Key Highlights

  • Micron Technology, Inc. trades on the Nasdaq Global Select Market under the ticker symbol MU; all financial data referenced is sourced from the company's Form 10-Q (Q2 FY2026) and Form 10-K (FY2025).
  • Q2 FY2026 Revenue reached $23.86 billion, a 196% increase year-on-year, as AI-driven data centre Demand for DRAM and NAND outpaced industry Supply.
  • The filing states explicitly: AI-driven memory and storage growth is outpacing industry supply, leading to supply allocation decisions that affect customers across end markets.
  • The Cloud Memory Business Unit (CMBU), which includes HBM for all data centre customers, grew revenue from $1.87 billion in FY2023 to $13.52 billion in FY2025 — a 623% increase in two fiscal years.
  • Average selling prices for DRAM rose in the mid-110% range year-on-year in Q2 FY2026, with NAND ASPs rising in the high-70% range — pricing signals of a market where demand is structurally outrunning supply.

 

There is a category of infrastructure that modern civilisations depend upon so completely that its absence is inconceivable, yet its presence is rarely discussed. Electricity is one such category. Semiconductor memory is becoming another. Micron Technology, Inc. (NASDAQ: MU), headquartered at 8000 S. Federal Way, Boise, Idaho, sits at the centre of this transformation. Every large language model running today — every query answered, every image generated, every AI-driven search result returned — is processed through memory. Not metaphorically. Physically. The DRAM and HBM that Micron manufactures is the medium through which artificial intelligence thinks.

The Architecture of Dependency

To understand why memory is indispensable to AI, one must understand how large-scale AI inference and Training actually work at the hardware level. A graphics processing unit — the chip that performs the billions of matrix multiplications required to run an AI model — is only as fast as its ability to move data in and out of memory. The bottleneck in AI computation is almost never the GPU's arithmetic capability. It is memory bandwidth. The speed at which data can be fed to and retrieved from the processor determines the practical throughput of the entire system.

High-bandwidth memory, or HBM, addresses this bottleneck by stacking DRAM chips vertically on top of the GPU using a technology called through-silicon vias. This physical proximity dramatically reduces the distance data must travel, increasing bandwidth by orders of magnitude compared to conventional DRAM. Micron began Volume production of its 8-high 24GB HBM3E in 2024 and by the fourth quarter of FY2025, HBM3E 12-high represented the majority of its HBM shipments. The company has delivered samples of HBM4 36GB 12-high to multiple key customers to power next-generation AI platforms. This is not a product roadmap item for the distant future. It is in production today.

The Supply Constraint That Changes Everything

Micron's Q2 FY2026 10-Q filing contains a statement that should be read carefully by anyone trying to understand the current state of the AI infrastructure market: AI-driven growth in the data centre has accelerated demand for memory and storage at a rate greater than our ability and the industry's ability to increase supply. This has led to decisions on supply allocation that may impact certain customers and end markets as the overall market demand for memory and storage exceeds overall industry supply.

This is not standard boilerplate language. This is a company telling its investors that it is rationing its most important product because demand exceeds what it and all of its competitors can collectively produce. In an industry historically characterised by overcapacity and price wars, this represents a structural inversion. The conventional memory cycle runs as follows: prices rise, all three major producers invest in new capacity, supply floods the market, prices collapse, producers cut Capital Expenditure, supply tightens, and the cycle repeats. What Micron's filings suggest is that the AI demand wave is arriving faster than new capacity can be constructed.

From Data Centre to Intelligent Edge

The demand is not confined to hyperscale data centres. Micron's 10-K for FY2025 describes a simultaneous demand wave across four distinct business units. The Cloud Memory Business Unit serves hyperscale cloud customers and all data centre HBM demand. The Core Data Centre Business Unit serves mid-tier cloud, enterprise, and OEM customers. The Mobile and Client Business Unit is responding to AI capabilities being embedded in smartphones, where on-device AI inference is driving higher DRAM content per handset. The Automotive and Embedded Business Unit is serving the growing memory requirements of autonomous driving systems and advanced driver-assistance applications.

The FY2025 10-K data illustrates this breadth precisely. DRAM revenue grew from $10.98 billion in FY2023 to $28.58 billion in FY2025. NAND revenue grew from $4.21 billion to $8.50 billion over the same period. These are not incremental improvements. They represent the financial expression of a fundamental shift in what computing infrastructure requires, driven by AI workloads that are memory-intensive in a way that no previous generation of applications has been.

Why Memory Is Different This Time

Previous periods of memory demand growth were driven by consumer device cycles — personal computers in the 1990s, smartphones in the 2000s. These were elastic demand cycles, meaning that when prices rose, consumers and manufacturers could delay purchases or reduce specifications. The demand was discretionary at the Margin. AI infrastructure Investment is not discretionary. When Microsoft, Google, Amazon, and Meta are building AI infrastructure to remain competitive in a market where falling behind carries existential consequences, their memory procurement decisions are driven by strategic necessity, not price sensitivity. The demand curve for AI memory does not slope downward in the conventional sense.

This structural distinction is what separates the current cycle from its predecessors and why Micron's Q2 FY2026 results — 196% revenue growth, gross margins of 74%, Operating Income of $16.1 billion in a single quarter — deserve to be read not as the peak of a familiar cycle, but as the early chapters of a structurally different story.

Disclaimer: This article is for informational and educational purposes only and does not constitute financial, investment, or any other form of professional advice. All data and figures are sourced exclusively from Micron Technology Inc.'s (NASDAQ: MU) Form 10-Q for the quarter ended February 26, 2026 and Form 10-K for the fiscal year ended August 28, 2025, as filed with the US Securities and Exchange Commission.