Why Billionaire David Tepper Just Sold 92% of UnitedHealth Group and Is Buying These 3 Stocks Instead

Appaloosa Management materially reallocated its Q3 2025 portfolio, selling roughly 2.25 million UnitedHealth shares (about 92% of its UNH stake after a Q2 add of ~2.28 million) and redeploying capital into Whirlpool (position up ~1,967%), an +85% increase in KraneShares CSI China Internet ETF (KWEB), a 895,000-share add to Qualcomm (position up ~256%), and a new sizable position in AMD. Tepper's moves appear driven by profit-taking in a healthcare blue chip and a shift toward value exposure in consumer durables and strategic bets on AI/edge chip leaders (AMD, Qualcomm) that he expects to benefit from AI adoption. These trades are notable for investors monitoring positioning and flows in large-cap healthcare, consumer cyclicals and AI semiconductor names.

Analysis

Market structure: Tepper’s rotation materially benefits AI semiconductor and edge-AI plays (AMD, QCOM, NVDA) and semiconductor capital goods while modestly penalizing healthcare beta (UNH) and cyclicals exposed to consumer demand (WHR). Expect 6–18 month demand concentration into datacenter GPUs/accelerators and edge SoCs; price power shifts to vendors with software/driver ecosystems (NVDA > AMD > QCOM) but share gains by AMD and QCOM can cap NVDA’s margin expansion.

Risk assessment: Tail risks include US/China export controls or an AI demand pullback that could erase 30–50% of projected incremental chip revenue in 6–12 months, and consumer recession risks that could knock WHR earnings by >20% over a year. Hidden dependencies: cloud hyperscaler procurement cadence and AI model efficiency materially alter component demand; catalysts to watch in next 90–180 days are NVDA/AMD/QCOM earnings commentary and Qualcomm AI200 commercialization milestones.

Trade implications: Direct tactical bias long AMD (conviction) and QCOM (edge-AI optionality) with options to size risk; trim/distribute UNH exposure into these tech names if UNH positions are >2% of portfolio. Cross-asset: stronger AI capex supports industrials/equipment and widens equity-bond spread (risk-on) while raising semi-equipment order books over 3–12 months.

Contrarian angles: The market underprices Qualcomm’s edge TAM and overprices cyclical risks in Whirlpool; conversely, consensus may understate regulatory/export risk to datacenter GPU flows. Mispricings: AMD/ QCOM downside capped vs NVDA’s convex upside—structured long-dated call exposure on AMD + protective hedges on WHR capture asymmetric payoff.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in AMD (ticker: AMD) with a 12–24 month horizon. Implement with 12–18 month LEAPS (e.g., Jan 2027 calls) or buy stock; take profits on +50% or trim to 1% weight, stop-loss at -30%.
• Establish a 1–2% tactical long in Qualcomm (QCOM) using 3–9 month call spreads (buy ATM, sell 20–30% OTM) to limit premium. Target +30–50% if Qualcomm announces AI200 design wins within 90 days; cut to flat if revenue guide misses by >3% sequentially.
• If UNH position shows >20% unrealized gain since Q2, reduce exposure by 40–60% and redeploy proceeds into AMD/QCOM; otherwise hold to 1–2% core weight. Re-add above 2% only on a >15% pullback with unchanged 2026 fundamentals or better guidance.
• Initiate a small 0.5–1% opportunistic long in Whirlpool (WHR) with a 6–12 month horizon, but buy a 6–9 month 10% OTM put as hedge. Only scale above 1% if two consecutive months show positive US Durable Goods/ Housing Starts momentum or WHR reports margin recovery +100bps.