When a Genius AI Fund Bet on Chips and Miners: The 13F That Told a Story Too Late

We didn't see the 13F until two weeks after the fire. That's the cruel irony of SEC disclosure rules—they give you a rearview mirror when you need a headlight. On August 15, 2026, the Situational Awareness fund, helmed by former OpenAI researcher Leopold Aschenbrenner, filed its quarterly 13F with the SEC. The snapshot, dated June 30, revealed a jaw-dropping $20.24 billion portfolio. But by the time the filing hit the public, the fund had already been gutted by a July liquidation crisis, with Citadel stepping in to take over its "problematic" positions. This isn't just a story of a blown-up hedge fund; it's a masterclass in how high conviction, concentrated leverage, and a narrative-driven bet on AI infrastructure can implode in slow motion. Let me rewind the context. The Situational Awareness fund was never a typical crypto fund. It was a thematic vehicle that placed a massive bet on the physical layer of AI—storage chips, power, and data centers. Its name itself, borrowed from Aschenbrenner's famous essay, signaled a worldview: that compute is the new currency of geopolitics, and whoever controls the hardware controls the future. But what the 13F revealed was a portfolio that looked less like a technology fund and more like a concentrated bet on a single industrial bottleneck. And that bottleneck, as I've seen in my own audits of AI supply chains, is both real and fragile. At the core of the filing is a staggering 55.5% allocation to just two storage chip makers: SanDisk (28.0%) and Micron (27.5%). That's over $11 billion tied to a single subsector of the AI supply chain. The rest of the top holdings paint a similar picture: Bloom Energy (9.4%) for fuel cells and power, Taiwan Semiconductor ADR (6.2%) for chip fabrication, and two cloud compute providers—Nebius and CoreWeave—together accounting for about 9.8%. Then there's the tail: a collection of Bitcoin mining companies that have pivoted into AI data centers, including Core Scientific, Applied Digital, IREN, Riot Platforms, and CleanSpark. The top seven holdings alone represent 84.3% of the portfolio. This is not diversification; it's a vertical integration of the AI compute stack, from sand to socket. Let me break down the technical logic, because it's worth understanding even if it failed. The thesis is that the bottleneck for AI scaling has shifted from pure GPU supply to the physical infrastructure around it: high-bandwidth memory (HBM), advanced packaging, and power. Micron and SanDisk (now a division of Western Digital) are the primary suppliers of HBM and NAND flash for AI servers. Bloom Energy provides fuel cells that can power data centers with less strain on the grid. CoreWeave and Nebius are GPU cloud providers that buy massive amounts of both storage and power. And the Bitcoin miners? They sit on physical assets—power purchase agreements, substations, and existing data center shells—that can be retrofitted to host AI workloads. The logic is elegant, on paper. But here's where the contrarian angle cuts in. The portfolio's concentration created a non-linear downside when the AI trade started to wobble in July 2026. As AI stocks corrected, the fund faced margin calls. The 13F doesn't show leverage, but we know from market reports that the fund was forced to sell most of its public holdings under pressure. My own experience auditing leveraged positions in 2020 taught me that when a portfolio is 84% concentrated in seven names, and those names are all in the same thematic basket, there is no diversification benefit. The correlation among them during a drawdown approaches 1.0. The fund didn't just lose money; it became a forced seller of everything, amplifying the very downturn it was trying to profit from. The Bitcoin miners in the portfolio added another layer of fragility. Stocks like Core Scientific and IREN are small-cap, high-volatility names. When the fund needed to liquidate, these positions likely suffered the worst price impact. More importantly, the miners' transition to AI hosting is still in its early stages. They are not yet generating stable, predictable cash flows from AI contracts. By including them, the fund doubled down on the AI narrative without any hedge against Bitcoin price risk or miner operational risk. One could argue that the miners were actually a bet on "power connectivity" rather than on crypto, but the market didn't treat them that way during the sell-off. Now, let's talk about the governance angle. Leopold Aschenbrenner is a brilliant AI researcher, but he's not a traditional hedge fund manager. His background is in AI safety and policy, not portfolio risk management. The 13F reveals a portfolio that is almost entirely long, with no apparent short positions or hedging instruments. The fund's collapse was not due to fraud or misrepresentation—it was a straightforward case of concentrated leverage meeting a narrative-driven market correction. Citadel's takeover suggests that the positions were transferred in a structured settlement, possibly via total return swaps, rather than a simple margin call. This is a pattern we saw in the 2022 crypto credit crisis: smart money with high conviction, limited risk management, and a narrative that works until it doesn't. What does this mean for the crypto and AI markets going forward? The 13F is a post-mortem, but it carries a forward-looking signal. The market has now learned that the AI infrastructure trade can deleverage just as violently as crypto. The exit of such a concentrated player will likely tighten funding conditions for similar thematic funds. For the Bitcoin miners involved, the overhang of liquidated positions may not be fully absorbed yet. And for the broader narrative, this event serves as a reminder that we are still in a bear market for risk assets—where survival matters more than conviction. We didn't need the 13F to know the fund blew up. But we did need it to understand the anatomy of the explosion. The next time you see a portfolio with 55% in two names, ask yourself: is this conviction or complacency? In a bear market, the difference is measured in losses.