Western Digital, Micron, Sandisk hover at all-time highs as AI trade reignites
1 min read
The story
The memory sector is trading at or near all-time highs as AI infrastructure buildout — particularly HBM and enterprise SSD demand — re-energizes the group. Micron (MU) and Western Digital (WDC) each posted ~50% YoY revenue growth in their most recent fiscal years with gross margins near 40%, signaling a genuine upcycle rather than a dead-cat bounce. Sandisk (SNDK), recently spun off from WDC, grew revenue only 10% YoY and is running a -22% net margin and -$11.32 diluted EPS, suggesting it remains in a restructuring/cost-absorption phase.
The setup is a classic upcycle-at-highs tension: MU and WDC have the revenue and margin momentum to justify elevated multiples if AI capex holds, but buying at all-time highs leaves little cushion for any demand-softening signal or inventory build. SNDK is the wildcard — if its cost structure normalizes post-spinoff, it could re-rate sharply, but the negative EPS makes it speculative. Watch for MU's next earnings print and any hyperscaler capex commentary as the key near-term catalysts.
The case — both sides
2 of 3 names have verified EOD history. The basket chart is hidden rather than showing illustrative data.Missing: SNDK
MU's 48.9% YoY revenue growth and near-40% gross margins, driven by HBM and data-center NAND demand, suggest the upcycle has fundamental support that could sustain elevated multiples if AI capex continues its current trajectory.
All three names are already at or near all-time highs, and memory cycles are historically prone to sharp reversals — any signal of hyperscaler capex moderation or NAND/DRAM oversupply could rapidly unwind the premium multiple the group currently commands.
The house read
Leans bullWith MU and WDC posting ~50% revenue growth and approaching all-time highs on AI momentum, the question is whether the memory upcycle has room to extend or is priced to perfection at current levels.
Wrong ifA broad memory-sector pullback or inventory-build warning from any hyperscaler would hit all three names simultaneously, collapsing the pair's relative performance thesis and turning the long legs into losses.
Published read · research, not advice