HBM market to hit $100B by 2027: why the AI memory cycle is different this time
Beth_Kindig · x · 2026-09-26
I/O Fund analyst Beth Kindig breaks down why the AI memory cycle may break its historical pattern:
- Exploding demand: HBM grew from a $4B market in 2023 to $34.6B in 2025, with Micron projecting $100B by 2027; SK Hynix and Micron shares rose roughly 400% and 500% in a year.
- Inference is the driver: model complexity and KV cache are the two main memory consumers, and context windows have ballooned — OpenAI's leading models went from 230-260X growth in three years to 1.05M tokens, with Meta's Llama 4 Scout at 10M.
- Supply stays tight: constraints are spreading from HBM to conventional DRAM and NAND, with new capacity possibly not arriving until 2028, keeping demand above available supply for years.
- But growth is cooling: despite the cycle looking far from topped, more memory firms are signing long-term agreements and reported growth rates are decelerating — an unusual setup.
More from Infra
- Altman says OpenAI aims to replace Nvidia in the medium to long term — zephyr_z9 · 2026-11-11
- Chinese University Bought Nvidia H200 Chips Despite US Blacklist — pstAsiatech · 2026-11-11
- Baidu's ERNIE Evolve tops public bests on NanoGPT Speedrun, brings RSI to real inference optimization — xiaohu · 2026-11-11
- Baidu says RSI auto-optimized ERNIE inference, 6.24x decode speedup on Kunxin P800 — xiaohu · 2026-11-11
- Nvidia's supply commitments jump from $119B to $279B in one quarter, locking up 2027-28 memory — tengyanAI · 2026-11-11
- Train 27B models on a 24GB GPU: ternary optimizer cuts memory to 1 byte/weight — oguzeldereli · 2026-11-11