Prediction: The Nasdaq's AI Stocks Will Outperform the S&P 500 Over the Next 12 Months. Here's What to Buy.

The key setup is not simply “AI is good,” but that earnings dispersion within large-cap tech should widen materially as capex intensity shifts from a cost center to a demand signal. Hyperscaler spend and neocloud buildout should create a reflexive loop: more infrastructure spend lifts semiconductor, networking, power, and cooling demand first, then monetization follows with a lag as enterprise AI usage converts into software ARPU and seat expansion. That favors the platforms with both distribution and balance-sheet flexibility; it is more fragile for capital-intensive newcomers that need flawless execution to finance growth.

CoreWeave is the clearest beneficiary on momentum, but it is also the most vulnerable to a reversal in financing conditions because its equity story is still tied to backlog conversion and continuous external capital. The market is likely underwriting a scarcity premium for AI compute capacity, yet that premium can compress quickly if hyperscalers internalize more capacity, if GPU supply normalizes, or if contract concentration becomes a focus. The second-order winner is the broader AI infrastructure stack: NVDA, networking, memory, and power-management names should see a cleaner, less binary earnings path than the pure-play neoclouds.

Microsoft is the higher-quality expression of the same theme because AI attach rates can lift revenue per user without the same balance-sheet strain. The market may still be underestimating how much of Copilot adoption is a seat-expansion story versus a productivity feature, which matters because the former is more durable and margin-accretive. Near term, the main risk is that AI software adoption looks strong in pilots but slower in enterprise-wide rollout, which would cap multiple expansion even if the narrative stays positive.