David Tepper's Appaloosa Management disclosed a 40 percent allocation to three names in its latest 13F filing: Amazon, Micron Technology, and Taiwan Semiconductor Manufacturing. The trio now represents $3.08 billion of the fund's $7.7 billion in reportable long equity positions. The concentration marks a deliberate shift away from diversified mega-cap exposure toward infrastructure that scales generative workloads—chipmaking capacity, memory bandwidth, and hyperscale compute.
The filing shows Appaloosa holding material positions in all three names during the quarter ending December 31, 2024. Amazon anchors the group as both a hyperscale cloud operator and semiconductor designer through its Graviton and Trainium lines. Micron supplies high-bandwidth memory for GPU clusters and inference accelerators. Taiwan Semi manufactures leading-edge logic for Nvidia, AMD, and the hyperscalers' in-house silicon. The combination covers three choke points in AI infrastructure: chip fabrication, memory supply, and hyperscale deployment.
The positioning reflects Tepper's view that AI capital expenditure remains in early innings despite elevated valuations across the semiconductor complex. Amazon Web Services expanded capital expenditure guidance to $100 billion for 2025, nearly double the prior year, with the majority allocated to GPU clusters and custom silicon deployment. Micron's high-bandwidth memory shipments are sold out through 2025, with pricing holding firm despite broader DRAM softness. Taiwan Semi's 3-nanometer capacity is fully allocated through mid-2026, primarily to AI accelerator orders. Tepper's concentration suggests he expects these supply constraints to persist longer than consensus estimates.
The allocation also signals skepticism toward software-layer AI valuations. Appaloosa exited or materially reduced positions in several application-layer AI names during the quarter, reallocating capital toward picks-and-shovels infrastructure. The move reflects a view that monetization risk sits with software vendors while margin expansion accrues to semiconductor and cloud operators who control scarce manufacturing and deployment capacity. Fund managers watching Appaloosa's book should note the absence of exposure to AI software platforms and the corresponding overweight to companies with multi-year backlog visibility.
Operators should track Amazon's Q1 2025 capital expenditure disclosure in late April, Micron's high-bandwidth memory pricing commentary in its March earnings call, and Taiwan Semi's 2-nanometer production ramp timeline updates. Any downward revision to hyperscale capex guidance or easing in memory pricing would challenge Tepper's thesis that infrastructure scarcity persists through 2026.
Appaloosa's 13F shows a fund willing to hold concentrated positions when supply-demand imbalances are visible and contractually locked. The 40 percent weighting says Tepper expects these three companies to capture a disproportionate share of the estimated $200 billion in global AI infrastructure spending forecast for 2025.