AI Memory Bottleneck Map | Capital Structure Maps #6
- Jenny LEE
- Mar 25
- 1 min read
AI doesn’t scale with compute alone.
It scales with memory bandwidth.
As models grow larger,
the constraint shifts from processing power
to how fast data can move in and out of memory.
Compute is abundant.
Bandwidth is not.
Key layers:
HBM (High Bandwidth Memory):
$MU $SKHYNIX $SSNLF
Memory Packaging & Integration
$AVGO $AMKR $TSM
Memory Controllers & Interconnect
$MRVL $NVDA
AI Systems Integration
$SMCI $DELL
About This Series: Capital Structure Maps is an ongoing research series examining the structural foundations behind major technology and industrial cycles. Each map outlines the constraint chains and capital pathways that connect infrastructure, semiconductors, energy systems, automation, and end-market applications. The goal is to provide a structural view of where capital must flow as new technologies scale into the real economy. A new map will be published weekly as part of the Equity Regime research framework.
