David Tepper's Appaloosa Management disclosed a $3.08 billion position across three names in its latest 13F filing — Amazon, Micron Technology, and Taiwan Semiconductor Manufacturing Company — representing 40% of the fund's $7.7 billion equity portfolio. The concentration is unusual for a fund historically known for event-driven opportunism and broad macro plays.
The filing shows Amazon as the largest position, followed by Micron and TSMC. The trio shares a common thread: all sit upstream or midstream in AI infrastructure buildout. Amazon controls cloud compute distribution through AWS. Micron supplies high-bandwidth memory for Nvidia's H100 and H200 chips. TSMC manufactures the logic dies for every major AI accelerator in production. Tepper's positioning is not a bet on consumer AI adoption. It is a bet on the capital expenditure cycle that precedes it.
The 40% concentration carries two implications. First, Tepper is expressing a view on timing — specifically, that the semiconductor cycle trough is behind us and that hyperscaler capex will continue expanding through 2025. Micron's November earnings call guided to sequential revenue growth in fiscal Q2 2025, citing AI server demand. TSMC reported $23.5 billion in Q4 2024 revenue, up 38.8% year-over-year, driven by 3nm node orders. Second, the position size suggests Tepper believes regulatory or geopolitical risks to Taiwan Semiconductor are priced in. The stock trades at 22x forward earnings despite controlling 60% of global foundry revenue and 90% of advanced-node capacity. That valuation implies the market is already discounting a non-zero probability of supply disruption.
The risk is twofold. Micron and TSMC share exposure to inventory cycles. If hyperscalers pause server deployments or extend refresh timelines, both names compress simultaneously. Amazon insulates partially through AWS revenue diversification, but the company's own capex forecast for 2025 — expected north of $100 billion — depends on enterprise AI adoption accelerating beyond current chatbot pilots. If adoption stalls, Amazon's infrastructure spending becomes a margin drag, not a moat.
Operators and allocators should watch three catalysts over the next 90 days. First, Micron's March earnings call will clarify whether HBM supply constraints are easing or tightening — a signal for memory pricing power. Second, TSMC's April earnings will disclose 3nm utilization rates and any commentary on N2 node ramp timing for 2026. Third, Amazon's Q1 2025 earnings in late April will report AWS revenue growth and any revision to full-year capex guidance. Each data point either confirms or contradicts the infrastructure-first thesis.
Tepper is not betting on Nvidia. He is betting on the suppliers Nvidia cannot build without and the distributor that monetizes what Nvidia enables. The filing is a thesis on picks-and-shovels scarcity, not AI application hype.