Billionaire Stanley Druckenmiller Sells Sandisk Stock and Buys an AI Stock Wall Street Says Is Deeply Undervalued

Druckenmiller sold his entire stake in Sandisk and more than tripled his position in Alphabet in Q4. Wall Street places a $385 median target on Alphabet (current $295), implying ~30% upside; Alphabet is forecasted to grow EPS ~15% annually through 2029 and trades at ~27x earnings, having beaten consensus by ~15% on average over the last six quarters. Sandisk reported a 404% non-GAAP earnings jump last quarter and is modeled to grow adjusted earnings ~73% annually to June 2029, but currently trades at ~95x adjusted earnings despite market-share gains in NAND flash.

Analysis

AI-driven demand dynamics are reshaping who captures storage and compute economics: custom inference/hosting stacks (TPU-like) create annuity revenue and higher gross margins for the owner, while commoditized DRAM/NAND suppliers remain exposed to volatile spot pricing and multi-year capex cycles. That bifurcation favors vertically integrated service providers that can monetize chips as a product and a service, and it increases the marginal value of long-term purchase agreements from large enterprise customers who will pay for predictable performance and integration.

The dominant macro risk for semiconductor suppliers is timing: fab buildouts are lumpy and have 12–36 month lead times, so capacity additions made on current pricing can cause a multi-quarter inventory glut and rapid price decay. For cloud/AI vendors, the near-term catalyst set is product monetization cadence (enterprise contracts, hosting agreements) and margin conversion — these can re-rate revenues within 2–12 quarters, but regulatory or competitive responses from other cloud providers can compress realized upside quickly.

Consensus is underweight the optionality of chip-as-service models but overconfident on memory pricing persistence. That implies asymmetric opportunities: long exposure to service providers that can sell compute+chip bundles (high margin, recurring) and short exposure to commodity memory suppliers lacking long-term contracted revenue. Execution matters — use calendar spreads and pairs to isolate the structural trade from market beta and the next earnings headline.

Key Decisions for Investors

• Long GOOGL via 9–15 month call-spread (buy 1x near-ATM call / sell 1x 30–40% OTM call) size 2–4% NAV — R/R skewed positive: limited premium risk vs 30–50% upside if TPU monetization and ad AI adoption accelerate; roll or take profits at 50% realized.
• Pair trade: long GOOGL equity (1.5–2% NAV) / short MU equity (1.5–2% NAV) — horizon 6–18 months. Rationale: capture cloud monetization vs memory-cycle downside; set hard stop-loss at 12% adverse move in pair relative performance.
• Buy MU 6–9 month puts ~25% OTM (or a 1x2 put spread to finance premium) sized to risk 1–2% NAV — target payoff 3–5x if NAND/DRAM pricing reverts. Close or hedge if fab utilization stays >90% for two consecutive quarters.
• Hedge NAND directional exposure: sell SNDK (or buy 6–12 month put spread) as a hedge against memory price deterioration; if SNDK shows sequential gross-margin expansion tied to contracted hyperscaler deals, reduce hedge by 50%.
• Monitor tech-sales cadence: set alerts for two triggers — (A) three consecutive quarters of >5pp cloud gross-margin expansion (trim long GOOGL) and (B) two quarters of inventory build + >5% sequential price decline in NAND/DRAM (add to MU shorts and SNDK hedges).