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

David Tepper’s Appaloosa piled into AI-linked names—holding Nvidia, Taiwan Semiconductor, and Amazon—so ~77% of its invested assets are tied to AI. However, it exited Sandisk in Q2 after Sandisk shares surged over 3,000% over the prior year, consistent with profit-taking and the stock’s historically cyclical memory/storage setup. The news is mainly positioning/13F signal-based and may influence sentiment around AI supply-chain beneficiaries, but it is unlikely to be broadly market-moving.

Analysis

The actionable read is not “Tepper hates storage”; it is that he is rotating away from the most reflexive, valuation-sensitive beneficiary of the AI buildout and toward the enablers with the clearest pricing power. That matters because memory/storage tends to be the first place capital gets yanked when AI capex expectations wobble: buyers can defer purchases, channel inventory can rebuild quickly, and the earnings leverage cuts both ways. By contrast, compute and foundry exposure still enjoy tighter demand/supply coupling, so NVDA and TSM remain the cleaner expression of the AI cycle.

The second-order risk is that SNDK’s blowoff move becomes a sentiment anchor for the whole storage complex, not just one name. If investors start treating SNDK as the “most expensive AI beneficiary,” then MU could face multiple compression even without a near-term fundamental miss, especially if management commentary points to normalizing lead times or moderating NAND pricing. The key time horizon is 1-3 months: 13F data is stale, but it can still influence factor flows and momentum traders ahead of earnings.

Contrarian view: the market may be overreacting to a fund-level liquidation that could simply reflect position sizing after a 3,000% run. If SNDK prints another clean quarter with no inventory build, the selloff thesis is likely just profit-taking noise, and a crowded short could squeeze hard. What would falsify the bearish storage read is continued ASP stability, resilient gross margin, and no increase in channel inventory; what would confirm it is any sign that enterprise AI storage demand is decelerating faster than broad AI capex.

Key Decisions for Investors

• Do not chase SNDK strength here; if you want to express the view, use a small tactical put spread into the next earnings window, with a stop if gross margin or inventory commentary stays tight.
• Prefer long TSM over SNDK on a 1-3 month basis: TSM has better visibility into AI capex conversion, while SNDK is more exposed to multiple compression if the market rotates away from momentum storage.
• Maintain or add to NVDA and AMZN as core AI exposure rather than rotating into high-beta storage names; the risk/reward remains better in names with structural demand and cash flow visibility.
• Watch MU as the cleaner read-through for storage pricing: if MU guides conservatively on NAND/DRAM or inventories rise, that would strengthen a bearish pair versus SNDK; if not, avoid pressing the short.
• If SNDK gives back 10-15% without a fundamental miss, consider a rebound trade only after earnings confirmation; otherwise treat any weakness as the start of a valuation reset, not an entry point.