Memory Shortage Crisis Keeps AI Capex Elevated; Mag 7 CEOs Warn on Supply

In two days last month, the CEOs of Microsoft, Meta, Alphabet, Amazon, and Apple each raised the same concern on earnings calls: memory supply is bottlenecked and the crunch is not ending soon. This coordinated message from the largest AI investors reflects a fundamental constraint in the build-out of AI infrastructure. Yet the market has barely priced this risk into memory chipmakers. Micron Technologies trades at 7x forward earnings, an unexpectedly low multiple for a commodity supplier of a critical input to the largest capex cycle in tech history.

Memory demand from data centers remains insatiable. High-bandwidth memory (HBM), critical for AI accelerators, faces multi-year lead times. The Mag 7's comments suggest they see no near-term resolution. Goldman Sachs, Bank of America, and other institutional investors began buying the dip on May 13, accumulating exposure to $MSFT, $GOOGL, $AAPL, and semiconductor equipment makers like $AVGO, signaling confidence that the supply crunch itself validates higher capex budgets and justifies premium valuations for infrastructure names.

Semiconductor equipment makers and substrate suppliers stand to benefit most. The narrative supports continued outperformance of $NVDA (which already hit a record $5.5 trillion market cap following CEO Jensen Huang's last-minute addition to President Trump's China trip), $AVGO (Broadcom), and memory producers who can expand capacity. Energy importers and data center operators face margin pressure from constrained supply, while AI chip designers and chipmakers with proven HBM roadmaps command pricing power.

Sceptics argue memory supply bottlenecks are temporary and will ease by 2026. If Samsung, SK Hynix, or Micron accelerate output faster than expected, spot prices could collapse, pressuring all three. The bulls counter that even a 20-25 percent drop in prices would still leave memory a profitable, capacity-constrained business given AI demand visibility.