Are Micron and Sandisk Stocks in a Bubble?

Micron and Sandisk are both benefiting from a memory-chip shortage, with Micron saying it can meet only about half to two-thirds of medium-term demand and projecting HBM addressable market growth from $35B in 2025 to $100B in 2028. Sandisk reported 251% year-over-year growth versus Micron’s 196%, and the article argues both stocks still trade at reasonable forward valuations despite sharp gains of 860% for Micron and 4,160% for Sandisk over the past year. The piece is broadly constructive on both names, but frames the move as cyclical rather than bubble-driven.

Analysis

The core setup is less about "memory is hot" and more about a pricing power inflection in an industry that historically destroys its own upside by overbuilding. The second-order effect is that foundry and equipment capacity discipline matters more than end-demand: if capex stays constrained, margins can remain elevated longer than the market typically expects, which would support both SNDK and MU despite their run-ups. The market is still underappreciating how AI data-center demand shifts the mix toward higher-value bits per dollar of wafer capacity, which structurally favors vendors with tighter product portfolios and better allocation discipline.

The bigger winner may be the ecosystem, not just the chipmakers. NVDA benefits if HBM and SSD constraints force AI builders to keep paying up for memory to preserve accelerator utilization, because memory scarcity effectively taxes AI deployments and reinforces NVDA's system-level pricing power. Conversely, INTC is exposed if the memory supercycle is read-through as a signal that data-center budgets are being reallocated toward accelerators and memory rather than general-purpose compute; that can keep pressure on CPU attach rates and extend the gap in AI share gains.

The main risk is timing, not direction: these are cyclical names, and the right short thesis is not "bubble" but "earnings peak arrives before investors expect it." If HBM supply catches up faster than expected or NAND demand normalizes after a front-loading phase, multiple compression could happen well before revenue rolls over. The market is also prone to extrapolate growth too far out, so the next 1-2 quarters may still look great even as forward returns worsen.

Consensus is likely missing that the best trade may not be outright long memory, but long the names with the best capital discipline and shortest supply response. If management teams use the upcycle to expand wafer starts aggressively, the current scarcity premium can vanish quickly; if they keep supply tight, the cycle can run for several more quarters. That makes guidance cadence and capex commentary more important than the reported growth rate itself.

Key Decisions for Investors

• Stay long MU and SNDK into the next 1-2 earnings prints, but treat rallies as a stair-step trade rather than a hold-and-forget position; upside remains if supply stays constrained, but trim 20-30% on any capex acceleration surprise.
• Pair trade: long MU / short INTC for the next 6-12 months to express AI memory scarcity versus legacy CPU exposure; the risk is MU valuation compression if the market starts pricing peak-cycle earnings.
• Use NVDA as a hedge for AI demand persistence: long NVDA against a basket of memory names if you want exposure to data-center capex without relying solely on the memory cycle; this should outperform if memory remains tight but AI spend broadens.
• Buy downside protection on SNDK via 3-6 month put spreads if implied vol remains below realized vol; memory names can gap sharply on any sign of inventory normalization or capex lift from competitors.
• Monitor management commentary on wafer starts and HBM/MIX shift as the key catalyst window over the next 1-2 quarters; if commentary turns expansionary, reduce exposure fast because the de-rating can begin before reported margins peak.