Micron: reassessing earnings duration and cash returns

Thesis and research conclusions

01

Micron research should shift from a single HBM quote to the sources and duration of group profits.

AI demand and the lag in supply expansion support the earnings baseline, but do not eliminate cyclical volatility. UBS’s bullish FY27 forecast relies first on conventional DRAM and NAND, then on further HBM gross-profit releases. Tracking this structure cannot be replaced by one HBM headline. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Operating view: rebuilding the UBS model attributes 66.7% of FY27 incremental gross profit to conventional DRAM, 24.4% to NAND and 8.8% to HBM; HBM rises to 33.7% of the FY28 increase. Per-share view: UBS’s FY27 EPS of $189.98 exceeds our conditional middle path of $160.40 mainly because of profit differences, with a lower share denominator adding further accretion. Valuation view: UBS raised earnings but reduced its multiple, leaving its target unchanged; earnings improvement does not guarantee an equivalent share-price gain.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Company actuals, company guidance, named sell-side forecasts and our own conditional scenarios are presented separately. We do not assign unestimated probabilities to scenarios or substitute a bank target for our valuation judgment. Available Samsung and SK hynix materials support a comparison of pricing, competition and valuation frameworks, but not a same-basis reconstruction of group net income. We therefore do not invent a three-company profit ranking from assumed market shares.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Financial baseline: growth and margin recognition

02

Micron reported FQ4 FY2026 revenue of $54.229 billion and non-GAAP diluted EPS of $33.42.

Next-quarter guidance midpoints are $61.50 billion of revenue and $38.15 of non-GAAP EPS. These reported figures establish a high earnings starting point, but the next-quarter margin guide is below the current quarter: revenue growth and margin changes need separate explanations. [P1][P2]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 1. Quarterly results and next-quarter guidance

Revenue and expenses: USD 100 million; EPS: USD/share. Guidance is a company midpoint or range, not an actual result. Micron announcement and prepared remarks, 30 September 2026.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Quarter length affects the apparent growth rate. Regulatory quarter-end dates imply 14 weeks in FQ4 versus 13 in FQ3. Revenue grew 30.8% sequentially, or 21.5% after simple weekly normalisation. The latter still indicates strong growth, but the extra week cannot all be attributed to pricing or demand, and weekly normalisation is not a rigorous organic-growth adjustment. [P1][P3] Gross-profit recognition is not the same as product pricing. Prepared remarks link part of next-quarter margin pressure to manufacturing incentive compensation capitalised into inventory and later recognised on sale. Price, unit cost, mix and accounting timing should be separated; a one-quarter margin change alone does not establish HBM4 yield or demand. [P2]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Historical earnings follow company disclosures; forecasts follow the adjusted basis stated by each model. Company FY26 non-GAAP EPS is $75.52. UBS’s historical GAAP line differs from the company on a standardisation basis, so we neither use that line for growth calculations nor splice it into non-GAAP forecasts. [P1][U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Profit sources: conventional memory versus HBM

03

UBS forecasts FY27 adjusted net income of $211.841 billion, up 144.2%, and adjusted EPS of $189.98, up 151.6%.

This is a bullish model materially above the contemporaneous market snapshot cited in that report, not company guidance. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 1. Profit declines and per-share cushioning

Conditional models and labelled figures in the original report; not live quotes · FY26=100 · 2026-10-04T00:00:00Z

FY26 adjusted baseline=100. FY27–FY29 are UBS forecasts, not realised performance. UBS p.17; AlphaGBM calculations.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 2. UBS group earnings forecasts and growth

Revenue and net income: USD 100 million; EPS: USD/share. UBS pp.1 and 17; growth recalculated by AlphaGBM.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 2. FY27 incremental gross profit: conventional memory leads

Conditional models and labelled figures in the original report; not live quotes · % · 2026-10-04T00:00:00Z

The other-business residual of USD 0.16 × 100 million is not included in the three bars. UBS p.8; AlphaGBM calculations.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

FY27 incremental group gross profit is $149.943 billion: conventional DRAM contributes $100.074 billion, NAND $36.626 billion and HBM $13.227 billion, with $16 million of other-business residual. Testing conventional-memory pricing and costs therefore covers more of near-term forecast earnings than checking only HBM4 quotes. HBM becomes more important later, contributing 33.7% of FY28 incremental gross profit. Yet in FY29, even with HBM revenue forecast to grow 24.1%, group revenue and adjusted net income decline 14.3% and 20.0%. Structural product growth cannot fully offset a downturn elsewhere. This is gross-profit attribution within a bank model, not causal identification of realised net income. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

HBM pricing: volume, price, cost and comparability

04

UBS’s HBM profit expansion depends on both volume and price: FY27 shipments grow about 54%, ASP rises 54.8%, and unit cost increases only modestly.

Price multiplied by volume drives revenue; gross-profit elasticity also depends on costs, yields and mix. A quoted price increase cannot directly establish group net income. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 3. UBS estimates for MU HBM volume, price and cost

All figures, including FY26, are UBS estimates, not separately disclosed company HBM actuals. USD/GB denotes capacity pricing; volume is million GB. UBS p.8.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 3. A wider unit spread does not guarantee rising margins

Conditional models and labelled figures in the original report; not live quotes · USD/GB · 2026-10-04T00:00:00Z

All UBS estimates, including the FY26 baseline. The difference between the lines is the unit spread; the unit is not USD/Gb.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Unit spread and gross margin are different measures. In FY29, the model’s absolute spread still expands, but costs rise faster relative to price. Recalculated HBM gross margin falls from 84.2% in FY28 to 80.4%. Research must track realised pricing and costs together, not just one product generation’s quoted price.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 4. HBM gross-profit bridge: pricing is the largest contribution

Conditional models and labelled figures in the original report; not live quotes · USD 100m · 2026-10-04T00:00:00Z

Web bars retain the original waterfall values: 73.00 + 39.34 + 93.94 - 1.05 + 0.04 = 205.27. The decomposition order allocates interaction effects; this is not causal identification.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

In this ordered decomposition, shipment growth contributes $3.934 billion, ASP adds $9.394 billion and unit costs subtract $105 million. Pricing accounts for about 71.0% of the reported gross-profit increase, including the volume-price interaction assigned to pricing by this order. A $4 million display-rounding residual is retained, not labelled an operating contribution. Fiscal years, calendar years and cross-company forecasts are not interchangeable. UBS forecasts about 76% growth in MU’s CY2027 blended HBM ASP; Goldman’s Korea team forecasts Samsung HBM pricing up 130%; TrendForce forecasts industry blended ASP up 121%. Different companies, product mixes and methods cannot be averaged or transplanted into MU group EPS. [U][G][T] Capacity units must be normalised first. IT Home’s 3 October relay of Korean reporting uses USD/Gb, not USD/GB. USD 5/Gb equals USD 40/GB only as a unit illustration, not a confirmed transaction. Comparing it with HBM3E at about USD 1.5/Gb is a cross-generation quote comparison, not the same product’s annual ASP. Per-pin speed in Gbps is also distinct from capacity pricing. [V][S]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Supply commitments improve revenue visibility, not a group-profit floor. Goldman’s Korean report records 26 SCAs, expected to cover over 35% of revenue through 2030. About 75% of that expected SCA revenue has explicit price frameworks, generally with floors and ceilings. The 50% figure is a long-term coverage objective, not an achieved level, and 75% does not refer to all group revenue. Some agreements extend into 2031. [G] The lag from factory construction to saleable bits supports duration, but a bank relaying management’s supply-demand outlook is not an additional independent source. UBS’s customer-exit economics and downside-margin estimates are not actual contract terms; delivery, realised prices and costs still require validation. [G][U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Independent scenarios and variable stress tests

05

Our model anchors FQ1 FY27 to company guidance.

Annual operating expenses use FY26 expenses plus the company’s expected increment; the tax rate is fixed at 15.5%, annual shares at 1.15 billion, and other net income is held unchanged. Q2–Q4 sequential growth and margin paths are research assumptions, not company guidance, consensus or confidence intervals. [P1][P2]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 4. AlphaGBM scenarios and the UBS model

Revenue: USD 100m; EPS: USD/share. Our model is a non-GAAP-equivalent bridge; UBS uses its adjusted basis. Definitions are not identical. UBS average shares are about 1.115bn.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 5. Key assumptions in our scenarios

Research assumptions, not company guidance or probability intervals.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 5. Quarterly paths matter; guidance is not simply annualised

Conditional models and labelled figures in the original report; not live quotes · Quarterly EPS / USD · 2026-10-04T00:00:00Z

The web chart retains the first-quarter guidance anchor and explicitly labelled final-quarter EPS. Intermediate assumptions remain in Table 5; unlabelled points were not estimated from pixels. Not actual history or probability bands.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

UBS’s contemporaneous consensus snapshot is approximately $159–160 for FY27 and $180–182 for FY28. The cover and p.7 differ slightly, so ranges are retained. Our middle path’s proximity to this range does not mean it was calibrated to consensus, and the stronger path is not an upper bound on market optimism. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 6. Equal ASP deviations: conventional DRAM has greater EPS impact

Conditional models and labelled figures in the original report; not live quotes · USD/share · 2026-10-04T00:00:00Z

Each segment’s price is 10% below the UBS FY27 baseline, with volume, costs and expenses fixed; 15.5% tax and 1.15bn shares. Not a complete group stress forecast.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

With other conditions held constant, the separate conventional DRAM, NAND and HBM price deviations change EPS by -$14.33, -$5.21 and -$1.95. DRAM’s larger impact mainly reflects its revenue base, not a claim that its price is more likely to fall. The shock is 10% below the baseline price, not an equal percentage-point reduction in year-on-year growth, and it does not jointly stress volume, cost and buybacks. The group model already includes HBM; no earlier uniform-industry-share demonstration profit is added again. Such industry examples explain volume-price-cost and share-transfer mechanisms, not full group earnings for the three companies.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Cash returns: net income, free cash flow and EPS

06

UBS treats repurchases as a second argument alongside profit duration.

Its cash-flow model expects operating cash flow to cover capital expenditure and leave substantial shareholder-return capacity. This is a bank’s capital-allocation forecast, not completed repurchases or a board-authorised amount. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 7. Cash uses: repurchases depend on free cash flow

Conditional models and labelled figures in the original report; not live quotes · USD 100m · 2026-10-04T00:00:00Z

Grouped bars retain the original chart’s labelled OCF and FCF. Net repurchases are 75.4% / 98.3% / 98.2% of FCF as stated below. No amounts are reverse-engineered from the unlabelled line. All UBS forecasts.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

In the model, FY27, FY28 and FY29 net repurchases equal 75.4%, 98.3% and 98.2% of FCF. These are not shareholder returns or authorisation-completion rates: the latter two years allocate nearly all forecast FCF to buybacks. Lower cash generation requires changes in actual repurchases or other uses. Average diluted shares are forecast at 1.115 / 1.001 / 0.847 billion; these are not year-end cumulative buyback percentages. Net and gross capex differ. UBS FY27 gross capex of $57.0bn less $3.308bn of partner and government contributions gives net capex of $53.692bn. The company’s first-half net capex of about $25bn, with more in the second half, is not a full-year guide and is not directly comparable with bank gross capex. [P2][U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 8. From our EPS to UBS: profit and denominator differences

Conditional models and labelled figures in the original report; not live quotes · USD/share · 2026-10-04T00:00:00Z

Profit is replaced first, then the effective denominator. Web bars preserve the original bridge, not independent causal identification; adjusted-profit definitions differ.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

UBS net income divided by our fixed share count gives EPS of $184.21. Replacing profit first adds $23.81; changing the effective denominator adds $5.77, linking our $160.40 to UBS’s $189.98. Price, execution timing and dilution affect realised accretion; annual average diluted shares are not year-end shares or cumulative repurchase rates. Buybacks cushion per-share metrics, but do not remove an operating downturn: UBS FY29 net income falls 20.0% while EPS falls only 5.4%. Earnings quality must be assessed through net income, FCF and shares together, not merely a smooth EPS curve. [U]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Valuation: earnings duration and multiple constraints

07

UBS retains Buy and a $1,625 target.

It starts with approximately $256 of CY2029 EPS, discounts one year at about 13.6% cost of equity to roughly $225 in CY2028, and applies approximately 7× P/E, previously about 8×. Higher earnings alongside a lower multiple and higher discount rate illustrate that operating optimism and valuation restraint can coexist. [U] This is not FY27 EPS multiplied directly by a multiple. Rounded published parameters do not reproduce the target exactly; we retain that precision limit rather than inventing an undisclosed multiplier. The following sensitivity instead multiplies our FY27 EPS by assumed multiples: the horizons differ.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 9. Earnings and multiple sensitivity; no single-point target

Conditional models and labelled figures in the original report; not live quotes · Conditional price / USD · 2026-10-04T00:00:00Z

Rounded prices from the original heatmap. Changes versus the 2 October 2026 close of $1,074.89 are in the table. Assumed multiples carry no probability or return promise.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Figure 9 data: conditional prices and changes versus reference

Prices and percentages retain original display rounding, not targets or expected returns.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

At the 2 October 2026 regular-session close of $1,074.89, our three paths imply P/E of about 8.01 / 6.70 / 6.24×, versus 5.66× on the UBS FY27 forecast. Every denominator is a future scenario or forecast, not realised earnings. Even if next-quarter results keep growing, shorter profit duration and multiple compression can jointly affect price. A public relay of Goldman’s US MU view retains Neutral and a $1,250 target, whereas the Korean report is constructive on Samsung and SK hynix. Companies, teams and evidence levels differ; these are not a single unified bank view. The original US MU model was not obtained, so the relay is only a disagreement signal, not an earnings-model input. [B1][G]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Samsung and SK hynix: shared upside and share competition

08

Industry revenue expansion underpins shared benefits; execution differs by company.

Micron’s original disclosures and full sell-side business model allow group earnings decomposition. Samsung HBM pricing needs qualification, yields and deliveries; SK hynix needs realised pricing, costs and share defence. Goldman identifies Samsung’s HBM progress as a risk to SK hynix, showing that industry growth and competitive pressure can coexist. [G]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 6. Company research focus and named bank valuation methods

Korean targets retain original currency, without unverified FX conversion. Ratings and targets are those in the 1 October 2026 reports; not all were necessarily raised. Absolute targets cannot be compared across companies.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

NAND also deserves tracking. Goldman records MU FQ4 NAND volume and ASP growth above its forecasts for the Korean companies’ quarterly average, a constructive read-across for later Korean results. But companies, time windows and forecast status differ; this does not establish that Samsung or SK hynix has already delivered the same growth. [G] Samsung includes other businesses, so HBM gross-profit growth cannot automatically equal group net-profit growth. The Korean materials do not provide fully reconstructable same-basis group-profit models. We therefore retain supported pricing, competition and valuation judgments without a false-precision three-company incremental-profit ranking.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Catalysts, risks and monitoring

09

Updates should focus on observable business variables, not unconfirmed event dates.

Positive news on pricing, cash returns or qualification should enter a group model only when it translates into realised revenue, gross profit and share counts.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Table 7. Catalysts and disconfirming evidence

Monitoring variables are not realised events or automatic trading instructions.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Other key risks include more efficient AI models, customers reducing memory capacity or stacking layers, technology migration and yield deviations, and competitive share shifts. Shareholder returns do not automatically offset all operating or valuation risks. [U][G][T]

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

The durability of high memory profits merits further testing, in this order: business volume-price-cost, group gross profit and expenses, cash flow and shares, then valuation. For Micron the near-term test remains non-HBM profit; for Samsung and SK hynix, share and delivery must be checked beyond the common cycle. Current evidence is insufficient to turn any single price headline into a certain share-price gain.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

Sources and important disclosures

10

This report was compiled, calculated and formatted by AlphaGBM Research with AI assistance.

Source list and evidence boundaries in the original report

Third-party originals remain internal; public references appear in the sources area. Web conversion is not an additional independent verification.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

It carries no named human analyst certification and invents no qualifications, compensation relationship or independence certification. AlphaGBM assigns no Buy/Hold/Sell rating and no target price. Named banks’ ratings and targets are quotations; scenarios and sensitivities are analytical tools, not recommendations or actual outcomes.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

1. Sources and nature. The report uses public company disclosures, bank reports provided for this research reading, and identified industry information and media relays. No non-public customer contracts were obtained and the companies did not review it. Source accuracy, completeness and ongoing validity are not guaranteed; possessing a bank original does not make its forecast an actual result. 2. Not investment advice. The report is for research, not an offer, invitation, recommendation, rating or trading advice. It is not tailored to recipients’ objectives, financial position or risk tolerance and does not replace independent judgment. 3. Forward-looking statements. Earnings, supply-demand, price and repurchase forecasts depend on assumptions and actual results may differ materially. Historical performance is not predictive; scenarios neither guarantee returns nor establish that developments are unpriced.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

4. Data boundaries. Cutoffs and source dates are on the cover and in the source list. Prices retain the original verified regular close, not intraday quotes. FY and CY are separated; gross profit, net income and EPS are not mixed. Actuals, guidance, bank forecasts and our assumptions remain distinct. USD/Gb is not USD/GB. 5. Commercial use and redistribution. Rights in third-party reports and materials remain with their owners; this report transfers no such rights. Without the relevant authorisation it may not be used commercially, including resale, sublicensing, inclusion in paid products, fundraising or AI model training, or copied at scale, redistributed or publicly circulated. Quotations must retain attribution, dates and risks without removing context.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally

6. Conflicts and responsibility. Holdings and conflicts of the research team and related persons were not checked, so no no-conflict assertion is made. The report neither checks nor changes user holdings and executes no trades. Research may be revised; recipients must verify key premises and must not equate tool validation with reliable investment outcomes.

[1] AlphaGBM Research · MU · 2026-10-04 · v4.0 · Author-owned research; original retained internally