Sandisk: The AI Memory Supercycle Is Just Getting Started

Management guides gross margins of 64.9–66.9% for Q3 FY26 (current gross margins >50%) and a guidance midpoint implying 170.6% YoY revenue growth. Sandisk is transitioning from volatile consumer demand to secular, higher‑margin data‑center/AI infrastructure sales supported by long‑term agreements and new products (Stargate SSDs). The combination of explosive NAND demand, strong margin expansion and durable contracts materially improves the company's revenue and profitability outlook and is likely to drive notable upside for the stock and memory-sector peers.

Analysis

Winners extend beyond the obvious NAND vendor: hyperscalers and AI OEMs lock in lower long-term TCO when vendors offer higher-density NVMe assemblies, which structurally favors suppliers that can deliver integrated firmware + hardware stacks (controller/IP owners and OSAT partners) over pure-play wafer suppliers. Second-order beneficiaries include packaging/test vendors and PCIe/NVMe controller designers whose revenue per bit rises as hyperscalers shift spend from raw wafers to finished, rack-ready storage modules. Conversely, firms that depend on consumer NAND ASPs or that lack direct hyperscaler contracts face margin compression and inventory write-offs faster than the market currently prices.

Key risks are classic semiconductor cyclicality and customer concentration. A meaningful capex re-response by incumbents or a pivot by a hyperscaler from training-heavy expansion to software optimization can depress NAND ASPs and margins within ~9–15 months; similarly, a single large hyperscaler inventory destocking event can create a 6–12 week demand hole that cascades through the supply chain. Near-term catalysts to watch are (a) the next quarterly bookings cadence from the top three hyperscalers and (b) any incremental long-term supply contracts disclosed by Micron or Samsung — either can validate or puncture the revenue durability story within 1–3 quarters.

From a positioning standpoint, trade ideas should capture asymmetric upside while protecting against short-cycle supply risks: favor structures that monetize the AI-driven re-rating but cap carry if the cycle rolls. Monitor implied vol around earnings and guidance dates — the market often compresses IV after the first two beat cycles, which makes selling premium into conviction a viable tactic. The contrarian read is that while secular demand is real, the market may be underpricing the speed at which suppliers can flex capacity and underestimating single-customer execution risk; margin durability is not guaranteed beyond the current contracting wave and should be treated as a priced, not permanent, asset.

Key Decisions for Investors

• Long SNDK equity exposure via staggered LEAPs: buy SNDK Jan-2027 LEAP calls sized for 3–5% portfolio exposure and sell nearer-term calls (6–9 months) to finance ~30–50% of premium. Timeframe: 9–18 months. R/R: captures multi-quarter structural rerating while reducing carry; downside limited to premium paid if NAND cycle reverts quickly.
• Paired trade (relative): long SNDK / short MU, equal dollar exposure. Timeframe: 3–9 months. R/R: benefits if NAND-led AI demand outpaces DRAM or if SNDK converts hyperscaler wins into higher ASPs; risk if DRAM also rerates or MU reports better-than-expected hyperscaler spend—set stop at 10–12% adverse move.
• Protective collar for core positions: buy 1-year 20% OTM put and sell 1-year 35% OTM call on SNDK to lock in downside while funding upside participation. Timeframe: 12 months. R/R: caps downside beyond strike while allowing participation in further multiple expansion from continued AI spend.
• Volatility sell into confirmed beats: if SNDK posts two consecutive quarters of guidance beats, sell 30–60 day strangles around earnings (size modest, <1% portfolio), collecting premium before the next guidance. Timeframe: event-driven (days-weeks). R/R: collects IV premium that historically compresses after confirmed secular narratives; risk of gap moves — use defined-risk wings or hard stops.