In the cold light of an SEC filing, I found the ghost of the architect.
On August 15, 2026, Situational Awareness LP, the fund helmed by Leopold Aschenbrenner—a name that echoes in the corridors of AI existential risk and geopolitical strategy—filed its quarterly 13F. The document revealed a radical shift by the end of Q2: a portfolio that had once balanced long-short positions with a hedge of put options had been transformed into a concentrated, almost monolithic long bet on the physical infrastructure of artificial intelligence. The filing was not a strategy update; it was a confession of conviction.
Micron Technology's holdings surged from $5.86 million to $5.574 billion. SanDisk's stake rose from $724 million to $5.674 billion. Together, these two memory chip giants accounted for 55% of the fund's publicly disclosed equity portfolio. Bloom Energy, TSMC ADR, Nebius, and a constellation of Bitcoin mining and AI data center operators—CoreWeave, Core Scientific, Applied Digital, IREN, Riot—filled the remaining weight. The put options that had protected the fund in Q1 against SMH, NVIDIA, Broadcom, AMD, Oracle, Micron, and TSMC? They were slashed to near zero. The hedge had been removed. The bet was naked.
As a Web3 Research Partner who has spent years auditing the narratives of protocols, I see this filing as a signal from the edge of a different kind of network—the financial network that underpins the AI arms race. This is not a story about a hedge fund manager's risk appetite. It is a story about the narrative of scarcity, the architecture of conviction, and the fragility of a market that is beginning to price in the physical limits of compute.
Context: The Narrative of Compute Scarcity
The AI boom of 2024-2026 taught investors one thing: the bottleneck is not code, but silicon. Training frontier models requires clusters of H100s, GB200s, and the next generation of accelerators. But the real constraint is not just the GPUs—it's the memory, the power, the cooling, the physical footprint. When Aschenbrenner, fresh from his departure from OpenAI and his manifesto on superintelligence, launched Situational Awareness LP, the market expected a focus on AI safety research. Instead, the fund's 13F reveals a bet on the raw materials of intelligence: memory chips, power cells, and mining rigs.
Why memory chips? Micron and SanDisk are the arteries of the AI data center. Every model inference requires bandwidth and capacity. As models scale, so does the demand for HBM (High Bandwidth Memory) and NAND flash. The narrative is seductive: AI is insatiable, and memory is the bottleneck within the bottleneck. Aschenbrenner's fund is not just buying chip stocks; it is buying a thesis that the physical world cannot keep up with the digital demand.
But there is a deeper layer. The fund's holdings in CoreWeave, Core Scientific, Applied Digital, IREN, and Riot are not just AI infrastructure plays. CoreWeave and Core Scientific are major Bitcoin miners that have pivoted to AI compute. Applied Digital and IREN operate data centers that can be repurposed for both mining and AI inference. Riot is a pure Bitcoin miner. In a bull market for crypto, these stocks become leveraged plays on energy and compute. By holding them alongside memory chip giants, Aschenbrenner is constructing a narrative that AI and crypto are not separate—they are competing for the same finite resources: electricity, chips, and cooling. The portfolio is a bet on a global compute shortage that will drive prices for all forms of computation upward.
Core: The Mechanism of Concentrated Conviction
To understand the risk, we must look at the mechanics of the filing. The 13F only shows long equity positions; it does not reveal leverage, derivatives, or margin. But the concentration itself is a signal. When a fund places 55% of its disclosed portfolio in two stocks—Micron and SanDisk—it is not making a diversified bet. It is making a statement.
During my time auditing smart contracts for the failed Project Aether in Zurich, I learned that technical correctness is not enough if the narrative trust is broken. Here, the technical correctness of the bet is that memory chip demand will outstrip supply. But the narrative trust is fragile. The memory chip industry is cyclical, and the current upcycle is driven by AI hype. If the hype falters, the correction is not gradual—it is violent.
From the two 13F filings, we can infer that the fund's Q2 positions were taken at elevated prices. The Philadelphia Semiconductor Index (SOX) had been rallying on AI optimism. But since July, the AI chip and storage sectors have faced multiple sell-offs. Micron, SanDisk, SK Hynix—all under pressure. The SOX recorded a rare monthly decline. The fund's portfolio, which had no puts, was exposed to the full drawdown. As August began, cooling inflation data and a recovery in AI earnings sentiment brought a rebound, but the damage was already done.
My own analysis of on-chain transaction patterns during the DeFi liquidity crisis of 2020 taught me a lesson: when the pool empties, only the intent remains. In traditional finance, the pool is liquidity. A concentrated portfolio with no hedges is a pool that can empty quickly if a single position triggers a margin call. The fund's holdings in Micron and SanDisk, combined with leveraged positions in mining stocks, create a correlation matrix that amplifies drawdowns. A drop in memory chip stocks could force liquidation of mining stocks, which in turn would pressure the entire portfolio.
Aschenbrenner's fund is not alone. Many AI-focused funds have piled into similar names. The narrative of compute scarcity is shared across Wall Street. But the concentration of conviction in this filing is extreme. It is the kind of bet that either makes history or breaks the fund.
Contrarian: The Hedgeless Bet and the Blind Spot of Narrative
The conventional wisdom would say that Aschenbrenner's removal of puts is a sign of supreme confidence. He is betting that the AI narrative is not just a bubble but a long-term structural shift. The contrarian view, however, is that the removal of hedges itself is a narrative signal—a sign that the fund's conviction has become a liability.
In my work as a Research Partner, I have seen how institutional investors often use hedges not just to protect against downside, but to maintain emotional discipline. When you hold a put, you are forced to periodically reassess the thesis. When you remove the put, you are free to ignore counter-signals. This is the same psychological trap that led to the collapse of Three Arrows Capital: overconfidence in a single narrative, coupled with leverage.
The fund's holdings in Bitcoin miners are particularly interesting. CoreWeave, Core Scientific, Applied Digital—these are companies that have been volatile due to the crypto cycle. In a bull market for crypto, they benefit from rising Bitcoin prices and AI demand. But in a bear market, they face a double headwind: falling crypto revenue and potential overcapacity in AI compute. The narrative that AI and crypto are complementary is not wrong, but it is fragile. If the AI narrative cracks, the mining stocks will fall faster than the memory chips, because they have less institutional support.
Moreover, the fund's bet on memory chips is a bet on a specific technology cycle. HBM is in high demand now, but new manufacturing capacity is coming online. Micron and Samsung are investing heavily. If supply catches up, the margin compression could be severe. The narrative of scarcity is a temporary truth; the narrative of abundance is always lurking.
Takeaway: The Next Narrative Shift
The Situational Awareness LP filing is a snapshot of a moment when the AI narrative is at its peak of conviction. But the market is already pricing in a correction. The question is not whether the bet will pay off, but what happens when the narrative shifts. The fund's portfolio is a seismograph of the next fault line: the transition from compute scarcity to compute abundance.
As I wrote in my private essays during the bear market solitude, the spiritual bankruptcy of speculative finance is that it treats conviction as a virtue without examining the foundation. The audit of this portfolio is not a check; it is a confession. The confession is that the AI narrative has become a religion, and Aschenbrenner is its high priest.
To own a piece of this bet is to inherit its narrative, but also its fragility. The next move in the market will be determined by whether the narrative of scarcity holds or breaks. And if it breaks, the liquidity crisis will be swift.
When the pool empties, only the intent remains. The intent here is clear: to bet everything on the physical limits of intelligence. But the physical limits may be more elastic than the market believes.