Sandisk Stock Is Still Down 50% From Its Highs. Is the Memory Winner Still a Screaming Buy Before September?

The article argues that Sandisk (SNDK) remains undervalued after a ~50% drawdown from its all-time high, citing sustained NAND/SSD pricing strength as AI data-center demand outstrips supply. It points to Amazon raising 2026 capex from $200B to $220B due to higher memory chip prices and claims Wall Street expects SNDK revenue growth of 151% in fiscal 2027 (ending June 30). Overall, the piece expects the memory-chip pricing imbalance to persist until well after 2027, which it says supports a stock rebound beyond the current sell-off.

Analysis

This is less a “AI demand” story than a pricing-power story inside a commodity chain. The key mechanism is that NAND scarcity is transferring value from buyers to suppliers, and the first-order winners are the names with enough scale to lock in contracts before capacity catches up; the second-order loser is the hyperscaler/customer base, which absorbs higher storage BOMs and may be forced to ration capex into the highest-ROI workloads. That matters because the market often prices AI spend as a pure demand tailwind, but memory inflation can quietly lower the number of deployed clusters per dollar of capital outlay.

The bigger cross-asset read-through is that AMZN is probably the cleaner canary than NVDA for this particular pressure point. Rising storage costs hit cloud economics through longer payback periods and higher depreciation, which can compress free-cash-flow conversion even if headline capex still grows; over the next 1-3 quarters that’s a sentiment risk, not an earnings-crater risk. For NVDA, the effect is more indirect: higher total server cost can delay some deployments, but it is not yet enough to change the GPU shortage narrative unless we see customers start cutting full-system orders.

Contrarian take: the market may be too eager to fade SNDK simply because memory has historically been cyclical. The cycle could stay tight longer if hyperscalers pre-buy inventory and suppliers remain disciplined, but the flip side is that the stock may already be discounting an extended supercycle, so the thesis breaks quickly if contract pricing stops rising or if utilization gets nudged up. The important falsifier is not macro noise; it is a 1-2 quarter stall in NAND pricing or margin guidance that signals supply is finally overtaking demand.

Key Decisions for Investors

• Long SNDK on pullbacks over the next 1-3 weeks; best entry is after any post-rally consolidation, with upside to re-rate toward prior highs if NAND pricing stays firm. Falsifier: flat-to-down contract pricing or gross margin guide compression on the next earnings call.
• Pair trade: long SNDK / short AMZN in a small size over 1-3 months to express the relative beneficiary vs. the capex consumer. Risk/reward improves if AMZN commentary shows another round of 2026 capex inflation without corresponding AWS margin protection.
• Do not chase NVDA as a direct beneficiary of this theme; treat memory inflation as a mild deployment headwind rather than an earnings driver. If you already own NVDA, use it as a hedge against any overweight in AI infrastructure beneficiaries, not as the primary expression.
• Set a watch item on SNDK spot/contract NAND pricing and customer inventory turns for the next 60-90 days; if pricing rolls over before capacity additions are visible, trim the long. If pricing accelerates again, add on strength rather than waiting for the next dip.