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 reported $5.5 billion in assets across 24 stocks, with a portfolio heavily concentrated in AI infrastructure names rather than the largest AI platforms. Top holdings include Bloom Energy at 16%, Lumentum at 8.7%, CoreWeave at 7.9%, Core Scientific at 7.4%, Iren at 6%, Applied Digital at 5%, and Sandisk at 4.5%, with several positions showing triple- to quadruple-digit gains over the past year. The piece is broadly bullish on AI infrastructure and highlights strong momentum in related stocks, but it is primarily a fund-holdings profile rather than a direct corporate catalyst.

Analysis

This is a clean read-through on AI capex diffusion rather than a renewed endorsement of the obvious semis trade. The highest-beta beneficiaries are the “picks-and-shovels for compute” names where revenue can reprice faster than investor models, especially power, optical interconnect, and purpose-built data center capacity. That matters because once AI deployment shifts from model training to always-on inference, the bottleneck moves from GPU scarcity to electrons, cooling, bandwidth, and memory persistence — a broader, less obvious spend stack with multiple winners.

The second-order effect is that the market is likely underestimating how quickly incumbents in adjacent industrial niches can re-rate when they become mission-critical to hyperscaler buildouts. Bloom and the converted crypto-to-AI data center names have the most convexity because they sit closest to constrained physical inputs: power, land, and permitting. That also means their upside is path-dependent on execution; any slip in delivery schedules or customer concentration issues will hit multiples harder than fundamentals because expectations have already sprinted ahead of trailing numbers.

The contrarian risk is crowded positioning, not thesis failure. If AI infrastructure spend decelerates even modestly for one quarter, the market will punish the highest-multiple “infrastructure convenience” names first, since they are being priced as if demand is linear and financing is frictionless. The more durable trade may be the enablers with downstream leverage to storage and networking demand, where capacity additions are slower and pricing power can persist through the next few quarters even if sentiment cools.

Key Decisions for Investors

• Long BE vs short a basket of high-duration AI winners: the setup favors an outright beneficiary of power constraints, but size as a 3-6 month tactical trade because the stock has already re-rated sharply and will be sensitive to any data-center order delay.
• Initiate a pair trade: long LITE / short AVGO for 6-12 months. LITE should benefit more from incremental optical intensity in AI buildouts, while AVGO remains a higher-quality but more fully-owned AI exposure with less incremental surprise.
• Buy CRWV on pullbacks, not breakouts, and use call spreads rather than stock. The risk/reward is attractive over 6-9 months if AI cloud demand keeps compounding, but valuation and financing risk make upside convexity preferable to linear exposure.
• Prefer SNDK as a relative long versus the broader infrastructure basket for 12 months. Memory is one of the few AI inputs where supply discipline can support pricing; if inference workloads keep rising, storage content per server should keep expanding.
• Avoid chasing APLD and IREN after multi-hundred-percent runs; if you want exposure, use defined-risk calls into earnings or contract announcements. These are execution stories with high optionality, but the market is paying for near-perfect delivery.