The unwind of Situational Awareness isn't the only reason the AI trade is back

Microsoft surged 15.5% last Thursday and added another 11% through Thursday’s close after strong Azure growth and Copilot traction, helping validate a rebound in the AI infrastructure trade. Amazon’s AWS revenue grew 37% year-over-year and CEO Andy Jassy defended aggressive AI capex, reinforcing confidence that spending won’t peak—supporting data center suppliers including Corning, Intel, Eaton, and Micron. The article links the shift to the forced unwind of the leveraged Situational Awareness hedge fund after a July margin-call/mass selling event, with the AI complex described as moving onto “sturdier ground” following earnings confirmations.

Analysis

The key read-through is not just that AI spending is back; it is shifting from a narrative trade to a cash-flow trade. The market is rewarding names where incremental capex has a visible revenue bridge or an obvious bottleneck position: hyperscalers with monetization proof, power/cooling vendors, and the chip layer with real order visibility. That re-ranks winners versus the first-half trade, where the market paid up for “AI exposure” without demanding evidence; the second-order loser is lower-quality software shorts that depended on a capex-slowdown story that now looks too early.

Near term, the biggest risk is that investors extrapolate one forcing event into a multi-quarter re-leveraging of the whole complex. If rates stay sticky or hyperscaler capex guidance pauses in the next earnings season, the highest-multiple names can retrace fast even if the long-term AI thesis remains intact. The cleaner setup is in the picks-and-shovels with backlog and power constraints, where demand is less sentiment-driven and more tied to physical buildout over 6-18 months; that favors infrastructure over pure compute beta.

The contrarian point is that the market may be underestimating how asymmetric the winners are inside AI. Not every beneficiary deserves the same multiple: cash-generative platforms with AI monetization should keep winning, while crowded “AI safety/cyber” beneficiaries and the most levered chip proxies can give back gains once headline risk fades. What would falsify the bullish regime is a capex rollback, a sudden yield spike that hits duration names, or evidence that memory/pricing and order lead times are rolling over.

Key Decisions for Investors

• Go long MSFT on any post-earnings pullback for a 1-3 month hold; it remains the cleanest expression of AI monetization with the best downside support from free cash flow. Invalidate if Azure growth decelerates or AI spend rises without visible revenue conversion.
• Initiate a relative-value long ETN / short NVDA pair for 3-6 months to express the view that power, cooling, and electrical bottlenecks are the more durable AI bottleneck than compute beta. Falsify if NVDA re-accelerates guidance materially above the rest of the stack.
• Add to AMZN on weakness, but only as a medium-term trade, not a chase. AWS plus management willingness to fund the buildout improves the capex durability story; exit if cloud growth stalls or margins compress more than expected.
• Treat MU as a watchlist long, not an outright buy, until memory pricing confirms the next capex leg. The stock has the best operating leverage if the buildout persists, but it is the first to reverse if capex softens.
• Fade PANW/CRWD strength tactically if AI-cyber headlines cool and the stocks keep outperforming fundamentals; this is a sentiment trade with a shorter half-life than the infrastructure leg.