Egan-Jones Examines the AI Race and Its Implications for Investors

Egan-Jones says AI competition among leading tech firms is intensifying and could reshape sector winners and losers as AI investment grows. It highlights likely beneficiaries across the AI stack—semiconductors, data centers, power producers, foundational model developers, and hardware—while flagging constraints in components and energy availability. The note emphasizes a more dynamic market driven by lower switching costs and increased competitors, with long-term leadership defined by intelligence, speed, and price, but warns that regulatory uncertainty remains high.

Analysis

This is not a fresh catalyst so much as a confirmation that the AI value chain is splitting into two businesses: compute supply and distribution. The near-term market implication is that capital will keep crowding into the same infrastructure names, but the next leg of outperformance should be driven by who has hard constraints on supply, energy, and customer lock-in rather than who has the best narrative. That favors the pick-and-shovel layer, but only selectively; many AI-linked names are already priced for sustained capex growth.

The more interesting second-order effect is margin compression in the model/application layer. Lower switching costs and faster release cycles make it hard for stand-alone AI developers to defend pricing, while platform owners with OS, browser, or device control can bundle AI into existing monetization funnels. That is structurally better for AAPL, MSFT, and GOOGL than for pure-play model vendors, and it argues for continued relative strength in networking, memory, cooling, and power infrastructure names such as ANET, AVGO, VRT, ETN, DLR, and select utilities.

Contrarian view: the consensus is probably still overestimating the durability of AI pricing power and underestimating how quickly inference costs fall. If cost per query keeps dropping, the winner is the owner of demand and distribution, not the model itself. The key falsifier over the next 1-3 quarters is hyperscaler capex guidance and backlog conversion; if spending growth slows, the infrastructure trade loses its fundamental support within months, even if AI enthusiasm remains high.

Key Decisions for Investors

• Stay overweight AI infrastructure on pullbacks: NVDA, AVGO, ANET, VRT as the cleanest expression of continued capex; use weakness from broad tech selloffs to add rather than chase strength.
• Prefer platform owners over stand-alone model exposure: long MSFT/AAPL/GOOGL on a 6-12 month horizon because distribution and bundling should defend monetization better than pure model vendors.
• If needing a relative-value hedge, pair long SMH against short a broad software basket (IGV) into any AI enthusiasm spike; thesis is that infrastructure revenue is more visible while software multiples are more vulnerable to price competition.
• Watch the next hyperscaler earnings cycle for capex guidance; if 2026 spend growth decelerates or power/interconnect commentary softens, trim AI infrastructure longs and reassess.
• Keep DLR/ETN/CEG/NEE on a watch list rather than a full recommendation until contracted load and backlog data confirm that AI demand is translating into actual utility and data-center economics.