Micron FY2026 Q4 Earnings: An $85B Year, With Cash to Match

Micron closed fiscal 2026 with $133.188 billion in revenue and $84.969 billion in net income, then guided the next quarter above an already record fourth quarter. The harder question is what remains when memory pricing eventually slows. Operating cash generation supports the profit today; customer deposits and new factories determine how much of that cash can safely be treated as surplus. The September 30 results make those distinctions unusually important.

The headline numbers: annual first, quarter separately​

Fiscal 2026 ended September 3, 2026. It contained 53 weeks, including a 14-week Q4; the previous fiscal year contained 52 weeks. That extra week matters when comparing quarterly growth. Micron disclosed the calendar in its Q3 Form 10-Q.

These annual GAAP results are the period modeled in the ledger below. The source is the full-year, unaudited statements furnished with the September 30 Form 8-K; the FY2026 10-K was not yet available in SEC filing history as of October 2.

The separate fourth-quarter comparison shows the exit rate. It should not be added to the annual results, which already include it.

Amounts and per-share figures come from the 8-K earnings exhibit; growth rates are calculated from those figures. Neither table substitutes adjusted earnings for reported profit.

The striking relationship is between revenue and cost. Annual revenue rose by $95.810 billion while cost of goods sold rose by only $3.179 billion. That arithmetic explains most of the earnings expansion before any debate about future AI demand. A manufacturing business with a largely committed production base can convert higher selling prices into extraordinary incremental profit. The same structure exposes earnings when prices fall.

Revenue deep dive: data centers dominate, but HBM is not a segment​

The earnings deck's business-unit table separates four businesses. Shares below use consolidated Q4 revenue as the denominator.

The four rows total $54,223 million, leaving a $6 million difference from consolidated revenue. Keeping that difference visible is preferable to assigning it to a business without disclosure.

Cloud Memory is central to the HBM thesis, but its revenue is not an HBM revenue figure. The two data-center units together represent 63.2% of quarterly sales. Treating that entire amount as accelerator memory would erase conventional server memory and storage demand and give investors a false measure of HBM concentration.

Core Data Center deserves separate attention. Its larger revenue base means the investment case cannot depend solely on advanced packaging. Micron describes SSD uses as “expanding the addressable market” in its prepared remarks. That is an opportunity statement, not proof that market expansion exceeded an earlier forecast. Adoption must ultimately appear in repeat orders and cash collections.

Mobile and Client provides a useful check on broad demand claims. Premium-device memory content can support revenue even when device shipments weaken. A reader should therefore separate demand for memory per device from demand for devices themselves. Revenue growth alone cannot establish that households bought proportionately more phones and PCs.

Automotive and Embedded adds another end market, but diversification by customer industry does not automatically diversify memory pricing exposure. These customers can face the same constrained supply pool. The relevant test is whether orders persist when price increases moderate, not simply whether every segment currently grows.

Management reports “robust demand trends” and “ongoing supply tightness”; it expects DRAM and NAND to “remain supply constrained” through calendar 2027 and 2028. Its Singapore HBM packaging preparation is “ahead of plan,” a capacity milestone rather than a quantified claim that an HBM product launch beat forecast. These are distinct assertions in the prepared remarks. Inventory and receivables can help test the narrative over subsequent quarters, but neither balance proves end-customer demand by itself.

The margin story: price matters more than the extra week​

All margins in this table are GAAP, calculated from the reported statements. Annual history comes from the FY2025 10-K, with FY2026 and Q4 from the current release.

An additional selling week can increase revenue and absorb some fixed costs. It cannot, by itself, explain a doubling of the annual gross-margin percentage. The magnitude requires examining what Micron earned on each unit of memory, alongside manufacturing costs and product mix.

For NAND, management states, “Prices increased approximately 30%,” against roughly 10% sequential bit-shipment growth. DRAM pricing rose in the high teens while bits grew in the mid-single digits. Those disclosures in the prepared remarks support pricing as the larger immediate driver. They do not eliminate product differentiation; they show why shipment growth alone understates both current profitability and future downside sensitivity.

Reported earnings also retain costs that an adjusted presentation removes. The annual reconciliation identifies $500 million of patent-license charges and $511 million of debt-prepayment losses. They remain visible in the ledger. Removing selected charges without their tax effects would not produce a defensible alternative net-income figure, so the headline stays with GAAP.

The next useful distinction is between a margin level and its direction. A slight decline from a very high margin can still produce greater dollar profit on growing sales. Conversely, maintaining an exceptional margin while spending aggressively on capacity can leave less distributable cash than the income statement initially suggests.

The one big question: do contracts change the memory cycle?​

Management expects supply-demand conditions to become “much tighter.” The more durable question is whether customers have committed enough cash and volume to improve the economics of the next downturn. Micron's strategic customer agreements, or SCAs, give that argument substance, while leaving important limits. The prepared remarks describe 26 agreements and approximately $150 billion of remaining performance obligations, measured using committed volumes and minimum prices for agreements with determined pricing frameworks.

RPO is future contracted business, not recognized revenue. A contract can improve planning without converting its entire face value into today's cash or profit. Price floors and ceilings can also distribute the benefits of scarcity differently from spot pricing. Investors need to evaluate the protection received alongside the upside surrendered.

The cash-flow statement makes a second separation necessary:

These figures and the non-GAAP reconciliation appear in the earnings deck. Net capex subtracts the two proceeds lines from gross investment; adjusted free cash flow subtracts net capex from operating cash. Customer deposits sit outside that calculation.

That classification supports a meaningful conclusion: the reported free-cash-flow result does not depend on counting the deposits as operating inflows. But financing cash is still available cash with associated obligations. Spending it on a factory does not extinguish those obligations, and adding it again to revenue or free cash flow would double-count the economic benefit.

My reading is that contracts improve the financing and planning of capacity more convincingly than they prove the end of cyclicality. Their real test comes when customers would prefer less memory or lower prices. Until that test arrives, a strong cash balance and exceptional current margins are evidence of resilience, not evidence that every future factory will earn the same returns.

Q1 FY2027 guidance gives the first test​

The company's guidance reconciliation keeps reported and adjusted forecasts separate:

The revenue midpoint is 13.4% above Q4, a calculated increase rather than a realized result. Management expects Q1 to be the year's gross-margin low, partly because manufacturing incentive compensation held in inventory will flow through costs. Its capex outlook is net of anticipated government incentives: approximately $25 billion in the first half, with a larger second half. These remain forecasts from the prepared remarks.

The combination sets a clearer standard than another revenue record alone. Rising sales must fund a substantially larger investment program while margins absorb costs already embedded in inventory. Comparing gross investment, incentives received and operating cash separately will show whether the expansion is self-funded on the terms management anticipates.

Tracking a $133B revenue year in plain text​

Double-entry makes the relationship between revenue, costs and profit inspectable. Our guide to modeling a public company explains the shared conventions. This is the full FY2026 income transaction from Micron's published ledger, in millions of US dollars; income credits are negative and expenses are generally positive.

; Check: -133188 + 25684 + 5650 + 1947 + 500 + 511 - 834 + 14761 + 84969 = 0
 
2026-09-03 * "Micron Technology, Inc." "FY2026 Income Statement"
  Income:Revenue                          -133188 MUSD
  Expenses:CostOfRevenue                    25684 MUSD
  Expenses:ResearchAndDevelopment            5650 MUSD
  Expenses:SellingGeneralAdministrative      1947 MUSD
  Expenses:OtherNet                          500 MUSD  ; patent license charges; release GAAP/non-GAAP reconciliation
  Expenses:OtherNet                          511 MUSD  ; loss on debt prepayments; same reconciliation
  Expenses:OtherNet                         -834 MUSD  ; remaining net other income, calculation above; combined OtherNet expense 177
  Expenses:IncomeTax                       14761 MUSD
  Equity:Adjustments                       84969 MUSD  ; net income offset; retained earnings set by balance assertion

The three OtherNet postings total a $177 million expense. The negative posting preserves the remaining net other income rather than concealing the patent and financing charges inside one small net number. The full source mapping includes other operating items, interest, other non-operating items and equity-method earnings.

The balance sheet reconciles to $195.888 billion of assets = $57.510 billion of liabilities + $138.378 billion of equity. Condensed categories remain condensed: the release does not separately disclose intangibles, so their carrying amount is included within other noncurrent assets rather than estimated from last year's number.

The most revealing asset is property, plant and equipment: $63.310 billion, up from $46.590 billion a year earlier. It captures a growing capital commitment, not instantly available capacity. Construction, equipment installation and commercially useful production occur at different times. The profit model needs to remain attractive across that delay.

The multi-year arc​

The FY2023 and FY2025 annual reports, together with the current release, show why the latest margins should be evaluated across a cycle. Amounts are millions of US dollars; gross margins are calculated.

The FY2023 loss demonstrates how little a substantial factory base guarantees about near-term profitability. The subsequent recovery demonstrates the earnings potential when the same capital-intensive model encounters favorable demand and pricing. Both observations belong in the investment case.

Today's larger PP&E base raises the consequences of the next cycle. Better products and firmer customer commitments may justify more investment, but the spending decision has to work across future price environments. Projecting the latest margin across every additional dollar of capacity would assume the very conclusion investors need to test.

The verdict: bull versus bear​

Bull case

• Annual operating cash exceeds reported net income. Continued strong conversion would support the quality of the profit while capacity expands.
• The two data-center units generate 63.2% of Q4 sales. Sustained orders across both would broaden the thesis beyond HBM alone.
• Contract commitments give management a basis for capacity planning. Converting RPO into recognized revenue without weakening margins would demonstrate their value.
• The Q1 revenue midpoint implies another sequential increase. Delivering it alongside the forecast margin would support management's near-term supply outlook.

Bear case

• Pricing contributes more to current growth than bit shipments. A slowing price cycle could affect profit faster than revenue, even if physical demand remains healthy.
• The 53-week year and 14-week Q4 complicate comparisons. Growth on an ordinary calendar must continue to justify extrapolating the exit rate.
• Capacity milestones do not establish product-launch outperformance. Future HBM shipments and margins must substantiate the broader technology claims.
• Deposits carry obligations, while new factories require capital. Cash accumulation alone cannot prove that the expanded asset base will earn adequate returns.

Our take: Micron's FY2026 profit is supported by substantial operating cash, and the customer-contract structure deserves weight in assessing its ability to invest through a cycle. The case is strongest when framed as a better-funded memory manufacturer with improving demand visibility. We would judge FY2027 by cash conversion after investment and by realized contract economics, rather than capitalize the latest gross margin as a permanent earnings rate.