The Top Of the AI TradeIsn't In Yet: Macro Man Podcast

The key read-through is that AI leadership is transitioning from a one-way momentum trade into a stock-selection market. When the “obvious” beneficiaries of AI capex start wobbling, capital usually rotates one level down the stack toward picks-and-shovels with cleaner cash conversion: power, networking, optics, memory, and semicap equipment with backlog visibility. The second-order effect is that the market begins rewarding revenue durability over narrative duration, which tends to compress multiples for the most crowded mega-cap compute names before it materially hurts the broader AI spending cycle.

The bigger risk is not a near-term collapse in AI demand, but a positioning air pocket. The tape can stay fragile for days to weeks if systematic funds are forced to de-gross and retail call-buying slows, because these names have become index-level beta proxies. Over 1-3 months, the more important catalyst is whether hyperscalers validate 2026 capex budgets on earnings calls; if they do, the trade likely stabilizes and leadership broadens. If they don’t, the market will start discounting a slower second-half spend cadence, which would pressure the highest-multiple semis first.

The contrarian view is that the recent pullback may be more of a valuation reset than a thesis break. AI monetization is still early, and the market has likely over-punished anything tied to near-term GPU scarcity while underappreciating beneficiaries with secular demand but less cyclicality. In prior momentum unwind episodes, the best risk/reward came not from catching the first bounce in the crowded leaders, but from owning enablers with cleaner balance sheets and shorting the most extended “AI-adjacent” names that need perfect execution to justify their multiples.