Ledger balances do not lie; they only wait.
On August 14, 2026, the Situational Awareness Fund—managed by former OpenAI researcher Leopold Aschenbrenner—filed its 13F with the SEC. The document is a snapshot of June 30 holdings: $202.4 billion in US equities. The portfolio is a monument to high-conviction AI infrastructure betting. Yet the fund had already collapsed by July, forced to liquidate most holdings under leverage pressure. The 13F is not a warning; it is a post-mortem.
Hype evaporates; receipts remain. The receipts show a portfolio that is 55.5% concentrated in two storage chip makers: SanDisk (28.0%) and Micron (27.5%). The remaining 44.5% includes GPU cloud providers CoreWeave and Nebius, fuel cell manufacturer Bloom Energy, TSMC ADR, and a cluster of Bitcoin miners—Core Scientific, Applied Digital, IREN, Riot Platforms, CleanSpark. The thesis: AI compute is the bottleneck, and the bottleneck is not just GPUs but storage, power, and physical infrastructure. The miners are rebranded as AI data center hosts.
From my forensic audits of crypto mining balance sheets, I have seen this narrative before. In 2024, every miner with a power contract pivoted to AI hosting. The pitch: we have power, land, and cooling. All we need is GPUs and a hyperscaler client. The Situational Awareness Fund bought that pitch wholesale. The 13F shows the fund held roughly 7-10% in miners, but the combination of leverage and illiquid mid-cap stocks created a powder keg.
Core Systematic Teardown
Let me dissect the holdings by liquidity layers. Layer one: SanDisk and Micron are large-cap, liquid, but still—selling 55% of a $200 billion portfolio in a single sector during a drawdown is not trivial. Layer two: Bloom Energy, TSMC, Nebius, CoreWeave—mid-cap to large-cap, still liquid but with higher beta. Layer three: the miners—Core Scientific (market cap ~$3B), Applied Digital (~$1.5B), IREN (~$2B), Riot (~$4B), CleanSpark (~$3B). These are small-cap stocks with thin order books. The fund held roughly $15-20 billion in this layer. When the AI stock selloff began in July, the leverage triggered margin calls. The fund had to sell into falling markets. The miners likely suffered the most severe price impact.
But the 13F does not tell the full story. It does not disclose leverage ratios, derivative positions, or short holdings. The fund’s “leverage” could be 2x or 10x. The fact that Citadel—a prime broker—had to take over the “problematic stock portfolio” suggests a structured unwind, likely a total return swap or a margin loan with a collateral shortfall. Volatility is not risk; opacity is. The true risk was hidden in the financing structure.
Contrarian Angle: What the Bulls Got Right
The bulls will argue that the thesis is sound. AI compute demand is real. Storage bandwidth (HBM) is a bottleneck. Power constraints for data centers are tightening. Bitcoin miners do have real assets: power purchase agreements, substations, and buildings. The fund’s vertical integration across the AI supply chain is logically consistent. Aschenbrenner’s own “Situational Awareness” paper argued that compute is the currency of AI dominance. The 13F is a perfect expression of that worldview.
But execution matters more than thesis. The fund ignored correlation risk. Storage, power, and GPU cloud are all driven by the same variable: AI capital expenditure. If AI CapEx slows—even briefly—every layer declines simultaneously. The fund also ignored the cyclicality of semiconductors. Micron and SanDisk are memory manufacturers; memory is a boom-bust industry. HBM demand is high now, but supply is catching up. By 2027, HBM oversupply is a real possibility. The fund had no hedge.
Furthermore, the miners are not pure AI infrastructure plays. They are still Bitcoin miners. If Bitcoin price drops, their core business cash flow suffers, and their AI pivot narrative loses credibility. The fund doubled down on a narrative that required both AI demand and Bitcoin price to remain stable. That is a fragile assumption.
Takeaway: The Accountability Call
The Situational Awareness Fund’s 13F is a lesson in how leverage transforms a sound thesis into a systemic risk. The fund’s collapse was not due to fraud or a rug pull—it was a textbook case of concentration plus leverage. The market absorbed the losses, but the episode should serve as a warning for any fund that believes “AI is different this time.”
Smart contracts are not the only things that can be liquidated. The same dynamics apply to traditional equities. The next time you see a 13F with a CR2 of 55%, ask: what is the leverage? Where is the liquidity? And if the answer is “we don’t know,” then the risk is not priced in—it is hidden.
Data does not forgive. The 13F is now a permanent record. It will be studied by regulators and risk managers. The question is not whether the thesis was wrong—it is whether the industry will learn that leverage and concentration are a toxic combination, regardless of the narrative.