Billionaire David Tepper of Appaloosa Is Overweight AI Stocks -- but He Recently Dumped Every Share of the Hottest AI Stock of 2026

The real signal here is not that AI is still favored; it is that the trade is becoming more discriminating. The strongest balance-sheet and capacity owners in the AI stack—NVDA, TSM, and to a lesser extent AMZN—remain the cleaner expressions of durable capex, while memory/storage is starting to look like the part of the chain where gains get competed away faster. That argues for a relative-value stance inside semis: the market may keep rewarding the scarce infrastructure names, but multiple expansion in the more cyclical nodes is likely to stall first.

SNDK’s exit matters because late-stage winners in commodity-adjacent hardware usually stop working before the fundamentals roll over. If managements confirm a few quarters of stable pricing and supply tightness, the stock can keep levitating, but the asymmetry gets worse when ownership becomes momentum-driven rather than earnings-driven. The second-order loser is any adjacent NAND/flash exposure that still trades on “AI storage” scarcity rather than on sustained free cash flow; those names tend to mean-revert hard once capacity normalizes.

The contrarian view is that Tepper’s sell is more likely portfolio hygiene than a negative fundamental read, so fading SNDK too early is risky. The better tell over the next 1-3 months is whether memory ASPs, cloud capex commentary, and HBM allocation remain tight; if those stay firm, the AI supply chain can keep running and SNDK may remain a momentum trade despite valuation. What would falsify a cautious view is another quarter of accelerating AI-related storage demand with no sign of inventory buildup or pricing compression.