This new AI trade is leaving Nvidia and Micron in the dust

AI-driven hardware rotation is broadening beyond Nvidia, with Western Digital and Seagate outperforming since ChatGPT launched in November 2022 and the storage rally accelerating in April 2025. Micron is up roughly 130% since the March 30 low and has added about $470 billion in market value, while Intel is up around 200% and Nvidia has added more than $1 trillion. The article argues the market is increasingly favoring companies that relieve AI infrastructure bottlenecks in memory, storage, networking, and legacy chips.

Analysis

The market is rotating from “pick-and-shovel” compute beneficiaries into the constraint layer of the AI stack, where pricing power is often better and earnings sensitivity is less crowded. Storage, memory, and legacy semis are now trading like operating bottlenecks, not mature cyclical hardware, which implies the next leg of the AI trade is about scarcity normalization rather than pure end-demand growth. That favors names with leverage to capacity tightness and disciplined supply rather than the highest absolute revenue growth.

The second-order effect is that this rally can persist even if hyperscaler capex growth moderates, because the market is repricing the mix of spend, not just the level. If AI workloads continue to shift toward persistent data retention, inference logs, and model checkpointing, storage intensity can stay elevated for several quarters after GPU enthusiasm peaks. That makes WDC and STX more exposed to a durable re-rating than a simple momentum burst, while MU remains the cleanest beneficiary if DRAM/NAND discipline holds.

The risk is that this becomes a consensus crowded trade in the “AI bottleneck” basket. Once investors conclude supply is catching up, the multiple expansion can reverse faster than fundamentals, especially for the lower-quality laggards and for names where the rally has outrun visible estimate revisions. Intel’s move looks particularly reflexive; if the Apple angle fades or execution slips, the stock can mean-revert sharply because the re-rating is not yet fully supported by free-cash-flow visibility.

The contrarian miss is that Nvidia’s underperformance relative to these second-order names may be temporary rather than structural. If AI capex broadens again into new training cycles or sovereign deployments, NVDA’s earnings power can re-accelerate faster than storage peers because it still sits closest to the budget authority. In other words, the trade is not “NVDA is done,” but “the market is paying extra for the bottlenecks until supply proves it can clear them.”

Key Decisions for Investors

• Long WDC/STX basket for 1-3 months as a scarcity trade; use a tight trailing stop because the move is largely multiple-driven, but expect upside if storage ASPs remain firm and sentiment keeps chasing bottlenecks.
• Pair long MU vs short INTC over 3-6 months; MU has clearer direct exposure to AI memory tightening, while INTC’s re-rating is more headline-sensitive and more vulnerable if the Apple narrative stalls.
• Buy NVDA on weakness only via call spreads for 6-9 months rather than outright equity; this preserves upside if the market rotates back to core compute while limiting downside if bottleneck names keep leading.
• Fade the most extended momentum names in the basket with put spreads after any 10-15% short-term spike, especially INTC and LITE, because these names have the highest risk of air pockets once flow-driven buying cools.