Billionaire David Tepper of Appaloosa Increased His Nvidia Stake by 533% and Completely Sold Out of a Hypergrowth Artificial Intelligence (AI) Stock

Appaloosa Capital's David Tepper materially increased his Nvidia position to 1.9 million shares as of Sept. 30 (a 533% rise over six months) after buying 1.45M shares in the June quarter and 150k in the September quarter, citing Nvidia's strong AI GPU demand, >70% gross margins and CUDA ecosystem, though valuation (P/S ~23.5) and competition are noted risks. Conversely, Tepper fully exited Oracle after peaking at 2.3M shares in Q1 2024, selling 550k in Q2 and the remaining 150k in Q3, even as Oracle reported a 359% yoy jump in remaining performance obligations to $455B and aggressive Oracle Cloud Infrastructure revenue guidance (FY26 +77% to $18B; FY27 +78% to $32B; FY28 +128% to $73B; FY29 +56% to $114B; FY30 +26% to $144B). The moves appear driven by an early‑April market dislocation and positioning around AI winners, presenting a cautious signal on valuation sensitivity despite strong fundamentals and guidance.

Analysis

Market structure: Tepper's re-accumulation of NVDA reinforces a winner-takes-most dynamic for AI compute — Nvidia, its CUDA ecosystem, and adjacent suppliers (Micron MU, ASML ASML) capture pricing power while customers and hyperscalers face rising total cost of ownership. NVDA's gross margins >70% and P/S ~23.5 imply limited short-term supply elasticity; a 10–20% selloff would more likely pull forward incremental sell-side liquidity than indicate demand destruction.

Risk assessment: Key tail risks are an AI funding bubble (30–50% drawdown scenario if capex is cut), U.S.–China export controls or customer on-prem chip development that reduces upgrade cadence, and execution misses (e.g., gross margin under 65%). Expect immediate momentum (days–weeks) driven by flows, short-term re-rating over quarters as guidance/data-center orders print, and structural adoption over years if CUDA remains dominant.

Trade implications: Favor concentrated, size-aware exposure to NVDA (small % of AUM) while using options to control drawdowns; overweight semiconductors and data-center infra, underweight overvalued AI-app names. Cross-asset: stronger tech rally tightens real yields and raises equity implied vols — use skew-aware option structures and limit duration in fixed income exposure if growth surprises persist.

Contrarian angles: Consensus underestimates hyperscaler vertical integration risk and RPO conversion lag (Oracle’s huge RPO may take quarters to convert to free cash). NVDA's momentum can be crowded; consider disciplined entry on 8–15% pullbacks or margin/guidance misses rather than chasing at current multiples.

Key Decisions for Investors

• Establish a tactical 1.5–3.0% long position in NVDA (size relative to liquid equity sleeve) on either a) a pullback of 8–15% from current levels or b) if P/S falls to ≤18; trim to half if NVDA reports gross margin <65% or FY revenue guide misses consensus by >5%.
• Implement a 3–6 month NVDA call spread to capture upside while limiting premium: buy ATM call and sell +15% strike call (1:1) sized equal to a 1% cash position; if IV collapses, convert to long stock on 10% realized drop.
• Pair trade: long NVDA (1%) / short INTC (0.75%) — thesis: NVDA benefits from software moat; short caps structural risk in Intel's AI rollout. Rebalance if INTC announces a win with a major cloud customer or NVDA margin compression occurs.
• Reduce exposure to high-valuation AI application names by 20–40% and rotate proceeds into semiconductor suppliers (MU, ASML) and data-center operators (AMZN, GOOGL) with a 3–9 month horizon to capture capex flow-through.
• Set specific stop/trigger rules: cut NVDA exposure if price falls >20% in 10 trading days or if US export restrictions materially expand within 30 days; consider buying ORCL on any >20% pullback from current levels given RPO visibility, with a 6–12 month hold to capture conversion to revenue.