All research
Equities & sectorsThemes Management

Exploring the Opportunity in Memory Stocks

SemiconductorsArtificial intelligenceAI infrastructure

Our summary

Memory stocks could become strategic beneficiaries of generative AI as constrained high-speed memory shifts from a commoditized component to a critical bottleneck in computing.

  • GPUs powering models such as ChatGPT, Gemini, and Claude require high-speed memory to process the data needed for real-time text, image, and code generation; insufficient memory creates the “Memory Wall.”
  • Global memory demand is projected to exceed supply through 2030, with wafer supply reportedly trailing demand by 20% as of March 2026.
  • Micron’s fiscal-year revenue rose to $133.2 billion for the year ended September 3, 2026, from $37.4 billion a year earlier, while SK Hynix is expected to reach approximately KRW 334 trillion in revenue this year versus KRW 97 trillion last year.
  • The industry may become less cyclical as suppliers sign multi-year agreements featuring cash prepayments, take-or-pay provisions, and pricing mechanisms; Micron has signed 26 Strategic Customer Agreements expected to represent 35% of revenue through 2030.
  • SanDisk has agreements with eight data-center and cloud customers, lasting up to five years with a weighted average duration above four years and expected gross margins of 80%.

Takeaway: Investors can access the theme through HBM suppliers such as Micron and SK Hynix or NAND suppliers such as Samsung and Western Digital, while a diversified memory ETF may reduce stock-specific risk.

Published by Themes Management. Summary by HF Best Ideas.

More from Themes Management

Latest research

Share this summary