche tab, §7 of that tab) rather than asserted:

The $100B RPO is $20.0B a year over the five-year SCA term. At a 13x through-cycle multiple and a 48% post-contract merchant margin, $900.20 requires $220.8B of permanent annual revenue. The entire contracted book is therefore 9.1% of what the price requires — one part in eleven. On gross profit the arithmetic is the same: the contracted tranche throws off $12.2B of gross profit a year at its floor, against the $106.0B the price needs — one part in 8.7. Against the current annualised run-rate of $165.8B it is 12.1%, which is the "about one-eighth" framing carried in the v1.4.0 memo.

Verdict on the prior estimate: it holds, and if anything was slightly generous. The three framings bracket the answer at one-eighth to one-eleventh depending on the denominator, and the strictest one — against the revenue the price requires, which is the relevant test — is one-eleventh. Every one of these figures is a live formula in the workbook (Two-Tranche rows 79–84), not a back-of-envelope.

Secondary sanity check (EV/TAM), with its elasticity stated: EV of $992.3B against a CY2028 HBM TAM of $100B is 9.9x — a meaningless-looking figure precisely because HBM is not Micron's market. Against a total DRAM+NAND industry revenue pool of roughly $130–160B in CY2026, EV/TAM is ~6–7x, i.e. Micron's enterprise value is six to seven times the entire annual revenue of the industry it competes in, from which it takes ~20–25% share. That denominator is itself elastic and vendor-sourced; it is reported here as a cross-check only, never as the headline, and never compared across sectors.

9. VALUATION SUMMARY (rebuilt under v1.5.0 — explicit two-tranche model)

9.1 What replaced what

v1.4.0 handled the SCA disclosure by bumping a single blended gross margin (Bear FY2028–31 34/30/33/35% → 40/38/40/41%; Base 47/44/46/46% → 50/48/49/49%) and lifting the normalised anchor 48% → 52%. That blurred exactly the variable that matters. v1.5.0 replaces it with a duration-matched two-tranche model on a new Two-Tranche tab, per references/regime-change-test.md:

Why straight-line and not a ramp: the disclosure gives a cumulative figure over the term with no annual schedule. Straight-line is the only allocation the Tier 1 evidence supports; a ramp would layer an unevidenced assumption on top of a contractual fact. A back-loaded ramp is shown at Two-Tranche row 100 for sensitivity — note that a ramp defers revenue and therefore lowers PV, so straight-line is the less conservative of the two and the modelled case is not flattering the contracted tranche.

Why FY2031 is zero: contracts end CY2030, so FY2031 strictly contains a 4-month CY2030 stub worth ~$6.7B. It is excluded, so FY2031 and the terminal value are 100% merchant. Duration matching: model the contracted view for exactly as long as the evidence reaches, not one year longer.

9.2 The tranche split, by year

Contracted share of revenue, all three scenarios (Exhibit 32):

The share does not decline monotonically inside the contract term — it rises, because the contracted tranche is fixed while the merchant tranche collapses. That is the floor doing precisely what a floor does, and it is why the Bear reaches 25.8% while the Bull never exceeds 10.2%. The decline that matters is the cliff to zero at expiry, and that cliff — not a smooth taper — is the honest expression of where the evidence stops.

9.3 Principle 5 — does the contracted tranche raise the FLOOR more than the CEILING?

Yes, decisively. The signature is produced by the model's structure, not asserted. Because contracted GM is MAX(floor, MIN(merchant, ceiling)), the floor only lifts the tranche when the merchant market is below it:

The floor rises 13x more than the ceiling. The Bull gets a token $0.65 — and only because merchant GM drifts to 58/57/56% in FY2029–31, marginally below the 61% floor; in FY2028 the Bull uplift is exactly zero. The regime-change reading survives the Principle 5 test intact. Had the contracted tranche lifted the Bull as much as the Bear, the argument would have been about demand wearing an evidence costume; it is not.

9.4 Where the contractual floor actually earns its keep — the deep tail

The modal Bear is not where the SCAs matter. Run the FY2023 analogue — merchant gross margin at the actual −9.1% trough — against Bear FY2028 revenue of $79.2B:

A negative-gross-margin year is contractually impossible while the SCAs run. That — not the 5% lift to the modal Bear — is the real Tier 1 effect, and it is why P(Bear) stays at 30% rather than reverting to the pre-SCA 35%. The contracts do not make the Bear less bad; they make the catastrophe unavailable.

9.5 Revised valuation

9.6 The honest attribution — and a correction to v1.4.0

The target falls from ~$510 back to ~$474, and the gap widens from −43% to −47%. This is not new bearishness; the contractual evidence is exactly as strong as it was. It is a correction of a duration error:

• Inside the contract term, v1.4.0 was well calibrated. The tranche build derives Bear FY2028–30 blended margins of 40.1 / 38.0 / 39.7% against v1.4.0's hand-set 40 / 38 / 40%, and Base of 49.1 / 46.7 / 48.1% against 50 / 48 / 49%. The judgement calls on the contracted years were essentially right.
• Outside it, v1.4.0 leaked. It also lifted FY2031 (Bear 41% vs. the merchant 35%; Base 49% vs. 46%) and the normalised anchor (52% vs. 48%) — both entirely post-contract periods. Because the DCF terminal is 9x FY2031 EBIT and the primary anchor carries 50% weight, those two leaks are the whole of the $36 difference.
• Net: the SCAs are worth ~$1.7/share on the weighted target (the $4.89 probability-weighted DCF uplift at 35% weight), not the ~$45 that v1.4.0 attributed to them. The contractual floor is real, dated, quantified and verifiable — and, against a $900.20 stock, economically small.

Rating: unchanged in direction, better understood in structure. The equity remains substantially overvalued against every method that assumes memory stays cyclical, and fairly valued only under an explicit regime-change assumption. The contractual floor truncates the catastrophic left tail without meaningfully touching the modal case or the right tail. Per the skill's hard rule, that is not by itself a directional call; Task 5 decides.

---

Feeds Task 5 (Trade Construction). Model: Micron_Financial_Model_2026-07-27.xlsx, Two-Tranche tab. All values read back from live Excel via AppleScript after recalculation, 2026-07-27, and cross-checked against an independent hand calculation. Chart: 32_contracted_share_by_year.png.