Artificial Intelligence (AI) Prodigy Leopold Aschenbrenner's Situational Awareness Fund Owns Only 24 Stocks. These Are the Top 7 Holdings.

Leopold Aschenbrenner’s Situational Awareness has grown to $5.5 billion in assets and holds 24 stocks, with a heavy emphasis on AI infrastructure rather than marquee AI names. Top positions include Bloom Energy (16%), Lumentum (8.7%), CoreWeave (7.9%), and Core Scientific (7.4%), with several holdings up 29% to 2,800% year to date. The piece is broadly bullish on AI-support infrastructure and highlights strong year-to-date performance across the fund’s core holdings.

Analysis

This basket is less a “pick-and-shovel” AI trade than a power-and-capacity constraint trade. The common denominator is that model scaling is now colliding with physical bottlenecks: grid interconnects, rack-level power density, cooling, optics, memory, and dedicated inference capacity. That makes the winners more cyclical than the AI megacaps, but also more leveraged to any incremental capex surprise from hyperscalers over the next 2-6 quarters.

The strongest second-order effect is that capital is rotating from software narratives into bottleneck suppliers with shorter feedback loops. If AI demand keeps accelerating, names tied to power generation and data-center buildouts should see the fastest estimate revisions; if spending pauses, these same names can de-rate violently because many still trade on forward scarcity rather than steady-state cash flow. The high-beta setup also implies crowdedness risk: these stocks are becoming a consensus “AI infrastructure” expression, which usually precedes a period of factor underperformance even if the fundamental story remains intact.

The contrarian read is that the market may be overpaying for scarcity and underestimating substitution. Hyperscalers can defer, internalize, or redesign around some of these bottlenecks, especially where customers demand flexibility or where leasing economics get stretched. The more durable edge likely sits in components with the least substitutability and highest qualification friction—optics, memory, and power delivery—rather than in the most obvious landlord-like data-center plays.

Near term, the catalyst set is mostly earnings and capex commentary over the next 1-2 quarters, plus any evidence of delayed interconnects, power shortages, or customer concentration. The main tail risk is that order growth remains strong but gross margin compresses as competition intensifies and financing costs rise, turning “AI infrastructure” into a crowded growth-at-any-price trade.

Key Decisions for Investors

• Long LITE / short APLD for 3-6 months: prefer the more essential, lower-substitutability optical bottleneck over lease-up/data-center exposure that is more financing- and customer-dependent; target 15-20% relative outperformance if AI capex stays strong.
• Initiate a tactical long BE with tight risk controls into the next earnings cycle: power availability is the cleanest constraint in the chain, but treat it as a high-volatility expression; use a 2-3 month horizon and take profits quickly on any post-print gap-up.
• Favor SNDK over the rest of the basket on a 6-12 month view: memory is one of the least avoidable AI inputs, and the market still underprices how much inference and retrieval architectures lift durable storage demand.
• Avoid chasing CRWV and CORZZ at current levels; if you want exposure, use call spreads rather than common stock to cap downside. These names are most exposed to any slowdown in hyperscaler leasing or a funding window reset.
• Pair long INTC or AVGO only if you need a hedge against AI infrastructure exuberance; otherwise stay neutral on the megacaps, since the incremental signal here is not GPU demand but the tightening of adjacent supply chains.