David Tepper Trimmed Nvidia and AMD, but Doubled Down on This AI Stock

Appaloosa Management cut NVIDIA by roughly 13% of its shares and AMD by roughly 32%, while nearly doubling Amazon, adding about 2.14 million shares for a roughly 98% increase in the quarter ended March 31, 2026. The filing reflects a notable shift in positioning toward Amazon and away from two major semiconductor AI beneficiaries, but it is routine portfolio disclosure rather than a company-specific catalyst. Market impact is likely limited, though the trades may influence sentiment around AI-related holdings.

Analysis

This is less about a single manager’s opinion than about a late-cycle AI positioning shift: capital is rotating from the most crowded, highest-duration semiconductor exposure into the hyperscaler that monetizes the compute buildout. If a prominent growth fund is reducing NVDA and AMD while adding AMZN, it implies the market is starting to distinguish between “picks-and-shovels beta” and the platforms that can translate capex into cash flow and margin expansion. That tends to matter most when AI infrastructure spending slows from hyper-growth to merely strong growth, because the multiple compression risk shows up first in the pure hardware beneficiaries.

Second-order, this is mildly negative for the semiconductor supply chain and positive for the cloud/software stack. Fewer incremental dollars into accelerators can ripple into networking, advanced packaging, and select memory names over the next 1-2 quarters, while the spend may instead consolidate into AWS where utilization, workload management, and enterprise AI attach rates drive operating leverage. The key tell is that AMZN benefits even if AI demand is not reaccelerating; it only needs spend to remain durable, not explosive.

The contrarian read is that semis may be getting de-risked too early. If enterprise AI inference and sovereign/enterprise buildouts keep capex elevated into the next budget cycle, NVDA and AMD can re-rate back quickly because their earnings sensitivity is still far greater than the cloud beneficiaries. Conversely, if GPU supply normalizes and customers start demanding efficiency over raw performance, AMZN’s positioning could prove the cleaner way to express AI monetization with lower multiple risk.

Key Decisions for Investors

• Over the next 2-6 weeks, favor long AMZN vs short a basket of AI hardware names (NVDA/AMD or SOXX) as a relative-value expression of AI monetization over AI spending; target 8-12% spread capture if hyperscaler cash-flow visibility improves while semis consolidate.
• Use any 5-8% post-news bounce in NVDA to initiate a tactical short or buy puts with 1-3 month tenor; risk/reward is attractive if the market starts pricing in capex normalization and multiple compression rather than earnings growth.
• Avoid outright short AMD unless paired; its downside is more idiosyncratic and a cleaner trade is long AMZN / short AMD to isolate platform-vs-component dispersion with lower thesis risk.
• For semis bulls, wait for a broader flush in the group before adding—this kind of flow often creates a 2-4 week window where negative sentiment overextends and entry improves materially.
• If AMZN outperforms for two consecutive reporting periods, add on pullbacks and extend horizon to 6-12 months; the trade works best if the market shifts from AI narrative to AI ROI, where AWS is structurally advantaged.