Financials

Financials in One Page

Micron is a deeply cyclical memory manufacturer (DRAM ~77% of product revenue, NAND the rest) that just lurched from a brutal FY2023 trough to record results. FY2025 revenue reached $37.4B (+49% year-on-year) with operating margin recovering to 26%, but the real signal sits one quarter past the fiscal year-end: in fiscal Q2 2026 (ended Feb 2026), revenue almost tripled to $23.9B in a single quarter and gross margin spiked to 74% as AI memory demand collided with constrained industry supply. The balance sheet is investment-grade clean (net debt/EBITDA 0.27x), but the company is running historically high capex (~$16B in FY2025, guided higher in FY2026) so reported free cash flow remains thin relative to net income — earnings quality has held up only because depreciation absorbs the heavy capex over time. With the stock near $640 and consensus FY2026/FY2027 EPS at $58 / $101, the forward multiples (~11x and ~6x) look optically cheap, but only if you believe the cycle stays up — that's the whole debate.

The single financial metric that matters most right now: forward gross margin trajectory. It is the cleanest read on whether the AI/HBM-driven memory super-cycle is intact or already peaking.

Quick definitions: DRAM = working memory (volatile); NAND = storage memory (non-volatile); HBM = high-bandwidth memory stacked DRAM used in AI accelerators (GPU/ASIC). Net debt = total debt minus cash. EBITDA = earnings before interest, tax, depreciation, amortization. FCF = operating cash flow minus capex. ROIC = after-tax operating profit divided by capital employed.

Revenue, Margins, and Earnings Power

How the income statement actually behaves

Micron's revenue line moves with the global memory price cycle, set by industry-wide DRAM/NAND supply versus demand from data centers, smartphones, PCs, autos, and (now dominantly) AI accelerators. Through-the-cycle revenue scaled roughly 5x from 2005 ($4.9B) to FY2025 ($37.4B), but the path is anything but smooth — peaks in FY2014, FY2018, FY2022, FY2025 were each followed by sharp drawdowns (FY2016 -39%, FY2020 -8%, FY2023 -49%). That cyclicality is not a bug; it is the business.

Operating income is much more volatile than revenue: at the trough (FY2023), Micron lost $5.7B operating; one year later it earned $9.8B. That is operating leverage on a fixed-cost manufacturing footprint.

Margins are the swing variable

Two observations:

1. Operating margin has a 75-percentage-point range across the cycle (-37% in FY2023 to +49% in FY2018). Memory pricing — not unit volume — drives it. When DRAM bit prices fall 30-50%, the cost line stays close to fixed and the operating line craters; the reverse is now happening.
2. The post-FY2025 inflection is steeper than the FY2017-18 prior peak. FY2025 already prints 26% operating margin; the trailing-quarter run-rate suggests significantly more.

Quarterly trajectory: the steepening

Revenue moved from a $3.8B quarterly trough (Q3 FY23) to $23.9B in Q2 FY26 — a 6x recovery. Gross margin has expanded from -47% to +74%. Cost of revenue was nearly flat sequentially in Q2 FY26 (~$6.1B) while revenue grew 75% — that is the supply-shortage signature the company itself now references ("customers getting only 50% to two-thirds of their requirements").

Earnings power judgment: Micron is in an early-cycle peak phase. Reported FY2025 numbers materially understate run-rate earnings. The forward question is not whether margins are improving — they clearly are — but how long this peak lasts before competing capacity (Samsung, SK hynix, CXMT in China) forces normalization.

Cash Flow and Earnings Quality

Free cash flow (FCF) is the cash a business generates after running its operations and reinvesting in property, plant, and equipment. For a capital-intensive memory maker, the FCF question reduces to: does operating cash flow comfortably exceed capex through the cycle, or is the company perpetually outspending what it earns?

Three patterns repeat:

• Operating cash flow is consistently above net income (often 2-4x) because depreciation on the fab footprint is a huge non-cash charge ($8.4B in FY2025). That is healthy and structural.
• FCF is much weaker than OCF because Micron reinvests most of OCF back into capex. FY2025: $17.5B OCF, $15.9B capex, only $1.7B FCF. The ratio of FCF to net income is 19% in FY2025 — low earnings-to-cash translation by any standard.
• FCF goes deeply negative at troughs when OCF collapses but capex doesn't (lead time and contracted equipment). FY2023: -$6.1B FCF on -$5.8B net income.

FCF margin and conversion ratios

Major cash-flow distortions to watch

Item	FY2025	FY2024	FY2023	What it means
Capex ($M)	-15,857	-8,386	-7,676	Capex jumped 89% YoY in FY2025 to chase the AI supply gap. Guided higher again for FY2026.
Depreciation & Amortization ($M)	8,352	7,780	7,756	D&A is the largest non-cash expense; it lags capex by years.
Stock-Based Compensation ($M)	972	833	596	SBC is rising fast — a real economic cost the GAAP FCF does not deduct.
Working capital build (receivables, $M)	9,265 vs 6,615 prior	6,615 vs 2,443 prior	—	Receivables more than doubled FY23→FY24 and grew another 40% into FY25 — the cycle moved cash to AR before cash.
Acquisitions ($M)	0	0	0	Micron has not made a material acquisition since FY2017 (Inotera).
Dividends paid ($M)	-522	-513	-504	Modest, started in FY2021.
Buybacks ($M)	0	-300	-425	Largely paused since FY2024 to fund capex.

Earnings-quality verdict: earnings are real (OCF > NI consistently) but cash conversion is weak because most of OCF is being plowed back into the fab footprint. In a capital-light franchise that would be a red flag; in memory it is structural. The honest metric is "free cash flow after a normalized capex cycle," not point-in-time FCF. By that measure, FY2025's $1.7B FCF understates real cash power — but only if you assume capex will eventually moderate.

Balance Sheet and Financial Resilience

Micron exits FY2025 with a balance sheet that supports — rather than constrains — the upcycle. The company has been deliberate about rebuilding cash and stretching debt maturities since the FY2018 deleveraging.

Leverage ratios

Liquidity and working capital snapshot

Resilience verdict:

• Net debt is $4.3B against $18.1B EBITDA — leverage of 0.27x. That is investment-grade-comfortable even at trough EBITDA (FY2023 leverage spiked to 2.16x but absolute debt did not spiral; the company funded the trough with new bond issuance, not desperation actions).
• $10.3B cash + $28.8B current assets vs. $11.5B current liabilities = 2.5x current ratio. Comfortable.
• Goodwill + intangibles total $1.6B against $54B equity — only 3% — meaning very little balance-sheet risk from past acquisition impairments.
• The pressure point is capex commitments, not the balance sheet itself: management has pre-committed multi-year fab construction (Idaho, New York US Cloud Memory expansion, India assembly) that must be funded from OCF + new debt regardless of the cycle. That converts cycle risk into balance-sheet risk on a 3-5 year horizon if OCF disappoints.

Returns, Reinvestment, and Capital Allocation

The cycle dictates returns. Through-the-cycle ROIC has averaged ~12-15% over the last decade, but the variance is huge. FY2018 hit 51% ROIC at peak; FY2023 went deeply negative.

Capital allocation: capex dominates everything

Over FY2018-FY2025, Micron spent $80.9B on capex versus $7.4B on buybacks and $2.0B on dividends combined. Buybacks were paused in FY2025 to redirect cash into AI capacity. The dividend ($0.115/quarter) is a token at this scale — the implied yield at $640 is below 0.1%.

Share count and dilution

Diluted share count has drifted in a 1,036M-1,152M band for a decade — there is no per-share value erosion from large issuance. SBC creates dilution but buybacks have offset it.

Capital-allocation judgment: Micron is reinvesting at attractive returns when the cycle cooperates, and the cycle is currently cooperating. The FY2025 decision to redirect buyback cash into capex is rational at the start of an upcycle (return on next dollar of capacity > return on next share repurchased), but it is the opposite of what a stable-margin compounder would do. Capital allocation has been disciplined (no foolish acquisitions, no leverage spiral), but the return on that allocation is set by the memory cycle, not by management skill.

Segment and Unit Economics

Micron began reporting under a four-segment structure aligned to end-market in fiscal 2025: Cloud Memory Business Unit (CMBU, hyperscale data-center memory including HBM); Core Data Center Business Unit (CDBU, traditional enterprise/server); Mobile and Client Business Unit (MCBU, smartphones and PCs); Automotive and Embedded Business Unit (AEBU). Detailed segment financials by quarter are in the 10-K and 10-Q filings; no clean machine-readable segment file is available in this run. Two facts are critical:

1. Cloud Memory (CMBU) is now the single largest segment and the highest-margin segment, driven by HBM3E shipped to NVIDIA, AMD, and other AI accelerator vendors. Management has stated HBM is fully sold out for calendar 2026 with pricing already negotiated for portions of 2027.
2. Mobile and Client (MCBU) is the most cyclical segment and has been the slowest to recover as the smartphone refresh cycle stayed weak through FY2024.

Geographically, Micron generates roughly two-thirds of revenue outside the US (Taiwan, Mainland China, rest of Asia, Europe), and has meaningful US fab exposure (Idaho and incoming New York fab) that ties directly to CHIPS Act incentives.

Valuation and Market Expectations

Trailing valuation versus history

The trailing P/E line is uninformative — it is negative or absurdly high at every cycle trough because earnings collapse. EV/EBITDA and P/B are the more durable lenses for a memory cyclical:

• EV/EBITDA peaks at troughs (40x in FY2023) and bottoms at peaks (3x in FY2018, FY2022). Investors pay up for trough earnings and demand discount on peak earnings.
• P/B has expanded from 0.5-0.9x in 2009-2017 to 2.5x at FY2025 year-end — and it has expanded much further since. A persistent P/B re-rating would mark the AI-memory thesis as a regime change, not a cycle.

The current setup at $640 (May 7, 2026)

Where consensus has moved

The 60-to-30-day window is where the dam broke: Q2 FY26 actuals (March 18, 2026) drove an immediate doubling of FY26 and FY27 EPS estimates. The 30-day count was 26 up / 0 down for FY2026 EPS.

Bear / Base / Bull framework

Published Street targets currently span $87 (deep bear) to $1,000 (Mizuho/D.A. Davidson bull), with an average around $551. That ~$900 spread on a single name is itself the headline — the disagreement is not about the next quarter, it is about whether HBM/AI demand is structural or cyclical.

Valuation judgment: at $640, MU is not obviously cheap on trailing numbers and not obviously expensive on forward numbers. The trailing 40x EV/EBITDA looks frightful; the 6x FY27 forward P/E looks like a steal. Both are right inside their respective frames. The honest read is that the stock is priced for the AI-memory-as-strategic-asset thesis to persist — not for a return to normal memory cyclicality. The gap between $420 (bear, peak-cycle normalization) and $1,500 (bull, regime change) defines the live debate, and the burden of proof now sits with the bull case to show that two consecutive cycles can avoid a full retracement.

Peer Financial Comparison

A few caveats:

• Peer FY periods don't align. MU is FY2025 (ended Aug 2025); SNDK and WDC fiscal years end in summer; NVDA FY2026 ended Jan 2026.
• The two closest economic peers — SK hynix and Samsung — are not shown with detailed financials in the data set. Both publicly disclosed record FY2025 results driven by HBM, qualitatively confirming MU's cycle position.
• Versus the public US cohort, MU sits between Sandisk (pure-play NAND, still loss-making — NAND lagging DRAM) and WDC (HDD-heavy storage, profitable but lower-margin). MU's gross margin (40%) is now higher than WDC (39%) and well above Sandisk's 30%, with the gap likely to widen at the run-rate.
• NVDA is the AI customer, not a competitor. The 71% gross margin and 60% operating margin are what an AI-bottleneck owner can earn. MU's recent quarter (74% gross margin) shows what AI-memory pricing can do for the supply side — briefly approaching NVDA-like economics.

The peer-gap takeaway: MU's premium versus pre-cycle valuations is justified if the AI-driven gross-margin expansion holds. It is harder to defend versus SK hynix on a pure economic basis (SK hynix has the larger HBM share and is the NVIDIA-preferred supplier).

What to Watch in the Financials

Closing read

What the financials confirm: Micron is in the early innings of the largest memory upcycle in its history. Operating margin recovery is real, the balance sheet is investment-grade clean, returns on capital are scaling, and Q2 FY26 confirms that AI-memory pricing has unlocked NVDA-adjacent unit economics — for now.

What the financials contradict: the "compounder" narrative. Free cash flow is still thin relative to EBITDA because capex commitments are enormous and growing, dividends are token, buybacks are paused. Per-share value compounding requires the cycle to stay up; it does not happen automatically through reinvestment.

The first financial metric to watch is gross margin in Q3 FY26 (results expected ~late June 2026). A second consecutive quarter above 70% gross margin would functionally validate the FY2027 $100+ EPS path the bull case requires; a sequential decline toward 60-65% would force a re-rating of the back half of consensus and would mean the cycle peak arrived in Q2 FY26.