Micron hits record 84.9% gross margin as memory shortage props up pricing power

Micron reported a gross margin of 84.9% in fiscal Q3 2026, more than doubling from 39% a year earlier and surpassing all major U.S. tech companies. The margin outpaces Meta (81.9%), Nvidia (75%), Broadcom (69.5%), Microsoft (67.6%), and Alphabet (62.4%). Revenue jumped to $41.46 billion in the quarter, up more than $20 billion from the prior period, and net income reached $28.24 billion, up over 100% from the previous record. Micron's stock rose 14% in extended trading and is up over 700% in the past year, pushing market cap well past $1 trillion.

The margin surge reflects a severe memory shortage driven by AI accelerator demand. Data center companies are competing for HBM to feed Nvidia Blackwell systems, Google TPUs, and AMD MI-series processors. Micron is locking in long-term strategic customer agreements (SCAs) with price bands that CEO Sanjay Mehrotra said keep "gross margin for Micron well above our peak quarterly margins in any past cycle." For fiscal Q4, Micron projects a margin of roughly 86%, with expectations that "the market to remain tight beyond 2027."

Customers face mounting pressure. Apple CEO Tim Cook told the Wall Street Journal the iPhone maker will need to raise prices due to what he called an "unsustainable" memory situation. Other device makers and AI-infrastructure vendors have no choice but to accept Micron's premium pricing. Sandisk, another memory vendor, recently raised quarterly gross margin to 78.4% from 51.1% — evidence the entire sector is benefiting from the shortage.

For architects, this marks a historic reversal: a commodity memory business has captured pricing power, shifting leverage to suppliers. The shift from spot pricing to multi-year SCAs suggests memory costs are becoming a fixed capital expense rather than a variable one. This has major implications for data center capex modeling and could accelerate adoption of alternative memory hierarchies or on-chip storage. The continuation of tight supply into 2027 suggests memory cost will remain a significant budget line in AI infrastructure planning.