Think the AI Bull Market Is Just Getting Started? These 3 ETFs Are Positioned for the Next Leg

Morgan Stanley estimates over 80% of the roughly $3T in expected AI infrastructure investment through 2028 is still ahead, supporting the view that the AI trade has further runway. The article highlights three diversified ETF vehicles—Vanguard VGT (0.09% expense ratio, $146.6B AUM, 326 holdings), VanEck SMH (0.35%, $70.8B AUM, 26 holdings; Nvidia 17.55%), and Global X AIQ (0.68%, $9.8B AUM, 93 holdings)—to gain exposure across AI infrastructure, semiconductors, and broader AI software/cloud/automation.

Analysis

This is more a confirmation of an already-established capex regime than a fresh catalyst, so the edge is in relative positioning, not outright beta chasing. The first-order winners remain the infrastructure toll collectors: NVDA is still the highest-quality exposure, but at this stage the better risk/reward is often in the less crowded second-order beneficiaries such as TSM, MU, and AVGO, where incremental AI spend can show up in utilization, mix, and pricing before it is fully reflected in consensus. MSFT benefits later in the cycle as inference and workflow monetization catch up to build-out.

The loser set is more structural than cyclical. INTC remains the clearest relative short because it needs execution to change the narrative, while the AI spend wave mostly rewards firms already embedded in the supply chain. Broad tech wrappers like VGT should outperform more concentrated semiconductor baskets on a risk-adjusted basis if AI adoption broadens beyond training hardware, because they capture both infrastructure and monetization with less single-name crowding.

The main risk is not that AI is "over" but that positioning is too one-way: any hint of hyperscaler capex moderation, export-control friction, or a pause in accelerator orders can trigger a sharp de-rating in the most crowded names within days. Over 1-3 months, the key falsifier is guidance from NVDA/TSM/AVGO/MU showing demand normalization faster than expected; over 6-18 months, the debate shifts from build-out to ROI, which should increasingly favor software and platform names over pure compute. The contrarian miss is that investors keep treating this as a semis-only trade when the more durable alpha may sit in diversified exposure plus select monetizers.

Key Decisions for Investors

• Long VGT / short SMH for 1-3 months as a relative-value pair: if AI spend broadens, the lower-fee, more diversified wrapper should hold up better; risk is a renewed NVDA-led squeeze that keeps semis outperforming for another leg.
• Build a basket long in MSFT + AVGO + TSM on pullbacks over the next 2-6 weeks: these names capture both infrastructure spend and later-stage monetization, with better durability than chasing the most crowded pure-beta AI trade.
• Use INTC as a relative short against the AI complex for 1-3 months: it needs a tangible execution inflection to rerate, and absent that, AI capex is more likely to leak away share than create a catch-up story.
• If already long NVDA/SMH, tighten risk and wait for earnings/capex confirmation rather than add here; the payoff is still positive, but the marginal return now depends on continued surprise, not the theme itself.
• Watch for any hyperscaler capex slowdown or TSM utilization commentary as the first falsifier; if those data points soften, rotate from SMH into VGT/MSFT within days.