Memory Chips Remain Severely Constrained; MU Trading at 7x Earnings Discount

Memory supply constraints have become the silent constraint on AI infrastructure buildout. Last month, the CEOs of Microsoft, Meta, Alphabet, Amazon, and Apple all signaled the same message on earnings calls: DRAM and storage capacity remain severely bottlenecked and show no immediate relief. This isn't speculation about chip demand; it's first-hand testimony from the largest consumer of AI infrastructure that supply remains inelastic.

The market's response has been paradoxical. While semiconductor equities have raced higher on AI enthusiasm, Micron Technology continues to trade at only 7x earnings, a steep discount relative to the supply-demand imbalance these executives describe. The disconnect suggests either skepticism about memory capex near-term, or a belief that Micron's operational execution will disappoint relative to the tailwind.

This matters for AI infrastructure investors because it implies memory supply will remain a binding constraint on deployment speed throughout 2026. Companies like Nvidia, AMD, and foundries are already allocating production around memory availability. The longer memory remains scarce, the higher the premium for next-gen accelerators with embedded storage, and the longer capex cycles extend.

Bears note that capex cycles eventually saturate; if demand for AI chips peaks before memory supply normalizes, Micron and its peers could face a sharp downgrade. However, the chorus of CEOs publicly acknowledging constraints is unusual enough to suggest the problem is structural, not cyclical.