Industries: Mag-7 Isn’t the Only AI Game in Town

AI-related capital spending remains a powerful growth driver, with hyperscaler cloud earnings continuing to beat and support the case for sustained semiconductor demand. The article says the Mag-7 now accounts for 30.6% of S&P 500 market cap, 25.1% of forward earnings, and 13.7% of forward revenues, while Mag-7 forward earnings growth is 25.4% versus 17.9% for the S&P 500 excluding Mag-7. Investor attention is broadening beyond the original seven as semiconductors, memory chips, and photonics continue to make new highs on AI infrastructure demand.

Analysis

The market is moving from a single-node AI narrative to a supply-chain bottleneck trade, and that broadening is the key second-order effect. The highest-conviction beneficiaries are not the platforms but the vendors whose revenues scale with every incremental dollar of capex regardless of whose model wins: memory, storage, custom silicon, and lithography. That shifts alpha from “who owns the AI franchise?” to “who controls the scarce inputs,” which is why names with direct exposure to HBM and enterprise storage can keep outperforming even if hyperscaler stocks merely tread water.

The setup also creates a subtle relative-value split inside semis. The market is already paying for structural demand, but the next leg likely depends on whether supply remains constrained into the next two quarters; if capacity starts to loosen, the most levered names will de-rate fastest even if end-demand stays healthy. Broadcom and TSM/ASML are higher-quality expressions of the theme because they sit closer to toll booths, while AMD/INTC are more exposed to product-cycle execution and share shifts, making them better for tactical rather than core exposure.

The consensus may be underestimating how much of this move is already in the price for the obvious winners, especially after a strong run in memory and optics. The better contrarian trade is not to fade AI outright, but to avoid the most crowded momentum in favor of laggards with optionality or cleaner balance-sheet support. A reversal would most likely come from one of two catalysts over the next 1-3 months: a capex pause from hyperscalers or evidence that lead times and order visibility are peaking, which would hit the most economically sensitive semiconductor subsectors first.

Inflation fears from energy are an additional macro tailwind for the AI supply chain only insofar as they keep nominal revenues elevated and push investors toward real-asset-like growth. But if oil continues higher, it can eventually compress multiples across the broader tech complex, especially for long-duration software and hardware enablers. That argues for preferring names with direct earnings visibility and near-term cash conversion over anything whose valuation depends on a multi-year terminal growth story.

Key Decisions for Investors

• Stay long SNDK / WDC on a 1-3 month horizon; use pullbacks to add. These are the cleanest direct beneficiaries of persistent AI storage demand, with the best near-term risk/reward while the market still rewards bottleneck exposure.
• Maintain a core long TSM / ASML basket versus a weaker AI beta basket (e.g., AMD/INTC) as a relative-value trade. The thesis is that chokepoint toll collectors should hold up better if AI capex merely normalizes rather than accelerates.
• Consider a pair trade: long AVGO, short a basket of the most crowded hyperscaler proxies. AVGO monetizes custom silicon and networking demand with less earnings-risk from AI capex payback debates; upside is steadier, drawdown risk is lower over the next 6-12 weeks.
• Use AMD only tactically, not as a core AI hold. The stock can outperform on inference-share headlines, but it has more execution sensitivity than the infrastructure toll collectors; prefer defined-risk call spreads rather than outright equity.
• Avoid adding to INTC as an AI winner until foundry proof points improve. It remains a longer-dated turnaround, but the near-term risk/reward is inferior versus names already converting AI capex into earnings.