Unpacking the Latest Pullback in Micron Stock

Micron shares have fallen 27% since mid-June, but the article argues the sell-off is driven by investor skepticism about AI capex rather than Micron fundamentals—despite fiscal Q3 sales rising 345% to $41.5B and EPS surging 1,200% to $25.11. It cites continued AI infrastructure spending (Alphabet AI capex up to as much as $205B in 2026 and Amazon capex ~ $220B, +66% vs 2025) and claims memory shortages may persist through 2027 (potentially to 2030 at SK Hynix). With Micron trading around 19x trailing P/E versus a ~35x tech average, the piece frames the pullback as a buying opportunity.

Analysis

The market is mixing up two different trades: AI infrastructure duration and memory cyclicality. If hyperscaler capex remains elevated, the near-term beneficiary set is broader than the usual compute names — DRAM/HBM pricing power can keep margin leverage high even if unit growth slows, which is why MU can outperform on a softer tape while NVDA/SMH remain hostage to valuation compression. The second-order winner is likely the memory supply chain (SKHYV-adjacent names, equipment, substrates) because the real bottleneck in 1H-2025 is not demand destruction but allocation discipline.

The risk is that memory is the first place a capex slowdown shows up once AI ROI debates turn from narrative to budget review. Over 1-3 months, watch whether hyperscaler capex commentary at the next earnings cycle confirms incremental spend or merely maintains it; a downgrade in 2026 plans would hit MU faster than software or networking because inventory and pricing both re-rate quickly. Over 6-18 months, the key falsifier is a flattening in HBM/DRAM contract prices or rising channel inventory — that would compress MU's multiple back toward a cyclical semiconductor discount.

Contrarianly, the selloff may be overdone in the very short run because positioning is crowded against semis, but that does not mean MU deserves a tech multiple. The right framing is not 'AI is over' but 'AI winners are shifting from compute scarcity to memory scarcity.' Apple passing through higher memory costs is a useful read-through: if consumer OEMs start resisting price increases, memory demand still exists, but elasticity and mix can cap upside for AAPL while reinforcing MU's pricing power.

Key Decisions for Investors

• Long MU vs. short SOXX/SMH for a 1-3 month mean-reversion trade: asymmetric if AI capex stays firm and memory pricing remains tight; cut if MU underperforms semis by ~8-10% after the next hyperscaler earnings cycle.
• Add MU on further post-earnings weakness only if management does not guide down 2026 margin assumptions; otherwise treat this as a tactical trade, not a structural re-rating.
• Watch/alert: if next-quarter capex commentary from GOOGL/AMZN implies flat-to-down 2026 spend, reduce MU exposure immediately — that is the first-order falsifier for the memory shortage thesis.
• Prefer a basket long in memory beneficiaries over broad AI beta: MU and SKHYV-related exposure over NVDA if the objective is to isolate pricing power rather than valuation expansion.
• For a cleaner hedge, pair long MU against long-duration AI hardware beta (NVDA) when semis are weak: MU has nearer-term cash-flow torque from pricing, while NVDA is more exposed to multiple compression.