How David Tepper's Latest Portfolio Moves Reveal the Real AI Infrastructure Winner

A Billionaire’s Bold Bet on GPU Dominance

The shift in Appaloosa Management’s holdings tells an important story: Hedge fund legend David Tepper has made significant portfolio adjustments that signal where smart money sees opportunity in the artificial intelligence buildout. In recent quarters, Tepper’s fund has reduced exposure to traditional enterprise software giants while dramatically increasing positions in AI chip manufacturers.

Understanding the Intel and Oracle Exit

The decision to reduce holdings in Intel (NASDAQ: INTC) and Oracle (NYSE: ORCL) may seem counterintuitive at first glance, but reflects a pragmatic reassessment of long-term growth trajectories.

Intel’s multi-year foundry transition has proven slower and more capital-intensive than many anticipated. Market share losses to competitors in both data center and PC segments have further complicated the narrative. While the valuation appears reasonable on surface metrics, Tepper’s investment philosophy—shaped by decades of spotting value traps—recognizes that cheap valuations without strong growth catalysts often continue deteriorating rather than recovering.

Oracle presented a different scenario. The company benefited substantially from cloud and AI database demand, resulting in impressive share price appreciation. However, recent concerns about future capital expenditure obligations, elevated debt levels, and stretched valuation multiples may have prompted profit-taking. Freeing capital to redeploy into higher-conviction positions aligns with Tepper’s concentrated investment approach.

The GPU Play: Why AMD and Nvidia Now?

The reallocation reveals a clear thesis: semiconductor capacity for AI workloads represents the genuine structural opportunity in this cycle.

AMD’s Emerging Position: Tepper initiated a nearly $154 million position in Advanced Micro Devices (NASDAQ: AMD), representing approximately 2% of Appaloosa’s portfolio. AMD’s significance lies in serving as the primary alternative GPU supplier to Nvidia in enterprise AI deployments. With global semiconductor revenue projected to exceed $1 trillion by 2030, AMD captures growing demand from hyperscalers and enterprises seeking cost-effective compute solutions.

The company’s MI300, MI325, and MI350 GPU families are experiencing robust adoption, complemented by strong data center CPU demand. Third-quarter data center revenues reached $4.3 billion, up 22% year-over-year. AMD’s upcoming MI400 series—backed by multi-year commitments from Oracle and OpenAI—indicates confidence in sustained long-term demand expansion.

Nvidia’s Reinforced Position: Tepper added approximately 150,000 shares to existing holdings, bringing Nvidia to 4.8% of the portfolio. The rationale is straightforward: Nvidia (NASDAQ: NVDA) remains the backbone of infrastructure buildout, with data center revenue surging 66% year-over-year to $51.2 billion in Q3. Management has disclosed $150 billion in already-shipped Blackwell and Rubin orders, with $350 billion remaining on the schedule through end of 2026.

Recent partnerships—including allocations to Saudi Arabia’s Public Investment Fund subsidiary Humain and AI research firm Anthropic—expand the addressable market. By 2030, the AI infrastructure opportunity alone could reach $3-4 trillion annually, positioning both companies for substantial gains despite near-term volatility concerns.

What This Means for Investors

The lesson from Tepper’s repositioning isn’t necessarily to mirror his exact trades, but to recognize the directional signal. Both companies face elevated valuation multiples—Nvidia at 44.6x earnings and AMD at 105.9x—meaning meaningful upside is partially priced in.

For investors who believe in multi-year AI infrastructure expansion, a dollar-cost averaging approach offers a disciplined entry mechanism into these positions rather than attempting to time the market at current levels. The fundamental demand drivers remain intact: enterprises and cloud providers require exponentially more compute capacity regardless of near-term sentiment fluctuations.

Tepper’s move suggests that within AI infrastructure, the capital winners are being determined not by narrative debates about AI bubbles, but by measurable capacity constraints and the companies positioned to solve them.