Sandisk and Micron Dominated the First Half of 2026. Which Is the Best Buy Now?

Micron (MU) is up ~325% YTD and Sandisk (SNDK) nearly ~900%, as the AI build-out drives unprecedented demand for DRAM and NAND memory tied to data center capex. The article cites Nvidia projecting global data-center capex of $3T–$4T annually by 2030 and forecasts Wall Street growth of +78% for Micron and +122% for Sandisk next year. Valuations are described as still inexpensive (MU ~8x forward earnings vs SNDK ~13x), with upside framed as Sandisk potentially doubling and Micron potentially tripling if they re-rate. Overall bias favors Sandisk as the more focused NAND play, despite higher potential upside for Micron.

Analysis

The market is likely underestimating how asymmetric the memory cycle is for DRAM vs. NAND. AI capex supports both, but the cleaner beneficiary is the name with the most exposure to advanced DRAM and high-bandwidth content; the NAND-only player has more end-market fragmentation and typically loses pricing power faster when enterprise SSD orders normalize. In other words, this is not a pure "AI demand" trade — it is a relative-scarcity trade inside memory, and the scarcity premium can unwind quickly once supply responses show up.

The immediate risk is that consensus is extrapolating peak margin momentum into a structurally higher multiple, when memory earnings are still cyclical and supply elastic with a lag. Over 1-3 months, the key catalyst is whether contract pricing and gross margin guides stay firm into the next print; over 6-18 months, the bigger issue is capex response from the broader memory complex, which can cap upside even if end demand stays healthy. If pricing flattens or inventory days rise, these stocks can de-rate before the earnings estimates roll over.

The contrarian view is that the AI build-out does not translate 1:1 into memory demand per dollar of spend, especially if hyperscalers optimize storage architecture or shift mix toward compute-heavy deployments. The current narrative also ignores that comparing these names to mid-20s forward P/Es is the wrong anchor; memory deserves a cyclical framework, not a software multiple. The better expression is relative value, not blind beta chasing.

Key Decisions for Investors

• Prefer long MU / short SNDK as a 3-6 month relative-value pair: MU has better product mix, less NAND-only cyclicality, and more direct AI content leverage; SNDK is more vulnerable if NAND pricing stalls.
• Use MU on pullbacks rather than strength-chasing: buy into broad semis weakness and hold into the next earnings cycle; upside is tied to sustained gross margin expansion, while the main falsifier is softer forward pricing guidance.
• For investors who want convexity, consider a modest 3-6 month MU call spread instead of outright equity: the trade captures continued estimate revisions while defining downside if the cycle peaks early.
• Set alerts on NAND ASP commentary and inventory days for both MU and SNDK; a flattening of pricing or a guide to slower margins would be the earliest sign the move is overextended.
• If you need sector beta, own MU over SMH rather than chasing the highest-run meme names: the risk/reward is cleaner because the market is still underappreciating the quality gap within memory.