
Buy, $1,085 target: the next memory downturn has a price floor under it
We initiate at Buy with a $1,085 target, a 64.9% total return: the price discounts $86bn of mid-cycle NOPAT, between our bear ($61bn) and base ($139bn) cases and closer to the bear.
- The price sits closer to our bear than our base: $86bn of mid-cycle NOPAT against $61bn and $139bn. Our base needs 26 supply agreements (over 35% of revenue to 2030) to hold.
- CY28 is tighter than CY27: we are 7.6% above consensus on FY28E revenue. $343.9bn against $319.7bn.
- Cash arrives before the downturn: $328bn of free cash flow in FY27E-FY28E, 44% of the market value. Capex is funded internally.
What the market is missing. Price-implied mid-cycle EBIT margin is 37.5%; our base is 62.0% with floors holding, and the FY31E trough is 55%. We agree on FY27E and differ on FY29E-FY31E.
Catalysts. Q1 FY27 results in December (guide $61.5bn), Q2 FY27 in March 2027 as repriced HBM bits land, and CY28 contract talks.
Valuation and target. 6.9x mid-cycle earnings at a 15.3% WACC on $143.3bn of weighted FY29E-FY36E NOPAT, plus FY27E-FY28E cash flow and $55.4bn net cash; $918 weighted, rolled to $1,085; DCF $876.
Risks. CXMT server DDR5 brings the downturn forward (−$29 a share), NAND cleanroom space from late CY28 (−$62) and NAND prices levelling (−$26).
| US$bn | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Revenue | 37.4 | 133.2 | 276.1 | 343.9 | 326.2 |
| Growth y/y | 48.9% | 256.3% | 107.3% | 24.6% | −5.2% |
| EBITDA | 18.1 | 108.8 | 240.7 | 304.4 | 282.9 |
| Margin | 48.5% | 81.7% | 87.2% | 88.5% | 86.8% |
| Net income | 8.5 | 85.0 | 195.3 | 246.0 | 225.8 |
| EPS, non-GAAP ($) | 8.29 | 75.52 | 171.21 | 215.36 | 197.83 |
| Growth y/y | n/m | 811% | 127% | 26% | −8% |
| Free cash flow | 3.7 | 62.3 | 135.6 | 192.7 | 192.8 |
| At $658.70 | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| P/E | 79.5x | 8.7x | 3.8x | 3.1x | 3.3x |
| EV/EBITDA | 38.1x | 6.3x | 2.9x | 2.3x | 2.4x |
| P/B | 13.8x | 5.4x | 2.2x | 1.3x | 0.9x |
| FCF yield | 0.5% | 8.4% | 18.2% | 25.9% | 25.9% |
| Dividend yield | 0.1% | 0.1% | 0.1% | 0.1% | 0.1% |
| ROE | 15.8% | 61.4% | 58.3% | 42.3% | 27.9% |
| Net debt / EBITDA | 0.2x | (0.5)x | (0.8)x | (1.3)x | (2.0)x |
| P/E at our $1,085 target | — | — | 6.3x | 5.0x | 5.5x |
Source: Company filings (A); Penmark Research estimates (E). Multiples at the 5 Nov 2026 price; EV/EBITDA uses today's EV of $689.6bn for every year; only the last row is at our target. FY25A EPS growth is not meaningful (no FY24 non-GAAP EPS). $bn except EPS ($). Negative net debt is net cash, after $12.9bn of customer contract liabilities.
Base $911, bull $1,653, bear $507: 1.7x upside to downside against a $659 price
Source: Exchanges; Penmark Research estimates. Scenario values at the 3 Sep 2026 valuation date, shown at the model values with no roll-forward. The $1,085 target is the weighted $918 rolled forward at the 15.4% cost of equity to 5 Nov 2027 and is not plotted
Tightness holds into CY29: HBM4E repricing keeps DRAM up 10% in FY28E, the downturn slips to FY30E; trough gross margin 73.9%, mid-cycle EBIT 74.8%.
Tight through CY28, then a downturn as cleanroom space arrives; floors hold. FY31E trough 55% EBIT margin (62.3% gross); mid-cycle 62.0% and $138.7bn of NOPAT, against $85.8bn in the price.
Pricing peaks in CY27 as CXMT ships server DDR5 and floors fail; a 70% DRAM price fall to FY31E leaves a 43.8% trough gross margin, 43.9% mid-cycle EBIT and $61.4bn of NOPAT.
| Income, US$bn | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Revenue | 37.4 | 133.2 | 276.1 | 343.9 | 326.2 |
| Gross profit | 15.3 | 108.0 | 240.4 | 297.2 | 268.1 |
| EBITDA | 18.1 | 108.8 | 240.7 | 304.4 | 282.9 |
| EBIT | 9.8 | 99.3 | 229.3 | 285.2 | 255.5 |
| Net financial items | (0.4) | 0.2 | 1.9 | 5.9 | 11.7 |
| Tax | 1.1 | 14.8 | 35.8 | 45.1 | 41.4 |
| Net income, non-GAAP | 9.3 | 86.3 | 196.9 | 247.7 | 227.5 |
| Net income, reported | 8.5 | 85.0 | 195.3 | 246.0 | 225.8 |
| Cash flow, US$bn | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| EBITDA | 18.1 | 108.8 | 240.7 | 304.4 | 282.9 |
| Working capital and other | 1.0 | (4.6) | (16.2) | (7.5) | 4.6 |
| Tax and interest paid | (1.6) | (14.5) | (33.9) | (39.2) | (29.7) |
| Operating cash flow | 17.5 | 89.7 | 190.6 | 257.7 | 257.8 |
| Capex, net of incentives | (13.8) | (27.4) | (55.0) | (65.0) | (65.0) |
| Free cash flow | 3.7 | 62.3 | 135.6 | 192.7 | 192.8 |
| Dividends and buybacks | (0.5) | (1.3) | (0.7) | (0.7) | (0.7) |
| Acquisitions and other | — | — | — | — | — |
| Balance sheet, US$bn | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Cash and investments | 11.9 | 73.5 | 208.3 | 400.3 | 592.4 |
| Net working capital | 7.0 | 20.6 | 38.6 | 48.1 | 45.7 |
| Net PP&E | 46.6 | 63.3 | 106.9 | 152.7 | 190.2 |
| Goodwill and other | 7.3 | 11.5 | 11.5 | 11.5 | 11.5 |
| Total assets | 82.8 | 168.9 | 365.4 | 612.7 | 839.8 |
| Debt | 14.6 | 5.2 | 5.2 | 5.2 | 5.2 |
| Net cash / (debt) | (2.8) | 55.4 | 190.3 | 382.3 | 574.4 |
| Shareholders’ equity | 54.2 | 138.4 | 334.9 | 582.1 | 809.3 |
| Per share, KPIs | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Diluted shares (m) | 1,125 | 1,143 | 1,150 | 1,150 | 1,150 |
| EPS, non-GAAP ($) | 8.29 | 75.52 | 171.21 | 215.36 | 197.83 |
| EPS, reported ($) | 7.59 | 74.33 | 169.87 | 213.91 | 196.31 |
| Dividend per share ($) | 0.46 | 0.53 | 0.60 | 0.60 | 0.60 |
| FCF per share ($) | 3.31 | 54.52 | 117.88 | 167.57 | 167.64 |
| Book value per share ($) | 48.15 | 121.06 | 291.18 | 506.20 | 703.71 |
| Non-GAAP gross margin | 40.9% | 81.1% | 87.1% | 86.4% | 82.2% |
| Capex / sales | 36.9% | 20.5% | 19.9% | 18.9% | 19.9% |
Source: Company filings (A); Penmark Research estimates (E). Non-GAAP adds back stock compensation to cost of sales and charges it below EBIT; capex is net of government incentives.
Returns peak at 174% on capital in FY27E, and the stock still trades at 3.8x earnings
| Growth, margins (%) | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Growth y/y | |||||
| Sales | 48.9 | 256.3 | 107.3 | 24.6 | −5.2 |
| EBITDA | 99.5 | 500.6 | 121.1 | 26.5 | −7.1 |
| EBIT | 649.2 | 916.8 | 130.8 | 24.4 | −10.4 |
| EPS | n/m | 811.0 | 126.7 | 25.8 | −8.1 |
| DPS | 0.0 | 15.2 | 13.2 | 0.0 | 0.0 |
| Margins | |||||
| Gross margin | 40.9 | 81.1 | 87.1 | 86.4 | 82.2 |
| EBITDA margin | 48.5 | 81.7 | 87.2 | 88.5 | 86.8 |
| EBIT margin | 26.1 | 74.6 | 83.1 | 82.9 | 78.3 |
| FCF margin | 10.0 | 46.8 | 49.1 | 56.0 | 59.1 |
| Cash conversion, FCF/net income | 43.6 | 73.3 | 69.4 | 78.3 | 85.4 |
| Net interest / EBIT | −3.7 | 0.2 | 0.8 | 2.1 | 4.6 |
| Capex / sales | 36.9 | 20.5 | 19.9 | 18.9 | 19.9 |
| Ratios and valuation | FY25A | FY26A | FY27E | FY28E | FY29E |
|---|---|---|---|---|---|
| Returns and leverage | |||||
| ROE | 15.8 | 61.4 | 58.3 | 42.3 | 27.9 |
| ROA | 10.3 | 50.3 | 53.5 | 40.2 | 26.9 |
| ROCE | 17.2 | 141.7 | 174.1 | 152.5 | 115.1 |
| Net debt / equity | 4.9 | (49.3) | (60.7) | (67.9) | (72.6) |
| Net debt / EBITDA | 0.2 | (0.5) | (0.8) | (1.3) | (2.0) |
| Valuation at $658.70, 5 Nov 2026 | |||||
| P/E | 79.5 | 8.7 | 3.8 | 3.1 | 3.3 |
| EV/EBITDA | 38.1 | 6.3 | 2.9 | 2.3 | 2.4 |
| P/B | 13.8 | 5.4 | 2.2 | 1.3 | 0.9 |
| FCF yield | 0.5 | 8.4 | 18.2 | 25.9 | 25.9 |
| Dividend yield | 0.1 | 0.1 | 0.1 | 0.1 | 0.1 |
| P/E at our $1,085 target | — | — | 6.3 | 5.0 | 5.5 |
Source: Company filings (A); Penmark Research estimates (E). Ratios in %. ROE on closing equity, ROA on closing assets, ROCE on equity plus debt less cash. Values in brackets are net cash. Multiples at the 5 Nov 2026 price except the last row; EV/EBITDA on today's EV.
| US$m, $ per share | Q1 FY27E | FY27E | FY28E |
|---|---|---|---|
| Revenue | |||
| Penmark estimate | 61,450 | 276,057 | 343,907 |
| Consensus | 61,830 | 275,510 | 319,680 |
| Prior-year period | 13,643 | 133,188 | 276,057 |
| Diff. vs consensus | −0.6% | +0.2% | +7.6% |
| Growth y/y, Penmark | 350.4% | 107.3% | 24.6% |
| Adjusted EPS, $ | |||
| Penmark estimate | 37.78 | 171.21 | 215.36 |
| Consensus | 38.20 | 176.50 | 207.80 |
| Prior-year period | 4.78 | 75.52 | 171.21 |
| Diff. vs consensus | −1.1% | −3.0% | +3.6% |
| Growth y/y, Penmark | 690.4% | 126.7% | 25.8% |
Source: Company filings; S&P Global consensus (29 Oct 2026), Yahoo Finance for Q1 (14 Oct 2026); Penmark Research estimates. Q1 FY27 guide beside consensus: revenue $60.0-63.0bn, EPS $37.15-39.15.
Source: Company filings; aggregators, 3 Sep to 3 Nov 2026. Forward P/E on consensus 12-month earnings
| Micron vs peers | Micron | Peer median | Diff. |
|---|---|---|---|
| Forward P/E | 3.7x | 3.3x | +11% |
| P/E on FY28E EPS | 3.1x | 3.3x | −6% |
| Op. margin, TTM (SanDisk) | 82.3% | 72.5% | +9.8pp |
| Beta, 2-year (Micron) | 2.66x | 2.30x | +0.36 |
| DRAM share: Samsung, SK hynix | 24% | 38%, 25% | −14, −1pp |
| HBM share: SK hynix, Samsung | 18% | 50%, 33% | −32, −15pp |
Source: Company filings; BofA for DRAM and HBM share (Oct 2026); Penmark Research estimates. Median of Samsung, SK hynix, SanDisk. Shares are each company's own, not medians. The 2.48x beta used in the model is the midpoint of 2.66x and 2.30x. Difference in % for multiples, pp for ratios.
Four dated events decide the call; NAND cleanroom space is the largest single risk
- Dec 26Q1 FY27 results. Guide $61.5bn ±$1.5bn and EPS $38.15 ±$1.00; we model $61.45bn and $37.78, with 86.25% gross margin as the FY27 floor.
- Mar 27Q2 FY27 results. CY27 HBM bits contracted at higher prices land in the P&L; we model 87.0% gross margin against Kioxia's view that NAND levels off.
- 2H FY27CY28 contract talks and first SCA extensions. New and extended SCAs at floor prices; we model DRAM ASP +5% against consensus revenue of $319.7bn.
Source: Company filings and Q4 FY26 call (30 Sep 2026); Penmark Research estimates.
- In the base, lost if it fails: $215 a share. SCA floor path $134 (DRAM ASP −20/−40/−25% in FY29E-31E against the bear path); CY28 tightness $38; HBM repricing $18; server units $17; data-centre SSD share $8.
- Not in the base, lost if it lands: $152 a share. NAND cleanrooms from 2H CY28 −$62; CXMT server DRAM −$29; NAND prices levelling −$26; slower cost per bit −$21; slower AI build −$14.
Source: Penmark Research model, Thesis tab (catalyst values per share against the base case). The five bear items overlap and do not add to the $404 gap between base and bear.
Source: Exchanges; company filings. Monthly, price x 1,131m shares over book value
Source: Penmark Research estimates. Weighted value per share at the 3 Sep 2026 valuation date; each 10% off mid-cycle earnings is about $65 a share
| Indicator | Thesis confirmed if | Thesis broken if |
|---|---|---|
| DRAM contract price, q/q | 4Q26 +10-15%, flat to up through CY27 | Down more than 5% q/q before CY28 contracts |
| SCA coverage | Above 35% of revenue to 2030, extended each print | No extension at Q2 FY27, or floors cut |
| HBM price vs commodity | Gap narrows as HBM4E ships through CY27 | HBM priced at or below commodity DRAM |
| Capex intensity | At or below 28% of revenue, matching bit growth | Above 35% without matching bit growth |
| Base-case floor path | FY31E EBIT margin holds near 55% | DRAM follows the bear price path (−70%) |
3. The thesis in charts
Margin, not bit growth, is what the share price trades on
Micron's share price has tracked its gross margin through the upcycle: the market is pricing the durability of the margin, not the volume of bits. Each chart tests one leg of the call.
Source: Company filings; exchanges. Quarterly non-GAAP gross margin against the average daily close in the same quarter
Gross margin explains the stock, and the model explains gross margin. From Q1 FY25 to Q4 FY26 the margin rose 47.5pp and the average price from $100 to $957. Bear and base differ by 18.5pp of gross margin at the base trough, 3pp in FY28E.
We differ from consensus in one year only. FY27E revenue is within 0.2% of consensus; FY28E is 7.6% above it, worth $6 a share against the market, with the trough debate worth $252. The chart below right shows the same point on value: the two methods bracket $900 while the price sits at $659.
Source: S&P Global consensus (29 Oct 2026); Penmark Research estimates
Source: Company filings (A); Penmark Research estimates (E)
Source: Company filings (A); Penmark Research estimates (E)
3. The thesis in charts (continued)
The bull and bear cases differ by the depth of one downturn, not by the next two years
Through FY28E the three cases sit within 30% of the base case on revenue; from FY29E they diverge by a factor of two, and that is what the $659 price argues about.
Source: Penmark Research estimates. Year-on-year change in DRAM average selling price
Source: Penmark Research estimates. Revenue, $bn
Source: Penmark Research estimates. Value per share at 3 Sep 2026
Source: Penmark Research estimates; Thesis tab of the model.
Source: Penmark Research model, Thesis tab. Top five: in the base, lost if it fails ($215). Bottom five: not in the base, lost if it lands ($152). Value per share against the base case; items overlap
4. Key debates
Three debates, and only the third one is worth $252 a share
The market and we agree on the next twelve months. The argument is about what happens after the peak: the price discounts $86bn of mid-cycle NOPAT, against $139bn in our base (62.0% EBIT margin) and $61bn in the bear.
| Debate | Market view | Penmark view | Evidence | What would change our view | Resolves |
|---|---|---|---|---|---|
| 1. FY27E exit pricing: HBM repricing against NAND levelling | FY27E EPS $176.50 with gross margin flat near 86% after the Q1 guide. | FY27E EPS $171.21; margin rises to 87.0-87.5% as CY27 HBM bits land. | Most CY27 HBM bits contracted at significant price rises (Q4 FY26 call); TrendForce sees HBM blended ASP up 121% in 2027; Kioxia says NAND is levelling (2 Nov 2026). | Q2 FY27 gross margin below the 86.25% Q1 guide. | Mar 2027 |
| 2. CY28 tightness and FY28E pricing | FY28E revenue $319.7bn, up 16% on FY27E. | $343.9bn, up 24.6%, on DRAM ASP +5% and 20% bit growth. | Management sees CY27-28 tighter than CY26, no line of sight to balance (30 Sep 2026); ID2, Japan and Singapore NAND land late CY28. | CY28 contracts signed at or below CY27 prices. | 2H FY27 |
| 3. Depth of the FY29E-FY31E downturn | Price-implied: $86bn of mid-cycle NOPAT, a 37.5% EBIT margin on base-like revenue of $271bn. | Base: 62.0% mid-cycle EBIT margin and $139bn of NOPAT; FY31E trough 55% EBIT, 62.3% gross. | 26 supply agreements cover over 35% of revenue to 2030 at floor prices above any prior peak; RPO about $150bn at minimum prices and $12.7bn of customer cash deposits. | A CY28 contract signed below the floor price. | Late CY28 |
Source: Company filings and Q4 FY26 call (30 Sep 2026); S&P Global consensus (29 Oct 2026); TrendForce (29 Sep, 20 Oct 2026); Kioxia (2 Nov 2026); Penmark Research estimates.
Source: Penmark Research; Market Check. Base-case value less the value at the market's view
Debate 3 carries 99% of the gap. On the first two we are within $6 a share of the market. The trough is different: it requires believing that contract floors hold through a downturn.
The evidence is contractual, not cyclical. Customers paid $12.7bn up front for supply at minimum prices, which makes the floor a liability on both sides if breached. We therefore weight base 45% against bear 35%.
What is already priced. 28 of 30 FY27E EPS revisions were up in 30 days and the stock is 46% below its June high: the debate is no longer demand, but the margin in the year after next.
Source: Company calendar; Penmark Research estimates
5. How the company makes money
$133bn of revenue from selling bits at contract prices, and $99bn of it reaches operating profit
Micron makes DRAM and NAND and sells them to four customer groups on quarterly contracts, increasingly with floor prices. A new fab takes four to five years, so price, not volume, decides profit.
Customers
Hyperscalers and CSPs$43.1bn Cloud Memory Server and storage OEMs
$37.6bn Core Data Center Phone and PC makers
$36.6bn Mobile and Client Auto and industrial
$15.9bn Automotive
Offerings
DRAM, incl. HBM$100.7bn, 76% of revenue NAND, incl. data-centre SSD
$31.8bn, 24% of revenue NOR and other
$0.7bn
Revenue streams
26 SCAs: >35% of revenue to 2030 at floor prices Quarterly contracts and spot for the balance RPO ~$150bn at minimum prices; $12.7bn of cash depositsChannels
Direct to hyperscalers and OEMs: ~85% of revenue Distributors and catalogues for auto and industrial Four business unitsCost to serve
Cost of goods $25.7bn, 19% of revenue, of which D&A $9.5bn R&D $5.7bn; SG&A and other $2.5bn Capex, net $27.4bn, 21% of revenueWhat is left
Operating profit $99.3bn, a 74.6% margin Free cash flow $62.3bn after $27.4bn of net capex Net cash $55.4bn after $12.9bn of customer depositsSource: Company filings (A); Q4 FY26 call and earnings deck (30 Sep 2026); BofA for HBM share (Oct 2026); Penmark Research estimates for the direct-channel share. FY26A, $bn.
Pricing is quarterly, and the contract book is the new part. Volume comes from Micron's own fabs; price is set quarterly. Since FY26 it has signed 26 supply agreements covering over 35% of revenue to 2030 at floor prices, with $12.7bn of cash up front and a ~$150bn backlog at minimum prices — part of a spot business is now contracted.
Costs are fixed, so incremental revenue falls to profit. Cost of goods is 19% of revenue in FY26 against 60% in FY25, not because costs fell but because price rose; cash cost per bit falls about 6% a year. Each 1pp of gross margin is worth $1.3bn of operating profit.
Profit and cash diverge through capex and working capital. FY26 converted 73% of net income to free cash flow; the gap is $27.4bn of net capex and a $4.6bn working-capital build as receivables grew with revenue. From FY27E capex rises to $55bn and then $65bn, which is why cash conversion falls to 69% before recovering to 85% in FY29E.
Concentration is by customer type. Data-centre customers — hyperscalers and server OEMs — are 61% of FY26A revenue and 63% in Q4, up from 57% eight quarters ago, and the largest single agreement covers a mid-single-digit share of revenue. That concentration is why the SCA floor matters: a handful of buyers can underwrite a fab, and they have.
5. How the company makes money (continued)
A fifth of revenue covers the fab; three-quarters reaches operating profit
Two charts carry the model: where $133.2bn of revenue goes, and what produces it. Revenue is bits × price, and FY26 growth came from price, so we forecast ASP first.
Source: 10-K FY26 (filed 9 Oct 2026). GAAP; $bn
Source: Company filings; Q4 FY26 call. Growth rates as disclosed, ranges not points
Price, not bits, produced FY26. DRAM bits grew in the mid-20s percent while DRAM revenue rose 252%. The model forecasts ASP first and calibrates bits to it, with cash cost per bit falling 6% a year.
The business units are a customer lens on the same fabs. Cloud Memory and Core Data Center sell the same DRAM at different prices, so their margins differ.
Source: Company filings; sum of FY26 quarters, $bn
5. How the company makes money (continued)
Core Data Center is 12x its FY25 trough, and every unit earns a 76-88% operating margin
The four business units are the same wafer sold to different buyers. In the upcycle all four earn extraordinary margins; what separates them is how quickly their prices reset when supply loosens, which is why the model weights data-centre revenue as the last to fall.
Source: Company filings; quarterly revenue by business unit, $bn
| Business unit | FY26 rev. $bn | Share | Q4 gross | Q4 op. |
|---|---|---|---|---|
| Cloud Memory | 43.1 | 32% | 83% | 76% |
| Core Data Center | 37.6 | 28% | 90% | 85% |
| Mobile and Client | 36.6 | 27% | 90% | 88% |
| Auto and Embedded | 15.9 | 12% | 84% | 79% |
| Total, as reported | 133.2 | 100% | 87.0% | 82.3% |
| Growth in the latest quarter, q/q | ||||
| Core Data Center | +56% | — | +3pp | +2pp |
| Cloud Memory | +18% | — | 0pp | −2pp |
| Mobile and Client | +14% | — | +3pp | +2pp |
| Auto and Embedded | +47% | — | +5pp | +4pp |
Source: Company filings; Q4 FY26 earnings deck (30 Sep 2026). Unit margins as disclosed; margins are Q4 FY26; the total is the reported Q4 non-GAAP gross and operating margin; revenue is the FY26 sum of quarters.
Source: Company filings; quarterly revenue by technology, $bn
Mix has not moved, and that is the point. DRAM has been 73-79% of revenue in every one of the last eight quarters. The shift is inside the mix: data-centre DRAM and HBM take a larger share of DRAM bits, and data-centre SSD is now more than two-thirds of NAND revenue at about $10bn in Q4 FY26.
The unit that resets first is the one to watch. Mobile and Client carries the highest operating margin at 88% because it reprices every quarter, so it is the first to turn in a downturn. Our FY29E revenue fall of 5% assumes mobile resets first and data-centre contracts hold through the year.
6. The market
A $420bn DRAM market in FY26, in month twelve of an upcycle with no new supply until late CY28
Memory is a bit-growth market priced quarterly. Industry DRAM bits grow in the low-20s percent a year, cost per bit falls about 6%, and price does the rest. FY26 was the year price ran ahead of bits: the market more than tripled to $420bn on a 24% share held by Micron.
Source: Company filings; Penmark Research estimates. Micron DRAM revenue grossed up at a 24% share (BofA, Q2 CY26), $bn
Demand is now a data-centre number. Server units are guided up high teens in CY26 and CY27, and memory is 68% of cloud capex in 2027 (TrendForce). Micron's data-centre business units are 63% of revenue against 57% eight quarters ago, and data-centre SSD is more than two-thirds of NAND.
Supply cannot respond before late CY28. A greenfield fab takes four to five years and a cleanroom about two. Micron's own additions — Singapore HBM packaging early CY27, ID1 and Tongluo mid-CY27, ID2 and Japan late CY28, Singapore NAND 2H CY28, New York CY30 — set the timetable for the whole industry's next oversupply.
Cycle position: late but not over. Inventory in the channel is at record lows, more than 75% of FY27 output is committed, and pricing is still rising quarter on quarter. The bear case does not dispute that; it argues the turn comes a year early.
Source: Company filings; TrendForce; Penmark Research. Stage detail by FY26A revenue or capex
The profit pool sits in the three suppliers. Memory makers keep 43% of the value of the chain and take almost all of the cyclical risk: equipment suppliers are paid whether the cycle turns or not, and systems vendors pass memory cost through. That is why a 20% price move changes Micron's profit by more than it changes Dell's.
Two threats are real and one is not. CXMT's 10% DRAM share and its server DDR5 push are real and drive our bear case. HBM 8-high stacks cutting HBM per GPU is real but margin-neutral for Micron, since it raises bit volume. Substitution of DRAM by any other technology is not a threat in this cycle.
7. Competitive position
Third of three in DRAM and HBM alike: the contract book is the only durable advantage
Micron competes on cost and node, and it is neither the largest nor the lowest-cost producer. What it has built instead is a customer relationship — 26 supply agreements with floor prices and cash deposits — that we have not seen matched in its rivals' public disclosure.
Source: BofA (Oct 2026); TrendForce. Share of industry revenue
| Choice criterion | Micron | Samsung | SK hynix |
|---|---|---|---|
| Position at Q2-Q3 CY26 | |||
| DRAM share | 24% | 38% | 25% |
| HBM share | 18% | 33% | 50% |
| Sources of advantage | |||
| Contract coverage | 26 SCAs, 35%+ | Not disclosed | Not disclosed |
| Evidence it lasts | |||
| Floor-price contracts | Floors on 35% | Not disclosed | Not disclosed |
Source: Company filings; BofA (Oct 2026); TrendForce; Penmark Research estimates. Samsung and SK hynix do not disclose contract coverage.
The advantage is contractual, and it is new. Scale and node leadership belong to Samsung and SK hynix; Micron's 24% DRAM share is the third largest. What changed in FY26 is the contract book: 26 supply agreements covering more than 35% of revenue to 2030 at floor prices, $32bn of customer commitments and $12.7bn of cash already received.
A floor price is only as good as the customer's incentive to honour it. Customers posted cash, which makes walking away expensive for them and turns the floor into a shared liability. That is a stronger mechanism than a long-term agreement without deposits.
The weak spot is HBM. At 18% of HBM against SK hynix's 50%, Micron is a price taker in the fastest-growing part of DRAM, and SK hynix leads on HBM share. Our bull case does not require Micron to gain HBM share; it requires HBM pricing to narrow toward commodity DRAM as supply catches up, which lifts Micron's blended ASP without a share gain.
CXMT is the structural threat, not the cyclical one. At 10% of DRAM and pushing server DDR5, CXMT adds supply where pricing is richest. Its cost base is state-supported, so it will not exit at the trough the way a listed producer would — which is precisely what our bear case capitalises.
8. Management, governance and capital allocation
Management has beaten its own guide four quarters running and spent five years ploughing cash into fabs
Sanjay Mehrotra is chief executive and has been chair since January 2025, Lynn Dugle is lead independent director, and Mark Murphy has been chief financial officer since 2022. On the record we can test, they deliver: the last four quarters beat the top of guidance on revenue, and the last two beat on gross margin and EPS as well. Capital has gone into capacity, not into the share count.
Source: Company filings; 10-K FY22-FY26 (five-year sums, FY25-26 capex net of incentives). $bn, linear scale
| Five years to FY26A | US$bn | Share of CFO |
|---|---|---|
| Operating cash flow | 132.4 | 100% |
| Capex (net of incentives, FY25-26) | 69.3 | 52% |
| Buybacks | 3.8 | 3% |
| Dividends | 2.6 | 2% |
| Retained for capacity and cash | 56.7 | 43% |
| Guidance record, Q4 FY26 | ||
| Revenue vs guide, Q4 | 54.2 vs 49.0-51.0 | +8.5% |
| EPS vs guide, Q4 | $33.42 vs $30-32 | +7.8% |
| Q1-Q3 FY26 revenue | Each above the top of guidance | |
| Returned from 9 December 2026 | ||
| Buyback authorization | 35.16 | New |
| Dividend, annualised | 0.60 | 0.1% yield |
Source: 10-K FY22-FY26; Q4 FY26 release and call (30 Sep 2026); Zacks. FY25-26 capex is net of incentives. Shares of operating cash flow are five-year sums.
The record is delivery, not promises. Four consecutive quarters beat the top of revenue guidance and the Q4 FY26 print beat on revenue, gross margin and EPS together. The FY27 guide extends the pattern: record revenue, sequential growth every quarter, and the Q1 gross margin of 86.25% set explicitly as the floor for the year.
Capital allocation has been pro-cyclical and is only now turning. Of $132.4bn of operating cash flow over five years, $69.3bn went into capex and $6.4bn into buybacks and dividends. The $35.16bn authorization that starts on 9 December 2026 is a new return commitment, and management's stated intent is to return 100% of excess cash over time.
Governance: the chair and chief executive are the same person, with a lead independent director. Mehrotra holds both roles, which concentrates accountability in a capital-intensive business where the board's main job is to police capacity spending. The counterweight is a balance sheet with $55.4bn of net cash and a BBB+ rating from all three agencies, so the capacity plan is not funded with leverage.
Insider activity is routine and small. The chief technology officer, Scott DeBoer, sold 28,444 shares at $900 on 16 October 2026 under a pre-arranged 10b5-1 plan and still holds 147,430 shares; there have been no open-market purchases.
9. What moves the stock
The stock trades on DRAM price and margin, not the contract book
We tested five candidate drivers against four tests: the move around the last eight results, the split of the twelve-month return, the variance explained by market, sector and peers, and the topics raised on the Q4 FY26 call. Two pass most tests; the contract floor, which carries most of our value, fails every price test today.
| Driver | What it does to the stock | Tests passed | Market vs our model | Debate | Next number to watch |
|---|---|---|---|---|---|
| DRAM contract price | Moves revenue and margin together; r = 0.88 with the average share price over eight quarters | 4 of 4 | Price-implied $86bn of mid-cycle NOPAT (37.5% EBIT margin); base $139bn (62.0%) | 1, 3 | 4Q26 contracts, +10-15% q/q (TrendForce), Dec 2026 |
| Non-GAAP gross margin | The line the market reprices on: the Q3 FY26 margin beat drove +15.4% | 3 of 4 | 86.25% Q1 guide against our 86.25%; FY27E 87.1% | 1 | Q1 FY27 print, Dec 2026 |
| Capex intensity | Sets free cash flow; explains the March 2026 drawdown | 2 of 4 | 21% of FY26 revenue (net); 20% in FY27E at $55bn | 2 | FY27 capex, Q1 and Q2 prints |
| SCA floor price | Sets the trough margin, so most of our value; no reaction has tested it | 1 of 4 | No SCA has yet been tested by a downturn | 3 | New SCA at the Q2 FY27 print, March 2027 |
Source: Company filings; exchanges; TrendForce; Penmark Research estimates. Tests passed out of four: results-day reaction, return decomposition, factor sensitivity, call Q&A.
Source: Exchanges; company filings. Close-to-close move from the session before the release to one session after
Source: Company filings; exchanges. Price at each date over trailing non-GAAP EPS; EV/EBITDA and P/B are not shown because the model holds only FY25 balance-sheet items at the earlier date
Source: Penmark Research model, Valuation and Market Check tabs. Mid-cycle NOPAT FY29E-FY36E; price-implied solved so the model reproduces the $659 price. Compared in NOPAT, not margin, because the price-implied 37.5% EBIT margin sits on base-like revenue of $271bn and the bear's 43.9% on $165bn
9. What moves the stock (continued)
Three drivers explain the price; the fourth is $134 a share of the gap and is untested
Source: Exchanges. 103 weekly returns to 5 Nov 2026; separate regressions of Micron on each factor
Source: Q4 FY26 earnings call Q&A (30 Sep 2026); five analysts, eight topics
Driver 1. DRAM contract price — the stock is a price instrument, and the price is still rising
What the market does: a 60% sequential DRAM ASP rise in Q3 FY26 produced the 176% twelve-month return, and the correlation between average quarterly share price and gross margin is 0.88 across eight quarters. Sector beta explains 58% of weekly variance and memory peers 47%, so Micron trades as a price proxy, not as a company.
What our model says: FY27E DRAM revenue of $205.5bn comes from that quarterly build. Built quarterly on ASP of +12%, +8%, +3% and +1% and bits of +1%, +3%, +4% and +5%, the year compounds to +104% DRAM revenue. In the model, FY28E DRAM ASP +5% against flat is worth +$37.84 a share and the CXMT server DRAM scenario −$29.34.
Debate 1 and 3. Next number: 4Q26 DRAM contract settlement, guided +10-15% q/q (TrendForce), December 2026. We expect +12%; the eight-event study is too small to size a price reaction.
Driver 2. Gross margin — the single line that reprices the equity
What the market does: the Q3 FY26 print (84.9% gross margin against 78% expected) drove +15.4% in two sessions, the largest reaction in eight quarters; the Q1 FY27 guide of 86.25% is now treated as a floor by the sell side.
What our model says: 87.1% for FY27E and 86.4% in FY28E, with cost per bit down 6% and a 68% incremental margin. The model values HBM repricing at +$18.19 a share and NAND prices levelling at −$26.06.
Debate 1. Next number: Q1 FY27 gross margin, December 2026: guide 86.0-86.5%, our model 86.25%, consensus 85.7%. A miss below 85% breaks the margin narrative and the stock.
Driver 3. The contract floor — $134 a share of the base-to-price gap
What the market does: nothing yet. No SCA has been tested by a downturn, only one analyst asked about coverage, and the price implies a 37.5% mid-cycle EBIT margin — the FY18 prior-cycle peak was 49.3% — against our base mid-cycle 62.0%.
What our model says: the floor path (DRAM ASP −20/−40/−25% in FY29E-31E against the bear path) is worth +$134.44 a share, 53% of the gap between our base value and the price.
Debate 3. Next number: the Q2 FY27 print in March 2027, when the first SCA renegotiation is disclosed: coverage above 40% of revenue confirms, a renegotiation at spot prices breaks it.
| Candidates tested and dropped | Why the market assumes it matters | Why it fails our tests |
|---|---|---|
| HBM market share | The fastest-growing product line | Share is 18% and has been stable; no results-day reaction to HBM share news in eight quarters |
| Hyperscaler capex | Sets memory demand | Explains demand, not the price; adds no variance once the sector factor is included |
| Buyback authorization | $35.16bn from December 2026 | Not yet in force; the model treats it as value-neutral |
| PC and phone units | The traditional memory demand driver | Client is 27% of revenue; unit growth has been flat while revenue tripled |
Source: Company filings; exchanges; S&P Global; TrendForce; Penmark Research estimates. Limits: the event study covers eight observations, so single-quarter reactions carry wide error bands, and no SCA has yet been tested by a downturn.
10. Thesis chapter 1 | Driver: DRAM contract price
Pricing is contracted now, and the contracts were signed at the bottom of the cycle
2. What the market believes
Consensus has FY27E revenue at $275.5bn against our $276.1bn — a 0.2% difference — and FY28E at $319.7bn against our $343.9bn, so the disagreement is entirely about FY28 and beyond (+7.6%). The price is consistent with $86bn of mid-cycle NOPAT, a 37.5% EBIT margin on base-like revenue; FY18 was 49.3% and only FY26 has been higher in the reported record we hold. Our base is 62.0%.
In other words the market accepts the FY27 price path and treats FY28 as the peak before a normal memory collapse. It does not give Micron credit for the mechanism that would make the collapse shallower.
4. What it is worth
The debate values are −$4.93 a share on FY27E exit pricing, +$6.10 on CY28 pricing and +$251.5 on the depth of the FY29E-FY31E downturn. FY28E DRAM ASP +5% against flat is worth $37.84.
3. The evidence, three independent sources
- Filings. The FY26 10-K discloses 26 supply agreements covering more than 35% of revenue to 2030 at floor prices, with $12.7bn of cash deposits received and about $150bn of remaining performance obligations at minimum prices.
- Pricing data. DRAM ASP rose 60% sequentially in Q3 FY26 and NAND 80%; 4Q26 contracts settled a further 10-15% higher (TrendForce). Price per bit is now 3.4x its FY25 average.
- Guidance. Q1 FY27 revenue is guided to $60-63bn with an 86.0-86.5% gross margin, and management called FY27 "another record year" on 30 September 2026. Guidance has been beaten at the top end four quarters running.
- Industry structure. Three suppliers hold 87% of DRAM and CY27 industry bit growth is guided to the low-20s, below the demand implied by 68% memory content in 2027 cloud capex.
Source: Company filings (A); Penmark Research estimates (E), Assumptions tab. Q4 FY26 on a revenue-per-bit basis
| Signposts | Date | Thesis confirmed if | Thesis broken if |
|---|---|---|---|
| 4Q26 DRAM contract settlement | Dec 2026 | +10% or more q/q | Flat or down |
| Q1 FY27 results | Dec 2026 | Revenue above $63bn and margin at or above 86.5% | Margin below 85% |
| Q2 FY27 SCA disclosure | Mar 2027 | Coverage rises above 40% of revenue | An SCA renegotiated to spot |
Source: Company filings; Penmark Research estimates. Dates are expected reporting dates, not company-confirmed.
11. Thesis chapter 2 | Driver: capex and supply timing
No meaningful new supply before late 2028, so FY28 is a growth year and not the peak
2. What the market believes
Consensus has FY27E EPS at $176.50 and FY28E at $207.80, a 17.7% increase, then assumes a fall. Our FY27E is $171.21, 3.0% below consensus, and our FY28E is $215.36, 3.6% above it — so we are more cautious on FY27 and more constructive on FY28.
The gap is not about demand. It is about the date supply arrives: consensus treats FY28 as the first year of oversupply, we treat it as the last year of tightness.
4. What it is worth
Our FY28E EPS is $7.56 above consensus. Debate 2 puts that at $6.10 a share against the market's path; FY28E DRAM ASP +5% against flat is worth $37.84. The supply calendar also protects FY29E: we model a 5% revenue fall that year, not the 30% a normal downcycle would deliver.
3. The evidence, three independent sources
- Disclosed capacity calendar. Singapore HBM packaging in early CY27, ID1 and Tongluo cleanrooms in mid-CY27, ID2 and Hiroshima in late CY28, Singapore NAND in 2H CY28 and New York in CY30 — a gap in commodity bit capacity until late CY28.
- Capex versus depreciation. FY27E capex of $55bn is about 3x depreciation, but 40% of it is HBM packaging, EUV and US greenfield — spending that adds value per wafer, not wafers. Industry bit growth is guided to the low-20s for CY27 and CY28.
- Sold-out position. Management has said more than 75% of FY27 output is committed and that FY27 will grow every quarter. A supplier with that visibility does not cut price to fill capacity.
- Behaviour of the other two. Samsung and SK hynix are prioritising HBM capacity over commodity DRAM, which removes commodity bits even as total wafer capacity rises.
Source: Company filings; Q4 FY26 call (30 Sep 2026). Announced capacity additions by first output
| Signposts | Date | Thesis confirmed if | Thesis broken if |
|---|---|---|---|
| Singapore HBM packaging output | 1Q CY27 | Starts on schedule and is HBM-only | Converted to commodity DRAM |
| Samsung and SK hynix capex guidance | Jan 2027 | 2027 capex growth below 10% | Above 25%, led by commodity DRAM |
| ID1 and Tongluo cleanroom outfitting | Mid CY27 | Tool move-in slips to CY28 | Wafer starts pulled into 2H CY27 |
Source: Company filings; Penmark Research estimates. Dates are our expectations from disclosures, not company-confirmed schedules.
12. Thesis chapter 3 | Driver: the SCA floor price
The base holds the next trough at a 55% EBIT margin; the floor path is worth $134 a share
2. What the market believes
The price is consistent with $86bn of mid-cycle NOPAT, a 37.5% EBIT margin, against our base of $139bn (62.0%) and a bear of $61bn (43.9%). In effect the market pays for a normal memory cycle with a normal trough, like FY23.
The 37.5% price-implied margin sits below the 49.3% operating margin Micron earned at the FY18 prior-cycle peak.
4. What it is worth
The floor price path against the bear path (DRAM ASP −20/−40/−25% in FY29E-31E) is worth $134.44 a share, 53% of the difference between our $911 base value and the price.
3. The evidence, three independent sources
- The contracts themselves. 26 SCAs covering over 35% of revenue to 2030 at floor prices, $32bn of customer commitments, $12.7bn of cash deposits and about $150bn of RPO at minimum prices. The deposits make walking away expensive for the buyer, not just the seller.
- Reported behaviour in the upcycle. With more than 75% of FY27 output committed, Micron has stopped selling on spot; the FY26 gross margin of 81.1% was earned with a contract book in place, not in a spot market.
- The historical counterfactual. In FY23, with no contract book, the operating margin fell to −37% and the company lost $5.8bn. Our FY29E-FY31E margins slide from 78% to 55%, and the bear case takes the trough to 44%.
- Where the risk sits. The floor covers DRAM sold to data-centre customers, not NAND and not client. The NAND cleanroom bear item is worth −$61.62 a share precisely because NAND has no equivalent floor.
Source: Company filings (A); Penmark Research estimates (E). Operating margin, GAAP
Source: Company filings; Penmark Research estimates. Market-implied margin solved from the $658.70 close
| Signposts | Date | Thesis confirmed if | Thesis broken if |
|---|---|---|---|
| First SCA renegotiation disclosed | Mar 2027 | Renewed at or above the original floor | Repriced to spot |
| CXMT server DDR5 qualification | 1H CY27 | Confined to China OEMs | Qualified by a US hyperscaler |
Source: Company filings; TrendForce; Penmark Research estimates. Dates are expected, not company-confirmed.
13. Forecasts
Revenue doubles in FY27 on the level of price, then falls 30% in FY30 when the cycle turns
We build revenue as bits × price for DRAM and NAND, with cash cost per bit falling 6% a year. FY27E revenue of $276.1bn doubles because the full year carries prices that only arrived in the second half of FY26, which is why the driver rows below begin in FY28E.
| Base case | FY26A | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|
| DRAM bit shipments, growth | — | Quarterly | +20% | +18% | +15% |
| DRAM ASP, change | — | Quarterly | +5% | −20% | −40% |
| NAND bit shipments, growth | — | Quarterly | +22% | +20% | +18% |
| NAND ASP, change | — | Quarterly | −1% | −20% | −40% |
| Cash cost per bit, change | −10% | −6% | −6% | −6% | −6% |
| Revenue, $m | 133,188 | 276,057 | 343,907 | 326,153 | 227,095 |
| Growth | +256% | +107% | +25% | −5% | −30% |
| Gross margin, non-GAAP | 81.1% | 87.1% | 86.4% | 82.2% | 70.3% |
| EBITDA, $m | 109,344 | 240,679 | 304,421 | 282,943 | 180,565 |
| Depreciation and amortisation, $m | 9,503 | 11,396 | 19,245 | 27,481 | 34,234 |
| Capex, net, $m | 27,367 | 55,000 | 65,000 | 65,000 | 63,587 |
| Free cash flow to the firm, $m | 62,308 | 132,137 | 185,720 | 180,832 | 108,165 |
Source: Company filings (A); Penmark Research estimates (E). FY27E is built quarter by quarter (DRAM ASP +12%, +8%, +3%, +1%; bits +1%, +3%, +4%, +5%), so no single annual change describes it. FY26A EBITDA and operating expense are on the non-GAAP basis.
Source: Company filings (A); Penmark Research estimates (E). Quarterly revenue, $bn
The path is a plateau, not a spike. FY27E revenue grows 5-13% a quarter and reaches $75.5bn in Q4 FY27E. FY28E adds 25% on bit growth of 18-20% before DRAM ASP falls 20% in FY29E, and the decline comes in FY29E-FY31E, when base DRAM ASP falls 20%, 40% and 25%.
Costs: fixed, depreciated and slow to fall
Opex grows from $6.8bn to $11.1bn over four years, a 13% compound rate on R&D. Depreciation triples to $34.2bn as the $185bn of FY27E-FY29E capex is put in service, which is why the gross margin falls to 70.3% in FY30E even with cost per bit down 6% a year. The trough is a price event, not a cost event.
13. Forecasts (continued)
We are 3.0% below consensus on FY27 EPS and 3.6% above on FY28: the argument is timing
Our revenue is within 0.2% of consensus in FY27 and 7.6% above in FY28, and our operating income is within 0.4% of consensus — yet our FY27 EPS is 3.0% lower. The gap sits below EBIT: we model a 15.5% tax rate and FY27E net interest of $1.9bn, held at $475m a quarter while cash grows from $73bn to $208bn. Stock compensation is not the cause, because non-GAAP adds it back. The disagreement that matters is FY28.
| US$m unless stated | Penmark | Consensus | Diff. |
|---|---|---|---|
| FY27E | |||
| Revenue | 276,057 | 275,510 | +0.2% |
| Operating income (EBIT) | 229,283 | 230,300 | −0.4% |
| EPS, non-GAAP ($) | 171.21 | 176.50 | −3.0% |
| FY28E | |||
| Revenue | 343,907 | 319,680 | +7.6% |
| EPS, non-GAAP ($) | 215.36 | 207.80 | +3.6% |
| FY29E EPS ($) | 197.83 | — | — |
Source: Company filings; S&P Global consensus, 29 Oct 2026 (41 estimates for FY27E EPS; 10 for Q1 EPS); Penmark Research estimates.
| Q1 FY27E | Guidance | Model | Consensus |
|---|---|---|---|
| Revenue, $m | 60,000-63,000 | 61,450 | 61,830 |
| Gross margin | 86.0-86.5% | 86.25% | 85.7% |
| Opex, $m | 2,000-2,120 | 2,060 | 2,040 |
| EPS, non-GAAP ($) | 37.15-39.15 | 37.78 | 38.20 |
| Diluted shares, m | 1,140-1,160 | 1,150 | 1,150 |
| Tax rate | 15-16% | 15.5% | 15.0% |
| Capex, $m | 11,000-12,000 | 11,500 | 11,400 |
Source: Q4 FY26 call and guidance (30 Sep 2026); S&P Global consensus (29 Oct 2026); Yahoo Finance and Zacks for Q1 (14 and 28 Oct 2026); Penmark Research estimates.
Source: Penmark Research model, Thesis tab. Value per share against the base case
Cash: $592bn by FY29E, and the model does not spend it. Free cash flow to the firm totals $499bn over FY27E-FY29E against $185bn of capex, taking cash and investments from $73.5bn to $592.4bn. We hold the share count at 1,150m; the model leaves out the $35.16bn authorization and treats it as value-neutral.
Balance sheet: net cash of $55.4bn after customer deposits. Debt stays near $5.2bn, interest income reaches $21.2bn by FY31E, and working capital absorbs $18.0bn in FY27E before releasing $2.5bn and $13.9bn in FY29E and FY30E as receivables shrink with revenue.
The sensitivities are asymmetric. The floor path, worth $134.44, is larger than any single bear item; the five bear items sum to −$151.91, and the bear case sits $404 below the base because the items compound.
14. Valuation
Two methods, 4.9% apart: $918 a share at the valuation date, $1,085 in twelve months
We value Micron on mid-cycle earnings because a single forecast year is meaningless when EPS ranges from $171 to $99 across four years. Averaging FY29E-FY36E NOPAT at a 6.9x multiple and adding the cash the company will hold gives $918.27; a full-cycle DCF gives $875.76.
| Build, US$bn unless stated | Value |
|---|---|
| Probability-weighted mid-cycle EBIT margin, FY29E-FY36E | 62.6% |
| Probability-weighted mid-cycle NOPAT | 143.3 |
| Justified multiple on mid-cycle NOPAT | 6.90x |
| Value of operations at mid-cycle | 988.9 |
| Discounted at the 15.3% WACC | 743.6 |
| PV of FY27E-FY28E free cash flow | 257.0 |
| Net cash after SCA deposits | 55.4 |
| Diluted shares | 1,150m |
| Value per share at the 3 September 2026 valuation date | $918.27 |
| Rolled forward 14 months at the 15.4% cost of equity | $1,085.40 |
Source: Penmark Research model. Mid-cycle is the weighted average of FY29E-FY36E; the justified multiple is (1 − g/RONIC) ÷ (WACC − g) = 6.90x.
| DCF, FY27E-FY36E | Value |
|---|---|
| Risk-free rate and equity risk premium | 4.9% / 4.23% |
| Beta, midpoint of own 2-year (2.66x) and peer median (2.30x) | 2.48x |
| Cost of equity | 15.39% |
| After-tax cost of debt, 0.7% weight | 4.7% |
| WACC | 15.32% |
| Terminal growth / return on new capital | 3.0% / 20% |
| Terminal value, share of enterprise value | 27.7% |
| Value per share | $875.76 |
| Gap to the primary method | −4.9% |
Source: Penmark Research model. Free cash flow to the firm FY27E-FY36E; terminal value on a RONIC-consistent fade.
| US$ per share | DCF | Primary | Weighted | Price, 5 Nov 2026 | Target, 5 Nov 2027 |
|---|---|---|---|---|---|
| Value | 875.76 | 918.27 | 918.27 | 658.70 | 1,085.40 |
| Return from the 5 Nov price | +32.9% | +39.4% | +39.4% | — | +64.8% |
| Implied FY27E P/E | 5.1x | 5.4x | 5.4x | 3.8x | 6.3x |
Source: Penmark Research model; exchanges. P/E on FY27E non-GAAP EPS of $171.21.
Why mid-cycle earnings, and why 6.9x. Our EPS falls 54% from FY28E to FY31E, so a forward multiple on any one year is a bet on cycle timing. Averaging eight years removes the timing call and leaves the franchise. With zero growth, a 15.3% cost of capital justifies 6.5x. We use 6.9x, which is (1 − 3%/20%) ÷ (15.32% − 3%): the growth that a 20% return on new capital supports, and nothing more.
Why the DCF corroborates rather than competes. The DCF uses the model's own cash flows to FY36E, a 3% terminal growth rate and a 20% return on new capital, and lands 4.9% below the primary method — inside the 15% tolerance, so the two methods agree on the answer. We would worry above a 75% terminal value share; it is 27.7%.
14. Valuation (continued)
Two of three cases clear the price; the bear is 23% below it
The scenarios are built from the drivers in section 9, not from a multiple range. Bear assumes the floor fails at CXMT's price and DRAM ASP falls 70% from the peak by FY31E; base assumes it holds, with a 62.3% trough gross margin and a 55% EBIT margin in FY31E; bull assumes DRAM stays tight into CY28 and the trough margin is 73.9%. The bear value of $507.25 is 23% below the current price, so the case for Buy rests on the 65% weight in base and bull, not on all three outcomes.
Source: Penmark Research model. Probability-weighted value; price at 5 November 2026
Source: Penmark Research model. Bridge from the share price to the base-case value, $ per share
| Scenario | Prob. | FY27E EPS | FY28E EPS | Trough gross | Mid-cycle EBIT | Value | Return |
|---|---|---|---|---|---|---|---|
| Bear: floors fail, CXMT | 35% | $152.15 | $158.48 | 43.8% | 43.9% | $507.25 | −23.0% |
| Base: floors hold | 45% | $171.21 | $215.36 | 62.3% | 62.0% | $911.39 | +38.4% |
| Bull: CY28 tighter | 20% | $181.63 | $241.91 | 73.9% | 74.8% | $1,653.04 | +151.0% |
| Probability-weighted | 100% | — | — | — | — | $918.27 | +39.4% |
Source: Penmark Research model, Scenarios tab. Value at the 3 September 2026 valuation date; return to the 5 November 2026 price of $658.70, before the roll-forward. Trough is the lowest non-GAAP gross margin in FY29E-FY36E.
What the price implies. Reverse-engineering the model at $658.70 gives $86bn of mid-cycle NOPAT at the model multiple, a 37.5% EBIT margin on base-like revenue of $271bn. That cannot be set beside the bear's 43.9% on $165bn, so we compare NOPAT: the price ($86bn) sits between the bear ($61bn) and the base ($139bn), closer to the bear. The FY18 prior-cycle peak margin was 49.3%.
Why we weight the bear at 35%. Two of the five bear items are live today: CXMT is shipping 10% of DRAM and NAND cleanroom additions are scheduled for 2H CY28. The probability is a judgement, not a measurement; the Exhibit 15 grid shows value at other multiples and earnings levels.
Reconciliation to page 1. The $918.27 value at 3 September 2026 rolls forward 14 months at the 15.4% cost of equity, less dividends, to $1,085.40: the 12-month target, 64.8% above the price. With the 0.1% dividend yield the expected total return is 64.9%, above the +10% Buy threshold.
What would change the rating. Two things: a bear-case outcome, with the mid-cycle margin near 44%, if a CXMT price war overwhelms the contracts, or a share price above $987, where the expected return to target falls under 10%. Neither is our base case, and the second is 50% above today's price.
14. Valuation (continued)
The base value rests on FY29E-FY36E: 78% EBIT margin in FY29E, a 55% trough in FY31E, then 58-60%
The central judgement is how long the contracts last. If the agreements end in calendar 2030, they cover at most one quarter of the FY31E trough year and none of FY32E-FY36E. Our base holds FY32E-FY36E EBIT margin at 58-60% because price and cost per bit both fall about 6% a year.
Source: Penmark Research model, Forecast tab (EBIT after stock compensation / revenue). Mid-cycle averages FY29E-FY36E: bear 43.9%, base 62.0%, bull 74.8%
Source: Penmark Research model, Forecast tab; NOPAT in $bn. Mid-cycle averages FY29E-FY36E: bear $61bn, base $139bn, bull $297bn
| FY to Aug | FY27E | FY28E | FY29E | FY30E | FY31E | FY32E | FY33E | FY34E | FY35E | FY36E |
|---|---|---|---|---|---|---|---|---|---|---|
| Base case: floors hold | ||||||||||
| EBIT margin | 83.1% | 82.9% | 78.3% | 64.4% | 55.4% | 58.4% | 59.6% | 59.0% | 58.5% | 58.2% |
| NOPAT, $bn | 193.7 | 241.0 | 215.9 | 123.6 | 93.9 | 114.8 | 129.9 | 136.4 | 143.5 | 151.4 |
| Bear case | ||||||||||
| EBIT margin | 81.5% | 78.2% | 62.4% | 33.6% | 32.7% | 41.0% | 44.2% | 43.7% | 43.3% | 42.8% |
| NOPAT, $bn | 171.8 | 176.0 | 97.8 | 34.3 | 34.7 | 52.1 | 63.2 | 66.3 | 69.6 | 73.0 |
| Bull case | ||||||||||
| EBIT margin | 84.1% | 84.7% | 84.2% | 78.9% | 73.0% | 74.7% | 74.6% | 73.0% | 71.8% | 70.9% |
| NOPAT, $bn | 205.7 | 271.3 | 315.5 | 252.6 | 219.4 | 272.8 | 309.5 | 321.1 | 334.7 | 350.2 |
Source: Penmark Research model, Forecast tab. EBIT after stock compensation; NOPAT after tax. Mid-cycle is the FY29E-FY36E average: EBIT margin 43.9% bear, 62.0% base, 74.8% bull; NOPAT $61bn, $139bn, $297bn.
Four names for four margins. Price-implied: 37.5% EBIT, $86bn of NOPAT. Base mid-cycle, floors hold: 62.0%. Base trough, FY31E: 55% EBIT, 62.3% gross. Probability-weighted mid-cycle: 62.6%, valuation only, as it includes the bull case. In NOPAT the price sits between the bear ($61bn) and base ($139bn), closer to the bear.
What the contracts are. 26 supply agreements cover over 35% of revenue to 2030, with about 50% targeted. Remaining performance obligations are about $150bn at minimum prices and customer commitments $32bn. Customers have paid $12.7bn of cash deposits; the model deducts the $12.9bn contract liability in net cash.
What the base already contains. All five bull catalysts and none of the bear ones. From the base, $215 a share depends on the bulls landing and another $152 is at risk from the bears; the only upside is the bull case. The floor path alone is $134 a share, 53% of the gap between base value and the price.
Price your own view. Each 10% taken off mid-cycle earnings costs about $65 a share (Exhibit 15).
15. Catalysts
Four dated events can move the rating in the next twelve months, and three are print dates
The calendar is unusually dense because the contract book resets on a schedule. The Q1 FY27 print in December carries the first full quarter of 86% gross margin; the Q2 print in March is the first test of an SCA renewal; the Q4 print in September sets FY28 guidance against a supply timetable that does not loosen until late CY28.
| Event | Date | Penmark expectation | Consensus or guide | Value at stake | Test |
|---|---|---|---|---|---|
| Q1 FY27 results and 4Q26 contracts | 17 Dec 2026 | Revenue $61.5bn, margin 86.25%, EPS $37.78 | Guide $60-63bn; consensus $61.8bn, $38.20 | ±4% on print, ±10% on guide | Yes |
| Q2 FY27 and first SCA renewal | 25 Mar 2027 | Coverage rises above 40% of revenue | No coverage estimate | ±$134 a share on the floor path | Yes |
| Singapore HBM packaging output | 1Q CY27 | On schedule, HBM only | Management target | ±$8 a share | No |
| Q3 FY27 and HBM4 second customer | 24 Jun 2027 | One additional qualified customer | Consensus: two | ±$18 a share | No |
| Q4 FY27, FY28 guide, CY28 contracts | 28 Sep 2027 | FY28 revenue $343.9bn, EPS $215.36 | Consensus $319.7bn, $207.80 | ±$38 a share on FY28E DRAM ASP | Yes |
| ID1 and Tongluo cleanrooms | Mid CY27 | Tool move-in slips into CY28 | No consensus | −$62 a share if pulled forward | Yes |
Source: Company filings; Q4 FY26 guidance and call (30 Sep 2026); S&P Global consensus (29 Oct 2026); Yahoo Finance and Zacks for Q1; Penmark Research estimates. Value at stake from the model's catalyst table (Exhibit 54): floor path $134, FY28E DRAM ASP $38, HBM $18, NAND cleanrooms −$62.
Source: Company filings; Q4 FY26 call (30 Sep 2026); Penmark Research. Expected reporting dates, not company-confirmed
The next number to watch for each driver. DRAM price: the 4Q26 settlement in December, guided +10-15% q/q by TrendForce, our base +12%. Gross margin: the Q1 FY27 print, guide 86.0-86.5%, consensus 85.7%. Capex: Q1 FY27 capex, guided $11-12bn, our model $11.5bn, with $55bn for FY27E. Contract floor: the first SCA renewal disclosed in March 2027.
What would change the rating. The call is robust to the discount rate: at +1pt the target is $1,024 and the return 55.5%; on the base case alone it is $1,077 and 63.6%. It would move to Hold only if the share price rose above $987, where the expected return to target falls below 10%, or if a contract signed below the floor moved weight to the $507 bear value.
16. Risks
The risk that matters is the floor, and the strongest evidence against us is the sell side's own optimism
Downside risks to a Buy rating, ranked by value at stake. The floor failing ends the thesis; the other risks shave value without changing the rating. All are measured against the base-case value of $911.39, and none is in the base.
| Risk | Likelihood | How it reaches the model | Value | First visible in | Effect on base value |
|---|---|---|---|---|---|
| Floor fails: DRAM moves to the bear price path | Medium | DRAM ASP follows the bear path (−70% to FY31E); the floor-path item is $134 | −$134 | Q2 FY27 renewal, Mar 2027 | Base value to $777 |
| NAND cleanroom additions from 2H CY28 | Medium | NAND ASP falls 20pp faster from FY29E | −$61.62 | Supplier capex, 1H CY27 | Base value to $850 |
| CXMT server DRAM at scale | Medium | FY29E DRAM ASP −40% vs −20% in base | −$29.34 | CXMT qualification, 1H CY27 | Base value to $882 |
| NAND contract prices level from Q2 FY27 | Medium | NAND ASP flat rather than +3% in FY27E | −$26.06 | Q1 FY27 print, Dec 2026 | Base value to $885 |
| Slower cost-per-bit decline | Medium | Cost per bit −4% not −6% a year | −$20.93 | FY27 gross margin walk | Base value to $890 |
| AI data-centre build slows, HBM despec | Medium | HBM bits per GPU cut; mix shifts to commodity | −$13.96 | Hyperscaler capex, Jan 2027 | Base value to $897 |
Source: Penmark Research model, Thesis tab and bear case. Value per share against the $911.39 base; likelihood is the team's judgement, not a modelled probability.
What would end the thesis
An SCA renegotiated at spot prices. The entire bull case rests on the idea that a contract signed with cash on deposit survives a downturn; if the first renewal in March 2027 resets to market, the floor is a marketing document and the stock is a cyclical priced on a bear-like trough.
The strongest evidence against the call
The sell side is already bullish: 45 Buy, 3 Hold and 1 Sell, with an average target of $1,562 — 137% above the price. If a bull case this crowded exists, the argument that the market is missing the floor is weaker than it looks. Our answer is that consensus and our FY27E EPS ($176.50 against $171.21) agree; the disagreement is the depth of the FY29E-FY31E trough, where the price implies $86bn of mid-cycle NOPAT against $139bn in our base.
The precedent we cannot dismiss
Memory supply agreements have a history of being renegotiated when spot prices fall, and the FY23 loss of $5.8bn came with no contract book to cushion it. What is different this time is the cash: $12.7bn of deposits makes a customer's exit a balance-sheet decision, not a procurement one. If that distinction fails, it fails here.
Other risks
Key-person risk on Sanjay Mehrotra, who is chair and chief executive; Lynn Dugle is lead independent director. Construction delays at ID1, ID2 and New York would push out supply, which helps price but raises capex per wafer.
17. Ownership and trading
The Street is already long: 45 of 49 ratings are Buy, with an average target 44% above ours
Micron is a $745bn equity and a sell side 45 to 1 positive. The ownership risk is not a squeeze; it is that a bullish consensus leaves little room for a miss.
Source: S&P Global, October 2026; Penmark Research estimates
Source: S&P Global; broker ratings and targets, October 2026
| Trading and ownership | Value | Read |
|---|---|---|
| Market capitalisation | $745.0bn | 1,131m shares at $658.70 |
| Consensus FY27E / FY28E EPS | $176.50 / $207.80 | Penmark $171.21 / $215.36 |
| Street ratings | 45 Buy, 3 Hold, 1 Sell | Average target $1,562, 137% above the price |
| Index membership | S&P 500, Nasdaq-100, SOX | 0.1% dividend yield |
Source: Exchanges; S&P Global (29 Oct 2026); Penmark Research. Market capitalisation at the 5 November 2026 close.
Source: Company filings (A); Penmark Research estimates (E); S&P Global consensus, 29 Oct 2026 (41 estimates for FY27E EPS)
Insider activity, last two filings
| Date | Insider | Transaction | Value |
|---|---|---|---|
| 16 Oct 2026 | S. DeBoer, CTO | Sale of 28,444 shares at $900 | $25.6m |
| 16 Oct 2026 | S. DeBoer, CTO | Holdings after the sale | 147,430 shares |
Source: Form 4 filed 20 October 2026; Signals. The sale was made under a pre-arranged 10b5-1 plan.
The insider signal is neutral. A 10b5-1 sale of 16% of a holding at $900, 37% above today's price, is scheduled diversification. No executive has bought in the open market in twelve months, which is the more telling absence.
The crowding risk is not in the short interest. With 45 of 49 analysts positive, the marginal buyer is convinced. Our target is 31% below the Street's average because we value mid-cycle earnings at 6.9x, not FY27E earnings at a peer multiple.
18. Financial model | Appendix A
Five reported years that bracket the cycle, and five forecasts that repeat it
The reported record shows what the franchise looks like at both ends: a $5.8bn loss in FY23 and $85.0bn of net income three years later. The forecasts apply the same price mechanism with the contract book in place.
| US$m unless stated | FY22A | FY23A | FY24A | FY25A | FY26A |
|---|---|---|---|---|---|
| Revenue | |||||
| DRAM | 22,386 | 10,978 | 17,603 | 28,578 | 100,679 |
| NAND | 7,811 | 4,206 | 7,227 | 8,503 | 31,785 |
| Other | 561 | 356 | 281 | 297 | 724 |
| Revenue | 30,758 | 15,540 | 25,111 | 37,378 | 133,188 |
| Growth | +11% | −49% | +62% | +49% | +256% |
| Costs | |||||
| Cost of goods sold | 16,860 | 16,956 | 19,498 | 22,505 | 25,684 |
| Gross profit | 13,898 | −1,416 | 5,613 | 14,873 | 107,504 |
| Gross margin | 45.2% | −9.1% | 22.4% | 39.8% | 80.7% |
| Research and development | 3,116 | 3,114 | 3,430 | 3,798 | 5,650 |
| Selling, general and administration | 1,066 | 920 | 1,129 | 1,205 | 1,947 |
| Other operating income and expense | 14 | 295 | −250 | 100 | 567 |
| Profit | |||||
| Operating income | 9,702 | −5,745 | 1,304 | 9,770 | 99,340 |
| Operating margin | 31.5% | −37.0% | 5.2% | 26.1% | 74.6% |
| Depreciation and amortisation | 7,116 | 7,756 | 7,780 | 8,352 | 9,503 |
| EBITDA | 16,818 | 2,011 | 9,084 | 18,122 | 108,843 |
| Net income | 8,687 | −5,833 | 778 | 8,539 | 84,969 |
| Per share | |||||
| Diluted shares, m | 1,122 | 1,093 | 1,118 | 1,125 | 1,143 |
| Diluted EPS, GAAP | $7.75 | −$5.34 | $0.70 | $7.59 | $74.33 |
| Diluted EPS, non-GAAP | — | — | — | $8.29 | $75.52 |
Source: SEC filings via Signals; Q4 FY26 results release (30 Sep 2026). GAAP unless marked. Non-GAAP EPS not shown for FY22-FY24.
| US$m unless stated | FY26A | FY27E | FY28E | FY29E | FY30E | FY31E |
|---|---|---|---|---|---|---|
| DRAM revenue | 100,679 | 205,496 | 258,925 | 244,425 | 168,653 | 145,463 |
| NAND revenue | 31,785 | 69,027 | 83,371 | 80,036 | 56,666 | 53,493 |
| Revenue | 133,188 | 276,057 | 343,907 | 326,153 | 227,095 | 200,821 |
| Gross profit, non-GAAP | 108,015 | 240,445 | 297,232 | 268,121 | 159,622 | 125,137 |
| Gross margin | 81.1% | 87.1% | 86.4% | 82.2% | 70.3% | 62.3% |
| Operating expense, non-GAAP | 6,841 | 9,341 | 10,088 | 10,593 | 11,122 | 11,678 |
| Operating income, non-GAAP | 101,174 | 231,104 | 287,144 | 257,528 | 148,500 | 113,458 |
| Stock-based compensation | 1,333 | 1,821 | 1,967 | 2,066 | 2,169 | 2,277 |
| Net interest income | 231 | 1,898 | 5,944 | 11,705 | 17,468 | 21,200 |
| Income tax | 14,761 | 35,833 | 45,124 | 41,411 | 25,389 | 20,519 |
| Net income | 84,969 | 195,348 | 245,997 | 225,757 | 138,410 | 111,862 |
| Depreciation and amortisation | 9,503 | 11,396 | 19,245 | 27,481 | 34,234 | 39,517 |
| EBITDA | 109,344 | 240,679 | 304,421 | 282,943 | 180,565 | 150,698 |
| Diluted EPS, GAAP | $74.33 | $169.87 | $213.91 | $196.31 | $120.36 | $97.27 |
| Diluted EPS, non-GAAP | $75.52 | $171.21 | $215.36 | $197.83 | $121.95 | $98.94 |
Source: Penmark Research model, Statements tab. 1,150m diluted shares in all forecast years; no buyback assumed.
18. Financial model | Appendix B
Cash flow turns $190bn a year by FY27E, and the balance sheet carries no net debt
Free cash flow is $135.6bn in FY27E and $499bn over three years, against $5.2bn of debt. The FY26 balance sheet below is as reported.
| US$m | FY24A | FY25A | FY26A | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|---|---|
| Net income | 778 | 8,539 | 84,969 | 195,348 | 245,997 | 225,757 | 138,410 |
| Depreciation and amortisation | 7,780 | 8,352 | 9,503 | 11,396 | 19,245 | 27,481 | 34,234 |
| Stock-based compensation | 833 | 972 | 1,333 | 1,821 | 1,967 | 2,066 | 2,169 |
| Change in working capital | −884 | −338 | −6,130 | −18,003 | −9,499 | 2,486 | 13,868 |
| Cash from operations | 8,507 | 17,525 | 89,675 | 190,562 | 257,710 | 257,788 | 188,681 |
| Capital expenditure | 8,386 | 13,804 | 27,367 | 55,000 | 65,000 | 65,000 | 63,587 |
| Free cash flow | 121 | 3,721 | 62,308 | 135,562 | 192,710 | 192,788 | 125,095 |
| Free cash flow margin | 0.5% | 10.0% | 46.8% | 49.1% | 56.0% | 59.1% | 55.1% |
| Dividends | 513 | 522 | 610 | 690 | 690 | 690 | 690 |
| Share repurchases | 300 | 0 | 650 | — | — | — | — |
| Change in cash and investments | −768 | 2,605 | 61,517 | 134,872 | 192,020 | 192,098 | 124,405 |
| Cash and investments, closing | 9,331 | 11,936 | 73,453 | 208,325 | 400,345 | 592,443 | 716,848 |
Source: SEC filings via Signals (A); Q4 FY26 results release for the adjusted capex and free cash flow basis (A); Penmark Research model (E). FY27E onwards excludes buybacks.
| Assets, US$m | FY26A | FY25A | Liabilities and equity, US$m | FY26A | FY25A |
|---|---|---|---|---|---|
| Cash and equivalents | 38,364 | 9,642 | Accounts payable | 22,605 | 9,649 |
| Short-term investments | 5,070 | 665 | Current debt | 491 | 560 |
| Receivables | 36,197 | 9,265 | Other current liabilities | 4,386 | 1,245 |
| Inventories | 10,372 | 8,355 | Long-term debt | 4,688 | 14,017 |
| Other current assets | 1,067 | 914 | Customer contract liabilities | 12,895 | 142 |
| Long-term investments | 30,019 | 1,629 | Unearned government incentives | 786 | 1,018 |
| Property, plant and equipment, net | 63,310 | 46,590 | Other noncurrent liabilities | 11,659 | 2,002 |
| Goodwill | 1,150 | 1,150 | Total liabilities | 57,510 | 28,633 |
| Other noncurrent assets | 10,339 | 4,588 | Total equity | 138,378 | 54,165 |
| Total assets | 195,888 | 82,798 | Total liabilities and equity | 195,888 | 82,798 |
Source: 10-K FY26 (filed 9 October 2026); 10-K FY25. Reported basis; the forecast model presents net operating working capital and a simplified liability structure.
18. Financial model | Appendix C
The quarterly build: revenue and EPS both grow every quarter of FY27E
The quarterly path is where the forecast is tested. Revenue rises 5-13% a quarter through FY27E, gross margin plateaus at 87.5%, and EPS rises from $37.78 in Q1 to $47.12 in Q4 FY27E as operating expense holds at $2.4bn a quarter.
| US$m unless stated | Q1 FY26A | Q2 FY26A | Q3 FY26A | Q4 FY26A | Q1 FY27E | Q2 FY27E | Q3 FY27E | Q4 FY27E |
|---|---|---|---|---|---|---|---|---|
| DRAM revenue | 10,812 | 18,768 | 31,328 | 39,771 | 44,989 | 50,046 | 53,609 | 56,852 |
| NAND revenue | 2,743 | 4,997 | 9,943 | 14,102 | 16,095 | 16,909 | 17,754 | 18,269 |
| Other revenue | 88 | 95 | 185 | 356 | 366 | 377 | 389 | 401 |
| Revenue | 13,643 | 23,860 | 41,456 | 54,229 | 61,450 | 67,332 | 71,752 | 75,522 |
| Growth, q/q | +11% | +75% | +74% | +31% | +13.3% | +9.6% | +6.6% | +5.3% |
| Gross margin, non-GAAP | 56.8% | 74.9% | 84.9% | 87.0% | 86.25% | 87.0% | 87.5% | 87.5% |
| Operating expense, non-GAAP | 1,334 | 1,421 | 1,518 | 2,568 | 2,060 | 2,427 | 2,427 | 2,427 |
| Operating income, non-GAAP | 6,179 | 16,195 | 33,643 | 44,611 | 50,941 | 56,152 | 60,356 | 63,655 |
| Operating margin | 45.3% | 67.9% | 81.2% | 82.3% | 82.9% | 83.4% | 84.1% | 84.3% |
| Capital expenditure | 2,400 | 4,600 | 7,900 | 10,400 | 11,500 | 13,500 | 14,500 | 15,500 |
| EPS, non-GAAP | $4.78 | $12.20 | $25.11 | $33.42 | $37.78 | $41.61 | $44.70 | $47.12 |
Source: SEC filings via Signals and Q4 FY26 release (A); Penmark Research model (E). FY26 quarterly capex is per the 10-K; FY27E per the model's quarterly build.
| FY26A, US$m unless stated | Value | Per share |
|---|---|---|
| Net income, GAAP | 84,969 | $74.33 |
| Add back: stock-based compensation | 1,333 | $1.17 |
| Tax effect of the adjustment at 15.5% | −207 | −$0.18 |
| Other non-GAAP items | 224 | $0.20 |
| Net income, non-GAAP | 86,319 | $75.52 |
| Diluted shares, m | 1,143 | |
| Quality of earnings | ||
| Non-GAAP as a share of GAAP net income | 101.6% | — |
| Cash conversion, free cash flow to net income | 73.3% | — |
| Stock-based compensation, share of revenue | 1.0% | — |
| Depreciation, share of revenue | 7.1% | — |
Source: 10-K FY26; Q4 FY26 results release; Penmark Research. The gap between adjusted and reported earnings is 1.6% of net income.
19. Reference
Definitions, peer set and method: everything needed to rebuild the numbers
A quarterly build with an annual bridge to FY36E. Two definitions matter: enterprise value treats customer deposits as a liability, and net cash is stated after those deposits.
Glossary
SCA. Supply and capacity agreement: a multi-year contract with a floor price, a volume commitment and a customer cash deposit. Twenty-six are in force, covering more than 35% of revenue to 2030.
HBM. High-bandwidth memory: DRAM die stacked and packaged for AI accelerators. Micron held 18% of the market in Q3 CY26 against SK hynix's 50%.
Bit and ASP per bit. Industry volume is measured in bits; revenue is bits multiplied by the average price per bit. A 20% bit increase at a 6% cost-per-bit decline is the model's base productivity assumption.
RPO. Remaining performance obligations: contracted revenue not yet recognised, about $150bn at the minimum prices in the SCA book.
Ratio definitions
P/E. Share price divided by diluted non-GAAP EPS for the stated year. EV/EBITDA. Today's enterprise value of $689.6bn divided by each year's EBITDA. FCF. Cash from operations less net capital expenditure. Net cash. Cash and investments less total debt less customer deposits. ROCE. NOPAT divided by net debt plus equity. Mid-cycle EBIT margin. Average EBIT margin over FY29E-FY36E: 62.0% in the base, 62.6% probability-weighted.
Methodology note
Business model: products and units, price-led. Revenue is built as bits × ASP per bit for DRAM and NAND separately, with cash cost per bit declining 6% a year. There is no backlog-conversion model and no subscription metric; the KPI set is ASP, bit growth, cost per bit and contract coverage.
Valuation. Primary method is a justified multiple on mid-cycle NOPAT over FY29E-FY36E, cross-checked by a full-cycle DCF to FY36E. Peer multiples are shown for context only, because all four peers trade on near-peak earnings and a forward multiple therefore measures cycle position rather than value.
Definitions. EV includes the $12.9bn of customer deposits as a liability, so EV/EBITDA is not comparable with screens that ignore them. The buyback authorised on 9 December 2026 is not modelled and is treated as value-neutral.
Data sources and as-of dates
Price $658.70 at the close on 5 November 2026. Valuation date 3 September 2026; target date 5 November 2027. Consensus from S&P Global, 29 October 2026 (41 estimates for FY27E EPS; 10 for Q1 EPS); Yahoo Finance and Zacks for Q1 FY27. Company data from the FY26 10-K (filed 9 October 2026), the Q4 FY26 release and call (30 September 2026) and SEC filings via Signals. Industry data: TrendForce and BofA (October 2026). Market data: exchanges.
| Company | Model | Why it is in the set | Difference from Micron | Covered? |
|---|---|---|---|---|
| Samsung Electronics | Memory and devices | Largest DRAM and NAND supplier, 38% and 30% | Captive demand; devices business | No |
| SK hynix | Memory | HBM leader at 50%; closest pure play | Higher HBM mix, Korean cost base | No |
| SanDisk | NAND and SSD | Pure NAND; the NAND cycle read | No DRAM, no contract book | No |
| Seagate | HDD | Nearline storage substitute | Benefits from NAND tightness | No |
| Western Digital | HDD | Data-centre storage exposure | No DRAM; HDD pricing is set separately | No |
Source: Company filings; TrendForce; BofA. Shares are Q2 CY26 for DRAM and NAND, Q3 CY26 for HBM.
Analyst certification and important disclosures
Ratings
Ratings reflect the expected total return to our 12-month price target, including dividends, weighted across bull, base and bear scenarios.
| Buy | Expected total return above +10% |
| Hold | Expected total return between −10% and +10% |
| Sell | Expected total return below −10% |
We review ratings at each set of results and whenever the facts change materially. A share price can move outside its band before the rating changes, and targets may not be reached.
Analyst certification
The Penmark Team certifies that (1) the views in this report accurately reflect its views about the subject company and its securities, and (2) no part of its compensation was, is or will be directly or indirectly related to the specific ratings, price targets or views in this report.
Contact
Penmark Research
Questions about our research:
research@penmarkresearch.com
All our research and coverage:
penmarkresearch.com
Basis of preparation
This report uses information publicly available up to its publication date. Share prices are closing prices on the trading day before the publication date, unless stated otherwise. Forecasts, valuations and price targets are Penmark Research’s own, and forecast periods are marked E. Sources are shown beneath each exhibit.
Valuation and price targets
Price targets are based on the valuation methods set out in each report and are weighted across bull, base and bear scenarios. The main risks to each rating and price target are described in the report. Penmark Research updates its coverage after each set of results but is not obliged to update a report at any other time.
Independence and conflicts of interest
Penmark Research is independent. It does not provide investment banking, brokerage, market-making or asset management services, does not trade securities, and does not accept payment from any company for research coverage. Companies do not see or review Penmark Research reports before publication.
No member of the Penmark Team holds a long or short position, directly or through derivatives, in the securities of any company Penmark Research covers.
Regulatory status
Penmark Research publishes general, impersonal investment research on a regular basis. It is not registered as an investment adviser with the US Securities and Exchange Commission or any state securities authority, is not a registered broker-dealer and is not a member of FINRA. Penmark Research does not give advice tailored to any person’s circumstances.
Not investment advice or an offer
This report is for information only. It is not personalised investment advice and is not an offer or solicitation to buy or sell any security or other financial instrument. It does not take into account any reader’s objectives, financial situation or needs. Readers should make their own investment decisions and seek independent professional advice where appropriate.
Forward-looking statements and risk
Forecasts, valuations and price targets are estimates based on assumptions that may prove wrong, and they may not be achieved. Opinions are current as of the publication date and may change without notice. Investing in equities involves risk, including the loss of capital. Share prices can be volatile, and past performance is not a reliable indicator of future results.
Data and sources
Company financials, operating data and filings come from company filings, earnings releases and earnings calls. Market data and any consensus estimates come from third-party sources named beneath the relevant exhibit; consensus figures from aggregators are unofficial. Penmark Research does not use other brokers’ research. Third-party data belong to their providers and may not be extracted from this report.
Accuracy and liability
Information is taken from sources believed to be reliable, but Penmark Research makes no representation or warranty as to its accuracy, completeness or timeliness. To the fullest extent permitted by law, Penmark Research accepts no liability for any loss arising from use of this report. Descriptions of litigation and regulatory matters reflect allegations and public records, not findings, unless stated. If a published report is corrected, the corrected version carries an “Updated” date at penmarkresearch.com.
Distribution and access
Penmark Research reports are available to registered users of penmarkresearch.com. Downloads and programmatic access are governed by the Penmark Research subscription terms. This report is not directed at, or intended for, any person in Singapore or India, and access from those countries is blocked. It is not intended for distribution to, or use by, any person in any jurisdiction where that would be contrary to law or regulation or would subject Penmark Research to registration or licensing requirements there.
Copyright and permitted use
© Penmark Research. All rights reserved. Penmark Research is a trade name of Brightriver Labs. Readers may share a link to this report but may not copy, reproduce, redistribute, sell or use its content, including for text and data mining, without written consent. Contact: research@penmarkresearch.com.