MU — Q3 2026 Financial Report Analysis

A price shock, not a volume boom

Micron's fiscal third quarter of 2026 — the three months ended May 28, 2026 — produced revenue of $41.46 billion, up 346% from $9.30 billion a year earlier and up 74% from the $23.86 billion it reported just one quarter before. Net income went from $1.89 billion to $28.24 billion, and diluted earnings per share from $1.68 to $24.67.

Those numbers are close to meaningless without the reason behind them, and the 10-Q is unusually direct about it. Almost none of this growth came from shipping more chips. Micron breaks its revenue change into two pieces: average selling price (ASP — what a unit of memory sells for) and bit shipments (how much memory it actually shipped, measured in gigabits). For the quarter, against the same quarter last year:

• DRAM revenue rose 343%, driven by a "low-260% range increase in average selling prices" and only a "low-20% range increase in bit shipments."
• NAND revenue rose 361%, driven by a "mid-310% increase in average selling prices" and a "low-double-digit increase in bit shipments."

In other words, Micron sold roughly 20% more DRAM than a year ago and charged roughly 3.6x as much for it. The company attributes this to supply falling behind AI demand: "AI-driven memory and storage growth is outpacing industry supply... demand for memory and storage at a rate greater than our ability and the industry's ability to increase supply."

("pp" means percentage points — the simple difference between two percentages. Gross margin is the share of revenue left after the direct cost of making the product; operating margin is what's left after research, sales and administrative costs too, but before interest and tax.)

The cost side barely moved

The margin expansion is arithmetic rather than mystery. Cost of goods sold rose from $5,793 million to $6,400 million — about 10% — while revenue quadrupled. Memory manufacturing is overwhelmingly fixed cost: the fab, the equipment and the people cost roughly the same whether the chips coming off the line sell for $2 or $8. So when price moves and volume doesn't, nearly the entire increment drops to the bottom line.

Gross margin went to 84.6% from 37.7% a year ago and 74.4% in the prior quarter. Management attributes the sequential step-up "primarily due to increases in average selling prices" with help from "continued strong execution and favorable mix," and the year-over-year improvement to ASPs "and, to a lesser extent, favorable mix and manufacturing cost reductions."

Operating expenses grew, but nowhere near in proportion: R&D rose to $1,316 million from $965 million and SG&A to $407 million from $318 million. Combined operating expenses of $1,738 million consumed 4.2% of revenue, against 14.4% a year ago.

One quality check worth making on a quarter this large: the gap between reported (GAAP) and company-adjusted (non-GAAP) results is small. Non-GAAP EPS was $25.11 against GAAP's $24.67 — a $0.44 difference, under 2%. There is no large adjustment doing quiet work here.

Every business unit re-rated, and mobile did it on price alone

Micron reports four units: Cloud Memory (CMBU — hyperscale cloud customers plus high-bandwidth memory, or HBM, for all data-center customers), Core Data Center (CDBU — mid-tier cloud, enterprise and OEM data center, plus data-center storage), Mobile and Client (MCBU), and Automotive and Embedded (AEBU).

The most revealing line is MCBU. Its revenue rose 254% year over year "primarily due to increases in average selling prices, partially offset by lower bit shipments" — Micron shipped fewer bits into phones and PCs than a year ago and still nearly quadrupled the revenue, then earned an 86% operating margin on it, the highest of any unit. The same pattern holds sequentially (MCBU up 49% with lower bit shipments).

That is what allocation looks like in a shortage. The filing says as much: the imbalance "has led to decisions on supply allocation that may impact certain customers and end markets." Bits that would previously have gone to a smartphone maker are going to a data center instead, and the phone maker is paying more for less.

It also means the data-center story, while real, is not the whole story. CDBU's 653% growth is the largest number in the table, but it is off the smallest base ($1.53 billion) and the year-ago quarter was a weak one for that unit at a 20% operating margin. Combined, the two data-center units are $25.3 billion of the $41.5 billion, or 61% of revenue, up from 53% a year ago.

The strategic customer agreements: selling the cycle's upside for a floor

The most consequential item in this filing is not a number in the income statement. Micron discloses that in FQ3 and FQ4 2026 it has been signing multi-year take-or-pay contracts (the customer must pay for a committed volume whether or not it takes delivery) with binding volume commitments. Most have fixed pricing or a price band.

The terms matter, and the 10-Q spells them out: "The largest agreements generally have a ceiling price for existing products that approximates the market price in the second calendar quarter of 2026, and a floor price through the term of the agreement."

Calendar Q2 2026 is essentially the quarter just reported. So for existing products, Micron has capped its own selling price at roughly today's peak-cycle level for the duration of these contracts, in exchange for a floor beneath it. That is a deliberate trade: give up the upside if memory prices keep climbing, in return for protection when they fall — and in this industry they always eventually fall. Management's own framing is that even at floor pricing, gross margins from these agreements would "yield gross margins well above our peak quarterly margins in any past cycle."

Two things temper how much weight to put on this today. First, the contracted amount is still small relative to the run rate: remaining performance obligations were "approximately $5 billion" as of May 28, 2026, of which about one-third is expected to be recognized as revenue over the next twelve months — against a single quarter of $41.5 billion. Micron notes the disclosure uses minimum committed volumes and minimum pricing and "is not expected to be indicative of future revenue," so the economic value is larger than $5 billion, but the covered share of the business is clearly still partial. Second, the associated cash is real and near-term: Micron expects "$22 billion" of deposits and related financial commitments from agreements concluded to date, roughly $18 billion of it in cash deposits — money that funds capacity before the revenue arrives.

Cash: debt repaid rather than shares bought

Operating cash flow for the first nine months was $45.70 billion, against $11.80 billion in the comparable period. Adjusted free cash flow in the quarter alone was $18.3 billion after $7.1 billion of net capital spending.

What Micron did with it is informative:

• Repaid $9.38 billion of debt in nine months, prepaying the 2028 Notes, 2029 Term Loan A, 2029 A and B Notes and 2030 Notes in full and partially prepaying six other series. Long-term debt fell to $5,140 million from $14,017 million at fiscal year-end.
• Bought back only $650 million of stock (2.5 million shares, an average of about $260 per share) under its $10 billion authorization, plus $762 million of share withholding on employee equity awards.
• Paid $437 million of dividends, holding the quarterly payout at $0.15 per share — roughly 0.6% of the quarter's EPS.
• Spent $19.60 billion on property, plant and equipment, offset by $2.99 billion of government incentive proceeds, against full-year fiscal 2026 guidance of "approximately $27 billion" net.

Cash and marketable investments reached $30.13 billion from $11.94 billion at fiscal year-end; total equity roughly doubled to $100.72 billion from $54.17 billion.

Deleveraging aggressively while buying back almost no stock at the top of a price cycle is the conservative choice, and a defensible one in an industry where earnings can fall by an order of magnitude when pricing turns — the year-ago quarter in this very filing, at a 23.3% operating margin, is what the other side of a memory cycle looks like.

Two balance-sheet lines deserve a flag. Inventories were $8,567 million versus $8,355 million at fiscal year-end — essentially flat while quarterly revenue quadrupled, meaning inventory measured in days of sales has collapsed. That confirms how tight supply is, but it also removes any buffer: Micron has little finished product to sell into a demand spike, and equally little cushion if pricing turns. And "estimates of consideration payable to customers, including pricing adjustments and returns" nearly tripled to $3.32 billion from $1.19 billion — a liability that grows with the size and price of the book, worth watching as a drag if prices reverse.

Tax: the rate is structurally higher now

The effective tax rate rose to 15.0% from 11.1%, which cost roughly $1.3 billion at this quarter's pre-tax income. The cause is specific and durable: the OECD's Pillar Two 15% global minimum tax, which Singapore enacted effective for Micron in fiscal 2026 and which "largely offsets the benefit from our Singapore tax incentive arrangements." Micron manufactures heavily in Singapore, so this is a permanent step-up rather than a timing item. Noncurrent income taxes payable jumped to $5.79 billion from $648 million.

Customer concentration also moved in a useful direction: one customer was 10% of revenue in the first nine months of fiscal 2026, down from 16% a year ago — not because that customer shrank, but because everything else grew around it.

Takeaway: Micron is earning an 80% operating margin on roughly 20% more DRAM bits than it shipped a year ago — this is a pricing event, and pricing events in memory reverse. The number that decides whether this quarter marks a durable change rather than a cycle peak is not revenue or EPS; it is how much of the book gets locked into take-or-pay contracts before prices turn, and today that is still only about $5 billion of disclosed minimum obligations against a $41 billion quarter.

Guidance and outlook

For fiscal Q4 2026 (the quarter ending around early September 2026), management guided to:

At the midpoint that is another 21% sequential revenue increase and a further ~1.4 point gain in gross margin, implying pricing had not yet peaked as of the late-June guide. CEO Sanjay Mehrotra framed the quarter and the guide as reflecting "the strategic value of memory in the AI era."

Our read. The operating result is not in question — margins of this level with costs flat are simply what happens when memory prices triple. The open questions are about duration, and there are three:

1. Supply is coming, just slowly. Micron's own capacity additions are dated in the filing and they are all beyond this cycle's near term: first DRAM wafer output from the new Boise fab in mid-calendar 2027, the Tongluo (Taiwan) fab acquired from Powerchip for $1.8 billion supporting "meaningful product shipments... beginning in mid-calendar 2027," Singapore HBM advanced packaging from the first half of calendar 2027, a second Idaho fab by late calendar 2028, New York from 2030. Nothing Micron builds relieves the shortage inside the next four quarters — but competitors are investing into the same signal, and industry supply arriving in 2027-2028 is the mechanism by which memory cycles have always ended.

2. The contracts cut both ways. Capping price at roughly calendar-Q2-2026 market levels means that if the shortage persists into 2027, Micron will be selling below spot on its contracted volume. That is the cost of the insurance, and it is a defensible trade for a company that has twice in the past decade swung from a record-profit year to an annual net loss — but it does mean the reported peak margin is likely near its ceiling even if the market tightens further.

3. Mobile and client is the pressure point. MCBU is now earning an 86% operating margin while shipping fewer bits. Phone and PC makers cannot absorb memory costs at this level indefinitely without either raising device prices or reducing memory content per device. Watch MCBU bit shipments in FQ4 and FQ1 2027: if they keep falling while ASPs plateau, that unit's revenue growth stalls quickly, and it is 28% of the total.

The realistic base case is that fiscal Q4 lands near guidance and fiscal 2027's first half stays strong on contracted and allocated volume, with the risk concentrated in the back half of calendar 2027 as new industry capacity arrives. A balance sheet carrying $30 billion of cash against $5 billion of long-term debt, plus $18 billion of customer deposits inbound, means Micron will enter that downturn in a far better position than it entered the last one — which may prove more valuable than the size of this quarter's profit.