Billionaire Israel Englander Sells Sandisk Stock and Buys Another AI Memory Stock (Hint: Not Micron)

Billionaire Israel Englander (Millennium Management) cut Sandisk by 24% to 1.1M shares while buying Everpure, increasing the position by 60% to 343K shares in Q1. Sandisk’s AI-driven enterprise SSD demand helped it surge revenue 251% to $5.9B in FQ3’26, but the article flags a late-cycle risk: analysts expect memory sales to decline in 2028 and the stock trades at 56x earnings vs Everpure’s 36x. Everpure reported FQ1’27 revenue up 35% to $1.1B and non-GAAP EPS up 62% to $0.47, with operating margin up 5 percentage points, supporting the valuation comparison.

Analysis

This is less about one hedge-fund filing and more about an AI-storage taxonomy: the market is starting to separate commodity memory beta from software-defined storage margin power. If enterprise buyers keep paying up for NAND, P can still win even if unit growth slows because its software layer captures more of the budget and raises switching costs; that makes it a higher-quality way to own the same AI capex theme. The second-order beneficiary is any vendor that monetizes orchestration, data management, or migration services around flash, while pure component names become more valuation-sensitive.

The main risk is timing. SNDK can stay vertical for several quarters if AI demand keeps pulling supply forward, so a short here is a fight against momentum and potentially against another round of upward revisions. The thesis only breaks if enterprise SSD ASPs, lead times, or guidance flatten; absent that, the bear case is really a 2026-2028 supply response story, not a near-term one. For P, the key falsifier is whether higher NAND costs finally leak through to margins or whether enterprise customers pause refresh cycles.

The contrarian read is that Englander may simply be rotating from the most reflexive, multiple-sensitive winner into the cleaner compounder, not calling an end to the memory cycle. That argues for relative value, not outright directional aggression: P should outlast SNDK structurally because its earnings are less hostage to spot pricing. If the consensus is too bearish on SNDK, it is because the market keeps underestimating how long AI capex can delay the usual memory downcycle; if it is too bullish, it is because current multiples already discount several more quarters of perfection.

Key Decisions for Investors

• Initiate a beta-neutral long P / short SNDK pair on a 5-10% pullback over the next 1-3 weeks; target 15-20% spread over 3 months, with a hard stop if SNDK guides up again or NAND pricing re-accelerates.
• Buy P on weakness for a 3-6 month hold; the setup is margin durability from software mix, and the thesis breaks if revenue growth falls sharply or operating margin stops expanding despite strong demand.
• Do not chase SNDK after the parabolic move; if expressing downside, use small put spreads only after a failed breakout or a guide-down, because the stock can remain momentum-driven for 1-2 quarters.
• If you need a broader memory hedge, use MU as the cleaner liquid proxy rather than a direct short in SNDK; the trade becomes more attractive only when spot NAND prices and lead times begin to flatten.