Key Takeaway
Former OpenAI researcher Leopold Aschenbrenner (fired from OpenAI at age 22) founded Situational Awareness Fund with remarkable results: $225M grew to over $5.5B in one year.
Latest 13F filings and community tracking data show the fund’s managed assets reached $8M (community tracking version), up +58.3% since launching Autopilot on March 5.
Aschenbrenner’s core investment logic: In the AI era, owning the power and hardware to run models is more valuable than owning the models themselves.
Fund Performance
| Metric | Value |
|---|---|
| Initial Size | $225M |
| Current AUM | $5.5B+ |
| Community Tracker AUM | $8M |
| Autopilot Launch | March 5, 2026 |
| Autopilot Return | +58.3% |
Investment Logic: Jensen’s AI Five-Layer Framework
| Layer | Content | Investment Examples | Shortage Status |
|---|---|---|---|
| L1: Energy | AI training campus power | Nuclear, natural gas, grid | Each campus needs 1GW+ |
| L2: Compute | GPU/TPU training chips | NVIDIA, AMD | Continuously undersupplied |
| L3: Optical | Data center interconnects | Optical modules, NVLink | Bandwidth demand exponential |
| L4: Memory | HBM and storage | HBM, SSD vendors | HBM capacity tight |
| L5: Chips | Advanced node foundry | TSMC, semiconductor equipment | Capacity expansion takes 2-3 years |
Aschenbrenner’s argument: Every layer is in shortage, every layer is repricing higher. The most upstream layers—energy and compute—are the most irreplaceable.
Holdings Analysis
Core holdings include:
- NVIDIA (NVDA): Absolute core of AI compute
- Vertiv (VRT): Data center power and thermal management
- Constellation Energy (CEG): Nuclear operator, AI data center power supplier
- Eaton (ETN): Power management equipment
- Arista Networks (ANET): Data center network infrastructure
Notably, Aschenbrenner does not directly hold AI model companies (like OpenAI, Anthropic). His bet is entirely on “the shovel sellers.”
Why It Matters
1. “Dimensional Strike” from a Former OpenAI Researcher
Aschenbrenner’s understanding of the AI technology roadmap far exceeds traditional fund managers. His “AI infrastructure first” strategy is based on engineering-level understanding of model training needs.
2. Strategy Being Rapidly Copied
Multiple “AI infrastructure themed funds” have emerged. Lumenai Innovation Fund (launching June 1, 2026) is one example.
Risk Warnings
- $5.5B AUM grew 24x in one year—much of the growth comes from performance, not new capital inflows
- The fund is highly concentrated in AI infrastructure, lacking diversification
- Aschenbrenner is only 27 years old with relatively limited experience managing this scale
Sources:
- Twitter community tracking @ai_finance
- SEC 13F filings