Billionaire David Tepper Sold Nvidia Stock and Piled Into This AI Stock That's Up Nearly 200% in the Past 5 Years

David Tepper's Appaloosa increased its Alphabet stake by 28.8% in Q4 while trimming Nvidia by ~10.5%. Alphabet shares have climbed 190% over the past five years versus a 75.7% gain for the S&P 500. The company is benefiting from AI (Gemini LLM and search AI features), faster-growing cloud sales (third-largest infra market share), expanding YouTube streaming and subscription offerings, and long-term optionality from Waymo, supporting diversification beyond advertising.

Analysis

Capital is rotating from hardware-levered AI exposure into platform owners and ad/analytics stacks; that shift re-prices where incremental margin accrues. Expect 500–1,000 bps of gross-margin migration toward software/platform operators over 12–24 months as inference economics commoditize and value accrues to model IP, data moats, and distribution.

Second-order beneficiaries include cloud-native middleware, ad-tech firms with first-party data, and API-layer inference providers — while foundry and spot-GPU pricing become more volatile, creating inventory and channel financing stress for parts of the semiconductor supply chain. If inference demand growth decelerates by even 20% vs. current consensus in the next 6–12 months, expect double-digit downdrafts in spot GPU replacement orders and margin pressure for smaller fabless suppliers.

Regulatory and ad-cyclicality are the largest near-term reversers: antitrust actions or a material ad-budget pullback could compress multiples within 3–9 months, while faster-than-expected model cost declines (30–50% inference cost deflation over 12–36 months) would steepen the rotation away from hardware. Position sizing should reflect a two-tier horizon: tactical (3–12 months) to capture re-rating vs. hedges against hardware drawdowns, and strategic (12–36 months) to capture recurring-revenue optionality that’s underpriced by some investors.

Key Decisions for Investors

• Long GOOGL (platform exposure): Buy a 9–15 month call spread — long ATM call, sell +20% OTM call — sized to 2% NAV. Target +30–40% upside if platform multiples re-rate; max loss = premium. Add a stop if spot falls 15% from entry to preserve capital on ad-cycle shocks.
• Pair trade: Long GOOGL / Short NVDA (equal notional) for 6–12 months to express software/platform capture vs. hardware cyclicality. Reward scenario: GOOGL +15% / NVDA -20% -> pair returns ~+25% net; risk scenario: NVDA outperforms -> hedge with 3–6 month NVDA puts sized 25% of the short notional.
• Tactical hedge for hardware shock: Buy 3–6 month NVDA puts (protective insurance) sized to cover 30–50% of portfolio exposure to semiconductors/cloud capex names. Cost should be treated as insurance during high positioning risk from inventory cycles.
• Contrarian satellite: Small long-dated INTC calls (18–36 months) for mean-reversion in fab/CPU share if GPU-driven capex normalizes — asymmetric payoff if Intel executes product cadence. Limit to 0.5–1% NAV as a low-probability, high-payoff diversification.