Phase Space AI

Micron Valuation Analysis 2026-07-29

Micron Technology [MU]

Micron Technology, Inc. [MU] — Valuation Analysis (re-underwrite)

Investing Hub Research | 2026-07-29 | framework v1.5.1 | supersedes archive/Micron_Valuation_Analysis_2026-07-27.md

Price $739.00 (Alpaca SIP, 2026-07-29 close). Prior document used $900.20 (07-27) and its migration note used $820.51 (07-28). MU has fallen 18.0% in two sessions; that alone moves the valuation conclusion materially and is the single largest driver of what changed.

---

0. WHAT CHANGED FROM THE PRIOR DOCUMENT, AND WHY

The prior memo was not stale on the headline operating data — a premise worth correcting explicitly, because the re-underwrite was commissioned on the assumption that it was. It already carried the $41.5bn quarter, the 84.6% gross margin, TTM revenue of ~$90bn, and the SCA duration-matching argument. What was wrong was narrower and mostly balance-sheet and price:

Retained unchanged: the cyclical framing, the SCA duration-matching argument (it is correct and it is the crux), the price-to-book observation, and the base-rate discipline.

---

1. THE UNPRECEDENTED FIGURE — VERIFIED, ADOPTED, AND STATED AS SUCH

MU's quarter ended 2026-05-28 (52/53-week filer; Alpha Vantage normalises this to 2026-05-31) printed:

This exceeds MU's prior best quarterly gross margin of 61.03% by 23.5pp, and its best annual gross margin (FY2018, 58.9%) by 25.7pp. It was initially flagged as a data error on precisely that basis. That flag was wrong.

Verification against a source Alpha Vantage did not produce: av_vs_edgar.py compared 6 quarters of AV normalized statements against EDGAR primary filings and returned disagree=0/6, with 52/53-week fiscal-calendar tolerance applied (max offset 3 days). EDGAR's own RevenueFromContractWithCustomerExcludingAssessedTax for 9M FY2026 is $78,959m, and GrossProfit $60,457m — a 76.57% nine-month gross margin. The figure holds. It is adopted.

Per the brief's operative distinction: this is historical implausibility, not arithmetic impossibility, and historical implausibility is never grounds for rejection. A sanity ceiling calibrated on MU's prior cycles would have rejected the largest earnings event in the sector's history as bad data.

The one figure that IS wrong — and it is a 2.94x error

AV's ebitda understates by $23,546m — a factor of 2.94x — and comes out below ebit, which is arithmetically impossible with positive D&A. D&A is taken from the cash-flow statement throughout this document. This is a genuine arithmetic violation and is rejected on that basis.

Also checked, per the coordinator's mid-run findings: - AV ebit vs operatingIncome — on MU the gap is 0.00pp in the latest quarter and never exceeds 0.72pp across the last five. Operating margin is nonetheless computed as operatingIncome / totalRevenue throughout, never from ebit. (On CRDO the same gap is 2.78pp — see that memo.) - AV sellingGeneralAndAdministrative — matches EDGAR exactly on MU: FY2025 $1,205m vs EDGAR $1,205m; 9M FY2026 $1,088m vs EDGAR $1,088m. R&D likewise ($3,798m / $3,737m). The AMAT SG&A understatement does not affect this name, so the opex bridge in §4 rests on verified data. - Splits — MU has had no split since 2015. All per-share figures are on one basis.

---

2. NET CASH — REBUILT, AND THE PRIOR FIGURE WAS $8.5bn TOO HIGH

EV in this document is LEASE-INCLUSIVE (finance and operating lease liabilities treated as debt).

All figures from EDGAR primary at 2026-05-28:

Management's own "record levels of cash and investments of $30.2 billion" ties to the $30,128m above.

The deposit adjustment — the largest defect found on this name

The CFO, in Q&A on the FQ3 call:

"we have $22 billion of deposits and financial commitments associated with the agreements signed to date… approximately $18 billion of that is cash deposits… They are held by us during the performance commitments of the agreements and, as those agreements are satisfied, those deposits will be returned over time, heavily weighted to the back half of the agreements."

"the deposits are unrestricted cash." — and "The cash flows associated with customer deposits appear in financing-related cash flows and will not affect our free cash flow."

So the deposits (a) sit inside reported cash, (b) are refundable liabilities, and (c) are largely still to arrive ("will show up on our balance sheet more in fiscal Q4").

The amount received to date is not separately tagged. CustomerAdvancesAndDeposits is stale (2012) and ContractWithCustomerLiability is only $422m. The deposits are landing in other liabilities, which moved as follows:

I deduct the full $7,783m as the deposit estimate. This is an upper bound on the deposit component (some of the increase is ordinary accrual growth on a 3.5x larger revenue base), so net cash of +$15.9bn is the conservative end of a +$15.9bn to +$23.7bn range. The prior memo's +$24.4bn deducted no deposit liability at all and therefore overstated net cash by ~$8.5bn ($7.39/share).

Forward-looking note: further deposits are net-neutral to net cash — they add cash and liability together. Only deposits already inside the cash balance require adjustment. But note that ~$18bn of gross deposits flowing back out in CY2029–30 is a real call on liquidity in the exact years the SCAs expire.

---

3. COMPANY STATE, AND THE STRUCTURAL-VS-CYCLICAL DECISION

3.1 Declared state: STATE B — cyclical, retained. With a qualifier that changes the parameters, not the letter.

I am not reclassifying MU to State A, and I am not mean-reverting it to history either. Both would be reflexes. The evidence supports a specific, bounded structural change.

Why State B is retained. Three tests, all failed for State A: 1. ~60–75% of revenue remains merchant/spot-priced. The SCAs cover ~20% of DRAM volume and ~30% of NAND volume — ~25% of revenue on management's own arithmetic ("about 25% of our revenue that you can project over the term of these agreements"), rising to ~40% of revenue under fixed or ceiling-protected pricing when all planned SCAs execute. The majority of the book is still exogenously priced. 2. The contracts have a hard end date. "Typically, these agreements have a five-year term from calendar 2026 through the end of calendar 2030." A 5-year DCF underwritten from FY2027 terminates in FY2031 — at or just past SCA expiry. The terminal period is post-contract by construction. 3. Supply is arriving, and a cycle ends when supply arrives. MU capex goes ~$27bn (FY2026, net of government incentives) → >$40bn (FY2027), with a multiyear EUV supply agreement signed with ASML and new HBM packaging capacity contributing from 1H CY2027. See §6.

3.2 The qualifier: HBM has changed the instrument for part of the book, for a defined period

This is the position the memo is required to take, and it is not "cyclical as before."

What is genuinely, disclosably different from every prior MU cycle:

Mention-frequency evidence for the shift (prepared remarks / Q&A, 10 quarters):

The language has migrated from scarcity to contract. "Sold out" — the classic spot-shortage framing — has gone to zero in the last two quarters, while "agreement"/"committed"/"supply agreement" rise to their 10-quarter highs exactly as the SCAs are signed. HBM mentions are falling (6/12 vs a 27/28 peak) not because HBM matters less but because the narrative has moved up a level to the contracting structure. Note also that in every quarter the Q&A count exceeds the prepared count on hbm — analysts pull it out rather than management pushing it, which is the stronger form of evidence.

3.3 CONCLUSION — stated plainly

HBM and the SCAs have changed the instrument for ~25% of revenue (heading to ~40% with price protection) over CY2026–CY2030, and have raised the whole book's mid-cycle margin permanently. They have NOT converted Micron into a non-cyclical business, and they expire precisely at the terminal date.

Therefore: normalise, but normalise to a mid-cycle that is far above history — not to history. My terminal EBIT margin of 38.0% sits 26.7pp above MU's 20-year median operating margin of 11.3%, and 11.3pp below its best year ever (FY2018, 49.3%). It is 27.8pp below the TTM actual of 65.8%.

3.4 What would FALSIFY this conclusion

Explicitly, in both directions:

Falsifies the structural half (→ I am too generous, terminal margin should fall toward 25–30%): 1. SCA revenue realising at or near the floor rather than above it. Management says "we expect revenue to well exceed associated RPOs." The 10-Q now discloses next-12-month revenue against each RPO cohort — the FQ3 cohort shows RPO >$5bn against NTM revenue of ~$1.8bn. Track realised revenue vs. that NTM disclosure each quarter. Convergence toward the floor is the tell. 2. Non-renewal. Any indication in CY2029–30 that SCAs are not being extended on comparable terms. 3. Supply arriving faster than the trade ratio absorbs it. If industry bit supply growth outruns demand and contracted customers begin negotiating relief, the floor's economic value is illusory even if legally enforceable. 4. A ceiling that bites. The ceilings are set "at the current CQ2 market price." If spot runs far above CQ2 levels, MU has capped its own upside on ~40% of revenue — the floor's mirror image.

Falsifies the cyclical half (→ I am too harsh, terminal margin should be 50%+): 1. SCAs extended beyond CY2030, or renewed at higher floors, converting a 5-year contract into a rolling structure. This is the single most important thing to watch. 2. HBM disclosed as a separate reportable segment with sustained >60% gross margin through a period of falling commodity DRAM price — i.e. demonstrated decoupling. 3. Commodity DRAM price troughing at a level above the prior cycle's peak, which would confirm the trade-ratio supply constraint is structural rather than transient.

---

4. TERMINAL MARGIN — THE CRUX PARAMETER

4.1 What I am terminalising, and why

I terminalise the FY2031 (post-SCA) mid-cycle EBIT margin at 38.0%. Not the trailing 65.8%, not the current-quarter 80.4%, not the historical median 11.3%.

Demonstrated record (all operatingIncome / totalRevenue, AV normalized, verified vs EDGAR):

4.2 The opex bridge — m_EBIT,T = m_gross,T − R&D − SG&A − other

Hard constraint satisfied: 38.0% ≤ 46.0%. ✓

Bridge validated against actuals — the same construction reproduces reported margins: - FY2025: 39.79% − 10.16% − 3.22% = 26.41% vs actual 26.14% (residual 0.27pp) - 9M FY2026: 76.57% − 4.73% − 1.38% = 70.46% vs actual 70.40% (residual 0.06pp)

Total terminal opex of 8.0% of revenue = ~$9.6bn on a $120bn base. FY2025 actual R&D+SG&A was $5.0bn; FQ4 FY2026 guided opex is $1.65bn (~$6.6bn annualised). So I am underwriting a ~45% absolute increase in opex dollars over the current run-rate. This is the conservative direction.

4.3 The terminal margin is BELOW the trailing actual. Why that is not the error the framework warns about.

The brief is explicit: "If your terminal margin is below the company's own trailing actual, that is almost certainly an error — say why it is not." Here is why it is not, as an explicit causal bridge (valuation.md rule 2):

1. The trailing 65.8% is earned inside an acute supply shortage whose relief is already funded and dated — MU's own capex more than doubles to >$40bn, and its own new HBM packaging capacity lands 1H CY2027.
2. The contractual support for that margin expires at the terminal date. SCAs run CY2026–CY2030; FY2031 is post-contract. Carrying a contracted margin into a post-contract year is the exact duration mismatch the prior memo correctly identified.
3. Management itself guides down from the peak. The CFO on FQ4: "Our fiscal Q4 gross margin outlook reflects a meaningful moderation in the rate of price increases," and "We are at margin levels where incremental price yields less gross margin expansion."
4. The direction of the error in this project's record is the opposite of the usual one. The stored terminal margin was 26.1% — FY2025's actual — which is 39.7pp below TTM. I am moving it up by 11.9pp, not down. The defect being corrected is under-, not over-, statement.

This is not a haircut for uncertainty. Per valuation.md rule 5, uncertainty is expressed in position size and evidence_grade, never in the operating assumption.

---

5. THE IMPLIED-PATH TEST — THE VALUATION CRITERIA

assets/reverse_dcf.py. Terminal value exceeds 60% of EV, so the reverse DCF is the mandatory primary long-horizon instrument.

Fixed inputs, all named: spot $739.00 · diluted shares 1,150m (FQ4 guidance basis; FQ3 weighted diluted 1,145m) · net cash +$15,899m · TTM revenue $90,274m · horizon 5 years · WACC 13.0% · terminal EBIT margin 38.0% · exit multiple 7.5x EBIT · tax 15%.

EV implied by today's price: $833,950m (9.2x TTM revenue).

5.1 The exit multiple is derived, not asserted

EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g):

Base 7.5x, the midpoint of the warranted range. The prior memo's 9.0x was asserted and sits above every point in this range. Note the low absolute level is a consequence of MU being capital-intensive (capex ~75–90% of CFO), which caps terminal ROIC and therefore the warranted multiple.

5.2 The result

VALUATION CRITERIA: FAIL, −10.4pp.

5.3 The most informative framing of the three

The CAGR framing is the framework's ranking metric, but it is the least informative here: a 43.0% CAGR from $90.27bn implies ~$538bn of FY2031 revenue, which would require the combined DRAM+NAND TAM to exceed $1tn. That path is not merely unlikely, it is close to arithmetically unavailable — which tells you the exit multiple and terminal margin are doing the real work.

The framing that actually decides the name:

At its own demonstrated 32.6% revenue CAGR and a derived 7.5x exit multiple, today's price of $739 requires a PERMANENT operating margin of 55.4%.

Judge that number: - It is above MU's best year ever (FY2018: 49.3%) by 6.1pp. - It is below what MU is earning right now (TTM 65.8%; current quarter 80.4%; FQ4 guide 82.7%). - It is 17.4pp above my underwritten terminal margin of 38.0%.

So the price is a bet that roughly two-thirds of the current margin regime is permanent. Not that it fully persists; not that it mean-reverts to history. That is a genuinely arguable bet — it is not the absurdity that a reflexive mean-reversion would make it look like, and it is not the free option that peak-extrapolation would make it look like. On my evidence it is modestly too optimistic, and the honest gap is −10.4pp on CAGR / −17.4pp on margin, not the prior document's −47.4%.

5.4 Sensitivity over the exit multiple (the highest-variance parameter)

Terminal margin 38.0%, TTM base:

And over terminal margin at 7.5x:

The break-even sits at a ~55% permanent EBIT margin, which is §5.3 restated.

---

6. CAPEX AND THE SUPPLY RESPONSE — stated for the sibling STX/WDC and KLAC/LRCX/AMAT read-across

Directly from the FQ3 FY2026 call. This is the input the WFE agents requested.

Read-across, stated plainly: MU capex goes from ~$27bn (FY2026, net) to >$40bn (FY2027), but more than half the increment is construction/cleanroom, not equipment. A WFE model that treats MU's capex step-up as fully addressable equipment spend will overstate the WFE opportunity. The equipment-relevant signal is the ASML multiyear EUV agreement at 1-delta and the HBM packaging (advanced packaging / TSV, hybrid bonding) capacity landing 1H CY2027.

For STX/WDC: MU NAND revenue was a record $9.9bn (24% of total), up 361% YoY, and ~30% of NAND volume is now committed under take-or-pay SCAs. The contracting structure is spreading beyond DRAM into NAND — that is the transferable finding.

---

7. SEGMENT AND HBM MIX — WHAT IS AND IS NOT DISCLOSED

HBM revenue as a share of total is NOT separately disclosed. This is a genuine gap and I will not manufacture the number. What is disclosed: - "we have already shipped over $1 billion in HBM4 revenue" (HBM4 specifically, cumulative). - Strategy: "we are choosing HBM share to be close to our DRAM share" — a deliberate cap, not a maximisation. Rationale given: "HBM consumes a significant amount of wafers and puts pressure on non-HBM supply," so MU is restraining HBM to serve diversified end markets. - Cloud Memory BU at $13.8bn (33% of revenue) is the closest disclosed proxy for the HBM-plus-high-capacity-DRAM complex, but it is not an HBM number.

Consequence for valuation: HBM cannot be carved out and multiple-differentiated on disclosed data. Attempting it would require an undisclosed split. I therefore value the consolidated entity and express the HBM-vs-commodity distinction through the terminal margin (§4) rather than through segment-specific multiples. Stated as a limitation.

Customer concentration

ConcentrationRiskPercentage1 is not tagged in MU's XBRL, and the FQ3 10-Q major-customer note was not retrieved within the time-box. This is an open item, not a finding. It matters: 16 SCAs with a handful of accelerator makers implies high concentration, and the FY2026 10-K (due ~October 2026) will carry the major-customers note. Flagged as INDETERMINATE.

---

8. TWELVE-MONTH TARGET

8.1 The multiple anchor is partially UNIDENTIFIED — and I declare it rather than substitute

MU's own P/S history, computed monthly from Alpaca SIP closes against TTM revenue, n=127 months from 2016-01:

This percentile is not usable as a valuation anchor, and I decline to use it. MU's historical P/S was earned at a 20–59% gross margin; the business now runs 72–85%. A sales multiple is not comparable across a doubling of gross margin — that is precisely the regime change valuation.md says must be declared rather than papered over. Applying the 2.77x median to FY2027 consensus revenue would give $575 (−22%) and would be an artifact of the regime break, not a valuation.

I therefore anchor on forward P/E, which is margin-neutral, and I flag the anchor as evidence grade C.

8.2 The target

Consensus (AV EARNINGS_ESTIMATES, revision history built in):

FY2026E consensus revenue of $129.78bn corroborates my independently built $129.0bn (9M actual $78,959m + $50bn FQ4 guide) to within 0.6%. Revisions are unanimously upward on both years.

MU trades at $739 / $153.74 = 4.81x FY2027E EPS.

12-MONTH TARGET: $922 — +24.8% ABOVE SPOT.

Basis: 6.0x FY2027E consensus EPS. MU has historically traded 4–8x forward EPS at cycle peaks (the low multiple is the peak signature — peak EPS in the denominator); 6.0x is the midpoint of that observed band. Evidence grade C on the multiple — the band is read from MU's cycle history rather than computed as a percentile of a clean forward-P/E series, because the P/S series is regime-broken and a forward-P/E series was not constructed within the time-box.

Named events inside 12 months that move estimates (detail in the catalyst calendar): 1. FQ4 FY2026 results (~late Sept 2026) — first quarter with RPO on 14 of 16 SCAs (~$100bn) disclosed, plus the next-12-month revenue against each RPO cohort. This is the falsification test in §3.4 becoming observable. 2. The ~$18bn of cash deposits landing on the balance sheet ("more in fiscal Q4"). 3. FY2027 capex guidance formalised (>$40bn) — read directly by the WFE names. 4. 1H CY2027 HBM packaging capacity starting to contribute. 5. FY2026 10-K (~Oct 2026) — major-customers note, closing the §7 concentration gap.

Divergence note: the prior document's $1,250 target was set against $820.51 (+52.3%). My $922 is 26.2% lower in absolute terms. The difference is almost entirely the exit/anchor discipline: $1,250 implied ~8.1x FY2027E EPS, the top of the band; I take the midpoint.

---

9. DOWNSIDE CASE WITH A NAMED CAUSE (MEASURED — does not reject the name)

Named cause: the CY2027 supply wave arrives into a decelerating AI-capex cycle, and the SCA ceilings — not the floors — turn out to be the binding constraint.

Mechanism, in order: 1. MU's own >$40bn FY2027 capex plus industry additions land from 1H CY2027. 2. Commodity DRAM/NAND spot prices break first, because ~60–75% of the book is unprotected. 3. The SCA ceilings ("at the current CQ2 market price") cap MU's participation on ~40% of revenue in the good state while the floors only protect it in the bad state — an asymmetry that is valuable in a downturn and costly in a melt-up. 4. Reported margin compresses toward the ~46% terminal gross margin faster than FY2029.

Quantification — hold the demonstrated 32.6% CAGR, cut the terminal margin to the FY2018 peak analogue of 25.0% and the exit multiple to 5.0x: the required CAGR to justify $739 rises to well above 55%, and the equity is worth materially less than half spot. Using the prior document's normalised-earnings anchor as the permanent-impairment marker (a 48% gross margin on $130bn of permanent revenue at 13x) gives ~$474/share, −35.9% vs $739 — note this is a smaller drawdown than the −47.4% it represented at $900.20, purely because the stock has already fallen 18%.

Type: MEASURED. Logged and scored; it does not reject the name.

---

10. CRITERIA SUMMARY

Evidence grade: B. Mechanism and accounting are grade A (contractual, EDGAR-verified). The terminal margin is grade B (bridged from verified actuals, but the post-2030 merchant margin is a judgement). The 12-month multiple anchor is grade C (regime-broken own-history series).

No position verdict is issued. The book decides. Per valuation.md rule 5, the grade-C anchor and the INDETERMINATE concentration should reduce position size, not the operating assumptions above.

---

Prepared 2026-07-29. Data: Alpha Vantage normalized statements (81 quarterly / 20 annual periods) cross-verified against SEC EDGAR companyfacts (CIK 0000723125); Alpaca SIP prices; MU FQ3 FY2026 earnings call transcript with speaker/title tags. Publication out of scope this run.