This Was Situational Awareness's Top 5 Holdings at the End of Q2. They're All on Sale Now

The article traces the collapse of hedge fund Situational Awareness in late July after highly leveraged bets on AI stocks triggered margin calls when the AI sector tumbled, and notes it sold over $10B of its AI portfolio to Citadel. It then highlights Q2 top holdings—e.g., Sandisk up 2,700% YoY and Micron up 600%—arguing memory tightness and AI demand could support upside, while flagging that the riskiest “neoclouds” (Nebius and CoreWeave) are growing with triple-digit revenue but carrying massive losses and billions in debt.

Analysis

The market implication is not “AI is over”; it is that the marginal buyer in the theme has become less important than the marginal financer. That matters because the most levered names can gap down for weeks after a deleveraging event even if end-demand is fine, while cash-generative infrastructure beneficiaries reprice more on fundamentals than flows. In that setup, CRWV and NBIS are the cleanest shorts on balance-sheet risk and refinancing sensitivity; BE is higher quality than the neoclouds but still trades like a crowded duration asset rather than a utility-like supplier.

The more durable winners are the bottleneck owners: TSM and MU. TSM monetizes every incremental AI capex dollar with less funding risk, and MU benefits if memory tightness persists, but the market will eventually separate HBM scarcity from commodity flash, which is why SNDK is a weaker expression than MU over 6-18 months. If AI spending stays intact, the next leg is likely driven by supply constraints and mix rather than broad multiple expansion.

Contrarian view: the consensus is treating the unwind as a temporary volatility event, but forced selling can create a 1-3 month air pocket in the highest beta AI proxies even without any fundamental break. The key falsifier is not a single bad print; it is either a sustained slowdown in hyperscaler capex or evidence that neocloud financing costs force equity dilution. If those do not appear, the dip is more likely a rotation within AI than the end of the trade.

Key Decisions for Investors

• Long MU / short CRWV for 1-3 months: express 'real bottleneck vs funded growth' with better downside protection if AI spending stays healthy; cover if CRWV regains momentum after a refinancing or backlog update.
• Long TSM / short NBIS on a 2-6 week horizon: TSM should outperform as a lower-leverage beneficiary of AI capex while NBIS remains exposed to funding-market volatility; thesis breaks if NBIS secures cheap incremental capital and narrows spreads.
• Buy MU on pullbacks after the next memory pricing datapoint: use weakness to build a core position for 6-18 months, but avoid chasing if HBM commentary weakens or channel inventory rises.
• Avoid initiating fresh longs in CRWV and NBIS until after their next financing/earnings cycle; if you need exposure, use small call spreads rather than equity because the downside from dilution is asymmetric.
• Watch BE as a secondary beneficiary, not a primary one: only add on confirmation of backlog conversion and margin durability; otherwise its valuation can compress faster than AI infrastructure peers.