Wall Street is betting on a new class of AI winners. Intel, AMD, and Micron surged double digits this week while Nvidia—the poster child of the AI chip boom—lagged behind, marking what analysts are calling a ‘changing of the guard’ in semiconductor investing. The rotation signals growing investor conviction that the next phase of AI infrastructure will demand more than just GPUs, opening opportunities for CPU makers and memory specialists that were left behind in the initial AI frenzy.

Intel, AMD, and Micron are suddenly the hottest names in AI investing. All three companies posted double-digit stock gains this week, outpacing Nvidia in a dramatic reversal that has strategists rethinking the semiconductor pecking order.

The surge comes as institutional investors shift their AI infrastructure bets beyond GPUs to the broader ecosystem of chips powering deployment and inference. While Nvidia’s dominance in AI training remains unchallenged, Wall Street is increasingly focused on what comes next—and that’s creating opportunities for chipmakers previously overshadowed by the GPU giant’s meteoric rise.

Intel saw particularly strong momentum, with shares climbing on renewed optimism around its Xeon processors for AI inference workloads. The company has been positioning its CPUs as essential for running AI models in production environments, where cost efficiency and power consumption matter more than raw training performance. AMD followed a similar trajectory, benefiting from growing adoption of its EPYC server chips in data centers expanding AI capabilities beyond training clusters.

Micron‘s rally reflects a different bet entirely—that memory bandwidth and capacity will become the limiting factor as AI models grow larger and more complex. High-bandwidth memory (HBM) has emerged as a critical component in AI accelerators, and Micron’s position as a leading supplier puts it at the center of this infrastructure build-out.

The rotation hasn’t gone unnoticed by market strategists. Investment firms are describing the shift as evidence that AI’s economic impact is broadening beyond the initial training infrastructure phase. Where 2024 and early 2025 saw capital flood into GPU capacity, the focus now appears to be shifting toward the chips needed to actually run AI applications at scale in enterprise environments.

This isn’t to say Nvidia is losing its crown—the company still commands over 80% of the AI accelerator market and continues to post record revenues. But the stock’s relative underperformance this week suggests investors are hunting for the next layer of AI infrastructure winners rather than continuing to pile into the most obvious trade.

The CPU makers are making aggressive pitches. Intel has been highlighting how its processors handle AI inference workloads more cost-effectively than GPU-based solutions for certain applications. AMD is pushing its Instinct accelerators and EPYC CPUs as a diversified AI platform that doesn’t lock customers into a single vendor. Both companies are betting that as AI moves from research labs to production systems, total cost of ownership will matter more than peak performance.

Memory is the other piece of the puzzle. As AI models scale beyond current architectures, memory bandwidth becomes a bottleneck that can’t be solved by adding more compute. That’s where Micron and other memory manufacturers come in, supplying the HBM that sits alongside GPUs and the DDR5 that powers CPU-based inference servers.

Analysts are divided on whether this rotation represents a fundamental shift or just a tactical trade. Bulls argue that AI infrastructure spending will broaden significantly as deployment accelerates, creating room for multiple winners beyond Nvidia. Skeptics point out that training remains the most capital-intensive part of AI and that Nvidia’s architectural advantages there aren’t going away.

What’s clear is that Wall Street’s AI thesis is evolving. The initial wave of investment focused almost exclusively on training infrastructure—building the massive GPU clusters needed to create frontier models. Now attention is turning to inference, edge deployment, and enterprise adoption, all of which require different chip architectures and create opportunities for a wider range of semiconductor companies.

The timing of this rotation is notable. It comes as major cloud providers and enterprises are beginning to deploy AI applications at scale, moving beyond pilot projects to production systems that need to run efficiently for years. That shift in focus from building models to running them creates natural demand for CPUs, memory, and specialized inference chips—exactly the products Intel, AMD, and Micron are selling.

For semiconductor investors, the question now is whether this is a short-term rotation or the start of a sustained trend. If AI infrastructure spending continues to broaden beyond GPUs, the current rally in CPU and memory stocks could have room to run. If training remains the dominant driver of AI investment, this week’s moves might just be a brief diversification moment before capital flows back to Nvidia.

Wall Street’s sudden embrace of Intel, AMD, and Micron signals a maturing view of AI infrastructure investing. While Nvidia’s GPU dominance in training isn’t threatened, the market is recognizing that deploying AI at scale requires a broader toolkit—CPUs for cost-effective inference, memory for bandwidth-hungry models, and diverse architectures for different workloads. Whether this rotation marks a sustainable trend or just a tactical pause in the Nvidia rally will depend on how quickly AI applications move from development to production. For now, investors are betting that the next phase of the AI boom will have multiple winners, not just one.