The 13F filing hit the SEC database on August 14, 2026. It showed a $202.4 billion portfolio as of June 30. The names were familiar: SanDisk, Micron, Bloom Energy, TSMC, CoreWeave, Nebius, and a handful of Bitcoin miners. The market reacted with a collective shrug. The crash had already happened in July. This filing was not a warning. It was an autopsy.
I have audited enough smart contracts to recognize a centralization of risk when I see one. In 2017, I spent 40 hours verifying the PotCoin ICO distribution logic and found an integer overflow that would have drained the wallet. The community dismissed it. The code did not. That experience taught me one rule: if the logic is fragile, the outcome is predetermined. The Situational Awareness fund’s 13F is a fragile logic. It is a portfolio masquerading as a thesis. And the market already executed the penalty.
Context: The Man Behind the Thesis
Leopold Aschenbrenner is not a trader. He is a former OpenAI researcher who wrote a viral essay called Situational Awareness about AI posing an existential risk. In 2024, he left OpenAI and started a hedge fund. The thesis was simple: AI compute is the bottleneck, and the bottlenecks are storage, power, and data centers. The 13F shows exactly that. SanDisk and Micron account for 55.5% of the portfolio. Bloom Energy provides fuel cells for data centers. CoreWeave and Nebius are GPU cloud providers. Core Scientific, Applied Digital, IREN, Riot, and CleanSpark are Bitcoin miners pivoting to AI hosting.
But the 13F is a snapshot. It was filed on August 14, covering positions held on June 30. The collapse happened in July. The fund was forced to liquidate due to leverage and AI stock declines. Citadel took over the “problem portfolio.” The filing is a post-mortem, not a live signal. Yet it is rich with lessons.
Core: The Compute Bottleneck Bet – A Quantitative Dissection
Let me walk through the numbers. The 13F discloses $202.4 billion in assets. The top two positions – SanDisk and Micron – represent 55.5% of the portfolio. The top seven positions account for 84.3%. That is not a portfolio. That is a single trade with seven names. It is a concentrated bet on the AI compute bottleneck.
I have seen this structure before. During DeFi Summer in 2020, I managed a €50,000 portfolio across Compound and Uniswap. I wrote an Excel tracker to monitor real-time yields. When I found a 15% arbitrage on cCOMPTOKEN, I rebalanced immediately. But the first rule of my strategy was diversification across protocols. This fund had no diversification. It was a high-conviction, high-leverage, high-concentration play. The Sharpe ratio looks great in a bull market. In a drawdown, it becomes a weapon of self-destruction.
Liquidity is the only truth in a fragmented chain. The miners – Core Scientific, Applied Digital, IREN, Riot, CleanSpark – are small-cap, low-liquidity names. They represent about 15% of the portfolio. But in a forced liquidation, they become the biggest drag. You cannot sell 10% of a small-cap stock without moving the price. The fund probably experienced severe slippage on those positions. The 13F does not show the execution prices. The market does not forget.
Now, look at the technical structure. The portfolio is a vertical stack of the AI compute supply chain:
- Storage layer: SanDisk (NAND flash) and Micron (HBM, DRAM) – 55.5%
- Foundry layer: TSMC ADR – 6.2%
- Cloud/GPU layer: CoreWeave and Nebius – 9.8%
- Power layer: Bloom Energy – 9.4%
- Miner/data center layer: Core Scientific, Applied Digital, IREN, Riot, CleanSpark – ~15%
The thesis is that AI compute demand will grow exponentially, and the bottlenecks are storage, power, and physical infrastructure. This is logical. But it is also fragile. The entire portfolio is long the same macro factor: AI CapEx. There is no hedge. No short positions. No software or application layer exposure. If AI CapEx slows, the entire stack collapses simultaneously.
I have seen this fragility in the 2022 Terra collapse. I held $30,000 in UST derivatives. When the algorithmic peg broke, I executed stop-losses across three exchanges within minutes. I saved 85%. But I also learned that a single point of failure – a stablecoin design, a levered portfolio – can wipe out years of gains. The Situational Awareness fund had a single point of failure: the assumption that compute bottlenecks are linear and permanent.
Contrarian: The True Failure Was Not Leverage – It Was the Narrative
The market narrative is that the fund blew up because of leverage. That is true but superficial. Every levered fund has leverage. The real failure is the narrative itself. The fund assumed that the compute bottleneck is a durable trade. But bottlenecks are temporary. Capacity catches up. The HBM supply is expanding. TSMC is building new fabs. Power grids are being upgraded. The fund’s thesis was a static bet on a dynamic system.
Beta is the tax you pay for ignorance. The miners are the worst part. They are not pure AI infrastructure. They are Bitcoin miners with a pivot narrative. Their revenue depends on both Bitcoin price and AI hosting contracts. The fund bundled them into the AI thesis, but they bring crypto volatility. In July, when AI stocks fell, the miners fell harder. The leverage amplified the move. The fund was forced to sell into a declining market.
Here is the blind spot: the portfolio had no exposure to the application layer. No OpenAI, no Anthropic, no AI software companies. The fund was betting on the “picks and shovels” of AI. But picks and shovels are only valuable if the gold rush continues. If AI spending slows, the entire infrastructure chain loses value. The fund did not hedge this risk. It was a pure directional bet.
I have seen this pattern in the 2024 ETF arbitrage trade. I built a Python script to track the Coinbase Premium Index. I found a 2% spread and executed a trade that generated €12,000. But I also set a stop-loss. The Situational Awareness fund had no stop-loss. It had a thesis. The thesis broke.
Takeaway: What the 13F Teaches Us About Risk
The 13F is a lesson in concentration risk. It is a warning about leverage. But more importantly, it is a reminder that narratives are not price levels. The market will eventually price in the bottleneck. The fund’s collapse shows that even the smartest thesis can be destroyed by poor risk management.
Sanity checks before sanity wins. I will watch the next 13F from Citadel. If they have reduced the miner positions, the liquidation is complete. If they are still holding, the overhang remains. The market is a learning machine. This filing is its latest lesson. The question is: have you learned it?
