This Memory Stock Has Soared From About $40 to More Than $2,300 in a Year. Is It Too Late to Buy?

Sandisk's fiscal Q3 revenue jumped 251% year over year to $5.95 billion, with data center revenue up 233% sequentially to about $1.5 billion and adjusted gross margin reaching 78.4%. Management also guided for adjusted EPS of $30 to $33 in fiscal Q4 and said five multi-year supply agreements cover more than a third of planned fiscal 2027 output. The article is bullish on the AI-driven NAND shortage and the stock's surge, but cautions that the business remains highly cyclical and valuation is already rich.

Analysis

The main second-order effect is that this is no longer just an AI-demand story; it is becoming a contract-structure story. If Sandisk’s multi-year commitments are real and enforceable, the market should start valuing its cash flows more like a specialty component supplier than a commodity memory house, which can justify a much higher multiple than history suggests. That re-rating can persist for months as investors extrapolate supply discipline, but it also creates a crowded-long setup if the next data point merely confirms rather than re-accelerates.

The cleaner relative beneficiary is MU, which retains optionality on the same pricing cycle without as much narrative overhang from a recent spinout and less of the “one-way bet” premium embedded in SNDK. In other words, MU can participate in the shortage while still trading with more skepticism, which usually offers better asymmetry if you believe pricing remains tight into next year. WDC is the obvious second-order loser if capital starts rotating away from legacy storage exposure and toward the purer NAND beneficiaries.

The key risk is not a near-term demand cliff but supply response with a lag. Once contract pricing and public commentary validate that NAND margins are extraordinary, wafer starts and capex tend to follow with a 2-4 quarter delay; that is the window where consensus usually gets caught over-earning and then over-discounting normalization. A sharper risk is customer concentration: if the AI buildout slows or inference economics improve via algorithmic efficiency, the most expensive capacity tier gets deferred first.

My contrarian read is that the move in SNDK is probably more about scarcity optics than durable earnings power. The market is likely underestimating how fast sentiment can swing if sequential growth merely decelerates from absurd levels, even while absolute numbers remain strong. That argues for owning the cycle, but through the least euphoric expression and with defined downside.

Key Decisions for Investors

• Prefer MU over SNDK on a 3-6 month view: long MU / short SNDK as a relative-value pair if you expect the NAND cycle to remain firm but the re-rating in SNDK to compress; target upside is driven by lower multiple risk rather than a better fundamental story.
• If seeking outright exposure, buy MU shares or 3-6 month call spreads rather than chasing SNDK after the vertical move; this keeps participation in the memory upcycle while reducing blow-off-top risk.
• Reduce or avoid WDC as a lagging beneficiary: use any strength to trim exposure, as legacy storage names typically underperform once investors rotate to the purest supply-constrained winner.
• For aggressive traders, sell put spreads on SNDK with 1-2 quarter tenor only on pullbacks, not strength; the setup is attractive only if you are paid for the risk of a sharp multiple reset.
• Set a catalyst watch for the next two quarterly prints: if SNDK commentary shows sequential growth normalizing faster than pricing, de-risk immediately; that is the first sign the shortage premium is peaking.