Micron crushed earnings but the stock stalled: inside the 54 billion dollar quarter

Micron reported record revenue for fiscal Q4 2026: 54.23 billion dollars, far above 41.46 billion in the prior quarter and 11.32 billion a year earlier. That is 379 percent year-on-year growth and a clear beat against the 51 billion Wall Street expected. Adjusted earnings per share jumped from about 3.3 dollars to 33.42 dollars, more than a tenfold increase, with adjusted net income of 38.4 billion dollars. These figures are confirmed by the official results published on investors.micron.com, where operating cash flow of nearly 44 billion dollars is disclosed in the same release.

Anatomy of a record quarter

The gross margin was the star of the report at 87 percent on an adjusted basis, with GAAP margin at 86.8 percent against 44.7 percent last year, showing the scale of the transformation. The jump comes from four forces stacking together: higher shipment volumes, dramatically stronger pricing, a richer mix weighted toward high-bandwidth memory, and AI demand flowing straight into profit. Gross profit reached 47.2 billion dollars and operating leverage expanded visibly. Yet the warning published on forbes.com belongs at this point, because the article argues that an 87 percent margin is an extraordinary level that invites huge capacity spending and raises durability questions.

Forward guidance

Management guides next quarter to 61.5 billion dollars of revenue and 38.15 dollars of adjusted earnings per share, with ranges of 1.5 billion dollars and 1 dollar respectively. The gross margin outlook at 86.25 percent stays in record territory, implying more than 53 billion dollars of gross profit. The company models about 2.06 billion dollars of operating expenses and a 15.5 percent tax rate on roughly 1.15 billion shares to reach that earnings figure. This forward framework appears in the official outlook text on investors.micron.com, and a range above prior analyst estimates shows how confident management feels about demand.

AI demand

Behind the growth sits secular demand from AI infrastructure: larger models, wider context windows, growing inference use and multiplying agents across enterprises keep raising memory needs. As concurrency rises and clusters scale, every accelerator wants more and faster memory attached to it. The speaker recalls telling viewers not to believe the slowdown rumors weeks earlier, with AI companies saying spending is accelerating rather than pausing. That demand picture is confirmed with the CEO own words on arstechnica.com, where management stresses that bigger models and wider agent deployment keep increasing both memory content and performance requirements.

Why HBM matters

High-bandwidth memory is the heart of the story, and the speaker analogy is that if the Nvidia GPU is the engine, HBM is the fuel feeding it. Even with the fastest accelerators, wasting expensive compute happens when data cannot be moved fast enough, which is why HBM became the defining component of modern AI systems. Micron is one of only a few companies producing it at scale, and HBM revenue grew faster than the company average in the quarter. This picture is supported by the analysis on techrepublic.com, which notes that HBM is outgrowing commodity DRAM while each new generation consumes more manufacturing capacity per unit.

Custom work with Nvidia

Micron is working with Nvidia on the industry first custom HBM4E implementation, called NV-HBM and aimed at next-generation GPUs and NVLink Fusion platforms. The step shows memory moving from an outside commodity into the performance core of the system, with deeper co-design between the two firms. Nvidia says the custom base die promises 30 percent higher bandwidth per stack and 15 percent lower power than commodity HBM4E, moving the controller into the stack and freeing up to 30 percent more compute area on the main die. This technical frame is described in the detailed review on tomshardware.com, which positions the block as a building step offered only to NVLink Fusion partners.

The 2027 and 2028 message

The toughest line from the earnings call is that the shortage is not ending: management expects the supply-demand balance to tighten further in calendar 2027 and 2028 compared with 2026, with no line of sight on when balance returns. New clean rooms opening in 2028 will ramp only gradually, and the rising trade ratio from HBM3E toward HBM4 and 4E constrains supply growth. Industry DRAM bit shipments are seen growing in the low twenty percent range with NAND in the mid twenty percent range, yet both markets stay constrained. This outlook is carried with the CEO direct quotes on arstechnica.com, where 2027 pricing is described as much higher than 2026 pricing.

Contract cover

Customers are trying to lock supply years ahead, and long-term agreements give visibility rarely seen in memory. Cash-backed customer commitments rose from 22 billion dollars in June to 32 billion, while remaining performance obligations grew from about 100 billion to 150 billion. More than 75 percent of planned 2027 output is already committed, and most new talks have shifted to 2028 volumes. This contract structure is confirmed by the summary on forbes.com, which notes that 32 billion dollars of deposits plus 150 billion of expected-revenue agreements create a protection layer missing in earlier cycles.

Is this time different

Memory history gives a harsh lesson: demand surges, prices rise, producers add capacity, supply catches up and prices collapse. That is why the market refuses Nvidia-like multiples for Micron, assumes peak earnings will fade, and analyst projections fall off after next year. A low forward price-earnings ratio can therefore signal peak profit rather than cheapness. Still, the current thesis published on seekingalpha.com argues this time can differ, claiming a 6x earnings multiple could rerate as agentic AI keeps memory demand durable and consensus estimates stay conservative.

Stock and strategy

The muted after-hours reaction with a red next day follows from expectations: tight memory, booming HBM and strong AI spending were already in the price after a monster run. The market did not grade whether earnings were good, but whether they cleared enormous embedded hopes, and the bar was extraordinarily high. Here the JPMorgan note carried on gurufocus.com is instructive, keeping an Overweight rating and lifting the target to 1540 dollars while TD Cowen holds a Buy at 1600. The speaker instead favors patience around the 1097 dollar area, buying near 950 on a pullback or taking theme exposure through an ETF such as DRRAM.