David Tepper's Appaloosa Management disclosed a 40% concentration in three names—Amazon, Micron Technology, and Taiwan Semiconductor Manufacturing—across its $7.7 billion equity portfolio in the latest 13F filing. The position sizes place roughly $3.1 billion into what the market calls AI infrastructure. Tepper does not call things. He sizes them.
The filing shows Amazon as the largest single position, followed by Micron and TSMC in amounts that turn sector bets into fund-level declarations. Appaloosa trimmed exposure elsewhere to make room. The moves happened in Q4 2024, a quarter when most allocators were still debating whether Nvidia's multiple was sustainable. Tepper went one layer down the stack—to the cloud operator spending the capex, the memory supplier choking on demand, and the foundry printing the chips no one else can print at scale. He bought the infrastructure, not the narrative.
This matters because Tepper trades liquidity, not technology. When he made $7 billion on financials in 2009, it was not because he loved banks. It was because he saw forced sellers and a Fed put. When he loaded Chinese ADRs in 2020, it was because the distress was structural and the reflexivity obvious. The AI concentration now is the same bet: enterprises will spend because they must, margins will compress because they cannot help it, and the three companies he owns will print revenue regardless of which model wins. Amazon captures the spend. Micron captures the bottleneck. TSMC captures the manufacturing monopoly that no amount of CHIPS Act funding will replicate before 2027.
The position sizing forces a question allocators should ask their own books: if Tepper is willing to let 40% of a multi-billion fund ride on three names, what does he see that the index huggers do not? The answer is probably capex visibility. Amazon's infrastructure spend is guided at $100 billion annually through 2025. Micron's HBM revenue is growing triple digits off a base that is no longer trivial. TSMC's N3 and N2 nodes are sold out into 2026, with Apple, Nvidia, and AMD all fighting for the same wafer starts. These are not bets. They are locks on the only three doors that matter.
Operators and allocators should watch Amazon's Q1 2025 earnings in late April for any AWS capex revisions. Micron's next earnings call in late March will clarify HBM3E supply commitments and whether Samsung's yield issues have extended DRAM pricing tailwinds into Q3. TSMC's March revenue disclosure will show whether N3 node loading remains above 90%, the threshold at which incremental pricing power appears. Tepper's position sizing suggests he expects all three data points to confirm the trend.
The 13F does not show exits, but it shows conviction. When a trader who made his name in distressed debt puts 40% of his fund into three companies, the market is not fragmented anymore. It is binary.