Micron’s Blockbuster Balance Sheet and the Flat After‑Hours Reaction: Has the Memory Cycle Really Changed?

Historic balance sheet, flat share price

Micron unveiled one of its strongest balance sheets ever, yet the share price refused to climb. In the fiscal Q4, revenue jumped to $54.2 bn – up 31% sequentially and nearly five times the year‑ago quarter. Gross margin hit 87%, pushing diluted EPS to $33.42, and management said every line beat expectations. For the full fiscal year, revenue reached $133.2 bn, a 256% YoY increase. Despite this, after‑hours trading showed the share slightly negative. The commentator’s question captures the tension: with numbers this good, why is the market not cheering?

The first part of the answer lies in the profit engine itself. DRAM revenue for the quarter was $39.8 bn, 73% of total, up 27% QoQ. Shipment growth was mid‑single‑digit, but average selling prices soared into the high‑teens percent range; essentially all the growth came from price. On an annual basis, DRAM exceeded $100 bn, accounting for 76% of FY26 revenue and up over 250% YoY. This shows how AI‑infrastructure demand has become a powerful price‑driver. The speed with which price pulls away from volume is the first spark in the duration debate.

The second engine, NAND , moved even faster. Quarter‑end revenue was $14 bn (26% of total), up 42% QoQ. Shipment volume rose about 10% while average selling prices climbed roughly 30% from the prior quarter. Annual NAND revenue came to almost $32 bn, a 274% YoY increase. However, a wrinkle appears. TrendForce’s PRNewswire outlook for 2027 suggests DRAM will remain tight while NAND supply could loosen in the second half as new capacity comes online and consumer electronics demand stays weak. The implication: today’s star could become tomorrow’s risk‑factor.

Breaking down by business unit shows the margin uniformity. Cloud memory (including HBM) generated $16.3 bn revenue at an 83% gross margin; it also contains high‑bandwidth memory. Core data center contributed $18 bn at a 90% margin. Mobile and client delivered $13.1 bn at a 90% margin. Automotive and embedded came in just under $7 bn at an 84% margin. All four units cleared the 80% margin mark, proving the tightness is not product‑specific but systemic. The commentator’s repeated point: this is not a niche shortage, it is the whole memory ecosystem.

The 2027 agreement language is critical. Management said the 2027 high‑bandwidth memory contracts contain significant price increases that narrow the gap with conventional DRAM. Normally a premium product’s price‑protection squeezes the spread, but here the spread is collapsing – a sign that conventional DRAM scarcity has reached premium‑product levels. TrendForce backs this: its research revised the 2027 HBM price outlook upward and projects the blended ASP to rise 121% YoY, noting growing HBM4 share and an early HBM4e ramp in H2 2027, with GPU/ASIC makers shifting to 8‑high stacks to save on wafer and cost.

Looking ahead to the coming quarter, guidance gave the second data point feeding the duration debate. Micron guided Q1 FY27 revenue to $61.5 bn and EPS to $38.15 (±$1); the Wall Street consensus was $57.4 bn revenue. Management did not just extrapolate the past – it lifted the future outlook. The guided gross margin of 86.25% was labeled the *floor* for FY27, with the expectation of higher margins later in the year as price increases flow through. Yahoo‑Finance‑aggregated consensus had EPS at $31.83; the actual print beat that by $1.59.

The market’s hesitation centers on capex guidance. Micron said FY27 Q1 capex would be about $11.5 bn, with H1 FY27 capex around $25 bn and H2 FY27 even higher, emphasizing a faster growth rate in construction‑related spend versus equipment. To an investor conditioned by the old memory cycle, this screams future oversupply and margin compression. Yahoo‑Finance’s after‑hours story tied the negative price directly to this capex outlook. Algorithmic selling may have added to the pressure. The fear is understandable, but pricing it without regard to the timing of when supply arrives is premature.

Because capex and useful supply do not arrive in the same year. Tom’s Hardware‑derived timetable shows Idaho Fab 1 starting wafer output mid‑2027, Idaho Fab 2 in late 2028, and the New York fab in 2030. Micron explicitly warned that new fabs need additional quarters before meaningful output hits the market. Thus today’s spending is matching 2028‑and‑later demand, not 2027. TrendForce confirms: capacity slated to come online in 2027 is not expected to contribute meaningfully before 2028. Capex‑driven sellers are, in effect, discounting 2028‑plus supply at today’s prices.

On the demand side, management’s message was tightening‑oriented. Micron expects the memory‑storage supply‑demand balance to be much tighter in 2027‑2028 than it was in 2026, for both DRAM and NAND. Moreover, CEO Sanjay Mehrotra stated that, even with additional industry clean‑room plans, there is no line of sight to when supply and demand will return to balance – a striking comment from a CEO amid a pricing boom. TrendForce’s research on AI‑server demand shows it is sucking up both advanced‑process and wafer capacity, keeping HBM and conventional DRAM in competition for limited resources, with tightness persisting through 2027.

The most concrete proof of change sits in the customer contracts . Micron has signed 26 strategic customer agreements running through 2030; these are said to cover more than 35% of the company’s estimated revenue through that year. The agreements are take‑or‑pay, with about three‑quarters having a defined pricing framework and the majority featuring price floors and ceilings. Customer financial commitments have reached $32 bn, largely as cash deposits on the balance sheet. The CFO disclosed that remaining performance obligations (RPO) have grown to roughly $150 bn. Two customers were even given one‑year extensions into 2031. As Reuters’ June piece noted, Micron, Samsung and SK Hynix are trying to turn these long‑term deals into a cycle‑mitigant. Memory is no longer spot‑market commoditiy; it is contracting like infrastructure.