Nvidia Gets All the Credit, but These 4 Stocks Are Quietly Capturing the $725 Billion AI Buildout

The article highlights four AI supply-chain beneficiaries—Taiwan Semiconductor, Broadcom, ASML, and Arm—with expected earnings growth of roughly 22%, 49%, nearly 30%, and almost 25% annually over the next three to five years. It cites Broadcom’s AI chip revenue potentially exceeding $100 billion next year and Arm’s new AI CPU chip entering full production later this year. The piece is broadly bullish on AI infrastructure spending, but it is largely commentary rather than new company-specific catalyst news.

Analysis

The key second-order trade is that the AI capex wave is broadening from “compute” into the industrial plumbing that monetizes every incremental wafer. That favors the toll-collectors with pricing power and tight bottlenecks more than the headline GPU vendor; if hyperscaler spend stays elevated, margins should migrate up the stack for foundry capacity, lithography tools, and custom silicon design, while generic server OEMs and second-tier chipmakers likely see less durable upside.

TSM and ASML sit on the highest-quality scarcity rents, but they are also the most exposed to a reversal in AI capex cadence: if hyperscalers stretch deployments or digest inventory, order growth can decelerate fast even if secular demand remains intact. ASML is the cleanest long-duration monopoly, yet export controls make the path lumpy; a policy shock can matter more than end-demand for a few quarters. TSM is more directly levered to AI mix, but its moat is narrower than ASML’s because capacity expansion eventually invites competitive normalization.

Broadcom looks like the best near-to-medium-term earnings inflection because custom accelerators reduce customer concentration risk for hyperscalers and expand silicon wallet share beyond GPUs; that is the clearest way for AI budgets to shift from a single-vendor narrative to multi-source procurement. The contrarian issue is that the market may be underestimating how much of AVGO’s upside is tied to a small set of hyperscaler programs, so the stock can still gap on any design delay. ARM is interesting as a later-cycle monetization play: if AI CPU deployment broadens at the edge and in inference, its royalty model could compound with little capital intensity, but the near-term product launch may disappoint if customers treat it as optional rather than mission-critical.

The cleanest risk/reward is to own the bottlenecks and fade the commoditized parts of the ecosystem. The overdone/underdone debate: the market likely still underprices ASML’s monopoly durability and AVGO’s content per AI dollar, while overpaying for “AI exposure” that lacks pricing power or is one product cycle away from competitive attack. A portfolio approach should prioritize names with structural scarcity and visible backlog over those dependent on one launch or one customer cohort.

Key Decisions for Investors

• Long AVGO into the next two earnings prints; use any 5-8% post-print pullback to add. Risk/reward is favorable because custom accelerator ramps can re-rate forward EPS faster than consensus, but trim if hyperscaler capex commentary softens.
• Pair long ASML / short a basket of lower-quality semiconductor hardware names over a 3-6 month horizon. Thesis: ASML’s monopoly should preserve pricing power through a cyclical wobble, while commoditized names are more vulnerable to order normalization.
• Buy TSM on 2-4 week weakness rather than chasing strength. Best setup is after any broad AI hardware selloff; downside risk is export-policy headlines, upside is continued mix shift to advanced nodes.
• Take a smaller, optionality-driven long ARM via call spreads 6-12 months out. This is a higher-beta royalty story with asymmetric upside if AI CPU adoption broadens, but it should be sized below TSM/ASML/AVGO because execution risk is higher.
• Avoid paying peak multiples for pure AI momentum names without supply-chain control; if you need AI exposure, prefer bottleneck owners over exposed beneficiaries. If sector breadth fades, rotate out of the more crowded AI beta names first.