David Tepper Dumps SanDisk After 591% Rally, Pivots Appaloosa Into AI Chip Stocks

The 591% run is over. SanDisk is out. Appaloosa is moving capital into AI chip names. David Tepper, the hedge fund manager known for aggressive macro bets, has rotated his portfolio away from the storage giant after a massive multi-year rally. The move, disclosed in a recent filing, signals a shift in conviction: storage cycles are cyclical, but AI compute demand is structural. The math holds until the incentive breaks. SanDisk, a leader in NAND flash memory, has been a beneficiary of the broader AI data boom. Storage demand for training datasets and inference logs has exploded. But Tepper is not buying the second derivative. He is selling the first. The 591% gain is in the rearview mirror. What matters now is the forward curve for AI accelerators. Let's be clear about what this rotation means. The filing, which shows a substantial reduction in the SanDisk position, is a forensic signal. Volume masks the insolvency structure, but here, the structure is simple: Tepper is moving from a high-certainty, high-cycle commodity to a high-certainty, high-growth franchise. Based on my experience analyzing institutional flows, this is not a tactical trade. This is a portfolio reconstruction. The core of the move lies in the capital expenditure cycle. AI chip demand is not a narrative; it is a line item. The top four cloud providers are budgeting over $200 billion annually for data center buildouts. Every dollar of that flows into NVIDIA, AMD, and a handful of others. SanDisk, while a beneficiary, is a supplier. The pricing power lies with the fab and the architecture, not the memory vendor. But I am not here to tell you this is a no-brainer. My job is to point out the structure. This is where the contrarian angle matters. Contrarian: The sell is not about SanDisk. It is about the opportunity cost. SanDisk's stock price reflects a certain level of future growth. AI chip stocks, specifically NVIDIA at a 60x trailing P/E and AMD at over 100x, are priced for perfection. Tepper is not buying value. He is buying momentum. The math holds until the incentive breaks. The incentive here is the need to outperform. When you are up big in a fund, you do not take risks; you take calculated risks. This is a calculated risk on the maintenance of the AI hype cycle. What is missing from this narrative is the storage angle. AI creates a massive need for storage. But it creates a bigger need for compute. The bottleneck is not NAND; it is CoWoS packaging capacity. The value accrues to whoever controls the scarcity. Tepper is selling the product and buying the means of production. Consensus is code, but code is fragile. The same logic applies to capital. The consensus is that AI chip stocks are the only game in town. That consensus is the risk. When the crowd is on one side, the liquidity is a one-way door. Liquidity is borrowed time. If the next earnings cycle shows even a slight miss in data center revenue, the markdown will be violent. SanDisk's 591% run had a similar profile. It goes up fast, but it goes down faster. Let's look at the financials. The P/E ratio differential is not the whole story. Tepper is a macro guy. He sees the world through the lens of interest rates and global liquidity. AI chips are a growth asset. In a rate-cutting environment, these names get a tailwind. Storage is a cyclical asset. It behaves differently. This is a beta decision, not an alpha decision. My experience in auditing financial systems tells me to focus on the incentive structure. Tepper's incentive is not to hold a winner. It is to find the next winner. SanDisk was a great trade. It is now a portfolio. The AI chip trade is a new trade with a similar or higher ceiling but a much more fragile floor. What I want to know is what happens next. The 13F filings will show the exact allocation. My prediction is not that he goes all-in. My prediction is that he rotates into a basket of names. He will not pick a single winner. He will buy the sector. This is a signal for the rest of us. The smart money is not chasing the past. It is financing the future. The past, in this case, is a 591% return. The future is the hardware that runs the algorithms. The question is not whether Tepper is right. The question is when the market decides that the future is already here. History repeats in the ledger, not in the news. The ledger shows a sell. The news shows a pivot. The logic is simple: storage is a feature. Compute is a religion. And Tepper is just a churchgoer. Risk is a feature, not a bug, until it isn't. And for SanDisk, the bug just bit.