Micron Technology, Inc. (MU) — Coverage File

Data as of: Q3 FY2026, quarter ended May 28, 2026 (last reported period) · market data 4–7 Sep 2026 File last updated: 7 September 2026

---

0. 30-second reopen

• Business: A scale producer of DRAM and NAND; sells bits it manufactures itself, priced off industry supply/demand, into cloud, data centre, mobile/client and auto/embedded.
• Debate: Is the current 85% gross margin a cycle peak that mean-reverts, or has memory become a contracted, supply-constrained AI input with structurally higher through-cycle economics?
• Base thesis: AI demand has outrun industry supply, and multi-year take-or-pay contracts with price floors convert some of that scarcity into durable margin — an economic model different from every prior Micron cycle.
• Bear case: 85% margins on a commodity are self-destroying; they fund Chinese and incumbent capacity arriving 2027–2030 and invite demand destruction and design-around, while $27bn/yr of capex is committed at the peak.
• Valuation: ~$1.15tn market cap, 22.9x trailing earnings but ~7x forward — the market is explicitly pricing a large earnings decline.
• What to watch: DRAM ASP direction q/q, bit-shipment growth, cash deposits actually received against the $22bn committed, and CXMT/YMTC supply news.

---

1. Business snapshot

Micron manufactures and sells semiconductor memory: DRAM (~76% of Q3 FY2026 revenue, $31.3bn), NAND (~24%, $9.9bn) and a small NOR business [Q3 FY2026 10-Q]. It owns its fabs, so revenue is bits shipped × average selling price, and profit is the spread between that price and a per-bit cost it lowers each node generation.

Four business units, all reported since FY2024: CMBU (hyperscale cloud, and HBM for all data-centre customers), CDBU (mid-tier cloud, enterprise, OEM data centre, plus all data-centre storage), MCBU (mobile and client), AEBU (automotive, industrial, consumer). Q3 FY2026 revenue split: CMBU $13.8bn, CDBU $11.5bn, MCBU $11.5bn, AEBU $4.6bn [Q3 FY2026 10-Q].

Manufacturing is concentrated in centres of excellence in the US, Japan, Taiwan and Singapore, with back-end assembly and test in Asia. Competitors are Samsung, SK hynix, Kioxia, Sandisk, and the Chinese state-backed CXMT and YMTC [FY2025 10-K].

2. Core debate

Every prior Micron cycle ended the same way: high margins funded capacity, capacity arrived, prices collapsed. FY2018 peaked at 59% gross margin and $14.1bn net income; FY2023 produced a negative 9% gross margin and a $5.8bn loss [FY2018/FY2023 10-Ks]. The debate is whether FY2026 is that pattern again at a larger amplitude, or a genuine regime change.

The regime-change argument rests on something new in the filings: multi-year take-or-pay strategic customer agreements with binding volume commitments and, for most, fixed pricing or floor/ceiling bands. Management states the largest agreements carry a ceiling near the calendar-Q2-2026 market price and a floor through the term, and that margins at floor pricing would still exceed Micron's best quarter in any past cycle [Q3 FY2026 10-Q]. If true and durable, the business stops being spot-priced and becomes a contracted supplier.

The counter is that contracts do not create supply discipline; they only redistribute who bears the price risk, and only while counterparties can perform.

3. Base thesis

AI infrastructure has shifted memory from a cost line to a bottleneck. HBM consumes materially more wafers and cleanroom space per bit than conventional DRAM [FY2025 10-K], so every HBM bit sold removes several conventional bits from supply — an internal supply constraint the industry has never had before. Industry capacity additions are physically slow: Micron's own new fabs deliver first wafers mid-CY2027 (Idaho 1), late CY2028 (Idaho 2) and 2030+ (New York) [Q3 FY2026 10-Q]. That leaves a multi-year window in which demand growth cannot be met.

The results are extreme. Q3 FY2026 revenue was $41.5bn, up 346% year-on-year, on DRAM ASPs up ~260% and NAND ASPs up ~310%, with bit shipments up only ~20% and low-double-digits respectively [Q3 FY2026 10-Q]. Cost of goods sold barely moved — $6.4bn versus $5.8bn — so gross margin went from 38% to 85% and net income from $1.9bn to $28.2bn.

Assumptions the thesis depends on:

1. HBM/AI data-centre demand keeps growing faster than industry bit supply through at least CY2027, before the new fabs land.
2. The strategic customer agreements are enforceable and honoured, and their floors hold — management's claim that floor pricing still yields above-peak-cycle margins is the single most load-bearing statement in the file [see §9, claims to verify].
3. Chinese suppliers (CXMT, YMTC) do not close the technology gap in leading-edge DRAM/HBM fast enough to add meaningful competitive supply before 2028.
4. Micron converts the cash into a permanently stronger position — cash and investments are already $30.1bn against $5.7bn total debt [Q3 FY2026 10-Q].
5. Demand is inelastic enough that customers absorb 3x prices rather than re-architecting to use less memory.

4. Main bear case

An 85% gross margin on a manufactured commodity is not an equilibrium; it is an invitation. It hands every competitor — including two Chinese firms with state funding and no cost-of-capital discipline — the strongest possible incentive to build. The industry is now committing capital at the top: Micron alone plans ~$27bn of net capex in FY2026 versus $15.9bn in FY2025 [Q3 FY2026 10-Q; FY2025 10-K], and it is the smallest of the three DRAM players. That capacity arrives 2027–2030 into demand that is currently being estimated from a period of acute shortage.

Demand destruction is already visible: MCBU bit shipments fell both year-on-year and quarter-on-quarter even as its revenue tripled on price [Q3 FY2026 10-Q]. Mobile and client customers are buying fewer bits, not more. If hyperscalers respond similarly — compression, tiering, alternative architectures — the bit growth underpinning the thesis does not materialise.

How it breaks: ASPs roll over first (a single quarter of negative DRAM ASP is the tell), contract floors become the realised price rather than a backstop, then customers slow-walk take-or-pay volumes or renegotiate. Margins compress toward the floor level while depreciation from the FY2026–FY2028 capex wave lands in cost of goods sold. Micron has run this film: FY2022's $30.8bn revenue and 45% margin became FY2023's $15.5bn and −9% in four quarters.

5. Business quality

Structurally better than its history, but not a high-quality business by nature. (Interpretation:) leading-edge DRAM is a consolidated oligopoly with severe capital and know-how barriers, and Micron is one of a small number of firms carrying both DRAM and NAND. (Fact:) it began shipping its first EUV DRAM node (1γ) in 2025, had HBM3E 12-high as the majority of HBM shipments by Q4 FY2025, and has HBM4 on schedule for calendar-2026 volume production [FY2025 10-K]; named competitors are Samsung, SK hynix, Kioxia, Sandisk, CXMT and YMTC.

Against that: Micron takes prices rather than setting them, and returns are dictated by aggregate industry supply behaviour, only a fraction of which it controls. Capital intensity is punishing — ~$27bn of planned FY2026 capex against $37.4bn of FY2025 revenue — and the assets are node-specific and depreciate whether or not prices hold. Customer concentration is rising with the AI mix: one customer was 17% of FY2025 revenue [FY2025 10-K].

6. Key value drivers

DRAM ASP is the dominant variable — a 260% ASP increase produced essentially all of the year-on-year profit change, with bits contributing ~20%. Bit supply growth versus industry demand sets that ASP. HBM share determines mix and how much conventional supply is absorbed. Per-bit cost reduction through node transitions (1γ, G9 NAND) is the one lever fully under management's control. Contract coverage — the share of volume under take-or-pay floors — is the new driver, and the one that could change the shape of the cycle. Net capex determines how much cash flow reaches shareholders.

7. Financial profile

Ten-year revenue and gross margin show the amplitude of the business [FY2016–FY2025 10-Ks; Q3 FY2026 10-Q]:

Balance sheet as of 28 May 2026: cash and investments $30.13bn, total debt $5.72bn (including $2.67bn finance leases), equity $100.7bn, total assets $134.1bn. Micron prepaid $8.5bn of principal across eleven instruments in the first nine months of FY2026 [Q3 FY2026 10-Q].

Two earnings-quality items to watch: receivables rose to $31.0bn from $9.3bn at FY2025 year-end — a normal consequence of the ramp, but a large credit concentration — and "consideration payable to customers, including pricing adjustments and returns" rose to $3.32bn from $1.19bn [Q3 FY2026 10-Q]. The effective tax rate stepped up to 15.0% from 11.1% as Singapore's Pillar Two implementation largely offset Micron's Singapore incentives — likely permanent.

8. Valuation

Company-side facts: nine-month FY2026 net income of $47.3bn on $79.0bn of revenue, a net cash balance sheet, and ~$27bn of planned FY2026 net capex. Nine-month operating cash flow was $45.7bn against $19.6bn of capex, partly offset by $2.99bn of government incentives received.

The valuation question is not what Micron earns now but what it earns at floor pricing, which the filings do not quantify — they state only that floor-price margins exceed prior-cycle peaks [Q3 FY2026 10-Q]. FY2018's peak annual gross margin was 59%. Taken at face value, the claim implies a trough margin above any historical peak. That is the crux of the case, and it cannot be verified from outside.

Market-data callout — as of 4–7 September 2026 (external, not from the sources folder) Share price $1,016.59 · market cap ~$1.15tn · enterprise value ~$1.12tn · trailing P/E 22.9x · forward P/E ~7.1x · EV/EBITDA 16.5x · P/B 11.4x · trailing FCF yield 2.3% · TTM revenue $90.3bn, TTM net income $50.5bn. Consensus FY2026 revenue $129.7bn and EPS $73.40; average price target $1,513 with a range of $361 to $2,200 across 48 analysts. Source: stockanalysis.com, retrieved 7 September 2026. Not company-sourced; treat as market context only.

The gap between a 22.9x trailing and ~7.1x forward multiple is the market saying earnings are near a peak. The $361–$2,200 target range is a more honest summary of the situation than any single multiple: the outcome distribution is genuinely bimodal.

9. Management & capital allocation

Sanjay Mehrotra has been CEO since May 2017 and Chairman since January 2025 — he has run Micron through a boom, the FY2023 bust, and this cycle. The senior team is heavily ex-SanDisk/Western Digital [FY2025 10-K]. On 26 August 2026 the company named two presidents — Manish Bhatia as President and COO, Scott DeBoer as President and Chief Technology and Products Officer — while Sumit Sadana moved from Chief Business Officer to Senior Advisor to the CEO [8-K 26 Aug 2026]. This reads as a succession structure being put in place. The board added Alexis Black Björlin in June 2026, bringing it to nine directors, eight independent [8-K 9 Jun 2026].

Capital allocation this cycle has been conservative: debt down $9.38bn in nine months, only $650mn of buyback executed ($7.84bn cumulative against a $10bn authorisation), and a small $0.15 quarterly dividend. Cash is going into capacity — Idaho, New York, Singapore and Taiwan (the $1.8bn Powerchip Tongluo fab, March 2026) — supported by up to $6.4bn of CHIPS grants, a 35% US investment tax credit, and a non-binding New York package of up to $5.5bn [Q3 FY2026 10-Q].

Compensation is better-designed than most: short-term incentive is 50% profitability (non-GAAP net income or operating margin — from FY2026, above-target payout requires both); PRSUs require 55th-percentile relative TSR versus the SOX index for target payout and are capped at target if absolute TSR is negative, with 25% weights each on HBM3E+ and data-centre SSD share [2025 DEF 14A]. CEO ownership guideline is 6x salary; Mehrotra beneficially owns 1,084,078 shares.

Claims to verify over time: (a) that floor pricing under the strategic agreements yields above-prior-peak gross margins; (b) that the $22bn of committed deposits and financial commitments, ~$18bn in cash, actually arrives; (c) that ~$27bn/yr of capex earns an acceptable return rather than becoming the next cycle's depreciation problem.

No Form 4 insider-trading data was collected for this issuer — not in sources.

10. Risks ranked by damage

1. Cycle reversion — ASPs normalise and the margin structure collapses

Damage rank: High · Permanence: Medium · Fixability: Low Mechanism: Supply catches demand; spot prices fall; contract floors become the realised price; new depreciation from the capex wave lands simultaneously. Evidence today: COGS is nearly flat while revenue quadrupled, meaning essentially all incremental profit is price — the most reversible profit there is. MCBU bit shipments are already declining on price [Q3 FY2026 10-Q]. What would confirm it: A negative DRAM ASP quarter, or industry capex announcements running well ahead of bit-demand forecasts. Impact on value: Severe. FY2023 showed the business can go from $8.7bn net income to a $5.8bn loss in four quarters.

2. Chinese supply build (CXMT, YMTC)

Damage rank: High · Permanence: High · Fixability: Low Mechanism: State-funded capacity added without return discipline permanently lowers industry pricing power, first in legacy nodes and mobile, then upward. Evidence today: Micron flags "DRAM and NAND oversupply due to significant investment… including by the Chinese government" [FY2025 10-K]; YMTC has filed multiple patent suits and a Lanham Act complaint against Micron. What would confirm it: CXMT credible DDR5/HBM qualification at a major customer; AEBU or MCBU pricing decoupling downward from data centre. Impact on value: Structural de-rating of through-cycle margins, not just a cycle trough.

3. Capital commitment at the peak

Damage rank: High · Permanence: Medium · Fixability: Low once committed Mechanism: ~$27bn of FY2026 net capex, against $15.86bn spent in FY2025, is committed on demand estimated during a shortage; output lands 2027–2030 [Q3 FY2026 10-Q; FY2025 10-K]. What would confirm it: Utilisation or write-down commentary; incentive clawback triggers. Impact on value: Converts a cyclical trough into a loss-making one via fixed depreciation.

4. Customer concentration and take-or-pay counterparty performance

Damage rank: Medium-High · Permanence: Low · Fixability: Medium Mechanism: One customer was 17% of FY2025 revenue, and the new contracts concentrate multi-year revenue in a handful of AI buyers whose own capex plans are unproven. Evidence today: Receivables of $31.0bn; remaining performance obligations of only ~$5bn recognised against $22bn of stated commitments [Q3 FY2026 10-Q]. What would confirm it: Deposits arriving materially below $18bn; rising consideration-payable balances; DSO extension. Impact on value: Removes the durability premium the thesis pays for.

5. Geopolitics, export controls and tariffs

Damage rank: Medium · Permanence: Medium · Fixability: Low Mechanism: China's CAC has barred critical information infrastructure operators from buying Micron products since May 2023 — disclosed as having "an adverse impact on our ability to compete effectively in China and elsewhere" [FY2025 10-K]. US export restrictions can also cut off customers, and tariffs raise cost. Impact on value: Permanent share loss in specific geographies rather than a valuation reset.

6. Technology execution — HBM4, 1γ, advanced packaging

Damage rank: Medium · Permanence: Low · Fixability: High Mechanism: HBM is share-based and qualification-gated; a missed generation forfeits a socket for years. HBM4 samples went to multiple customers with volume production stated as on schedule for calendar 2026 [FY2025 10-K], and the compensation plan puts 25% of PRSU value on HBM3E+ share. Impact on value: Loss of the highest-margin mix, not of the company.

7. Tax and government-incentive conditionality

Damage rank: Low-Medium · Permanence: Medium · Fixability: Low Pillar Two has already lifted the effective rate to 15.0% from 11.1%, largely offsetting Singapore incentives, and CHIPS funding is subject to conditions, reduction, termination or clawback [Q3 FY2026 10-Q].

11. Metrics to monitor

12. Thesis tripwires

13. Open questions

1. What gross margin is implied at contract floor pricing, and what share of FY2027–FY2029 volume sits under floors versus market pricing? The filings quantify neither.
2. Why is only ~$5bn of remaining performance obligation recognised against $22bn of stated commitments? The disclosure basis may understate it, but the gap needs closing.
3. What incremental bit growth and per-bit cost does the ~$27bn of FY2026 capex buy, and by when?
4. How much of the 2026 price spike is consumption versus customer inventory building ahead of contracted supply?
5. What is the succession timetable behind the August 2026 two-president structure?
6. What is management doing with its own shares? No Form 4 data was collected — worth adding to the source base.

14. Update log

15. Source map