Micron has sold out its entire high-bandwidth memory output for 2026 — a supply shortage now shaping AI infrastructure timelines across North America, Europe, and Asia.1 The constraint is structural: AI inference workloads run continuously at scale, and every query to a deployed model requires memory bandwidth. Production cannot currently match that demand.
Micron's Q3 FY2026 guided revenue of approximately $33.5B would exceed any full prior fiscal year.1 The company has begun shipping HBM for Nvidia's Vera Rubin platform, confirming the next generation of AI infrastructure is already ramping.1
The demand catalyst is inference, not training. Sandisk's CEO identified inference and reasoning workloads as the fastest-growing segment of global data center demand.1 Unlike training runs — which are finite — inference scales with every user query, every hour, across every timezone.
Sandisk shares surged over 700% year-to-date in 2026.1 Micron crossed a $1 trillion market capitalization.1 Both trade at forward price-to-earnings multiples of approximately 10-11x despite those gains. Earnings growth is outrunning stock appreciation — an unusual signal in any market cycle.
For global operators — from hyperscalers in Virginia to cloud buildouts in Singapore and Frankfurt — the implication is the same: HBM availability is now a scheduling variable, not a budget line. Infrastructure teams planning AI inference capacity must treat memory supply as a hard constraint alongside compute allocation.
The critical variable is supply. With output sold before the year begins, pricing power has shifted entirely to the supplier. Elevated HBM average selling prices flow directly to gross margins. Multi-year supply contracts at current price levels would lock in the supercycle thesis through 2027 and beyond.
Quarterly gross margin trends over the next four quarters will be the clearest signal of whether this cycle persists.
The memory layer — long treated as commodity infrastructure in every major industrial economy — has become the rate-limiting factor in global AI scaling.
