MSFT, GOOGL, Meta, AAPL, AMZN All Signal Memory Shortage; Micron at Only 7x Earnings

In consecutive weeks last month, the chief executives of Microsoft, Google, Meta, Amazon and Apple each cited identical concerns on their earnings calls: memory capacity is constrained and there is no near-term relief in sight. This coordinated signal from the largest cloud-infrastructure buyers in the world has crystallized a critical gap in the AI capex narrative.

Historically, when demand for computing infrastructure accelerates, semiconductor supply follows within 12 to 18 months. But memory makers have flagged persistent bottlenecks extending into 2026 and beyond, suggesting supply has lagged more severely than consensus expected. The issue spans both volatile memory (DRAM) and non-volatile storage (NAND), creating a structural underinvestment cycle.

Micron Technology (MU), the largest US memory producer, trades at approximately 7x forward earnings despite this secular backdrop. Competitors like SK Hynix and Samsung have seen their valuations reprice higher on similar commentary. The valuation disconnect suggests either (a) the market is skeptical that memory scarcity persists, or (b) investors are underweighting Micron's exposure to the highest-growth end-markets (AI training, data-center inference). Energy-constrained AI deployments and longer model-training cycles have pushed memory density requirements beyond typical refresh rates.

Sceptics point out that capex cycles have surprised before and that Micron's historical return on invested capital has been weak relative to more fabless peers. However, the unanimous testimony from five mega-cap customers on the same topic within days suggests this is not isolated commentary but rather a sector-wide structural shift that could justify multiple expansion for pure-play memory names.