The 13F filing is supposed to be a window into genius. A quarterly snapshot of what the smartest minds in capital allocation are betting on. The Situational Awareness fund’s 13F, filed on August 14, 2026, is something else entirely. It’s a crime scene photo. The portfolio, as of June 30, showed $20.2 billion in assets—a concentrated, levered bet on the AI infrastructure narrative, with SanDisk and Micron alone making up 55% of the fund. By July, that bet had blown up. The fund was forced to sell at a loss, and Citadel stepped in to take over the “problematic portfolio.” The 13F arrived two weeks after the explosion. The market already knew the story. But the filing reveals the structural rot that made the collapse inevitable. And it’s not about AI. It’s about leverage, concentration, and the illusion of conviction.
Leopold Aschenbrenner is not your typical hedge fund manager. He’s a former OpenAI researcher who left the alignment team in 2024, disillusioned with safety culture. His widely-read essay, “Situational Awareness,” argued that compute is the ultimate geopolitical asset—control the chips, control the future. His fund was the practical embodiment of that thesis. The 13F shows a portfolio that reads like a supply chain map for AI training: storage (SanDisk, Micron), fabrication (TSMC), cloud compute (CoreWeave, Nebius), power (Bloom Energy), and a handful of Bitcoin miners that had pivoted into AI data centers (Core Scientific, Applied Digital, IREN, Riot Platforms, CleanSpark). No software. No AI models. No diversification. Just a vertical slice of the physical layer.
Let me be clear: this is not a crypto-native fund. But the crypto market is deeply intertwined with its fate. The miners in the portfolio are publicly traded companies that originally derived their value from Bitcoin mining. Their pivot to AI hosting is the only reason they survived the 2022-2023 bear market. By 2026, the market had re-rated them as “AI data center plays,” not Bitcoin proxies. The 13F captures that re-rating. But it also captures the fragility.
I’ve been auditing crypto and tech portfolios since 2017, when I reviewed 40+ ERC-20 whitepapers and found critical reentrancy bugs in supposedly secure payment gateways. The pattern I see in this 13F is the same pattern I saw in those ICOs: high conviction, minimal hedging, and a structural blind spot to liquidity shocks. The fund’s concentration ratio is CR2 = 55.5%, CR7 ≈ 84.3%. For context, a typical institutional fund runs a CR10 of 20-40%. That’s not conviction—that’s a single point of failure masked as a thesis.
The core of the bet was simple: storage and power are the new bottlenecks. AI training requires massive amounts of high-bandwidth memory (HBM) and NAND flash. SanDisk and Micron are the primary suppliers. Bloom Energy provides fuel cells for data centers. The miners own power contracts and existing infrastructure. The logic is sound at the surface. But the execution ignores the second-order effect: when the AI CapEx narrative turns, every piece of the stack gets hit simultaneously. There is no buffer. The fund didn’t own any AI application stocks—no OpenAI, no Anthropic, no software layer. It was a pure “picks and shovels” play, but with a twist: the shovels were all levered to the same buyer.
The leverage is the hidden killer. The 13F doesn’t disclose margin, derivatives, or term sheets. But the market reporting from July confirms that the fund was forced to sell due to “leverage pressure” after AI stocks declined. This is not a unique story. In crypto, we’ve seen it with Three Arrows Capital, with Celsius, with every levered fund that thought the trend would last forever. Liquidity doesn’t care about your thesis. The fund’s holdings included small-cap miners with thin order books. When the forced selling started, those positions likely suffered the most severe price impact. The auditor blinked; the market didn’t.
Let me offer a contrarian angle: the market is framing this as an “AI reckoning” or a “crypto contagion.” It’s neither. It’s a classic macro leverage blow-up that happened to be themed around AI. The miners were not the cause; they were the victims. The real lesson is that the “AI trade” is now crowded and levered, and the next shock will come from the same structure. The Situational Awareness fund was a canary in the coal mine, but the coal mine is the entire infrastructure sector. If you strip away the narrative, this is a portfolio that would fail any basic risk management review. The fact that it was managed by a former AI researcher—not a seasoned risk manager—is not a coincidence. Conviction without mechanism is just gambling with extra steps.
What does this mean for the crypto market? The miners in this portfolio—Core Scientific, IREN, Riot—are now branded as “AI data center” companies. Their valuations partly depend on the AI narrative. If the AI CapEx cycle slows, these stocks will be hit harder than pure-play crypto miners because they have lost their “crypto premium” without gaining a stable “AI premium.” They are caught in a narrative no-man’s land. The 13F shows that as of June 30, smart money was still betting on the pivot. By July, that bet was unwound. The paradox is that the 13F itself is a lagging indicator: it shows what the fund owned before the crash, not what it was forced to sell. But the composition tells us the vulnerability. The highest concentration in the most liquid names (SanDisk, Micron) provided a veneer of safety, but the tail of illiquid miners was the real time bomb. When the margin call came, the liquid assets were sold first to cover the lever, but the illiquid tail took the biggest relative loss.
I’ve seen this movie before. In 2022, I analyzed the Terra collapse through the lens of shadow banking. The same pattern applies: a high-conviction, levered structure that looks stable in a rising market but becomes a death spiral when liquidity contracts. The Situational Awareness fund is not a crypto story, but it follows the same logic. The difference is that the collateral here is not algorithmic stablecoins but blue-chip tech stocks. The mechanism is the same: leverage amplifies both the upside and the downside. The upside was captured in the 13F. The downside is why we are reading about it now.
Smart contracts don’t commit fraud, but concentration does amplify ruin. The 13F is a perfect illustration of why diversification is not just a risk management tool—it’s a survival mechanism. The fund’s bet on AI infrastructure was not wrong. It was just too concentrated, too levered, and too late to exit. The filing is a post-mortem, not a roadmap. But for the rest of us, it’s a warning. The next time you see a 13F with 55% in two names, ask yourself: what happens when the music stops? The auditor blinked; the market didn’t.
I’m watching the secondary effects. The miners in this portfolio may face a new round of fundraising challenges as investors grow wary of the “AI pivot” narrative. The collapse of a $20B fund—even one that is not directly crypto—will tighten credit conditions for all levered thematic plays. That includes crypto miners, AI infrastructure startups, and even some DeFi protocols that rely on similar leverage structures. The macro environment is already fragile. The Federal Reserve’s quantitative tightening continues. The AI trade was the last pocket of exuberance. This fund’s failure is a signal that the exuberance is over.
My forward-looking judgment: the next 12 months will see a consolidation of AI infrastructure plays. The survivors will be those with real cash flow, low leverage, and diversified customer bases. The losers will be the ones that rode the narrative without building the buffers. The 13F from Situational Awareness is not just a historical document. It’s a template for what to avoid. Lock in your portfolio construction. Hedge your tail risks. And remember: liquidity doesn’t care about your thesis.
