Sandisk has rallied about 600% since January as AI-driven demand for memory hardware boosts SSD pricing and margins. Fiscal Q3 revenue rose 251% year over year to $5.95 million, gross margin expanded 55.9 points to 78.4%, and operating income jumped 319% to $4.11 billion. The article remains constructive near term but warns the memory cycle is highly volatile and that investors may want to take profits as AI spending and pricing power normalize.
SNDK is the cleanest public-market expression of the AI storage bottleneck, but the real trade is not simply “more AI = higher memory prices.” The stronger second-order effect is that a constrained memory supply chain forces hyperscalers to design around efficiency rather than brute-force capacity, which favors vendors with the highest mix of premium SSDs and the most pricing power. That makes the rally more durable than a pure momentum move in the near term, but also means the margin peak is likely set by supply discipline, not demand alone.
The market is likely underestimating how fast this can unwind once foundry and packaging capacity catches up. Memory has a long history of overshooting on both sides; the risk is not a gradual normalization but a sharp step-down in pricing when customers finish front-loading inventory and procurement teams pivot from urgency to cost control. That inflection could arrive before the AI capex cycle itself rolls over, so the stock can correct well in advance of any broad slowdown in AI spending.
WDC remains the cleaner relative short if the thesis is that AI demand is not a permanent step-function for the entire storage complex. HDDs do not participate equally in the near-term AI buildout, and any easing in enterprise storage budgets would likely compress the valuation gap faster on the lower-quality asset mix. NVDA and the broader compute stack are still the better structural longs because they sit closer to the control point of the spend; SNDK is more exposed to a commodity-like pricing regime once the supply shock fades.
Consensus is focused on how big the AI opportunity is, but the miss is that memory vendors often capture the least durable economics at the peak of a cycle. The right posture is to respect the trend but fade it with asymmetric downside protection rather than outright fighting it. If the market starts to price in 2027-2028 normalization now, the multiple can compress before fundamentals do.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Overall Sentiment
mildly positive
Sentiment Score
0.48
Ticker Sentiment