Article by Xiao Bing
In one quarter, revenue reached $41.4 billion, a staggering 346% year-over-year increase. Gross margin was 84.9%, with net profit of $28 billion.
These figures would be staggering for any tech company; for Micron, they are almost absurd. Two years ago in the same quarter, the company reported revenue of $9.3 billion, a gross margin of 39%, and net profit of less than $1.9 billion.
But the 15% post-market surge in stock price indicates that market excitement has shifted away from Q3 itself. What truly ignited sentiment was the Q4 guidance: $50 billion in revenue (midpoint), a gross margin of approximately 86%, and earnings per share of $31.
The storage supercycle is currently underway.
Behind the Numbers: An Anatomy of a Money Printer
Breaking down Micron's earnings report, every business segment is speaking with exponential growth.
DRAM generated $31.3 billion in revenue, accounting for 76% of total revenue, with an average selling price that rose more than 60% quarter-over-quarter. NAND reached $9.9 billion, also significantly exceeding expectations. However, the most explosive growth was within the business segments: core data center revenue exceeded $25 billion in a single quarter, annualizing beyond $100 billion—an increase of more than sevenfold compared to $1.53 billion in the same period last year. Data center SSD revenue surpassed $5 billion, doubling quarter-over-quarter. Automotive and embedded businesses also reached $4.63 billion, growing more than threefold year-over-year.
Operating cash flow of $25.39 billion, adjusted free cash flow of $18.3 billion. Cash and investments on hand total $30.2 billion, net cash of $24.4 billion. Debt was reduced by $4.4 billion this quarter, with all three major rating agencies simultaneously upgrading the rating to BBB+.
This earnings report makes the old saying "storage is a cyclical industry" obsolete.
16 SCA: Micron is rewriting its business DNA
Digital prices can be explained by supply and demand imbalances, but not by sudden shifts in business models.
Micron announced on the earnings call that it has signed 16 strategic customer agreements (SCAs). These are binding, take-or-pay contracts spanning from 2026 through the end of 2030, covering approximately 20% of DRAM shipments and one-third of NAND shipments. The agreements include four hyperscale customers and three mid-sized customers, with the remainder coming from the automotive industry.
CFO Mark Murphy disclosed a new metric for the first time: Remaining Performance Obligations (RPO). As of the end of Q3, RPO stood at $5 billion; including newly signed agreements after the quarter ended, this figure surged to approximately $100 billion. Management explicitly stated that actual revenue will "significantly exceed" the contract values underlying the RPO.
More importantly, the goal is for Micron to increase the percentage of revenue covered by SCA to over 50%.
What does this mean?
For the past 40 years, the storage industry’s business model has been built on spot pricing and short-term contracts, with wild price fluctuations being the norm—leading to a persistent "cyclical discount" on stock valuations. SCA fundamentally transforms memory chips from commodities into pre-sold infrastructure resources, exactly like cloud providers signing long-term contracts to purchase electricity or fiber optics.
If Micron can achieve a 50% SCA coverage rate by 2027, its revenue predictability will approach that of an enterprise software company, yet its current forward P/E ratio is only slightly above 10x—making it the cheapest among all trillion-dollar market cap companies.
Supply is unable to keep up with demand
A statement by CEO Sanjay Mehrotra during the earnings call is worth reflecting on: Micron currently "does not see a point at which supply will catch up to growing demand." He expects shortages in DRAM and NAND to persist beyond 2027.
This is not empty talk. Micron has already sold out its entire 2026 HBM capacity at fixed prices and quantities. The production ramp-up speed of HBM4’s 12-layer stacking is twice that of the previous generation, HBM3E, and it has already generated over $1 billion in revenue. Management forecasts the total HBM market to grow at a compound annual rate of approximately 40%, rising from $35 billion in 2025 to $100 billion in 2028—two years earlier than previously expected.
Supply-side constraints are physical. Building an advanced memory wafer fab requires 3–4 years and tens of billions of dollars in investment. Micron has raised its fiscal 2026 capital expenditure to approximately $27 billion (after government subsidies), with new facilities in Idaho and Japan ramping up, adding $100 million to $200 million in startup costs per quarter. However, capacity expansion is slow, while demand for memory from AI data centers is exponential.
Global hyperscale cloud providers will collectively spend over $725 billion on AI data center capital expenditures by 2026, and all of this spending will ultimately flow through memory chips.
The timeline of Anthropic's trading is intriguing.
Two days before the earnings report, Micron announced a strategic agreement with Anthropic, encompassing joint design of storage architectures, a multi-year supply contract, deployment of Claude within Micron’s infrastructure, and a strategic investment in Anthropic’s Series H funding round. With this, all three global leading HBM suppliers—Samsung, SK Hynix, and Micron—have become strategic investors in Anthropic’s Series H round.
No AI lab has simultaneously secured all three manufacturers of the global HBM supply chain. This deal’s significance extends far beyond the supply contract itself—it marks the moment when AI companies began treating the memory supply chain as a strategic asset. As the efficiency of model training and inference becomes increasingly dependent on the performance of the memory subsystem, the bargaining power of memory manufacturers will only continue to rise.
Concerns remain, but the weights are shifting.
The historical lesson from the storage industry is clear: every super cycle has been followed by a crash triggered by overcapacity. Will the synchronized expansion by Samsung, SK Hynix, and Micron ultimately replay the stories of 2018 or 2022?
The difference lies in structural changes on the demand side. In the past, storage cycles were driven by consumer electronics; once smartphone and PC shipments peaked, demand plummeted sharply. However, storage demand from AI data centers continues to compound—each new model is larger, each inference request consumes more tokens, and each agent requires a longer context window. Added to this, AI PCs are raising the standard memory specification from 16GB to 32GB, while flagship smartphones simultaneously drive up DRAM demand, structurally elevating the floor of storage demand.
The SCA framework is Micron's institutional hedge against cyclical risk. Even if demand growth slows, take-or-pay contracts ensure a floor for revenue. This does not eliminate cycles, but it significantly narrows their amplitude.
Micron’s earnings season has ended, but the questions it raised have only just begun: As storage evolves from a commodity into a strategic resource requiring advance booking, does the industry’s valuation language need to be rewritten?
Based on an annualized EPS of $31 according to Q4 guidance, Micron’s current forward P/E ratio is around 10x. NVIDIA, which also benefits from AI and faces similar supply-demand imbalances, has a forward P/E ratio exceeding 30x. This gap reflects both market pricing of the belief that “cycles will eventually return” and an outdated narrative that the SCA framework is gradually disproving.