Are we still in the early innings of the AI trade?

The article argues AI remains in the early innings, citing Dan Ives’ view that chip demand is still ahead of supply and enterprise adoption is limited. It highlights a bullish AI basket including Microsoft, Alphabet, Palantir, Nvidia, TSMC, CrowdStrike, Palo Alto Networks, AMD and Tesla, while noting Nvidia’s recent earnings as a key catalyst. It also flags a reawakening IPO market, with OpenAI and SpaceX generating speculation about future listings.

Analysis

The market is still pricing the AI buildout as a broad-duration theme, but the second-order winner remains the infrastructure layer, not the application layer. If compute demand is still outrunning supply, pricing power should stay concentrated in the semiconductor and wafer-fab ecosystem for longer than consensus expects, while software names will increasingly be judged on actual budget displacement rather than narrative. That tends to favor the names with near-term bottlenecks and visible backlog conversion, while creating fragility in higher-multiple “AI story” stocks that need multiple expansion to keep up.

The more interesting implication is competitive pressure inside the stack: as AI capex expands, hyperscalers and model builders will likely force vendor concentration around the most reliable suppliers, which can strengthen a few incumbents but hurt laggards that miss node transitions or packaging bottlenecks. Cybersecurity is a quieter beneficiary because AI adoption widens attack surfaces and raises data-governance costs, creating a budget line that can be funded even in a slower enterprise-spend environment. Conversely, the “physical AI” narrative is likely further out on the curve; it can drive sentiment now, but monetization is slower and more execution-sensitive than pure software or chip plays.

The key risk is timing: the theme can stay strong for quarters, but returns may compress if the market starts discounting 2026-2027 earnings too aggressively today. The most plausible reversal catalyst is not a collapse in demand, but a capex digestion phase where order growth normalizes while multiples remain elevated. On IPOs, the risk is that shiny listings reset public comps lower if they price off peak narrative rather than cash-flow durability, which can bleed into late-stage private AI names and weaken sentiment across the complex.

Key Decisions for Investors

• Stay overweight NVDA/TSM on a 3-6 month horizon; prefer TSM as the lower-beta way to express persistent compute scarcity, with NVDA reserved for stronger momentum but higher event risk around guidance and export headlines.
• Add CRWD and PANW on pullbacks as a secondary AI beneficiary trade; 2-4 month view, because security spend should be one of the few budget lines that keeps growing even if enterprise AI pilots slow.
• Use AMD as a relative-value hedge against NVDA strength: long NVDA / short AMD if you want to isolate share-gain and ecosystem leadership, or long AMD only on evidence of sustained design-win conversion; otherwise it remains the most vulnerable to narrative compression.
• Take profits selectively in PLTR and TSLA into strength; both have longer-dated optionality, but near-term upside is more dependent on multiple expansion than on near-term fundamentals, making them poor vehicles if the market gets more selective.
• If pursuing IPO exposure, favor a basket approach and smaller sizing; avoid concentrated day-one longs in the hottest private AI names, since post-listing volatility can easily create a 20-30% drawdown before fundamentals are established.