Billionaire David Tepper Sells Oracle, Micron, and Intel, and Buys an AI Stock Up 31,000% Since Its IPO

David Tepper's Appaloosa trimmed positions in AI-related winners—Intel, Oracle and Micron—locking gains after recent share-price rallies and redeployed capital into Qualcomm, which he sees as an underappreciated AI chip supplier across devices and data centers. Qualcomm's roadmap (AI200/AI250 inference solutions in 2026–27), strength in automotive and on-device AI demand, and a forward P/E of ~13 underpin the bullish case, while Intel's government stake, Oracle's $300 billion OpenAI-linked exposure and Micron's cyclical NAND-driven pricing power explain Tepper's risk-managed exits.

Analysis

Market structure: Tepper’s rotation — trimming MU, ORCL, INTC and redeploying into QCOM — signals a shift from pure data‑center semiconductor cyclic winners to diversified, device‑plus-inference plays. Qualcomm directly benefits (on‑device LLMs, AI200/AI250 racks, automotive compute) while pure-play memory (MU) and lagging fabs (INTC) face valuation contraction if supply normalizes or execution stalls; expect relative revenue re‑mixing over 4–18 months as OEMs evaluate on‑device vs cloud inference cost curves.

Risk assessment: Key tail risks are AI regulatory/export restrictions (US‑China chip controls), a sharp DRAM/NAND capacity rebuild that deflates MU pricing within 6–12 months, and OpenAI/Oracle counterparty or payment shortfalls that would stress ORCL’s balance sheet. Near term (days–weeks) sentiment swings from earnings/government headlines will dominate volatility; longer term (12–36 months) product ramps (QCOM chips 2026–27) and foundry execution (INTC) drive fundamentals.

Trade implications: Tactical plays favor underweighting memory cyclicals and adding diversified AI exposure via QCOM: low forward P/E (~13) provides asymmetric upside if on‑device LLM adoption accelerates. Use relative-value pair trades (long QCOM vs short INTC or MU) and limited-duration options to define risk around 6–18 month catalysts (earnings, product launches, NAND capacity signals).

Contrarian angles: Consensus overweights NVDA/data‑center monoculture; underpriced is Qualcomm’s cross‑form‑factor nexus (phones, autos, edge racks) which can capture secular OEM demand if vendors prioritize latency/cost. Conversely, the market may be underreacting to ORCL’s balance‑sheet funding risk tied to a single large AI customer — a concentrated counterparty risk that could cause outsized drawdowns if payments slip.

Key Decisions for Investors

• Establish a 2–3% long position in QCOM within the next 2–4 weeks as a diversified AI exposure; add up to +1% if upcoming quarterly revenue from non‑baseband segments shows >25% y/y growth or forward P/E stays <15.
• Implement a pair trade: long QCOM (2%) vs short INTC (1.25%) sized dollar‑neutral, horizon 6–12 months; cover or rebalance if QCOM outperforms INTC by >20% or Intel announces credible foundry revenue visibility within 3 months.
• Buy a defined‑risk options position: Jan 2027 QCOM call spread (long 20% OTM call, short 60% OTM call) sized to 0.5–1% portfolio risk to capture 12‑18 month upside from on‑device LLM adoption while capping premium paid.
• Trim existing MU exposure by 30–50% if position >2% immediately; if not held, avoid initiating >1% exposure. Reconsider adding only if NAND/DRAM industry supply‑to‑demand tightness persists past Q3 2026 or if Micron announces multi‑quarter pricing resilience.
• Do not initiate new ORCL longs until management discloses OpenAI payment/capex financing cadence in next 60–90 days; consider a targeted short or put spread (3–6 month) if disclosures show material debt‑funded build‑out (>$10–20bn incremental near‑term).