Is the AI Trade Entering Its Second Wave? This ETF Could Be a Major Beneficiary.

The key second-order dynamic is that the AI capex cycle is broadening from a pure design-led story into a capital-intensity story, which usually shifts alpha from the obvious winner to the upstream enablers. That favors TSM, ASML, LRCX, and AMAT because a sustained build-out requires more lithography, deposition, and etch intensity even if unit growth moderates; it also means the market is likely underestimating how much of the next leg comes from tool order durability rather than headline GPU demand. NVDA still anchors the trade, but the more asymmetric exposure may be the picks-and-shovels names with leverage to multi-quarter WFE revisions.

Consensus appears to be treating SMH as a clean way to own AI, but the structure embeds a hidden concentration risk: when leadership narrows, the ETF behaves less like a diversified basket and more like a megacap momentum product. If GPU demand pauses or hyperscaler spending shifts from training to inference optimization, the group can de-rate quickly even if secular growth remains intact. That creates a window where the ETF can underperform the underlying industry breadth because the market is paying for a continuation of the current leader set, not just sector growth.

The weakest name in the complex remains INTC, which functions more as a sentiment overhang than a direct beneficiary of the AI capex cycle. Any incremental progress there is likely to be capital-market driven rather than fundamental, so it can lag even in a strong tape if investors keep rotating toward higher-quality operating leverage. The contrarian setup is that the non-megacaps may have already cleared the worst of the skepticism, while the largest names still need perfect execution to justify their implied growth rates.

Near term, the biggest risk is not a collapse in demand but a timing mismatch: if orders are front-loaded and then digest for 1-2 quarters, the market could punish the whole group despite healthy 12-24 month fundamentals. Watch for guidance from cloud and foundry customers, because a small change there can move the semicap complex more than headline AI enthusiasm. On the upside, any evidence that 2026 capex plans are being revised higher would likely trigger a second leg in the entire chain, with the equipment names repricing fastest.