From Memory Headwind to Structural Tailwind

Research evidence excerpt

From Memory Headwind to Structural Tailwind

Asia Pacific InsightM

Investment Thesis

Chipflation – Navigating A Memory Crisis

AI is turning memory into a structural bottleneck. Over the past several months, our

Global Tech team's work on the memory supply chain (see Chipflation – Navigating a

Memory Crisis) has led us to reassess our view on the broader hardware sector. In

hindsight, had we reached these conclusions before publishing our November 2025 report

(see Global Technology Hardware: Memory Takes A Bite Out of Hardware Margins), we

likely would not have downgraded Lenovo.

Our previous thesis assumed that the current memory cycle would resemble prior

upcycles, during which elevated component costs ultimately compressed OEM margins.

We now believe that assumption no longer holds. Unlike previous cycles, AI-driven

demand is tightening supply across HBM, DRAM, and enterprise SSDs simultaneously,

while new capacity requires years, rather than quarters, to build, qualify, and ramp. As a

result, we view the current environment less as a conventional semiconductor cycle and

more as a structural shift in the industry's supply-demand balance.

This distinction has important implications for OEM profitability. In previous memory

cycles, customers generally viewed higher memory prices as temporary and often delayed

purchases in anticipation of lower prices. This dynamic limited OEMs' ability to pass

through higher component costs and often resulted in margin pressure. Today, customer

behavior appears meaningfully different. With limited expectations for a meaningful

decline in memory prices over the foreseeable future—and the possibility of further

increases—customers have become more willing to accept higher system prices or