We watched the leverage unwind in July, but we missed the infection spreading through the settlement layer. The SEC’s 13F filing for the Situational Awareness Fund, submitted on August 14, 2026, offers a chilling snapshot of a portfolio that was already in flames. The file reveals a $20.24 billion concentration in AI infrastructure—55.5% in storage chips alone—and a tail of Bitcoin miners that were supposed to be the next AI data centers. But the numbers are a historical artifact; the real story is the systemic contagion that turned a high-conviction macro bet into a forced liquidation, now managed by Citadel.
This is not a crypto-native event, but it is a textbook case of how macro leverage, concentrated positions, and a single narrative (AI as the new oil) can create a fragile system that breaks when the liquidity tide turns. The bubble burst, and the lessons remain for anyone who thinks crypto is the only arena where smart money can blow up.
Context: The Man Behind the Thesis
Leopold Aschenbrenner is not a typical hedge fund manager. He was a researcher on OpenAI’s “superalignment” team, leaving in 2024 after disagreements over AI safety. His public essay, “Situational Awareness,” argued that compute power is the new geopolitical currency—that controlling AI hardware is the key to national power. The fund was his attempt to quantify that worldview: a portfolio built around the belief that AI infrastructure bottlenecks (storage, power, and compute) would be the hardest assets to scale, and thus the most valuable.
By June 30, 2026, the fund held 202.4 billion dollars in disclosed US equities. But the 13F, filed on August 14, is a snapshot from 45 days earlier. The market already knew something was wrong: in July, AI-related stocks suffered a sharp drawdown, and the fund was forced to sell most of its public holdings due to “leverage pressure.” Citadel stepped in to take over the “problematic portfolio.” This 13F is not a warning—it’s an autopsy.
Core: The Anatomy of a Bottleneck Bet
I’ve spent years modeling liquidity flows across crypto and traditional markets, and this portfolio structure is one of the most concentrated I’ve seen outside of pure crypto funds. The fund’s technical architecture is a vertical stack of AI infrastructure, mapped across five layers:
- Storage Layer (55.5%): SanDisk (28.0%) and Micron (27.5%). High-bandwidth memory (HBM) and NAND flash are critical for AI training and inference. The thesis: HBM supply is constrained, and these companies are the bottleneck.
- Foundry Layer (6.2%): TSMC ADR. The sole manufacturer of advanced AI chips (NVIDIA, AMD, etc.).
- Cloud/GPU Layer (9.8%): CoreWeave and Nebius. GPU cloud providers that lease compute to AI labs. CoreWeave is a major NVIDIA customer; Nebius is a pure-play AI cloud.
- Power Layer (9.4%): Bloom Energy. Fuel cells for data center electricity—a bet on power as the next physical constraint.
- Miner Transition Layer (~7%+): Core Scientific, Applied Digital, IREN, Riot Platforms, CleanSpark. Bitcoin miners pivoting to AI data center hosting, leveraging their existing power contracts and infrastructure.
The top two holdings alone (SanDisk and Micron) account for 55.5% of the portfolio. The top seven (including Bloom, TSMC, Nebius, CoreWeave, Core Scientific) total ~84.3%. This is a concentration ratio 2-3x higher than a typical institutional fund (CR10 usually 20-40%). In crypto terms, this is like a protocol where the top 5 wallets hold 80% of the supply—except here, the “supply” is real operating businesses with genuine cash flows.
The fund’s technical assumption is that the AI bottleneck is shifting from GPU compute to physical resources: storage, power, and rack space. This is a valid macro observation—HBM is indeed in short supply, and data center power is a growing constraint. However, the duration of these bottlenecks is often shorter than markets assume. Once capacity expands (new fabs, new power plants), the same stocks can face valuation compression and earnings downgrades. The fund had no hedge against that cyclical risk.
Moreover, the inclusion of Bitcoin miners as “AI infrastructure” is a narrative overlay. Miners own power and rack space, and many have signed AI hosting contracts. But they also carry Bitcoin price exposure, mining economics, and operational leverage. When the AI narrative weakens, these stocks lose both the AI premium and the crypto floor. The double hit is brutal.
Contrarian: The Decoupling Fallacy
The conventional wisdom in crypto circles is that “this doesn’t affect us—it’s a traditional fund.” I argue the opposite: this is a canary in the coal mine for any asset class that depends on concentrated, leveraged bets on a single macro narrative. The mechanism is identical to what we saw in DeFi in 2020 and Terra in 2022: high conviction + high leverage + low liquidity tail assets = nonlinear downside.
What’s missing from the 13F is the leverage structure. The filing doesn’t require disclosure of margin loans, total return swaps, or short positions. The fund’s “leverage pressure” suggests a margin call or a TRS termination. The fact that Citadel took over the portfolio, not just liquidated it, implies a structured unwind—possibly a prime brokerage agreement that collapsed. Algorithms don’t fail; models do. The model here was that AI infrastructure would only go up, and that the fund could carry concentrated positions without a liquidity buffer.
Another blind spot: the fund had zero exposure to AI application layers (no OpenAI, no Anthropic, no software companies). This is a “picks and shovels” bet—but if the gold miners (AI labs) cut their capex, the shovel makers (storage, power, cloud) all suffer simultaneously. The portfolio had no diversification across the value chain, and no offsetting short positions. It was a pure directional long on AI infrastructure.
From a macro perspective, the timing is critical. The July drawdown occurred during a period of quantitative tightening and rising volatility. The fund’s leverage was exposed to the same liquidity contraction that hits all risk assets. The lesson for crypto is clear: macro liquidity cycles dominate all narratives. The next crypto bull run will not be immune to a similar unwind if leveraged funds pile into a single theme.
Takeaway: Positioning for the Aftermath
The Situational Awareness Fund is now a case study in how powerful macro narratives can lead to dangerous portfolio structures. The 13F gives us a clean before-and-after: the before is a $20.24 billion bet on AI bottlenecks; the after is a portfolio in Citadel’s hands, likely being unwound block by block.
For those of us who track cross-border capital flows and systemic risk, this is a reminder that the same patterns of contagion exist in all markets. The bubble burst, but the lessons remain. The next cycle will be built on more resilient infrastructure—or it will repeat the same mistakes. The question is not whether leverage will return, but whether we will recognize the warning signs before the next collapse.
Composability is a double-edged sword. In AI infrastructure, it means all positions are correlated; in crypto, it means DeFi protocols cascade. The macro watcher’s job is to trace the linkages, not to predict the next price move. The Situational Awareness Fund’s 13F is a map of the linkages that broke. We should study it carefully.
